Variability of the human gut microbiome across populations and over time

The human gut microbiota is seeded during birth and mainly develops over the first 3 years of life⁽ Reference Koenig, Spor and Scalfone ^{24 )}. From birth, neonates are exposed to microbes from a variety of sources, and the initial colonisation of their guts depends on the microbes first encountered. The initial composition of the gut microbiota depends on the mode of delivery: babies delivered vaginally harbour gut microbiota resembling microbial communities found in their mothers' vaginas, whereas those born via Cesarean section apparently acquire microbes from the skin, dominated by taxa such as Propionibacterium and Staphylococcus ⁽ Reference Dominguez-Bello, Costello and Contreras ^{25 )}. The feeding mode also influences the infant gut microbiota. Breast milk contains nutrients, maternal antibodies and also diverse commensal maternal bacteria including bifidobacteria and lactobacilli⁽ Reference Jost, Lacroix and Braegger ^{26 )}. Compared with formula-fed infants, breast-fed infants have lower levels of Atopobium and higher levels of Bifidobacterium ⁽ Reference Bezirtzoglou, Tsiotsias and Welling ^{27 )}. The diversity in the baby's gut is initially low and increases during development. Over the first 3 years of life, the composition of the microbial community becomes more adult-like, and major microbial shifts are associated with key events such as the introduction of solid food⁽ Reference Koenig, Spor and Scalfone ^{24 )}.

Although the adult gut microbiota are relatively stable within a person (in that serial samples from the same person typically resemble each other more closely than do samples from different people) and are typically dominated by members of the Bacteroidetes, Firmicutes and Actinobacteria, they show great interpersonal and intrapersonal variability. For example, fewer than 50 % of bacterial taxa are shared between monozygotic twins at the species level⁽ Reference Turnbaugh, Quince and Faith ^{28 )}. We are starting to appreciate this variability to a greater extent because of the numerous studies facilitated by the methods developed in large-scale projects such as the Human Microbiome Project⁽ Reference Turnbaugh, Ley and Hamady ^{7 , 8 )} and the Metagenomics of the Human Intestinal Tract (MetaHIT)⁽ Reference Ehrlich and Nelson ^{29 )}. It has been argued that the gut microbiota can be classified into three broad clusters, or ‘enterotypes’, based on the dominant presence of Bacteroides, Prevotella or Ruminococcus genera⁽ Reference Arumugam, Raes and Pelletier ^{30 )}, although statistical support for the three-cluster model is weak and has generally not been replicated in later studies. The idea is controversial, especially because the model of discrete clusters has a clear alternative in terms of continuous gradients in the major taxa that are well supported⁽ Reference Wu, Chen and Hoffmann ^{31 –} Reference Koren, Knights and Gonzalez ^{34 )}. Intriguingly, despite the considerable variability in the composition of the gut microbiota, metagenomic shotgun sequencing demonstrates that these diverse communities share a core set of gene functions in the microbiome⁽ Reference Turnbaugh, Hamady and Yatsunenko ^{35 )}.

The gut microbial community changes with age. In cross-sectional studies, we see an apparently smooth change in the gut microbiota in the first 3 years of life, followed by relatively subtle changes thereafter. Although few studies have specifically examined the gut microbiota in old age, recent studies on elderly population confirm that they harbour distinct gut microbial communities from younger adults⁽ Reference Claesson, Cusack and O'Sullivan ^{36 )}, and the composition of the microbiota in the elderly subjects correlates significantly with measures of frailty, co-morbidity and nutrition⁽ Reference Claesson, Jeffery and Conde ^{37 )}. Lachnospiraceae and microbial genes involved in the production of SCFA were enriched in elderly subjects living in the community compared with those in long-term care. Another study performing functional profiling of the gut microbiota found an increase in proteolytic function and a decrease in saccharolytic function, in addition to the loss of genes for SCFA production associated with ageing⁽ Reference Rampelli, Candela and Turroni ^{38 )}. However, more studies, and especially prospective longitudinal studies, are needed to clarify how diet and the gut microbiota affect the ageing process and how the gut microbiota are in turn affected by aging itself.

Understanding the variability of the microbiota between people and over time is a prerequisite to discriminating disease states from normal perturbations. Caporaso et al. ⁽ Reference Caporaso, Lauber and Costello ^{39 )} densely sampled microbiota from three body sites (tongue, gut and skin) of two healthy subjects for 6 and 15 months. This time series shows the remarkable temporal dynamics of the gut microbiota of individuals. Although the difference between gut samples from an individual is smaller than those from different individuals or different body sites, there seems to be no core set of taxa present at all time points. However, Faith et al. ⁽ Reference Faith, Guruge and Charbonneau ^{40 )} found that in faecal samples sequenced much more deeply, on an average 60 % of strains in the gut can persist over a long period up to 5 years when only a few time points are evaluated, suggesting that either the dynamic range of abundance of particular taxa is large (requiring very deep sequencing to recover them) or strains can be lost but recolonise the host. This temporal consistency assumes that other factors that could affect the microbiota, such as diet and disease status, remain relatively constant.

Most studies of the gut microbiota have been focused on Western populations in Europe, the USA and Canada. It is important to expand these studies to non-Western diet populations in order to fully understand the range of variation of the gut microbiota and how gut microbes have co-evolved with humans. For example, the Japanese population has a unique gene coding for the enzyme porphyranase in the gut bacterium Bacteroides plebeius, potentially transferred from marine Bacteroides spp. that naturally degrade seaweed, to facilitate the digestion of this component of the Japanese, but not, historically, the US diet⁽ Reference Hehemann, Correc and Barbeyron ^{41 )}. Gut microbiota of children in a rural African village in Burkina Faso showed an increase in Bacteroidetes and a decrease in Firmicutes relative to European children⁽ Reference De Filippo, Cavalieri and Paola ^{12 )}. This pattern was also found in a broader cross-cultural comparison of the gut microbiota of people living in the Venezuelan Amazon, rural Malawi and the USA, suggesting that carbohydrate-rich diets increase the amount of Prevotella ⁽ Reference Yatsunenko, Rey and Manary ^{2 )}. The Hadza hunter-gatherers in Tanzania also harbour unique compositions of microbiota, presumably as a result of their foraging lifestyle⁽ Reference Schnorr, Candela and Rampelli ^{42 )}. All these studies demonstrate the heterogeneity of gut microbiota across geographically and culturally diverse populations, although to date not enough populations have been studied to resolve which of these differences are due to diet, which to host genetics and which to environmental exposures, all of which differ among these populations.

Dietary effects on the gut microbiome

The quantitative contributions of host genetics, environmental factors and diet to shaping the gut microbiota remain largely unknown. Significant associations between host genotypes and their gut microbiota composition have been reported in both human and mouse studies (see Spor et al. ⁽ Reference Spor, Koren and Ley ^{43 )} for a comprehensive recent review). The environment also plays an important role in establishing and modifying the gut microbiota. For example, the close physical contact between humans and companion animals facilitates the acquisition and exchange of microbes. A study of skin, oral and gut microbiomes showed that the microbial communities of cohabiting family members resemble each other and that adults share more microbial taxa with their own dogs than they do with others' dogs⁽ Reference Song, Lauber and Costello ^{44 )}. Having a dog, and the resulting exposure to a diverse microbial community, can be beneficial. Pet ownership in the early, but not later, years of life⁽ Reference Wegienka, Johnson and Havstad ^{45 )} is associated with a significant decrease in the risk of allergic disease⁽ Reference Fujimura, Johnson and Ownby ^{46 )}.

As discussed earlier, dietary effects on the gut microbiota and health are often confounded by variation in host genotypes and environmental exposures. Nevertheless, accumulating evidence has suggested that long-term diet is a primary driver of the gut microbiota. As an extreme example, two co-evolution studies of mammals and their gut microbiota has found that both gut microbiota composition and functions are adapted to their diet (herbivorous, carnivorous and omnivorous)⁽ Reference Ley, Hamady and Lozupone ^{47 ,} Reference Muegge, Kuczynski and Knights ^{48 )}. In another more recent study, the convergently evolved composition of the gut microbiome in ant-eating mammals, despite their phylogenetic diversity, strongly suggests that diet is the major force shaping their microbiota⁽ Reference Delsuc, Metcalf and Wegener Parfrey ^{49 )}. Wu et al. ⁽ Reference Wu, Chen and Hoffmann ^{31 )} showed that the overall structure of the human gut microbiota, at least in a Western population, seems to be driven mainly by long-term dietary effects: Prevotella was enriched in people with high-fibre and high-carbohydrate diets, whereas Bacteroides was associated with a typical ‘Western’ diet high in protein and fat. The studies highlighted in the previous section also suggest that differences between the gut microbial communities of rural populations and Westerners are largely linked to differences in diet⁽ Reference Yatsunenko, Rey and Manary ^{2 ,} Reference De Filippo, Cavalieri and Paola ^{12 ,} Reference Schnorr, Candela and Rampelli ^{42 )}.

Fortunately, we can examine dietary effects more directly with techniques that constrain other interacting factors. Studies of the faecal microbiomes of human monozygotic and dizygotic twins have been helpful in identifying the effects of diet on gut microbiota controlling for host genetics⁽ Reference Smith, Yatsunenko and Manary ^{50 )}. In addition, gnotobiotic mice (initially germ-free mice colonised with either a defined set of microbial strains or with a donor faecal sample) provides a well-defined, powerful animal model of the human gut ecosystem to elucidate the effects of nutrients on the microbiome, including taxonomic composition, microbial and host gene expression, and, ultimately, host physiological status in response to controlled changes in diet⁽ Reference Turnbaugh, Ley and Mahowald ^{15 ,} Reference Turnbaugh, Ridaura and Faith ^{51 )}. Using the gnotobiotic mouse model, a recent study found that obese or lean phenotypes can be transmitted to germ-free mice by inoculating them with the gut microbiota from human twins discordant for obesity and that the obese phenotype of the mouse transplanted with the obese microbiota can be rescued by the invasion of a particular species of Bacteroidetes from the mouse transplanted with lean microbiota when they are co-housed⁽ Reference Ridaura, Faith and Rey ^{14 )}. This rescue and invasion is diminished when the mice are fed with a diet high in saturated fats and low in vegetables and fruit, confirming the dependency of the gut microbiota on diet.

Arguably, diet is among the most easily controlled factors that can potentially manipulate the gut microbiota. Studying the resilience of the microbiota and its patterns of change, during and after dietary intervention, could allow the design of effective nutrition therapy. Ley et al. ⁽ Reference Ley, Turnbaugh and Klein ^{52 )} have shown that the ratio of Bacteroidetes:Firmicutes increases in the gut microbiota of people consuming either a fat-restricted or a carbohydrate-restricted low-energy diet for a year and that people who lost more weight showed larger changes in the ratios of these taxa. Two other short-term controlled feeding experiments found that the changes in microbiota composition can be rapid (within 24 h), reproducible and can overwhelm interpersonal variability with a sufficiently extreme diet⁽ Reference Wu, Chen and Hoffmann ^{31 ,} Reference David, Maurice and Carmody ^{53 )}. A dietary intervention study of obese or overweight individuals found that the gene richness (the number of gut microbial genes) is increased and the metabolic status is improved in individuals with low gut microbial gene richness, after a 6-week low-energy, high-protein diet followed by another 6-week weight maintenance diet⁽ Reference Cotillard, Kennedy and Kong ^{54 )}, but this improvement is very limited for individuals with already high gut microbial gene richness. Dietary interventions in malnourished humans are not always successful, and the variation may be explained by individual differences in the gut microbiota. Studies of Malawian and Bangladeshi subjects with severe acute malnutrition have shown that their gut microbiota are resilient and often regress back to the state before the food intervention⁽ Reference Smith, Yatsunenko and Manary ^{50 ,} Reference Subramanian, Huq and Yatsunenko ^{55 )}, suggesting that longer intervention or a repair of the gut microbiota is needed to effectively treat malnutrition in these cases.