In the past 10 years, there has been a wealth of studies in which the relationship between the human gut microbiota and human health has been investigated. Moreover, recently there have been several human health-related microbiota studies with partly contradictory results regarding e.g. obesity-related microbiota and abundance of bifidobacteria in the faecal microbiota of babies. As it is likely that at least some of the differences may be explained by the methodology applied, it is of utmost importance that when reading articles related to human gut microbiota studies the most important confounding factors are known.
Human gut microbiota
In an adult human individual, resident bacteria outnumber human cells by a factor of ten; each adult harbours on average 1013 mammalian cells and 1014 microbial cells( Reference Savage 1 ). Most of the microbes, typically 1011–1012 microbes/g, can be found in faeces and from the large intestine( Reference Finegold, Sutter, Mathisen and Hentges 2 – Reference Harmsen, Raangs and He 4 ), which is considered to be a complex fermentor with a metabolic potential to rival that of the liver( Reference Marchesi and Shanahan 5 ). The environmental determinants, namely temperature, pH, redox potential, atmospheric composition, water activity, salinity and light, within each region of the human oro-gastrointestinal (GI) tract are very different, and therefore each region has its own distinctive microbiota( Reference Wilson 6 ). Since digestive enzymes are not secreted by the mucosa of the large intestine, further breakdown of dietary constituents is carried out by the resident microbiota( Reference Wilson 6 ). Carbohydrates are mainly fermented in the proximal colon, whereas the fermentation of proteins takes place mainly in the distal colon. Transit time of digesta through the colon strongly influences the activities of the gut microbiota. The mean transit time of the oro-GI tract has been reported to be approximately 70 h in UK people consuming a normal daily diet( Reference Cummings and Macfarlane 7 , Reference Stevens, VanSoest and Robertson 8 ). The primary activity of the caecum and colon microbiota is the breakdown of carbohydrates not digested in the ileum to SCFA, which are then rapidly absorbed. The principal products of carbohydrate fermentation are SCFA (acetate, propionate and butyrate), hydrogen and CO2 gasses and bacterial cell mass (biomass)( Reference Cummings and Macfarlane 9 ). The amount of energy derived from SCFA accounts for up to 10 % of the total energy requirement of human subjects( Reference McNeil 10 ). From a nutritional point of view, the SCFA are important since they not only provide the body with energy but are also metabolised in different tissues( Reference Cummings and Macfarlane 9 ).
The microbiota in the colon and faeces is extremely diverse and on the basis of estimations from culture-based and molecular studies more than 1200 prevalent bacterial species altogether reside there. Each individual harbours at least 160 such species( Reference Rajilić-Stojanović, Smidt and De Vos 11 , Reference Qin, Li and Raes 12 ). Under normal circumstances, predominant intestinal microbiota of an adult individual is fairly stable. However, in studies where the long-term temporal stability of the predominant microbiota has been assessed from healthy subjects, the number of subjects has been limited( Reference Zoetendal, Akkermans and De Vos 13 – Reference Seksik, Rigottier-Gois and Gramet 16 ). The human GI-tract, although harbouring a vast number of microbes, has only a limited diversity at the phylum level. Microbes from seven bacterial phyla (Firmicutes, Bacteroidetes, Actinobacteria, Proteobacteria, Fusobacteria, Verromicrobia and Cyanobacteria-like) and one archael phylum (Euryarchaeota) have been detected in the human intestine( Reference Eckburg, Bik and Bernstein 17 – Reference Turnbaugh, Hamady and Yatsunenko 21 ). However, the majority of the GI-tract population are representatives of three phyla: the Firmicutes (Families Lachnospiraceae and Ruminococcaceae), the Bacteroidetes (Bacteroidaceae, Prevotellaceae and Rikenellaceae) and the Actinobacteria (Bifidobacteriaceae and Coriobacteriaceae).
Faecal microbiota is dominated by bacteria from Families Lachnospiraceae (Eubacterium rectale group) and Ruminococcaceae (Clostridium leptum group) and Bacteroidaceae/Prevotellaceae( Reference Franks, Harmsen and Raangs 3 , Reference Harmsen, Raangs and He 4 , Reference Eckburg, Bik and Bernstein 17 , Reference Suau, Bonnet and Sutren 18 , Reference Sghir, Gramet and Suau 22 ), which together account for 60–80 % of the faecal bacteria of healthy adults( Reference Sghir, Gramet and Suau 22 – Reference Mueller, Saunier and Hanisch 24 ). The bacteria belonging to the Family Lachnospiraceae comprise on average 10–45 % of the total faecal bacteria as detected with quantitative hybridisation-based methods( Reference Franks, Harmsen and Raangs 3 , Reference Harmsen, Raangs and He 4 , Reference Sghir, Gramet and Suau 22 – Reference Marteau, Pochart and Doré 30 ) and the bacteria within Ruminococcaceae 16–27 %, thus co-dominating with the bacteria within Lachnospiraceae( Reference Sghir, Gramet and Suau 22 – Reference Maukonen, Satokari and Mättö 25 , Reference Marteau, Pochart and Doré 30 – Reference Saunier, Rouge and Lay 32 ). Many members of the Lachnospiraceae and Ruminococcaceae are polysaccharolytic and produce butyrate. In addition, they are obligately anaerobic. Members of the Bacteroidetes phylum comprise 12–60 % of the total faecal bacteria depending on which method has been used for the quantification of this group( Reference Sghir, Gramet and Suau 22 – Reference Mueller, Saunier and Hanisch 24 , Reference Doré, Schir and Hannequart-Gramet 33 , Reference Rigottier-Gois, Rochet and Garrec 34 ). Bacteroides spp. are saccharolytic bacteria that contain a wealth of polysaccharide-degrading enzymes. They have an excellent ability to ferment simple and complex sugars and polysaccharides, producing acetate and succinate as major metabolic end products. In addition, most species are weakly proteolytic( Reference Shah, Balows, Trüper, Dworkin, Harder and Schleifer 35 ). Most of the bacteria within Prevotellaceae and Porphyromonadaceae have been isolated from the oral cavity and are commonly found in oral sites in culture-based studies, whereas with DNA-based molecular studies members of Prevotellaceae and Porphyromonadaceae have also been found in faeces.
Bifidobacterium spp., Lactobacillus-group and bacteria within the Family Coriobacteriaceae although found in most human subjects, comprise smaller populations among faecal bacteria in adults (bifidobacteria 0·5–6 %, lactobacilli <1–3 % and Coriobacteriaceae 1–5 % of the total faecal bacteria)( Reference Franks, Harmsen and Raangs 3 – Reference Harmsen, Raangs and He 4 , Reference Sghir, Gramet and Suau 22 – Reference Mueller, Saunier and Hanisch 24 , Reference Marteau, Pochart and Doré 30 ). Moreover, there are several bacteria/bacterial groups, which usually comprise <1–2 % of the total faecal microbiota of healthy adults, such as Akkermancia muciniphila, ‘true clostridia’ (Family Clostridiaceae), enterobacteria, streptococci, bacteria within the Families Peptostreptococcaceae, Erysipelotrichaceae, Veillonellaceae, Eubacteriaceae, in addition to sulphate-reducing bacteria and/or methanogens( Reference Franks, Harmsen and Raangs 3 , Reference Sghir, Gramet and Suau 22 – Reference Lay, Rigottier-Gois and Holmstrøm 23 , Reference Rigottier-Gois, Le Bourhis and Gramet 27 , Reference Collado, Derrien and Isolauri 36 – Reference Gibson, Macfarlane and Cummings 39 ). Besides bacteria, low levels of viruses, archaea (<1 %) and eukaryotes (≤2 %) are also found from the large intestine( Reference Gibson, Macfarlane and Cummings 39 – Reference Doré, Gramet and Goderel 41 ).
The effect of sampling and used analysis methods on the results obtained from gastrointestinal-tract samples
Conditions during sample transportation have a major impact on sample quality( Reference Roesch, Casella and Simell 42 – Reference Molbak, Sommer and Johnsen 44 ) and therefore the time between sampling and further processing and storing should be limited to the minimum. It should be kept in mind that regardless of how sophisticated the applied method for characterisation of the microbial community is, the result can be only as good as the sample that arrives in the laboratory. In addition, when molecular techniques are applied to study microbial communities of the GI tract, oral bacteria may be detected with molecular techniques from faeces as well, since the human oro-GI tract is an open system. In studies with snapshot samples, it is difficult to confer that the detected bacteria are resident and not just transient. Therefore to be able to establish resident microbiome, follow-up samples would be needed.
Most of the steps in the processing of faecal samples may affect the microbiological results obtained. The sampling and storage conditions affect the results( Reference Roesch, Casella and Simell 42 – Reference Molbak, Sommer and Johnsen 44 ), although not as extensively as the following steps, especially the DNA extraction( Reference Kennedy, Walker and Berry 45 – Reference Salonen, Nikkilä and Jalanka-Tuovinen 48 ). One week storage at −20 °C decreases the detected numbers of Gram-negative Bacteroides spp. by over one logarithmic unit as compared with the fresh samples, whereas the storage temperature does not have such a significant effect on the Gram-positive bacteria (e.g. Firmicutes and Actinobacteria)( Reference Maukonen, Simões and Saarela 47 ). Since the analysis of fresh faecal samples, especially in the case of large cohorts, is impractical and mostly also impossible, the effect of the storage conditions on the results obtained should be taken into account when the results are discussed.
When DNA extraction is performed using solely chemical treatments combined with heat treatments, the results for some bacterial groups may be as much as four logarithmic units lower as compared with a DNA extraction method in which rigorous mechanical disruption is applied( Reference Maukonen, Simões and Saarela 47 ). Actinobacteria, in particular, are underestimated with many currently used techniques, most probably due to their high guanine–cytosine content( Reference Krogius-Kurikka, Kassinen and Paulin 49 ). In a study of Nakamura et al.( Reference Nakamura, Gaskins and Collier 50 ) it was shown that when the same samples were examined with quantitative PCR (enzymatic DNA extraction) and fluorescent in situ hybridisation (FISH), the difference in results was enormous. The proportions of bifidobacteria with quantitative PCR was 0·1–1·7 %, whereas the proportion with FISH was 28–84 %( Reference Nakamura, Gaskins and Collier 50 ). Moreover, the number of bacteria within Family Coriobacteriaceae has been shown to be 5–6 logarithmic units higher in a sample that was initially stored at +4 °C for 2 d and thereafter at −70 °C and when DNA extraction was performed mechanically than on the same sample that was stored at −70 °C and the DNA extraction was performed enzymatically( Reference Maukonen, Simões and Saarela 47 ). This may also explain why in some studies members of Coriobacteriaceae are not considered to be the part of the normal dominant microbiota( Reference Eckburg, Bik and Bernstein 17 ), whereas in others, especially those conducted with FISH, bacteria within Coriobacteriaceae constitute 1–8 % of the total population of the human gut microbiota( Reference Lay, Rigottier-Gois and Holmstrøm 23 , Reference Mueller, Saunier and Hanisch 24 , Reference Harmsen, Wildeboer-Veloo and Grijpstra 38 , Reference Matsuki, Watanabe and Fujimoto 51 ).
Another methodological aspect that may cause confusion is that different probes/primers target different bacterial populations, such as in the detection of Bacteroidetes and Firmicutes. In metagenomic/16S rRNA-based next-generation sequencing studies the authors usually refer to Bacteroidetes, whereas in studies conducted with FISH/PCR, the studied bacterial population is usually narrower (e.g. Bacteroides spp. or Bacteroides-Prevotella group)( Reference Hoyles and McCartney 52 ). The same controversy applies to Firmicutes. In many next-generation sequencing studies, Firmicutes are discussed as a single group, whereas in fact it contains several different bacterial classes e.g. Bacilli, Clostridia, Erysipelotrichia, which all have different metabolic properties. Furthermore, when PCR/FISH is used, each of these groups is usually studied separately. Moreover, since the phylum Bacteroidetes is more thoroughly characterised than the phylum Firmicutes, it is possible that the representation of Bacteroidetes in metagenomic studies is overestimated.
When microbial communities are examined, especially using several different techniques and/or next-generation sequencing, the quantity of data generated is enormous. Therefore the use of suitable algorithms e.g. measurement of microbial diversity and correct interpretation of the results are of utmost importance( Reference Walker, Duncan and Louis 53 – Reference Achtman and Wagner 61 ). It should also be borne in mind that even if statistical significance is achieved, the observed changes may not necessarily be of biological significance.
Factors that affect the gut microbiota
There are several factors that affect the composition of the human GI microbiota, such as genetics, sex, ethnicity, age, medication, diseases/disorders and last but not least the diet.
Genetics of the host and geography
Twins, especially monozygotic (MZ) twins, have been reported to have more similar interindividual faecal microbiota than unrelated people. In addition, twins and their mothers have a more similar microbiota than unrelated individuals. These findings have led to the conclusion that the host genotype affects the development of the gut microbiota and gut bacterial composition( Reference Zoetendal, Akkermans and Akkermans-van Vliet 14 , Reference Turnbaugh, Hamady and Yatsunenko 21 , Reference Turnbaugh, Quince and Faith 62 , Reference Stewart, Chadwick and Murray 63 ). However, the aforementioned studies have mainly been conducted with MZ pairs concordant for leanness or obesity. Simoes et al. ( Reference Simoes, Maukonen and Kaprio 64 ) studied MZ twins concordant and discordant for obesity, and found that the concordant normal-weight MZ twins had more similar bacterial populations than the MZ twins discordant for obesity. These findings also address the importance of the diet in addition to the genetic drivers( Reference Simoes, Maukonen and Kaprio 64 ). Although the genetics or the shared environmental factors during upbringing result in more similar bacterial populations, viromes have been shown to be unique to individuals regardless of their degree of genetic relatedness( Reference Reyes, Haynes and Hanson 40 ).
Besides genetics, the effect of the geographic origin (which may also include genetic differences and environmental sources of variation) has also been studied in relation to the gut microbiota composition. Even in Europe some differences may be found between different countries, e.g. proportions of bifidobacteria have been found to be 2- to 3-fold higher in an Italian adult study population than in other European study populations( Reference Lay, Rigottier-Gois and Holmstrøm 23 , Reference Mueller, Saunier and Hanisch 24 ). In addition, when 6-week-old infants across Europe were studied, geography was a more prominent factor than delivery mode, breast-feeding and antibiotics. In infants, children from Northern European countries had more bifidobacteria, whereas the Southern European infants were associated with more diverse microbiota and higher numbers of Bacteroides than the Northern European children( Reference Fallani, Young and Scott 65 ). In the studies involving people from different continents, the bacterial population differences have been more distinct. In a few recent studies, there have been consistent results that the European and North-American faecal microbiota differ significantly from the Southern-American, African and Chinese faecal microbiota( Reference Yatsunenko, Rey and Manary 66 – Reference De Filippo, Cavalieri and Di Paola 68 ). However, in all of these studies it is not possible to exclude the impact of other possible confounding factors, and therefore dietary habits and genetics may also contribute to the differences.
Bacterial colonisation of the infant GI tract is influenced by e.g. mode of delivery, prematurity, type of feeding (breast feeding v. formula feeding), antibiotic treatment of the child or the mother, lifestyle and geographics( Reference Fallani, Young and Scott 65 , Reference Fallani, Amarri and Uusijarvi 69 – Reference Bezirtzoglou, Tsiotsias and Welling 74 ). The earliest colonisers are usually facultative anaerobic bacteria such as Enterobacteriaceae, streptococci and staphylococci, whereas later colonisers tend to be strict anaerobes e.g. bifidobacteria, clostridia and Bacteroides spp. regardless of the infant's geographical origin and methods used for the detection( Reference Fallani, Young and Scott 65 , Reference Favier, De Vos and Akkermans 75 – Reference Wang, Ahrne and Antonsson 80 ). Immediately after birth, the rectal microbiota of vaginally delivered babies resembles their own mother's vaginal microbiota, whereas the rectal microbiota of babies delivered by Caesarean section resembles that of the skin( Reference Dominguez-Bello, Costello and Contreras 81 ). The gut microbiota of preterm infants is less diverse than those of full-term babies( Reference Rougé, Goldenberg and Ferraris 70 , Reference Schwiertz, Gruhl and Löbnitz 73 , Reference Magne, Abély and Boyer 82 – Reference Hall, Cole and Smith 84 ). There are numerous studies in which the predominance of bifidobacteria in exclusively breast-fed infants has been found( Reference Fallani, Young and Scott 65 , Reference Yatsunenko, Rey and Manary 66 , Reference Bezirtzoglou, Tsiotsias and Welling 74 , Reference Bullen, Tearle and Willis 85 – Reference Magne, Hachelaf and Suau 88 ). At age 3–6 weeks, exclusively breast-fed infants harbour higher numbers of bifidobacteria, whereas formula-fed babies have more diverse microbiota, lower numbers of bifidobacteria and higher numbers of Bacteroides, Lachnospiraceae, Lactobacillus group, Clostridium difficile and Coriobacteriaceae( Reference Fallani, Young and Scott 65 , Reference Penders, Thijs and Vink 71 , Reference Bezirtzoglou, Tsiotsias and Welling 74 , Reference Harmsen, Wildeboer-Veloo and Raangs 87 , Reference Penders, Vink and Driessen 89 , Reference Stark and Lee 90 ). By the end of the first year of life, when the child has already started to eat the same foods as the adults, the gut microbiota starts to converge towards a profile characteristic of the adult microbiota( Reference Palmer, Bik and DiGiulio 76 , Reference Stark and Lee 90 ). However, the faecal bacterial diversity is still lower. By the end of the second to third year, the phylogenetic composition evolves towards the adult-like composition( Reference Yatsunenko, Rey and Manary 66 , Reference Hentges 91 ).
The GI microbiota evolves with age( Reference Gorbach, Nahas and Lerner 92 ). Dental deterioration, salivary function, digestion, slower intestinal transit time and changes in diet and physical activity may affect the GI microbiota of ageing people. Interest, as well as the number of studies, in the GI microbiota of elderly people has grown as life expectancy in the Western world has rapidly increased. The elderly have been reported to have relatively stable microbiota( Reference Rajilic-Stojanovic, Heilig and Molenaar 93 – Reference Woodmansey, McMurdo and Macfarlane 95 ). However, the microbiota of the elderly has been reported to be more diverse and to contain partly different core microbiota as compared with younger adults( Reference Rajilic-Stojanovic, Heilig and Molenaar 93 , Reference Claesson, Cusack and O'Sullivan 94 , Reference Blaut, Collins and Welling 96 , Reference Maukonen, Mättö and Kajander 97 ). Moreover, inter-individual variation is greater in elderly people as compared with younger adults( Reference Finegold, Sutter, Mathisen and Hentges 2 , Reference Claesson, Cusack and O'Sullivan 94 , Reference Claesson, Jeffery and Conde 98 ).
In addition to dietary components, the other important external factor affecting the microbiota is antibiotic use. However, different types of antibiotics have different types of action mechanisms and thus different effects on the human microbiota( Reference Kohanski, Dwyer and Collins 99 , Reference Sullivan, Edlund and Nord 100 ). In addition, individual human responses may be different. Bifidobacteria are typically susceptible to the majority of clinically relevant antibiotics such as penicillins (β-lactam antibiotics), cephalosporins and macrolides( Reference Delgado, Flórez and Mayo 101 – Reference Huys, D'Haene and Cnockaert 106 ), and most of the commensal gut bacteria (e.g. bifidobacteria, Bacteroides spp.) to amoxicillin (β-lactam antibiotic) and clavulanate( Reference Jousimies-Somer, Summanen and Citron 102 ).
The effect of diet on human gut microbiota
The importance of the metabolic activities of the gut microbiota from the host's perspective
From the host's perspective, there are numerous activities of the commensal gut microbiota that are of great importance to health. Carbohydrates are mainly fermented in the proximal colon, whereas the fermentation of proteins takes place mainly in the distal colon. However, the metabolic output of the microbial community depends not only on available substrates, but also on the gut environment, with the pH playing a major role. For example, at pH 6·7 Bacteroides spp. predominate, whereas at pH 5·5 bacteria related to E. rectale predominate( Reference Louis, Scott and Duncan 107 – Reference Walker, Duncan and Carol McWilliam Leitch 109 ). The main saccharolytic genera in the human GI-tract are Bacteroides, Bifidobacterium, Clostridium, Eubacterium, Lactobacillus and Ruminococcus. The saccharolytic genera are able to produce SCFA, which have both local and systematic beneficial biological effects. A wide range of bacteria have proteolytic activities, such as clostridia, and species within genera Propionibacterium, Prevotella, Bifidobacterium and Bacteroides. Protein metabolism, however, is not as favourable to the host as carbohydrate metabolism; some of the end-products of amino acid metabolism may be deleterious to the host, e.g. ammonia, amines and phenol compounds. Some species of the genera Bacteroides, ruminococci and Akkermansia are able to break down mucin. Moreover, several Eubacterium spp. and Clostridium spp. are able to dehydroxylate bile acids, some Clostridium spp. transform conjugated bilirubin, Eubacterium coprostanoligenes is able to convert cholesterol to coprostanol and Bacteroides spp. inactivate tryptic activity, have dipeptidase activity and play a key role in the enterohepatic circulation of bile acids( Reference Cummings and Macfarlane 7 , Reference Shah, Balows, Trüper, Dworkin, Harder and Schleifer 35 , Reference Wade, Stackebrandt, Jones, Holzapfel, Reboli and Farrar 110 – Reference Derrien, Vaughan and Plugge 114 ).
In addition to the individual activities, cross-feeding between the gut microbes and the metabolic networks thus created are also of great importance. For example, it has been shown in vitro that lactate produced by Bifidobacterium adolescentis as a fermentation product from fructo-oligosaccharide (FOS) and starch was further utilised by butyrate producers, which were not able to grow solely on FOS and starch( Reference Belenguer, Duncan and Calder 115 ). In addition, Roseburia intestinalis and Anaerostipes caccae were able to grow with Bifidobacterium longum using FOS; R. intestinalis was able to grow on the FOS-supplemented medium when acetate, a major fermentation product of B. longum, was added to the medium, whereas A. caccae was able to utilise fructose that was released during the bifidobacterial fermentation of FOS( Reference Falony, Vlachou and Verbrugghe 116 ). Besides other survival mechanisms, gene transfer within and from outside the gut microbiota has also been shown to occur. In Japan, where consumption of marine algae is high, a Japanese gut bacterium (Bacteroides plebeius) has acquired genes coding for porphyromonases, agarases and associated proteins and thus the ability to utilise marine algae. These algae are not readily fermentable by Western gut microbiota( Reference Hehemann, Correc and Barbeyron 117 ).
Dietary interventions v. habitual diet
The main external factors that can affect the composition of the microbial community in generally healthy adults include major dietary changes and antibiotic therapy. Changes in some selected bacterial groups have been observed due to controlled changes to the normal diet e.g. high-protein diet( Reference Hentges, Maier and Burton 118 , Reference Russell, Gratz and Duncan 119 ), prebiotics( Reference Maukonen, Mättö and Kajander 97 , Reference Tuohy, Kolida and Lustenberger 120 – Reference Langlands, Hopkins and Coleman 122 ), probiotics( Reference Rinne, Kalliomaki and Salminen 123 – Reference Lyra, Krogius-Kurikka and Nikkila 125 ), weight-loss diet( Reference Ley, Turnbaugh and Klein 20 , Reference Duncan, Lobley and Holtrop 126 , Reference Nadal, Santacruz and Marcos 127 ) and berries( Reference Puupponen-Pimiä, Seppänen-Laakso and Nohynek 128 ). More specifically, changes in the type and quantity of non-digestible carbohydrates in the human diet influence both the metabolic products formed in the lower regions of the GI tract and levels of bacterial populations detected in faeces( Reference Duncan, Belenguer and Holtrop 129 ). The interactions between dietary factors, gut microbiota and host metabolism are important for maintaining homeostasis and health( Reference Flint, Duncan and Scott 130 ).
The impact of habitual diet on faecal microbiota has been studied for decades. In older culture-based studies, it was found that numbers of bacteroides were lower and numbers of enterococci and Escherichia coli higher in Ugandan, Indian and Japanese people on a high-carbohydrate diet as compared with people on a Western diet( Reference Drasar, Crowther and Goddard 131 , Reference Finegold, Attebery and Sutter 132 ). However, when English people on a strictly vegetarian diet were studied, their microbiota resembled those of people on a Western diet more than the microbiota of other vegetarian people from different continents( Reference Drasar, Crowther and Goddard 131 ). Similarly, numbers of bacteroides and clostridia were lower in the Nigerian Maguzawa tribal people (predominantly cereal diet) than in the other dietary groups( Reference Drasar, Montgomery and Tomkins 133 ). In more recent molecular studies, in which gut microbiota from different parts of the world has been compared, a significant correlation between habitual diet and faecal microbiota has also been found. African children consuming a diet low in fat and animal protein and rich in starch, fibre and plant polysaccharides (predominantly vegetarian) had significantly more Bacteroidetes and Actinobacteria and less Firmicutes and Proteobacteria than European children, who had a diet high in animal protein, fat, sugar and starch, but low in fibre. Moreover, members of the genera Prevotella and Xylanibacter were found exclusively from the African children( Reference De Filippo, Cavalieri and Di Paola 68 ). Partly similarly to these findings, Prevotella enterotype was previously found to be associated with high consumption of carbohydrates and with both vegetarian and vegan habitual diets, whereas Bacteroides enterotype was associated with high consumption of animal protein, amino acids and saturated fats( Reference Wu, Chen and Hoffmann 134 ). Strict vegetarians have higher percentages and/or numbers of Lachnospiraceae and Clostridium ramosum group (Clostridial cluster XVIII) bacteria than omnivorous people( Reference Hayashi, Sakamoto and Benno 135 , Reference Hippe, Zwielehner and Liszt 136 ), whereas vegans have lower faecal numbers of Bacteroides spp., bifidobacteria and Enterobacteriaceae( Reference Zimmer, Lange and Frick 137 ). In older culture-based studies, no fusobacteria were found in faecal samples of vegetarian people. In addition, the total count of anaerobic streptococci, Peptostreptococcus spp., Actinomyces spp. and Lactobacillus spp. were higher in strict vegetarians as compared with people on a traditional Western diet( Reference Finegold, Sutter, Mathisen and Hentges 2 , Reference Finegold, Attebery and Sutter 132 , Reference Finegold, Flora and Attebery 138 , Reference Finegold, Sutter and Sugihara 139 ).
Hospitalisation of Scottish elderly people has resulted in a decrease of faecal Bacteroides–Prevotella group and Ruminococcus albus prevalence, while increasing the Enterococcus faecalis prevalence, as compared with healthy elderly people. Moreover, when the hospitalisation was combined with antibiotic treatment, reductions in other bacterial groups were also observed( Reference Bartosch, Fite and Macfarlane 140 ). Long-term residential care of Irish elderly people has been associated with a higher proportion of bacteria belonging to the phylum Bacteroidetes and several genera from other phyla, whereas community-dwelling elderly people had higher proportions of bacteria belonging to the phylum Firmicutes. In addition, the microbiota of elderly people in long-term care was significantly less diverse. However, the results also correlated with the diet; if the elderly people were grouped according to their diet, the clustering was similar to that based on the residence location. After 1 month residential care, the diet was converged to ‘long-term diet’, but it took 1 year for the microbiota to clearly cluster within the long-term residential type. Collectively this indicates that the composition of the faecal microbiota is determined rather by the composition and diversity of the diet than by the location of residence( Reference Claesson, Jeffery and Conde 98 ).
Since it has not always been clear which part of the habitual diet induces the changes observed in the GI-tract microbiota, numerous dietary interventions have been conducted to elucidate the changes induced by specific nutritional substances, e.g. different types of fibres, proteins, fats and polyphenols. In addition, the effects of probotics and prebiotics on the GI-tract microbiota have been widely studied. Therefore, the effects of the aforementioned nutritional substances on GI-tract microbiota are addressed in more detail.
Fermentable dietary carbohydrates
Dietary components that escape digestion by endogenous enzymes in the upper GI tract become available substrates in the large intestine( Reference Cummings and Macfarlane 7 ). Dietary fibre (DF) is a normal constituent of most foods derived from plants( Reference Blaut 141 ). These ‘non-digestible’ dietary carbohydrate substrates include resistant starch, plant cell-wall material (non-starch polysaccharides) and oligosaccharides( Reference Cummings and Macfarlane 7 ). In the human colon, DF is metabolised by the microbiota to SCFA, comprising mainly acetic, propionic and butyric acids. SCFA have been implicated to have both local and systemic beneficial biological effects in the human body; acetate is readily absorbed and transported to the liver; propionate is a substrate for hepatic gluconeogenesis; butyrate is the preferred fuel of the colonocytes and also plays a major role in the regulation of cell proliferation and differentiation( Reference Cummings and Macfarlane 7 ). Several excellent review-papers already exist( Reference Flint, Duncan and Scott 130 , Reference Flint, Scott and Duncan 142 – Reference Flint 147 ), and therefore we will not go into detail with the dietary interventions with fibre.
In vitro and in vivo evidence indicate that a bacterial group related to Faecalibacterium prausnitzii, Roseburia and E. rectale plays a major role in mediating the butyrogenic effect of fermentable dietary carbohydrates( Reference Louis, Scott and Duncan 107 , Reference Aminov, Walker and Duncan 148 – Reference Pryde, Duncan and Hold 150 ). In addition, it has been shown in numerous studies that as dietary carbohydrate content is reduced in the diet the count of F. prausnitzii declines, respectively( Reference Duncan, Belenguer and Holtrop 129 , Reference Benus, van der Werf and Welling 151 ). However, some lactate-utilising bacteria within Lachnospiraceae produce less butyrate in the presence of lactate-utilising sulphate-reducing bacteria. Moreover, in the presence of higher abundance of lactate, the formation of butyrate was reduced even more and the formation of hydrogen sulphide was promoted( Reference Marquet, Duncan and Chassard 152 ).
Cereal grains are a good source of DF. The main DF components of cereal grains are arabinoxylan, cellulose, β-glucan, fructan, resistant starch and lignin( Reference Karppinen, Liukkonen and Aura 153 ). The gut microbiota stimulating activities of arabinoxylan (stimulates Bacteroides spp. and Roseburia spp.), resistant starch (stimulates bifidobacteria, Bacteroides spp., Ruminococcus bromii, E. rectale and Roseburia spp.), β-glucan (stimulates bifidobacteria) and fructan (stimulates bifidobacteria, Bacteroides spp., lactobacilli and butyrate-producers) are well recognised( Reference Flint, Bayer and Rincon 144 , Reference Hughes, Shewry and Gibson 154 – Reference Abell, Cooke and Bennett 161 ). In addition, arabinoxylan-oligosaccharides, which are enzymatic hydrolysis products of arabinoxylan, have been shown to stimulate the growth of bifidobacteria in some studies( Reference Broekaert, Courtin and Verbeke 156 ). The effects of cellulose and lignin on gut microbiota are less well known( Reference Ehle, Robertson and Van Soest 162 , Reference Robert and Bernalier-Donadille 163 ). However, it should be noted that e.g. the size of the grain flakes may lead to different bacterial responses: i.e. smaller-sized whole-oat grain flakes (0·53–0·63 mm) resulted in a significant increase in the numbers of Bacteroides–Prevotella group bacteria, whereas in a fermentation with larger oat flakes (0·85–1·00 mm) bifidobacterial numbers increased( Reference Connolly, Lovegrove and Tuohy 164 ). It has also been shown in in vitro model studies using different substrates that the majority of the bacteria attached to wheat bran belonged to Lachnospiraceae and some bacteria were Bacteroides spp., whereas R. bromii, B. adolescentis, Bifidobacterium breve and E. rectale were found attached to starch. When mucin was used as a substrate, the most commonly found bacteria were Bifidobacterium bifidum and an uncultured relative of Ruminococcus lactaris ( Reference Leitch, Walker and Duncan 165 ).
Prebiotics are non-digestible (by the host) food ingredients that have a beneficial effect through their selective metabolism in the intestinal tract. The prebiotics that currently fulfil the prebiotic criteria are inulin, FOS, galacto-oligosaccharides and lactulose( Reference Gibson, Probert and Van Loo 166 ). The best sources of naturally occurring prebiotics may be found in vegetables such as artichokes, onions, chicory, garlic and leeks( Reference Macfarlane and Cummings 167 ). There are numerous studies in which the bifidogenic properties of prebiotics are shown( Reference Walton, van den Heuvel and Kosters 121 , Reference Langlands, Hopkins and Coleman 122 , Reference Gibson, Probert and Van Loo 166 , Reference Tannock, Munro and Bibiloni 168 – Reference Roberfroid, Gibson and Hoyles 171 ). In addition, increase in abundance of lactobacilli( Reference Walton, van den Heuvel and Kosters 121 , Reference Langlands, Hopkins and Coleman 122 ) and F. prausnitzii ( Reference Ramirez-Farias, Slezak and Fuller 170 , Reference Louis, Young and Holtrop 172 ) has been shown. Moreover, in infant formulas, galacto-oligosaccharides + FOS supplementation of cow's milk-based formula has led to a bifidobacterial population which resembled more that of breast-fed infants than purely formula-fed infants( Reference Rinne, Gueimonde and Kalliomaki 173 ). FOS have also positive effects on the intestinal barrier function( Reference Cani 174 ).
Endogenous protein sources make up approximately one-third of the exogenous dietary protein pool ( Reference Alberts, Johnson and Lewis 175 , Reference Silk 176 ). Bacterial amino acid catabolism in the human gut occurs via a number of mechanisms involving either deamination or decarboxylation reactions. The types of SCFA produced from amino acids are dependent on the chemical compositions of the substrates. In addition to SCFA, branched chain-fatty acids and aromatic compounds, namely phenol, indole and a range of phenolic and indolic substituted fatty acids derived from phenylalanine, tyrosine and tryptophan may be formed. Moreover, branched-chain amino acids are slowly fermented by colonic bacteria, with the main acidic products being branched chain-fatty acids one carbon atom shorter than the parent amino acid( Reference Smith and Macfarlane 177 ). Many metabolites produced by amino acid fermentation are harmful to the host. Phenolic and indolic compounds are also thought to act as co-carcinogens, while amines serve as precursors of nitrosamine production( Reference Smith and Macfarlane 177 ).
Culture-based studies have shown( Reference Hentges, Maier and Burton 118 ) that the counts of Bacteroides spp. and clostridia increased significantly, whereas counts of B. adolescentis decreased significantly during a high-beef diet as compared with a meatless diet (fat and fibre contents were essentially the same in both diets). In addition, sulphide concentrations were high on a high-beef diet( Reference Magee, Richardson and Hughes 178 ). Since hydrogen sulphide is toxic to the colonic epithelium and sulphide inhibits butyrate oxidation, dietary sulphide may selectively stimulate the growth of a single group of bacteria, namely sulphate-reducing bacteria, with potentially harmful effects on the epithelium( Reference Cummings and Macfarlane 9 , Reference Pitcher and Cummings 179 ). In addition, an intervention diet with a high protein and low carbohydrate content reduced the numbers of Roseburia/E. rectale group, while increasing proportions of branched-chain fatty acids and concentrations of phenylacetic acid and N-nitroso compounds( Reference Russell, Gratz and Duncan 119 ). Moreover, it should be noted that the World Cancer Research Fund released in May 2011 a report based on 1012 clinical trials, in which red and processed meat were convincingly associated with increased risk, whereas foods containing DF, in particular cereal fibre and whole grains, were associated with decreased risk of colorectal cancer( 180 ).
Fats are composed of fatty acids that are divided into SFA and unsaturated fatty acids. Dietary SFA are mainly obtained from animal products, such as meats and dairy foods, but may also be obtained from some plant sources, such as coconut, cottonseed and palm kernel oils. The major dietary MUFA is oleic acid. Oleic acid is the primary component of olive oil, but may also be found in hazelnut, rapeseed and peanut oils. A carbon chain that contains two or more cis double bonds characterises the families of n-3 or n-6 PUFA. These families cannot be synthesised by the human body( Reference Ortega, Varela and Bermudez 181 ). Linoleic (n-6 PUFA) and α-linolenic (n-3 PUFA) acids form the majority of PUFA in most Western diets. The long-chain n-3 PUFA EPA and DHA are found in seafood, especially oily fish( Reference Calder, Ahluwalia and Brouns 182 ).
High intake of dietary fat may increase the quantities of bile acids and fat that reach the colon. It has been suggested that the gut microbiota may metabolise dietary fats (producing diacylglycerols from polyunsaturated fats), convert primary bile acids into secondary bile acids and impact on the enterohepatic circulation of bile acids and fat absorption from the small intestine( Reference Fava, Gitau and Griffin 183 ). However, there are only a few human studies in which the effect of high-fat diet on the human intestinal microbiota has been investigated, and especially those in which the correlations between the different types of dietary fat and intestinal microbiota have been investigated. In a study of Brinkworth et al. ( Reference Brinkworth, Noakes and Clifton 184 ), it was shown that a very low-carbohydrate, high-fat diet resulted in a significant reductions in bifidobacterial numbers, concentrations of butyrate and total SCFA, defecation frequency and faecal excretion as compared with isoenergetic high-carbohydrate, high-fibre and low-fat diet.
High MUFA-containing dietary intervention that lasted 4 weeks reduced the total bacterial numbers but did not affect the specific bacterial groups( Reference Fava, Gitau and Griffin 183 ). Conversely, high habitual intake of MUFA has been associated with lower numbers of bifidobacteria and slightly higher numbers of Bacteroides spp.( Reference Simoes, Maukonen and Kaprio 64 ). In a recent metagenomic study in healthy volunteers, the Bacteroides enterotype was found to be highly associated with the consumption of MUFA and SFA( Reference Wu, Chen and Hoffmann 134 ). These observations suggest that the consumption of fat and animal-derived products, typically present in the Western diet, are associated with increased Bacteroides spp. prevalence in the human gut microbiota.
Habitual n-3 PUFA intake has been shown to have a significant positive association with Lactobacillus group abundance( Reference Simoes, Maukonen and Kaprio 64 ). The increase in Lactobacillus group bacterial numbers in stool after n-3 PUFA intake has also been reported in a mouse study( Reference Pachikian, Neyrinck and Portois 185 ). In addition, in a human study by Santacruz et al. ( Reference Santacruz, Marcos and Wärnberg 186 ) the numbers of lactobacilli remained at the same level, even though the ingested amount of total PUFA was greatly reduced. The increase in n-3 PUFA is effective in supporting epithelial barrier integrity by improving transepithelial resistance and by reducing IL-4-mediated permeability( Reference Willemsen, Koetsier and Balvers 187 ), and several lactobacilli enhance the function of the intestinal barrier( Reference Anderson, Cookson and McNabb 188 , Reference Donato, Gareau and Wang 189 ). Maternal salmon (marine n-3 PUFA) consumption before delivery has also lowered the number of Coriobacteriaceae in bottle-fed infants( Reference Urwin, Miles and Noakes 190 ).
Higher habitual n-6 PUFA intake has been associated with decreased numbers of bifidobacteria( Reference Simoes, Maukonen and Kaprio 64 ). It has also been reported that high n-6 PUFA intakes decrease certain immune functions, such as antigen presentation, adhesion molecule expression, proinflammatory cytokines and T-helper 1 and T-helper 2 responses( Reference Harbige 191 ). Furthermore, genomic DNA of some bifidobacterial strains is able to stimulate the production of T-helper 1 and proinflammatory cytokines, interferon-γ and TNF-α( Reference Medina, Izquierdo and Ennahar 192 ). Overall, these results indicate an association between dietary fat types and their distinct effect on the faecal microbiota. As a consequence, it seems that balanced diet with regard to fat consumption is critical not only for the host's health, but also for the gut microbiota.
Plant foods contain significant amounts of phenolic compounds( Reference Aura 193 ). Plant polyphenols are a class of chemically diverse secondary metabolites that possess many different biological activities both within the plant and in the human subjects eating these plants. Plant polyphenols have the potential to affect certain risk factors of CVD, as well as being antioxidants, have antimicrobial properties and possessing inherent free radical scavenging abilities( Reference Tuohy, Conterno and Gasperotti 194 ). The main dietary sources of polyphenols are berries, fruits, beverages (e.g. coffee, tea and wine), chocolate, whole-grain cereals, vegetables and legume seeds( Reference Aura 193 ).
The human gut microbiota has extensive hydrolytic activities and breaks down many complex polyphenols into smaller phenolic acids, which can be absorbed across the intestinal mucosa( Reference Tuohy, Conterno and Gasperotti 194 ). Daily consumption of red wine polyphenols for 4 weeks significantly increased numbers of bacteria within genera Enterococcus, Prevotella, Bacteroides, Bifidobacterium, Eggerthella and Family Lachnospiraceae( Reference Queipo-Ortuno, Boto-Ordonez and Murri 195 ), whereas consumption of high cocoa flavanol drink for 4 weeks significantly increased the bifidobacterial and lactobacilli numbers but significantly decreased clostridial counts( Reference Tzounis, Rodriguez-Mateos and Vulevic 196 ). Human dietary intervention with ellagitannins, which are polyphenols abundant in strawberries, raspberries and cloudberries, induced changes in the composition of Lachnospiraceae and Ruminococcaceae( Reference Puupponen-Pimiä, Seppänen-Laakso and Nohynek 128 ). Tea phenolics (e.g. epicatechin, catechin, gallic acid and caffeic acid) significantly repressed certain bacteria such as Clostridium perfringens and C. difficile and members of the Bacteroides spp., whereas bifidobacteria, Lactobacillus spp. and nonpathogenic Clostridium spp. were less severely affected( Reference Lee, Jenner and Low 197 ). Moreover, many phenolic compounds have in vitro antimicrobial activities towards pathogenic bacteria, such as Salmonella spp., C. perfringens, C. difficile, E. coli and Staphylococcus aureus ( Reference Lee, Jenner and Low 197 – Reference Requena, Monagas and Pozo-Bayón 199 ).
Probiotics are live micro-organisms which when administered in adequate amounts confer a health benefit on the host, according to the widely accepted definition by Food and Agriculture Organisation WHO( Reference Araya, Morelli and Reid 200 ). Most of the currently used probiotics belong to the genera Bifidobacterium and Lactobacillus. However, probiotic preparations containing species of the genera Enterococcus, Pediococcus, Streptococcus, Lactococcus, Propionibacterium, Bacillus and Saccharomyces are also used( Reference Champagne, Ross and Saarela 201 ). The past two decades have seen a marked increase in the inclusion of probiotic bacteria in various types of food products, especially in fermented milks( Reference Daly and Davis 202 ). During recent years probiotics have also been increasingly incorporated into non-dairy foods such as fruit and berry juices and e.g. cereals( Reference Champagne, Ross and Saarela 201 ). In good quality products, the daily dose should be approximately 109 colony-forming units/d( Reference Donnet-Hughes, Rochat and Serrant 203 ). Probiotics do not usually colonise the GI tract, and therefore the products should be consumed daily for the health benefits( Reference Macfarlane and Cummings 167 ). In most of the studies, probiotics have not caused any significant changes in the predominant faecal microbiota of healthy adults. However, there are very few studies in which e.g. Lactobacillus rhamnosus GG has modulated the faecal microbiota and increased overall bacterial diversity in infants( Reference Rinne, Kalliomaki and Salminen 123 , Reference Cox, Huang and Fujimura 204 ). In addition, Bifidobacterium animalis subsp. lactis Bb12 has reduced the numbers of Enterobacteriaceae and Clostridium spp. in preterm infants( Reference Mohan, Koebnick and Schildt 124 ).
To answer the question posed in the title: does diet matter in regard to human microbiota? Yes, it does. From the host's perspective, there are numerous activities of the commensal gut microbiota that are of great importance to health. Moreover, by choosing what we eat, we can decide which bacteria we feed. However, even though diet matters, the results from dietary interventions are not always straight forward. It should be remembered that the detected effect is dependent on the study subjects, study protocols, used DNA-extraction techniques, used methodologies, and in case of next-generation sequencing also the algorithms used for cleaning and analysing the data. In addition, individual variation in human intestinal microbiota is so wide that the subtle changes may not be detected if the study cohort is not large enough. Therefore, studies in which different habitual diets have been compared with each other, usually get clearer correlations with nutrients v. bacteria than those observed in dietary interventions.
In order to get as much information from future dietary interventions there would be need for big enough group sizes, long enough trials, long enough wash-out periods in cross-over designs, sufficient background information (i.e. baseline 7-d food records and clinical parameters) and last but not least more interdisciplinary research across microbiology, nutrition, immunology, genetics, epigenetics, proteomics, transcriptomics, metabolomics and human physiology.
This work was supported by the EU-funded projects TORNADO (grant no. FP7-KBBE-222720) and ETHERPATHS (grant no. FP7-KBBE-222639). The funders had no role in the design, analysis or writing of the present paper.
Conflicts of Interest
J. M. drafted the manuscript and undertook the literature searches. M. S. reviewed the literature on polyphenols, proteins and fats. J. M. and M. S. reviewed and revised the manuscript.