Introduction

Gastrointestinal (GI) microbiota play an important role in the health and wellbeing of the host1. Several studies have shown that the intestinal microbiota fluctuates in response to a variety of intrinsic and extrinsic factors, such as host health2, genetic composition3, age4 and diet5. Among all factors, genotype and diet have been suggested to be the main components that exert a significant influence on the balance of GI microbiota.

Mongolian nationality originates from a tribe that was located in Northern China during the seventh century6,7. The Mongol Empire, one of the world's largest contiguous empires, exerted a major influence that greatly enhanced the cultural exchange between China and the occident that took place during the Middle Ages. In Mongolia today, more than 40% of the population lives in typical pasture areas (such as Khentii Province) and maintains a traditional nomadic lifestyle and diet. In contrast, many Mongolians living in Ulan Bator (the capital of Mongolia) and TUW Province (the suburbs of the capital) have adopted an urban lifestyle because of modernisation and economic development. However, little is known about the structure of Mongolian gut microbiota or how their microbial community is affected by such changes.

The typical Mongolian diet is characterised by a high and frequent consumption of fermented dairy products, red meat and liquor6. In the pastures of Khentii, locals exhibit a distinct seasonal variation in their food consumption. Meat and meat products are the main sources of energy during winter and spring (November to April), whereas dairy products are the main source during summer and autumn (May to October)8. However, in Ulan Bator, food is abundant and diverse; therefore, the diet in this city exhibited limited changes throughout the year. Given these divergent dietary lifestyles, Mongolians are excellent candidates to study the effects of seasonal dietary changes on intestinal microbiome compositions.

In a previous study, we described the profiles of the gut microbiota of Chinese Mongolians living in Inner Mongolia province by denaturing gradient gel electrophoresis (DGGE) and quantitative polymerase chain reaction (q-PCR) techniques9. However many Chinese Mongolians have inter-married with Han nationality race which is rare in Mongolia. Thus Mongolians in Mongolia were more authentic at the gene level. Moreover, the pyrosequencing has been suggested a more appropriate approach for intestinal microbiota diversity analysis than DGGE.

In the present study, 320 faecal samples were collected from 64 Mongolians distributed in three areas (Ulan Bator, TUW and Khentii) at five time points (January, March, June, September and November). 454 pyrosequencing combined with q-PCR technology were applied to explore the structure of Mongolians' gut microbiota and the effects of seasonal dietary changes on their intestinal microbiota.

Results

Sequencing coverage and estimation of bacterial diversity

In this study, the microbiotic compositions of the faecal samples were examined using a high-throughput 454 pyrosequencing technique. We generated a dataset consisting of 3,795,726 filtered high-quality and classifiable 16S rRNA gene sequences and an average of 11,843 sequences was obtained for each individual (range: from 2,780 to 30,480). All sequences were clustered with representative sequences and a 97% sequence identity cut-off was used. The number of OTUs per sample ranged between 118 and 1,815 (Table S3). The Simpson index, Chao1 index, Shannon index and observed number of species were estimated using the QIIME platform (Table S3).

The composition of intestinal microbiota in Mongolians

At the phylum level (Fig. 1A and 1B), Bacteroidetes, Firmicutes, Proteobacteria and Actinobacteria constituted the four most dominant bacterial phyla (contributing 55.56%, 39.53%, 2.68% and 0.85% of the total amount of sequences, respectively). For all participants, the average ratio of Firmicutes to Bacteroidetes (F/B) was 0.711 (range: from 0.006 to 2.253, Fig. 1C). At the genus level (Fig. 1B), Prevotella of the Firmicutes phylum was the most abundant genus (contributing to 36.31% of the total number of sequences) and the amounts of Bacteroides, Faecalibacterium, Oscillibacter, Roseburia, Clostridium, Coprococcus, Ruminococcus, Alistipes, Parabacteroides, Catenibacterium, Subdoligranulum and Eubacterium all exceeded 1%. Correlations among the genera that contributed more than 0.1% of the total number of sequences in Mongolians were determined based on Spearman's rank correlation (Fig. S1A). A general negative correlation was found between Prevotella and other genera. Using genus-specific primers, we quantified the predominant microbiota in the human gut (Fig. 1D). The amounts of Bacteroides, Bifidobacterium, Enterobacter, Prevotella, Lactobacillus and Faecalibacterium genera were 9.61 ± 0.13, 8.02 ± 0.85, 7.59 ± 0.21, 9.66 ± 0.17, 6.53 ± 0.18 and 10.34 ± 0.71 in log-transformed 16S rDNA gene copy number per gram of sample, respectively (Fig. 1D).

Figure 1
figure 1

The composition of intestinal microbiota of Mongolians.

(A) Inter-individual variation in the proportion of major phyla. (B) Box-plots showing bacterial compositions at genus and phylum level; maximum and minimum values are indicated using whiskers. (C) Inter-individual variation in the proportion of the genus Prevotella and the ratio of Firmicutes to Bacteroidetes (F/B). (D) The amounts of Bacteroides, Bifidobacterium, Enterobacter, Prevotella, Lactobacillus and Faecalibacterium as quantified using q-PCR.

Full size image

The core intestinal microbiota of Mongolians

A major aim of the present study was to determine whether a common core microbiota is shared among all or the vast majority of the Mongolian participants. Using a detailed OTU analysis, we were able to assign 22 core OTU candidates (out of 19,451 OTUs identified in this study) (Fig. 2A); each of these candidates exhibited an average frequency of occurrence higher than 90% over all samples. These core OTUs primarily belonged to the genera Faecalibacterium, Bacteroides, Dorea, Collinsella, Oscillibacter, Ruminococcus, Subdoligranulum, Coprococcus and Prevotella. Correlations among these OTUs were determined based on Spearman's rank correlation (Fig. 2A). Additionally, 9 of the 22 core OTU candidates (OTU ID: 32177, 17033, 6459, 15289, 16937, 26107, 903, 5801 and 30719) primarily belonged to the genus Prevotella, Bacteroides, Faecalibacterium, Ruminococcus, Subdoligranulum and Coprococcus. These candidates were stably detected in almost every sample (Fig. 2B). Furthermore, a small proportion of OTUs (1.58%) that contributed 49.53% of the sequences were present in 78.26% of samples (Fig. 2B). In addition, the relationship between the 9 core OTU candidates and all samples was revealed by visualising a large network (Figs. 2C and S1B).

Figure 2
figure 2

The core intestinal microbiota of Mongolians.

(A) Correlation matrix showing the Spearman's rank correlation among the 22 most abundant OTUs. The Spearman's rank correlation coefficient ranges from 1.0 to −1.0, corresponding to a strongly positive to a strongly negative correlation. (B) Fraction of OTUs shared across samples of the total OTUs within these samples (blue line) and the proportion that these OTUs represent of the total sequences obtained (red line) for the participants. (C) Bipartite network diagrams of evenly sampled bacterial 16S rRNA–derived top 9 core OTUs. Edges connect genus-level OTUs (purple points) to participant nodes (white points).

Full size image

Differences in gut microbiota between Mongolians from the Khentii pasturing area, TUW province and Ulan Bator

A diversity analysis based on Simpson, Chao1, Shannon and observed species indices (Fig. S2A–S2D) revealed that the alpha diversity of the intestinal microbiota was greatest in the Khentii pasturing area but least in Ulan Bator. Additionally, we compared the composition of the intestinal microbiota of Mongolians from Khentii, TUW and Ulan Bator. A PCoA based on the unweighted (Fig. 3A and 3B) and weighted (Fig. S3A and S3B) Unifrac distances was performed using the obtained pyrosequencing data. An apparent clustering pattern was identified for the participants from different locations. Points representing the intestinal microbiota composition of Khentii, TUW and Ulan Bator residents clustered at the top right, the centre and the bottom left, respectively. The enterotype analysis provided a clear visualisation of the relationships among the different sample groups (Figs. 3D and 3E). The silhouette index was more than 0.6. All samples clustered into one of two groups. Cluster 1 primarily comprised Ulan Bator residents and cluster 2 primarily comprised Khentii pasturing area and TUW province residents.

Figure 3
figure 3

Differences in gut microbiota among Mongolians from Khentii, TUW and Ulan Bator.

(A) A principal component (PCoA) score plot based on unweighted UniFrac metrics for all participants. Each point represents the composition of the intestinal microbiota of one participant. (B) PCoA score plot based on unweighted UniFrac metrics. Each point represents the mean principal component scores of all volunteers at one location at one time point and the error bar represents the standard deviation. (C) Sampling location and time point-driven unweighted UniFrac distances. (D) The silhouette index reflected the isolation degree of two enterotypes. (E) Enterotype analyses of the intestinal microbiota; cluster 1 contained participants primarily from Ulan Bator and cluster 2 contained participants primarily from the Khentii pasturing area and TUW province. (F) Heatmap constructed using the amount of significantly different genera among participants in the Khentii pasturing area, TUW province and Ulan Bator city.

Full size image

After establishing an intrinsic difference between the compositions of the gut microbiota of Mongolians living in different areas, we further identified differences in the specific bacteria of individuals that were principally responsible for the differences found using the Kruskal-Wallis test. The significant genera found (p < 0.05) are listed in Table 1 and the values were transformed into a heatmap (Fig. 3F). According to the heatmap, the genera Prevotella, Solobacterium, Succinivibrio, Escherichia coli/Shigella group, Olsenella, Oribacterium and Lactobacillus were abundant in Khentii residents and the genera Bacteroides, Oscillibacter, Roseburia, Alistipes, Coprococcus, Parabacteroides, Subdoligranulum, Barnesiella, Odoribacter, Parasutterella, Butyricimonas, Coprobacillus, Victivallis, Anaerosporobacter and Akkermansia were abundant in Ulan Bator residents.

Table 1 Significantly different genera among Mongolians in the Khentii pasturing area, TUW province and city of Ulan Bator
Full size table

Seasonal changes in the Mongolians' intestinal microbiota

Based on Fig. 3B and 3C, we found that the changes in the range of intestinal microbiota of Mongolians from Khentii, TUW and Ulan Bator were discrepant and exhibited seasonal alternation. Combining the results from five sampling points (January, March, June, September and November), we noted that the seasonal changes in the range of intestinal microbiota were more distinct in Khentii residents than in TUW and Ulan Bator residents. We therefore analysed the data according to the sampling location.

For the Khentii residents, the results of a PCoA based on unweighted Unifrac and A partial least squares discriminant analysis (PLS-DA) (Fig. 4A–4C; weighted, Fig. S4) indicated that their intestinal microbiota compositions in June and September were similar and close to those observed in March but were significantly different from those observed in January and November. The changed genera representing more than 1% of the total number of sequences were listed in Table 2 and confirmed using the q-PCR data (Fig. S7). Further genus-level analysis revealed that Faecalibacterium, Eubacterium, Dorea, Collinsella, Enterococcus, Solobacterium, Caldimonas, Escherichia coli/Shigella group and Subdoligranulum levels were altered significantly (p < 0.05), exhibiting a lower contribution from March to September (Table 3); however, the abundance of Prevotella, Bacteroides, Clostridium and Oscillibacter remained stable throughout the year. The changes in the intestinal microbiota of the TUW residents were not as profound as those of the Khentii residents. The results shown in Fig. 4D–4F indicate that the intestinal microbiota compositions in June, March and January were similar to each other but distinct from those observed in September and November (weighted Unifrac distances and an enterotype analysis are listed in Fig. S5). At the genus level, Faecalibacterium, Anaerosporobacter, Butyricimonas, Collinsella and Roseburia changed significantly (p < 0.05) with season (Table 3), but the abundances of Prevotella, Bacteroides, Clostridium and Oscillibacter remained stable. However, for the Ulan Bator residents, little change was noted in their intestinal microbiota composition throughout the year (Fig. 4G–4I, weighted Unifrac distances and an enterotype analysis are listed in Fig. S6) and only the genera Eubacterium, Dorea and Collinsella differed among sampling points (Table 3).

Table 2 The changed genera (relative amounts >1%) in Mongolians from Khentii, TUW and Ulan Bator due to seasonal change
Full size table
Table 3 Significantly changed genera in Mongolians from Khentii, TUW and Ulan Bator due to seasonal change
Full size table
Figure 4
figure 4

The changed range of the intestinal microbiota of Mongolians from the Khentii pasturing area, TUW province and Ulan Bator city was discrepant with the seasonal alternation.

(A, D and G) Principal component (PCoA) score plots based on unweighted UniFrac metrics of Mongolians in the three different locations. (B, E and H) The unweighted pair-group method with an arithmetic means (UPGMA) cluster analysis based on the distance metrics of Mongolians in the three different locations. (C, F and I) Partial least squares discriminant analysis (PLS-DA) based on the species abundance of Mongolians in the three different locations.

Full size image

Concordance of diet and intestinal microbiota

The traditional Mongolian diet is characterised by a high and frequent consumption of fermented dairy products, red meat and liquor. Currently, because of modernisation and economic development, many Mongolians living in Ulan Bator (the capital of Mongolia) and in TUW province (the suburbs of the capital) have gradually adopted an urban lifestyle and only a few Mongolians, who mainly live in pasturing areas, retain the traditional diet. A partial least squares discriminant analysis (PLS-DA) based on participants' food types and weights also demonstrated this tendency (Fig. 5A). The red points in the figure represent the responses obtained from Khentii Mongolians using the food frequency questionnaire and are clustered at the right of the figure. To the left of this cluster are the responses obtained from TUW and Ulan Bator residents.

Figure 5
figure 5

Concordance of diet and intestinal microbiota.

(A) Partial least squares discriminant analysis (PLS-DA) based on the participants' food types and weights. (B) Heatmap based on food frequency questionnaire data recording the food types and weights of all participants. (C–E) Procrustes analysis combining unweighted UniFrac PCoA of the microbiota, including a food type PCoA.

Full size image

Based on the heatmap (Fig. 5B), we found differences in the food types chosen by Mongolians. Food type diversity was the greatest among Ulan Bator residents and their dietary structure remained stable throughout the year. However, food type diversity was much less in TUW residents and the dietary structure of these residents was not stable. Notably, Khentii residents lacked food type diversity and their dietary structure changed significantly with season. Based on the results presented above, the food types that enabled the greatest discrimination were vegetables, fruit, red meat and kumiss, the consumption of which differed between the three groups of Mongolians.

To construct a concordance relationship between diet and intestinal microbiota, a procrustes analysis of the food frequency questionnaire and the microbiota β-diversity based on sampling locations was used to co-visualise the data (Fig. 5C–5E). Separations based on either diet or microbiota co-segregated along the first axis of both data sets (weighted UniFrac, Fig. S8A–S8C). Based on the figures (Fig. S8A–S8C), we observed a strong correspondence between diet and intestinal microbiota (the p values for Khentii, TUW and Ulan Bator were < 0.001, 0.008 and 0.017 respectively Rev-4).

Discussion

The phyla Firmicutes and Bacteroidetes predominated and together, represented an average of 91.6% of the sequences identified, in agreement with previous studies, which attributed the majority of human gut microbiota to these two phyla. A noteworthy feature of the faecal bacteria structure of Mongolians in our study was that the Firmicutes to Bacteroidetes (F/B) ratio was low, only 0.71. The F/B ratio relates to dietary habit and host physiology10,11. Those with high-fat western diets, the obese and young adults (versus the elderly) tend to exhibit higher F/B ratios. De Filippo et al.5 concluded that a more westernised diet (higher fat and meat consumption and lower vegetable and legume consumption) causes a higher F/B ratio. Due to their nomadic lifestyle, some Mongolians have adapted to the classical diet including a high consumption of meat, alcohol and fermented milk, which more resembles typical western diets than that of rural African areas studied by De Filippo. Notably, the F/B ratio of Mongolian adult samples (0.71) calculated in this study was at the low end of the range obtained by De Filippo et al. (from 0.47 in rural Africa to 2.81 in urbanised Italian children). However, members of the Korean population were reported to have a high F/B ratio of 2.95, even though Korean diets contained a relatively high fibre content (19.8 g/day versus 15.1 g/day for Americans) primarily from kimchee and steamed rice12. The average age of our Mongolian participants was 34, which was closer to the adult group reported in Marion et al. Marion et al. reported adults to have a higher F/B ratio than the elderly (10.9 and 0.6, respectively). Our results and other findings suggest that the age, Westernised diet and lifestyle (rich in fat and meat, low in vegetables and legumes) of the participants may not be determining factors for gut microbiota composition indicators, such as the F/B ratio. Other dietary components and factors, such as host genetics, may exert considerable influence3.

At the genus level, Prevotella and Bacteroides were found to predominate in the Mongolian samples, contributing 47.11% and 6.33% of the total sequences, respectively. This result was further supported by a high proportion of Prevotella and Bacteroides among the core OTUs (34/67, >50%). The genus Prevotella contains a wide array of carbohydrate- and protein-fermenting and acetate- and H2-producing bacteria such as Prevotellaruminicola13 and the genus Bacteroides has been mainly associated with the metabolism of animal proteins, a variety of amino acids and saturated fats14. The traditional Mongolian diet is characterised by a large amount of fried wheaten food, red meat and fermented dairy products with low quantities of vegetables and fruits. It is unsurprising, therefore, that these two genera dominated the microbiotic composition of Mongolian guts. In our previous study of Mongolians living in Inner Mongolia of China, the most abundant genus in intestinal tract was Phascolarctobacterium9. The Mongolians in Inner Mongolia of China usually live together with Han race which leads to a similar life-style and dietary habit with Han, but the Mongolians in Mongolia still keep the relatively traditional dietary habit, thus explains their differences at the genus level. Rev-1.

Nine OTUs, primarily belonging to Bacteroides, Faecalibacterium, Ruminococcus, Subdoligranulum and Coprococcus, were stably detected in nearly every Mongolian sample. Therefore, we defined these OTUs as core OTUs in Mongolians. Previous studies of the core microbiota in other nations have been widely reported. Martınez et al. characterised the faecal microbial communities of three young Americans over a one-year period by 454 pyrosequencing 16S rRNA tags to investigate the temporal characteristics of their bacterial communities15. The authors detected 16 stable core OTUs close to the genera of Bifidobacterium, Bacteroides, Faecalibacterium, Ruminococcus, Blautia, Dorea, Eubacterium and Coprococcus. Ling et al. investigated the faecal core microbiota of ten healthy Chinese undergraduates16. The dominant taxonomic groups in these faecal samples were Faecalibacterium, Coprococcus, Blautia, Bacteroides, Roseburia, Ruminococcus, Subdoligranulum, Sporacetigenium, Oscillibacter, Dorea, Phascolarctobacterium and Prevotella. Huse et al. explored the core faecal microbiota of more than 200 individuals from the NIH Common Fund Human Microbiome Project and 7 OTUs representing the genera Faecalibacterium, Oscillibacter and Bacteroides were identified as core OTUs17. Based on these findings, a surprising consistency in core intestinal microbiota was found among the nations. Previous studies at the functional and metabolic levels indicated that these genera play a key role in the synthesis of basic metabolites in the human gastrointestinal tract. Therefore, the core intestinal microbiota in all humans might vary within a limited range.

Due to modernisation and economic development, many Mongolians living in Ulan Bator and TUW have gradually adopted an urban lifestyle and only a few Mongolians, mainly in pasturing areas, maintain the traditional diet and lifestyle. Accordingly, the intestinal microbiotic compositions of Mongolians from the three areas differed. An analysis at the genus level revealed that the difference was primarily reflected in the populations of Solobacterium, Olsenella, Oribacterium and Lactobacillus, which were abundant in Khentii Mongolians. Previous reports indicated that Oribacterium and Olsenella are closely related to high incidences of periodontitis and gingivitis18 and that Solobacterium is considered a major cause of bromopnea19. These distinctions were related to dental hygiene. The participants from Khentii brushed their teeth less frequently than did those from Ulan Bator and TUW and some herdsmen never brushed their teeth. Lactobacillus is widely distributed in fermented foods (such as fermented dairy products), which are habitually consumed by Mongolians residing in pastoral areas. The versatile adaptation and remarkable colonisation ability of Lactobacillus in the human gut has been well demonstrated20,21. Therefore, it is not surprising that the faecal samples of the Khentii Mongolians consisted of a high amount of Lactobacillus.

Diet played an important role in shaping the intestinal microbiota of our subjects. In our research, we analysed the concordance of diet and intestinal microbiota by combining the data obtained using a food frequency questionnaire (FFQ) and pyrosequencing data at five sampling points (January, March, June, September and November). In Ulan Bator, food is plentiful and diverse, so limited seasonal changes were observed in the dietary structure of local residents. Accordingly, the composition of their intestinal microbiota was relatively stable. However, in the Khentii pasturing area, food is scant and simple and the dietary structure of the local residents changed significantly with the season. Thus, the intestinal microbiota of local residents distinctly changed from season to season. This study suggests that seasonally different components of the Khentii diet, such as vegetables, fruits, red meat and/or kumiss, could directly or indirectly modulate the intestinal microbiota profile. Among the microbes that varied with season, Faecalibacterium, Eubacterium and Subdoligranulum produce butyrate22,23,24,25 and may exert anti-inflammatory effects26,27. Escherichia coli/Shigella group is a potentially pathogenic bacterium that possesses pro-inflammatory properties. The seasonal variation of intestinal microbiota should therefore be further investigated due to its health implications.

Changes in the dietary composition have been associated with changes in the composition and metabolism of gut microbial populations. Long-term dietary intake influences the structure and activity of human intestinal microbiota, but it remains unclear how rapidly and reproducibly the human gut micro-biome responds to short-term macronutrient change. Recent research28 confirmed that dietary interventions in humans can alter gut microbial communities only 1 day. In addition, an animal-based diet had a greater effect on the microbiota than a plant-based diet. The study of Cotillard et al.29 on diet-induced weight-loss and weight-stabilisation interventions on obese and overweight individuals concluded that dietary intervention improves low gene richness and clinical phenotypes but appears to be less effective at improving inflammation variables in individuals with lower gene richness.

In this study, 454 pyrosequencing combined with q-PCR technology was applied to examine the diversity of the intestinal microbiota of Mongolians at different phylogenetic levels. In addition, we explored the effects of the adoption of an urban lifestyle and seasonal dietary changes on Mongolians' intestinal microbiota. This basic research will bring a new understanding to the human gut microbiota of different countries and how they are affected by diet.

Methods

Participant recruitment

In this study, 64 healthy Mongolian adults with no history of gastrointestinal-related diseases were recruited (the participants' information is listed in Table S1). Among these participants, 36 volunteers lived a typical modern lifestyle in Ulan Bator, the capital of Mongolia. Twelve volunteers were recruited from the Khentii pasturing area, a typical Mongolian grassland. The local residents maintain a traditional nomadic lifestyle and diet. Sixteen volunteers lived in the TUW province, which contains the suburbs of Ulan Bator. The living standards and experienced scale of urbanisation of these residents were lower than those of Ulan Bator residents but higher than those of Khentii pasture residents. Faecal samples were collected from these volunteers at five time points (January, March, June, September and November). After obtaining written and informed consent, we collected habitual long-term dietary information from all participants using a food frequency questionnaire (the dietary information is shown in a supplementary file). The study protocol was approved by the Ethical Committee of the Inner Mongolia Agriculture University (Hohhot, China).

Stool sample processing and DNA extraction

DNA was extracted from faecal samples using a QIAGEN DNA Stool Mini-Kit (QIAGEN, Hilden, Germany) in combination with a bead-beating method30. Isolated faecal DNA was then used as a template for further analyses.

PCR amplification, quantification, pooling and pyrosequencing

The V1–V3 region of 16S ribosomal RNA (rRNA) genes were amplified as described previously31. The PCR products were quantified using an Agilent DNA 1000 Kit using an Agilent 2100 Bioanalyser (Agilent Technologies, America) according to the manufacturer's instructions. The amplification products were pooled together in equimolar ratios with a final concentration of 100 nmol/L each. These pools sequenced using pyrosequencing with a Roche GS FLX.

Quantitative PCR analysis

Real-time quantitative PCR amplification was performed using an ABI Prism® 7500 Real Time PCR System (Applied Biosystems, California, USA) using the Maxima SYBR Green/ROX qPCR Master Mix (2×) (Thermo Scientific, Massachusetts, USA). The gene-targeted primer sequences, amplicon sizes and annealing temperatures used for each bacterial group are presented in Table S2.

Bioinformatic analyses

Low-quality sequences were removed based on the following criteria: a raw read shorter than 110 nucleotides, a sequence displaying an imperfect match to the barcode or a fuzzy match to at least one end of the 16S rRNA primers based on a standard BLAST search, a variable region shorter than 100 nucleotides, or more than 7% of the bases demonstrated a quality score of less than 20 in the raw read.

Bioinformatic analyses were performed using QIIME (v1.2.1)32 on the extracted high-quality sequences. Briefly, the sequences were aligned using PyNAST33 and clustered under 100% sequence identity using UCLUST34 to obtain the unique V1–V3 sequence set. After representative sequences were selected, the unique sequence set was classified into operational taxonomic units (OTUs) with a 97% threshold identity using UCLUST. ChimeraSlayer35 was employed to remove any potentially chimeric sequences in the representative set of OTUs. The taxonomy of each OTU representative sequence was assigned using the Ribosomal Database Project (RDP)36 classifier with a minimum bootstrap threshold of 80%. OTUs that occurred only once or twice were discarded. A de novo taxonomic tree was constructed using a chimera-checked OTU representative set in FastTree37 for downstream analyses, including alpha and beta diversity calculations. To evaluate alpha diversity, the Shannon-Wiener and Simpson's diversity indices and the Chao1 and rarefaction estimators were calculated. UniFrac38 metrics were calculated to evaluate beta diversity. Both weighted and unweighted calculations were performed prior to a principal coordinate analysis (PCoA).

Statistical analyses

Differences in alpha diversity and the relative abundance of the families and genera in each sample were computed using Mann-Whitney and Kruskal-Wallis tests. The gut microbiota were clustered among the different groups using a multivariate analysis of variance (MANOVA) test on a PCoA based on weighted and unweighted Unifrac metrics. The aforementioned statistical analyses were conducted using Matlab® (The MathWorks, Natick, MA, USA). A partial least-squares discriminant analysis (PLS-DA) was used to identify any correlation between food intake and subjects location. The network was constructed using the software Cytoscape (version 2.6.0). Data from the food frequency questionnaire and the microbiota β-diversity were analysed using the procrustes routine in QIIME (V1.5).

Additional information

Nucleotide sequence accession numbers: The sequence data reported in this paper have been deposited in the MG-RAST database (Project No. 8437).