You are here

Article Text

Article menu

Download PDFPDF

Gut microbiota
Original article
The effect of heritability and host genetics on the gut microbiota and metabolic syndrome
  1. Mi Young Lim1,2,
  2. Hyun Ju You1,2,3,
  3. Hyo Shin Yoon1,
  4. Bomi Kwon1,
  5. Jae Yoon Lee1,
  6. Sunghee Lee1,
  7. Yun-Mi Song4,
  8. Kayoung Lee5,
  9. Joohon Sung6,
  10. GwangPyo Ko1,3,7
  1. 1Department of Environmental Health Sciences, Graduate School of Public Health, Seoul National University, Seoul, Republic of Korea
  2. 2Institute of Health and Environment, Seoul National University, Seoul, Republic of Korea
  3. 3Center for Human and Environmental Microbiome, Seoul National University, Seoul, Republic of Korea
  4. 4Department of Family Medicine, Samsung Medical Center, Sungkyunkwan University School of Medicine, Seoul, Republic of Korea
  5. 5Department of Family Medicine, Busan Paik Hospital, Inje University College of Medicine, Busan, Republic of Korea
  6. 6Department of Epidemiology, Graduate School of Public Health, Seoul National University, Seoul, Republic of Korea
  7. 7N-Bio, Seoul National University, Seoul, Republic of Korea
  1. Correspondence to Dr GwangPyo Ko, Department of Environmental Health Sciences, Graduate School of Public Health, Seoul National University, 1 Gwanak-ro, Gwanak-gu, Seoul 151-742, Republic of Korea; gko{at}snu.ac.kr

Abstract

Objective Metabolic syndrome (MetS) arises from complex interactions between host genetic and environmental factors. Although it is now widely accepted that the gut microbiota plays a crucial role in host metabolism, current knowledge on the effect of host genetics on specific gut microbes related to MetS status remains limited. Here, we investigated the links among host genetic factors, gut microbiota and MetS in humans.

Design We characterised the gut microbial community composition of 655 monozygotic (n=306) and dizygotic (n=74) twins and their families (n=275), of which approximately 18% (121 individuals) had MetS. We evaluated the association of MetS status with the gut microbiota and estimated the heritability of each taxon. For the MetS-related and heritable taxa, we further investigated their associations with the apolipoprotein A-V gene (APOA5) single nucleotide polymorphism (SNP) rs651821, which is known to be associated with triglyceride levels and MetS.

Results Individuals with MetS had a lower gut microbiota diversity than healthy individuals. The abundances of several taxa were associated with MetS status; Sutterella, Methanobrevibacter and Lactobacillus were enriched in the MetS group, whereas Akkermansia, Odoribacter and Bifidobacterium were enriched in the healthy group. Among the taxa associated with MetS status, the phylum Actinobacteria, to which Bifidobacterium belongs, had the highest heritability (45.7%). Even after adjustment for MetS status, reduced abundances of Actinobacteria and Bifidobacterium were significantly linked to the minor allele at the APOA5 SNP rs651821.

Conclusions Our results suggest that an altered microbiota composition mediated by a specific host genotype can contribute to the development of MetS.

  • ENTERIC BACTERIAL MICROFLORA
  • ENERGY METABOLISM
  • GENETIC POLYMORPHISMS

https://doi.org/10.1136/gutjnl-2015-311326

Statistics from Altmetric.com

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

    Significance of this study

    What is already known on this subject?

    • Gut microbiota composition is associated with metabolic diseases such as obesity and type 2 diabetes.

    • Host genetics influence the abundance of human gut microbes.

    • Variation in specific host genes can contribute to alteration of the gut microbiome, leading to increased disease susceptibility.

    What are the new findings?

    • MetS has a marginal effect on the overall gut microbial community structure, but is associated with alterations in the abundance of specific gut microbes.

    • Several MetS-related gut microbial taxa are heritable, and their abundances are associated with the APOA5 single nucleotide polymorphism rs651821, which is a known genetic risk factor for MetS.

    How might it impact on clinical practice in the foreseeable future?

    • Our results emphasise the need for consideration of host genotype when developing future microbiome-targeted therapeutic strategies for metabolic disorders such as MetS.

    Introduction

    Due to recent dramatic shifts in human lifestyle, metabolic syndrome (MetS), which consists of multiple clinical factors that lead to a high risk of type 2 diabetes (T2D) and cardiovascular disease,1 has become increasingly prevalent. It was estimated that approximately 20–30% of adults suffer from MetS worldwide.2 The hallmarks of MetS include abdominal obesity, dyslipidaemia, hypertension and hyperglycaemia. These symptoms are closely associated with insulin resistance and chronic low-grade inflammation, which lead to the development of T2D and cardiovascular disease.3 Although several factors have been implicated in the development of MetS—including host genetic factors (eg, single nucleotide polymorphisms (SNPs) in or near the genes encoding lipoprotein lipase, cholesteryl ester transfer protein and apolipoprotein A (APOA)-cluster), dietary habits (eg, a high-fat/high-calorie diet) and sedentary lifestyle (eg, physical inactivity)4 ,5—the pathogenesis of MetS has not yet been fully elucidated.

    Recently, numerous studies have indicated that the gut microbiota is involved in the energy metabolism of the host, regulating nutrient absorption, gut hormone production, fat storage or gut permeability.6–8 In addition, accumulated data have suggested that dysbiosis of the gut microbiota is closely related to the development of obesity and, subsequently, other hallmarks of MetS.9 ,10 These data imply a critical role of the gut microbiota in the development of MetS in humans, but few studies have investigated the gut microbiota of individuals in relation to MetS.

    The composition of the gut microbiota is modulated by multiple factors, including the age, sex and diet of the host.11 Additionally, increasing evidence from mouse models has suggested that host genetics also contribute to shaping the gut microbiota.12 ,13 Recently, Goodrich et al14 found that a number of gut microbial taxa, including Christensenellaceae, were heritable in humans, which strongly suggests that the human gut microbiota can be influenced by host genetics. At the level of specific loci, several studies have demonstrated that variations in several host genes, such as nucleotide-binding oligomerisation domain 2 and fucosyltransferase 2, may contribute to alterations in gut microbial community structure, and consequently, affect Crohn's disease (CD) susceptibility in humans.15 ,16 Therefore, it can be hypothesised that MetS, gut microbial composition and host genotype are inter-related. However, to our knowledge, no studies to date have investigated which specific gut microbes are responsible for MetS, and whether such microbes are related to SNPs, in particular host genes.

    The APOA5 gene encodes apolipoprotein A-V, a component of chylomicrons, very low-density lipoprotein, and high-density lipoprotein (HDL) particles.17 This gene, which is located proximal to the APOA1/C3/A4 gene cluster on chromosome 11, appears to be involved in the regulation of triglyceride metabolism. Human APOA5 transgenic mice had significantly reduced plasma triglyceride levels compared with their wild type counterparts, whereas mice lacking the APOA5 gene had triglyceride levels fourfold higher than control mice.18 Moreover, it has been shown in humans that SNPs in the APOA5 gene are associated with increased levels of plasma triglyceride. The rs651821 SNP, located 3 bp upstream from the start codon of APOA5, showed significant associations with increased risk of hypertriglyceridaemia or MetS in various populations.19–21 Because APOA5 is also expressed in the intestine, although at a considerably lower level than in the liver,22 we hypothesised that there would be a relationship between the APOA5 SNP rs651821 and the gut microbiota, and that such relationship might be associated with MetS.

    Here, we investigated the gut microbiota of 655 Korean twins and their families to identify specific gut microbes associated with MetS status, estimate heritability of the gut microbes, and determine the associations of MetS-related and heritable gut microbes with an APOA5 genetic variant.

    Materials and methods

    Description of study population and their specimens

    A total of 655 participants in the Healthy Twin Study in Korea23 were included in this study. The study population consisted of 153 monozygotic (MZ) twin pairs (n=306), 37 dizygotic (DZ) twin pairs (n=74), and their parents and siblings (n=275) (table 1). All subjects had not taken antibiotics or cold medicine within 3 months prior to sampling. Overall, the mean subject age was 47.01 (±12.19 SD) years, with a range of 20 years to 81 years, and 58.47% of the subjects were female.

    View this table:
    Table 1

    Characteristics of participants in this study

    Faecal samples from the participants were obtained at home and immediately frozen in a home freezer. Then, frozen faecal samples were transferred to clinics and stored at −80°C until analysis. All participants answered questionnaires covering lifestyle, medication and disease history, and underwent clinical tests, biochemical tests and anthropometrical measurements. Genotyping of the participants was performed using the Affymetrix Genome-Wide Human SNP Array V.6.0 (Affymetrix, Santa Clara, California, USA).

    Measures of MetS components and definition of MetS

    Waist circumference was measured at a level midway between the lowest lateral border of the ribs and the uppermost lateral iliac crest. Blood pressure was measured using a standard manual sphygmomanometer. Blood samples were drawn from an antecubital vein after overnight fasting for at least 8 h. Triglyceride and HDL cholesterol were assayed using homogeneous and enzymatic methods, respectively. Fasting blood sugar (FBS) was measured using a hexokinase enzymatic method.

    MetS was defined following the revised National Cholesterol Education Program Adult Treatment Panel III criteria24 with the Korean-specific waist circumference cut-off values for abdominal obesity.25 The subjects were considered to have MetS if they had three or more of the following five criteria: (1) waist circumference ≥90 cm in men or ≥85 cm in women, (2) blood pressure ≥130/85 mm Hg, (3) triglycerides ≥150 mg/dL (1.7 mmol/L), (4) HDL cholesterol <40 mg/dL (1.03 mmol/L) in men or <50 mg/dL (1.3 mmol/L) in women, and (5) FBS ≥100 mg/dL (5.6 mmol/L).

    Analysis of the gut microbiome

    DNA was extracted from the faecal samples. The V4 region of the 16S rRNA gene was amplified with the Illumina-adapted universal primers 515F/806R26 and sequenced on the MiSeq platform (Illumina, San Diego, California, USA). Sequence data generated were analysed using QIIME software (V.1.7.0).27 Overall flow of the analyses is shown in figure 1.

    Figure 1
    Figure 1

    Schematic diagram of the data analysis. MetS, metabolic syndrome; SOLAR, Sequential Oligogenic Linkage Analysis Routines; SNP, single nucleotide polymorphism.

    A full description of the methods and related references are provided in online supplementary data.

    Results

    Association of gut microbial composition with MetS status

    Among the 655 individuals in our study, the subjects who fulfilled the MetS criteria represented 18.47% of the total, 12.75% of MZ twins, 13.51% of DZ twins and 26.18% of family members (table 1). The gut microbiota of the subjects was characterised by 16S rRNA gene sequencing. In total, more than 31.3 million quality-filtered sequences were obtained (47 743±18 711 sequences per sample). We first analysed the relationship between MetS status and α-diversity indexes (Chao1 richness, observed species and Shannon diversity index), and found that all of the three α-diversity indexes of the MetS group were significantly lower than those of the healthy group (see online supplementary figure S1). To test for significant differences in the overall community structure according to MetS status, we next performed analysis of similarities on the β-diversity distances (unweighted/weighted UniFrac and Bray-Curtis distances). The results showed that the overall gut microbial communities of MetS individuals were almost indistinguishable from those of healthy individuals (unweighted UniFrac: r=0.051, p=0.038; weighted UniFrac: r=0.050, p=0.024; Bray-Curtis: r=0.057, p=0.010): the r values were <0.25, even though the p values were <0.05.28 Also, on a two-dimensional non-parametrical multidimensional scaling plot, the gut microbiotas from these two groups were not clearly distinguished (see online supplementary figure S2).

    Supplemental material

    We next evaluated the associations between the abundances of specific gut microbial taxa and MetS status using multivariate association with linear models (MaAsLin).29 The results from the analysis of the filtered set of taxa excluding low abundance (<0.1% mean relative abundance) and unclassified taxa indicated that Lactobacillus (q=0.0556), Sutterella (q=0.0735) and Methanobrevibacter (q=0.0706) were increased markedly in MetS individuals, while Parabacteroides (q=0.1321), Bifidobacterium (q=0.2369), Odoribacter (q=0.1659), Akkermansia (q=0.1129) and Christensenellaceae (q=0.1371) were over-represented in healthy individuals (figure 2 and see online supplementary table S1). Additionally, we tested the associations between MetS status and the full set of taxa, including low abundance and unclassified taxa, to compare the result from the full set with that from the filtered set, and found that the taxa identified as significant features in the filtered set were in the upper ranks of p values in the full set (see online supplementary table S2). The analytical outline employed in this analysis is addressed in detail in the Discussion section.

    Figure 2
    Figure 2

    Differences in the gut microbiota composition between the healthy and metabolic syndrome (MetS) groups. Bar plot of γ-coefficients from MaAsLin analysis assessing associations between microbial taxa and MetS status, with adjustments for age, sex and family structures. Positive (red bar) and negative (blue bar) coefficient values represent taxa enriched in the MetS group and healthy group, respectively. Error bars represent SEs.

    For the MetS-related taxa identified by MaAsLin in the filtered set, we confirmed correlations between the gut microbial abundances and each MetS component, including waist circumference, triglycerides, FBS, blood pressure and HDL cholesterol, using Spearman's rank correlation test. We observed that the taxa enriched in MetS individuals were positively correlated with waist circumference, triglycerides, blood pressure or FBS, but negatively correlated with HDL cholesterol; in contrast, an opposite trend was observed for the taxa enriched in healthy individuals, with few exceptions (see online supplementary figure S3). For example, Lactobacillus, enriched in the MetS individuals, was positively correlated with FBS and waist circumference, but negatively correlated with HDL cholesterol. In contrast, Odoribacter and Rikenellaceae, over-represented in healthy individuals, were positively correlated with HDL cholesterol, and negatively correlated with all other MetS components. We also performed univariate analysis with the filtered set using the linear discriminant analysis effect size,30 and confirmed that most taxa identified as differentially abundant taxa between the healthy and MetS groups were consistent with the results of the MaAsLin analysis (see online supplementary figure S4). In contrast, there was no significant association between the presence/absence of taxa and MetS status, indicating that MetS is more closely related to the relative abundance of gut microbes than to their presence/absence.

    Heritability of MetS-associated gut microbial populations

    Since it is known that MetS is significantly influenced by host genetics,31 ,32 we evaluated the heritability of MetS and its individual traits with adjustments for age and sex in this study population. Consistent with a previous study conducted with Korean twins,31 MetS exhibited high heritability (70.3%). In addition, all of the MetS components exhibited significant heritability, ranging from 23.9% for diastolic blood pressure (DBP) to 73.5% for waist circumference (figure 3A).

    Figure 3
    Figure 3

    Heritability of metabolic syndrome (MetS) phenotype and gut microbiota. (A) Bar graph of the heritability estimates (H2r) of MetS status as well as each MetS component (DBP, diastolic blood pressure; SBP, systolic blood pressure; triglyceride; FBS, fasting blood sugar; waist: waist circumference; HDL, high-density lipoprotein cholesterol). Error bars represent the SEs of the estimates. (B) Dot plot of H2r values of gut microbial taxa. Red dots and blue dots denote H2r values of taxa that were enriched in MetS individuals and healthy individuals, respectively, in MaAsLin analysis. Error bars represent the SEs of the estimates. Only taxa with q values <0.05 are shown.

    To assess the influence of host genetics on the gut microbiota, we compared the β-diversity distances between the gut microbial communities of MZ twin pairs, DZ twin pairs, family members and unrelated individuals. The overall gut microbial community structures were not more similar between MZ twin pairs than between DZ twin pairs (unweighted UniFrac: p=0.146; weighted UniFrac: p=0.915; Bray-Curtis: p=0.907; two-sample t-test, 1000 Monte Carlo permutations), although MZ twin pairs had more similar gut microbial communities compared with those from members of a family or unrelated individuals (p<0.05 for unweighted/weighted UniFrac and Bray-Curtis distances) (see online supplementary figure S5).

    Because MetS is highly heritable (figure 3A) and the gut microbiota is related to MetS status (figure 2), we calculated heritability estimates (H2r) for each taxon after adjusting for MetS as well as age and sex to assess the genetic influence on the abundance of each taxon, independently of MetS status using a variance component method in Sequential Oligogenic Linkage Analysis Routines (SOLAR).33 Among 85 taxa in the filtered set, 50 taxa (58.8%) were significantly heritable, with H2r values ranging between 13.1% and 45.7% (see online supplementary table S3). Focusing on the taxa associated with MetS status in MaAsLin analysis, we found that 17 (63.0%) of the 27 taxa were heritable (figure 3B). Among the taxa enriched in the MetS group, Methanobrevibacter (H2r±SE; 20.9±5.9%) and Lactobacillus (14.8±5.9%) were significantly heritable. Among the taxa enriched in the healthy group, the Actinobacteria phylum (45.7±5.5%), Bifidobacterium (37.4±9.4%), Christensenellaceae (30.6±5.9%) and Odoribacter (19.2±5.8%) exhibited significant heritability. We reanalysed the heritability of microbial taxa after adjusting for age and sex only. These two results were highly correlated regardless of the inclusion of MetS status as a covariate (Pearson correlation coefficient=0.9978, p<2.2e-16).

    To determine whether these results are in line with a previous study carried out in a UK population,14 we additionally estimated the heritability of each taxon for a merged data set of Korean (this study) and UK twin pairs14 using another heritability estimation method: the heritability was estimated as twice the difference between the MZ and DZ intraclass correlation coefficients (ICCs) for the relative abundance of each taxon. In this analysis using ICCs, the families Bifidobacteriaceae and Christensenellaceae showed high heritability estimates (see online supplementary figure S6), which is consistent with the analysis using SOLAR. These results indicate that host genetic factors are significantly associated with the relative abundances of specific gut microbes related to MetS.

    Association of APOA5 SNP rs651821 with MetS-related gut microbial taxa

    To investigate whether a specific host gene can be associated with the abundances of specific gut microbial taxa, we performed a targeted analysis of APOA5 SNP rs651821. Because this SNP has been previously associated with triglyceride levels and MetS,19–21 we first tested whether such relationship existed in this study population with a total of 351 individuals (275 healthy and 76 MetS) whose rs651821 genotype data were available (minor allele frequency=33.1%). We observed that the prevalence of MetS was higher in carriers of the minor C allele than in non-carriers (p=0.049; χ2 test) (figure 4A). Moreover, we found that each minor copy of rs651821 was significantly associated with a 24.65 mg/dL increase in triglyceride (pointwise empirical p value =0.0011) using the family based association tests for quantitative traits (QFAM) procedure in PLINK.34 There were trends of greater increases in DBP, systolic blood pressure (SBP), FBS and waist circumference, and of greater reduction in HDL cholesterol in carriers of the minor allele, although statistically insignificant (figure 4B). These findings confirmed that minor alleles of APOA5 SNP rs651821 are closely linked to increased risk of high triglyceride levels as well as MetS in our cohort.

    Figure 4
    Figure 4

    Associations of APOA5 single nucleotide polymorphism (SNP) rs651821 with metabolic syndrome (MetS). (A) Prevalence of MetS according to APOA5 SNP rs651821. χ2 test: p=0.0488 in the comparison of MetS prevalence among the genotypes. (B) Box plots of each MetS component according to APOA5 SNP rs651821. Triglyceride increased significantly in carriers of the minor allele (pointwise empirical p value (EMP1)=0.0011 in the QFAM procedure). The box plots indicate the median (horizontal solid line), mean (diamond), IQR between the first and third quartiles (box), minimum and maximum values excluding outliers (whiskers), and the number of outliers. DBP, diastolic blood pressure; FBS, fasting blood sugar; HDL, high-density lipoprotein cholesterol; SBP, systolic blood pressure.

    Next, we tested for associations between APOA5 SNP rs651821 and each of 17 MetS-related and heritable gut microbial taxa (listed in figure 3B) using the QFAM module. After adjustment for age and sex, five significant associations were detected between SNP rs651821 and gut microbial taxa. Each additional copy of the minor C allele at APOA5 SNP rs651821 decreased the abundances of the Actinobacteria phylum, the Actinobacteria class, Bifidobacteriales, Bifidobacteriaceae and Bifidobacterium (see online supplementary table S4). Notably, although these taxa were found to be significantly related to MetS status (figure 2), SNP rs651821 remained associated with the abundances of these taxa even after adjusting for MetS status, age and sex (see online supplementary table S5 and figure 5). The β-coefficients suggest that the associations of the SNP with these taxa are of moderate effect size. For example, the β-coefficient for SNP rs651821 with Bifidobacterium abundance (the standardised residuals of the arcsine square-root-transformed abundances regressed against age, sex and MetS status) was approximately −0.25 (see online supplementary table S5). The associations between these taxa and SNP rs651821 were observed even after additional adjustment for triglyceride. Collectively, these data demonstrate that APOA5 SNP rs651821 might be an important host genetic factor in determining the abundances of the health-promoting Bifidobacterium bacteria and the taxa to which this genus belongs.

    Figure 5
    Figure 5

    Associations of APOA5 single nucleotide polymorphism (SNP) rs651821 with specific members of the gut microbiota. Box plots of the standardised residuals of the arcsine square-root-transformed microbial abundances regressed against age, sex and metabolic syndrome (MetS) according to APOA5 SNP rs651821. The box plots indicate the median (horizontal solid line), mean (diamond), the IQR between the first and third quartiles (box), minimum and maximum values excluding outliers (whiskers), and the number of outliers.

    Discussion

    We here report that host genetics is significantly associated with the abundance of specific gut microbial taxa related to MetS. In particular, we found that the Bifidobacterium genus is strongly associated with MetS status and that its abundance is significantly reduced with each additional copy of the minor allele at APOA5 SNP rs651821, leading to an increased risk of MetS.

    In this study, we applied multiple analytical steps to obtain reliable results regarding the associations between host genetics and gut microbial taxa (figure 1). First, we filtered out unclassified taxa from the analysis, because the main goal of the current study was to identify the specific taxa related to host genetics in terms of MetS. The taxa with <0.1% mean relative abundances were also excluded to acquire the results for the taxa that met the statistical conditions for subsequent heritability estimations. Next, we identified the candidate taxa associated with MetS status using MaAsLin pipeline with q<0.25, a threshold employed in previous microbiome studies that allows compensation for the large number of microbial taxa and multiple comparison adjustment.29 ,35 We also identified significantly heritable microbial taxa with q<0.05 and combined these results with the MetS-related taxa obtained from the MaAsLin analysis to identify MetS-related and heritable taxa. Additionally, we investigated the associations between these 17 microbial taxa and a specific SNP using the QFAM procedure with q<0.05. Therefore, we initially applied q<0.25 in the MaAsLin analysis to preclude false negative results arising from an overly stringent false discovery rate threshold for candidate taxa screening, and employed a considerably stricter statistical cut-off in the subsequent analyses to avoid false positive results.

    Decreased gut microbiota diversity has been reported in subjects with obesity.10 ,36 In this study, we also observed that the gut microbial diversity of MetS individuals was significantly lower than that of healthy individuals (see online supplementary figure S1). This result suggests that gut microbiota diversity may act as an important factor in modulating host metabolism, and thus it can be used as a simple indicator of health status. In our additional analysis of the heritability of diversity, we found that all of the tested α-diversity indexes (Chao1 richness, observed species and Shannon diversity index) were not significantly heritable. Thus, the overall diversity of an individual's gut microbiota is less likely to be related to host genetics.

    We identified a specific set of gut microbial taxa that may be beneficial or harmful to MetS-related host metabolism (figure 2). Numerous studies have focused on microbial associations with obesity and T2D in humans, and the results of these studies have been reviewed.37 ,38 We observed that the patterns of gut microbial alterations in MetS individuals were overall highly consistent with the results of previous studies on obesity and T2D (see online supplementary table S6). Specifically, the reduction in Bifidobacterium and the expansion of Betaproteobacteria and Alcaligenaceae in MetS individuals have also been reported to be associated with obesity or T2D in diverse populations.39–41 According to a meta-analysis of obesity-associated gut microbiota alterations,42 Bifidobacterium has been shown to have a consistent antiobesity effect in humans. We also found that Akkermansia, whose antiobesity functions were demonstrated in diet-induced obese mice,43 was enriched in healthy individuals. Additionally, we newly identified several gut microbial taxa that may contribute to the improvement of host metabolism, such as the Porphyromonadaceae family, including the genera Parabacteroides, the Rikenellaceae family and the genus Odoribacter. Notably, the relative abundances of these taxa were enriched in the healthy individuals, and associated with improving at least one of the MetS features (see online supplementary figure S3). Previous studies have reported inconsistent results on the association between Lactobacillus spp and obesity or T2D.44–46 Our results show increased abundance of Lactobacillus in the gut microbiotas of MetS individuals, which is consistent with several previous reports,44 ,46 while one group reported a relative reduction in Lactobacillus paracasei and Lactobacillus plantarum abundance in obese individuals.45 Such discrepancies are likely due to the strain specificity of Lactobacillus spp. Seventeen species of Lactobacillus are associated with the human gut,47 and different functional capacities for attenuating obesity-related complications were Lactobacillus strains specific to high-fat-diet-fed mice.48 These characteristics of Lactobacillus spp may have caused the considerable controversy over the effect of Lactobacillus on the control of adiposity. Therefore, although the mechanisms of interaction between these specific gut microbes and the host metabolism require further elucidation, our findings suggest that specific modulations of the gut microbiota might be an effective strategy for managing MetS.

    Our data show that host genetics is significantly associated with the abundances of specific taxa. The overall gut microbial community structures of MZ twin pairs harboured near-identical degrees of similarity to those of DZ twin pairs (see online supplementary figure S5), as shown in previous studies.10 ,49 This may be due to early life environmental exposures or limited statistical power, as recently pointed out by Goodrich et al.14 Nevertheless, we could not exclude the possibility that specific taxa are influenced by host genetic factors. By applying variance components methods, we found that more than half of the tested taxa were significantly heritable, and the higher taxonomic levels that encompass the Bifidobacterium (Actinobacteria, Actinomycetales and Bifidobacteriaceae) had the highest H2r among the taxa related to MetS status (figure 3B). Consistently, the heritability of Bifidobacteriaceae was also reported in the twins in the UK.14 Because Bifidobacterium dominates the gut microbiota of breastfed infants and decreases with maturation,49 it can be assumed that the abundance of this genus in adult twins could be affected by early shared environment. Yet, we detected no significant influence of the common environments in our study (C2=0.0198, q=0.4620) (see online supplementary table S3). Goodrich et al14 also demonstrated that the Christensenellaceae family was the most heritable taxon in the UK twins, and this family was associated with lean host phenotype. Consistently, we observed a high heritability of the Christensenellaceae family, which was enriched in healthy Korean individuals. By calculating MZ/DZ ICCs in a merged data set of Korean and UK twin pairs, we also found Bifidobacteriaceae and Christensenellaceae to exhibit high heritability (see online supplementary figure S6). Therefore, Bifidobacterium, its parent taxa, and Christensenellaceae are likely to play important roles in improving host metabolism by interacting with host genetics regardless of ethnicity, diet or geographical region. Additionally, we found that the phylum Tenericutes had a similar level of heritability in Korean and UK populations. This phylum was more abundant in the healthy group than in the MetS group, which is in line with a UK study in that the abundance of this phylum was higher in the lean group than in the obese group. On the other hand, the two populations showed some notable differences in heritable taxa as well. The genus Methanobrevibacter, which was enriched in the MetS group, exhibited significant heritability in the Korean population but not in the UK population. Also, the class Gammaproteobacteria and its related families Enterobacteriaceae and Pasteurellaceae, the first of which was enriched in the healthy group, were significantly heritable only in the Korean population. The levels of heritability of the taxa with significant heritability only in the Korean population (H2r<0.3) were lower than those of the taxa with significant heritability in both populations (H2r>0.3), with the exception of Tenericutes, which had an H2r of 0.23. Thus, taxa that are weakly heritable only in one population may be heavily influenced by unique individual environment in other populations, but taxa that are highly heritable in one population may also show significant heritability in other populations as well, even though the relative abundances of those taxa are significantly different between the two populations (see online supplementary figure S6C). These results suggest that there might be common host genetic factors that exert their influence on gut phenotypes and the abundances of highly heritable taxa such as Bifidobacteriaceae and Christensenellaceae. The effect of such host genetic factors would be sufficient to offset any differences in ethnicity, diet or other lifestyle factors. Further studies are needed to determine whether the highly heritable taxa in both the Korean and UK populations also show significant heritability in other diverse populations.

    Through a targeted analysis of the association of APOA5 SNP rs651821 with MetS-related and heritable taxa, we observed that minor alleles of this host genetic variant were significantly associated with decreased abundance of Bifidobacterium and its parent taxa, independent of the individual's MetS status (figure 5), which were simultaneously the health-associated and most heritable taxa (figure 3B). This finding suggests that the APOA5 variant may contribute to the compositional change of the MetS-related gut microbiota. The APOA5 gene is predominantly expressed in the liver, but is also produced at a low level in the intestine.22 The regulatory role of APOA5 in triglyceride metabolism has been demonstrated extensively in mice and humans, especially in liver and plasma.18 ,50 Recently, a novel role of APOA5 in the gut was demonstrated in APOA5-deficient mice—APOA5 regulates lymphatic transport of dietary lipids by modulating the production of chylomicrons in the gut.51 Based on the results of this recent study and our findings, it is likely that the gut environment of carriers of the minor allele of APOA5 SNP rs651821 is different from that of non-carriers in terms of quantity and types of lipids, which may lead to unfavourable conditions of gut ecology for the growth of Bifidobacterium. In high-fat-fed mice, a reduction in Bifidobacterium spp abundance was observed, together with characteristics of metabolic disorders, such as increased endotoxaemia and low-grade inflammation.52 Thus, Bifidobacterium could be a member of the gut microbiota that reacts to the lipids in the gut. Bifidobacterium exerts beneficial effects on host health by producing several bioactive metabolites. For example, Bifidobacterium breve produces conjugated linoleic acid, a group of positional and geometric isomers of linoleic acid, which has been reported to have antiobese, antidiabetogenic and antiatherosclerotic capacities.53 Also, Bifidobacterium is known for its ability to modulate metabolic endotoxaemia.54 Therefore, the reduction in the abundance of health-promoting Bifidobacterium in the MetS group may be significantly associated with the development of MetS.

    Interestingly, in a recent CD patient-based study combining host transcriptomic and microbial profiling, it was found that APOA1, APOA4 and APOC3 gene expression levels within the CD ileum were downregulated and associated with abundances of specific Firmicutes and Bacteroidetes taxa.55 Although the expression of APOA5 gene was not determined, our results are in line with their findings in that apolipoproteins in the APOA1/C3/A4/A5 gene cluster are associated with specific gut microbial taxa, and that this association is important for gut microbiota-related diseases.

    Our study has the following limitations: First, we performed taxon-based analysis at all taxonomic levels from genus to kingdom when identifying gut microbes associated with MetS status and host genetics. Although operational taxonomic units based analysis has the advantage of providing high-resolution characterisation of microbial composition, we examined the data using a taxon-based method because it improves statistical power by reducing the number of multiple comparison adjustments, while preventing false negative results. Second, a targeted association analysis for a single MetS-related SNP with MetS-related and heritable taxa was employed to compensate for the limited statistical power. The influence of host genetics on the gut microbial composition can arise from various and multiple genetic variants such as SNPs, copy-number variations or mutations. For a more thorough understanding of the associations between host genetic variants and microbial taxa, it would be helpful to conduct a genome-wide association study. Third, our study falls short of discerning the sequential causal relationship. Future studies using knockout animal models are warranted to confirm causal links among APOA5 genotype, alteration of gut microbiota and MetS development.

    In summary, we identified a specific set of gut microbial taxa that were significantly associated with MetS status and their relationship with host genetics. Our results suggest the possibility that alteration of the gut microbiota, which may be mediated in part by specific host genetic variants, can contribute to the development of MetS, and that development of microbiome-targeted therapeutic strategies for metabolic disorders should take into account host genetics.

    References

    1. ↵
      1. Wilson PW,
      2. D'Agostino RB,
      3. Parise H, et al
      . Metabolic syndrome as a precursor of cardiovascular disease and type 2 diabetes mellitus. Circulation 2005;112:3066–72. doi:10.1161/CIRCULATIONAHA.105.539528
      OpenUrlAbstract/FREE Full Text
    2. ↵
      1. Grundy SM
      . Metabolic syndrome pandemic. Arterioscler Thromb Vasc Biol 2008;28:629–36. doi:10.1161/ATVBAHA.107.151092
      OpenUrlAbstract/FREE Full Text
    3. ↵
      1. Esposito K,
      2. Marfella R,
      3. Ciotola M, et al
      . Effect of a Mediterranean-style diet on endothelial dysfunction and markers of vascular inflammation in the metabolic syndrome: a randomized trial. JAMA 2004;292:1440–6. doi:10.1001/jama.292.12.1440
      OpenUrlCrossRefPubMedWeb of Science
    4. ↵
      1. Kaur J
      . A comprehensive review on metabolic syndrome. Cardiol Res Pract 2014;2014:943162. doi:10.1155/2014/943162
      OpenUrlPubMed
    5. ↵
      1. Kraja AT,
      2. Vaidya D,
      3. Pankow JS, et al
      . A bivariate genome-wide approach to metabolic syndrome STAMPEED consortium. Diabetes 2011;60:1329–39. doi:10.2337/db10-1011
      OpenUrlAbstract/FREE Full Text
    6. ↵
      1. Samuel BS,
      2. Shaito A,
      3. Motoike T
      . Effects of the gut microbiota on host adiposity are modulated by the short-chain fatty-acid binding G protein-coupled receptor, Gpr41. Proc Natl Acad Sci USA 2008;105:16767–72. doi:10.1073/pnas.0808567105
      OpenUrlAbstract/FREE Full Text
    7. ↵
      1. Bäckhed F,
      2. Ding H,
      3. Wang T, et al
      . The gut microbiota as an environmental factor that regulates fat storage. Proc Natl Acad Sci USA 2004;101:15718–23. doi:10.1073/pnas.0407076101
      OpenUrlAbstract/FREE Full Text
    8. ↵
      1. Cani PD,
      2. Bibiloni R,
      3. Knauf C, et al
      . Changes in gut microbiota control metabolic endotoxemia-induced inflammation in high-fat diet–induced obesity and diabetes in mice. Diabetes 2008;57:1470–81. doi:10.2337/db07-1403
      OpenUrlAbstract/FREE Full Text
    9. ↵
      1. Zupancic ML,
      2. Cantarel BL,
      3. Liu Z, et al
      . Analysis of the gut microbiota in the old order Amish and its relation to the metabolic syndrome. PLoS ONE 2012;7:e43052. doi:10.1371/journal.pone.0043052
      OpenUrlCrossRefPubMed
    10. ↵
      1. Turnbaugh P,
      2. Hamady M,
      3. Yatsunenko T, et al
      . A core gut microbiome in obese and lean twins. Nature 2009;457:480–4. doi:10.1038/nature07540
      OpenUrlCrossRefPubMedWeb of Science
    11. ↵
      1. Delzenne NM,
      2. Neyrinck AM,
      3. Bäckhed F, et al
      . Targeting gut microbiota in obesity: effects of prebiotics and probiotics. Nat Rev Endocrinol 2011;7:639–46. doi:10.1038/nrendo.2011.126
      OpenUrlCrossRefPubMed
    12. ↵
      1. Zhang C,
      2. Zhang M,
      3. Wang S, et al
      . Interactions between gut microbiota, host genetics and diet relevant to development of metabolic syndromes in mice. ISME J 2009;4:232–41. doi:10.1038/ismej.2009.112
      OpenUrlPubMed
    13. ↵
      1. Benson AK,
      2. Kelly SA,
      3. Legge R, et al
      . Individuality in gut microbiota composition is a complex polygenic trait shaped by multiple environmental and host genetic factors. Proc Natl Acad Sci USA 2010;107:18933–8. doi:10.1073/pnas.1007028107
      OpenUrlAbstract/FREE Full Text
    14. ↵
      1. Goodrich JK,
      2. Waters JL,
      3. Poole AC, et al
      . Human genetics shape the gut microbiome. Cell 2014;159:789–99. doi:10.1016/j.cell.2014.09.053
      OpenUrlCrossRefPubMedWeb of Science
    15. ↵
      1. Rehman A,
      2. Sina C,
      3. Gavrilova O, et al
      . Nod2 is essential for temporal development of intestinal microbial communities. Gut 2011;60:1354–62. doi:10.1136/gut.2010.216259
      OpenUrlAbstract/FREE Full Text
    16. ↵
      1. Rausch P,
      2. Rehman A,
      3. Künzel S, et al.
      Colonic mucosa-associated microbiota is influenced by an interaction of Crohn disease and FUT2 (Secretor) genotype. Proc Natl Acad Sci USA 2011;108:19030–5. doi:10.1073/pnas.1106408108
      OpenUrlAbstract/FREE Full Text
    17. ↵
      1. O'Brien PJ,
      2. Alborn WE,
      3. Sloan JH, et al
      . The novel apolipoprotein A5 is present in human serum, is associated with VLDL, HDL, and chylomicrons, and circulates at very low concentrations compared with other apolipoproteins. Clin Chem 2005;51:351–9. doi:10.1373/clinchem.2004.040824
      OpenUrlAbstract/FREE Full Text
    18. ↵
      1. Pennacchio LA,
      2. Olivier M,
      3. Hubacek JA, et al
      . An apolipoprotein influencing triglycerides in humans and mice revealed by comparative sequencing. Science 2001;294:169–73. doi:10.1126/science.1064852
      OpenUrlAbstract/FREE Full Text
    19. ↵
      1. Tan A,
      2. Sun J,
      3. Xia N, et al
      . A genome-wide association and gene–environment interaction study for serum triglycerides levels in a healthy Chinese male population. Hum Mol Genet 2012;21:1658–64. doi:10.1093/hmg/ddr587
      OpenUrlAbstract/FREE Full Text
    20. ↵
      1. Hong K-W,
      2. Jeong SW,
      3. Chung M, et al
      . Association between expression quantitative trait loci and metabolic traits in two Korean populations. PLoS ONE 2014;9:e114128. doi:10.1371/journal.pone.0114128
      OpenUrlCrossRefPubMed
    21. ↵
      1. Yamada Y,
      2. Ichihara S,
      3. Kato K, et al
      . Genetic risk for metabolic syndrome: examination of candidate gene polymorphisms related to lipid metabolism in Japanese people. J Med Genet 2008;45:22–8. doi:10.1136/jmg.2007.052415
      OpenUrlAbstract/FREE Full Text
    22. ↵
      1. Guardiola M,
      2. Alvaro A,
      3. Vallvé J, et al
      . APOA5 gene expression in the human intestinal tissue and its response to in vitro exposure to fatty acid and fibrate. Nutr Metab Cardiovasc Dis 2012;22:756–62. doi:10.1016/j.numecd.2010.12.003
      OpenUrlCrossRefPubMed
    23. ↵
      1. Sung J,
      2. Cho SI,
      3. Lee K, et al
      . Healthy Twin: a twin-family study of Korea—protocols and current status. Twin Res Hum Genet 2006;9:844–8. doi:10.1375/183242706779462822
      OpenUrlCrossRefPubMed
    24. ↵
      1. Alberti KG,
      2. Eckel RH,
      3. Grundy SM, et al
      . Harmonizing the metabolic syndrome a joint interim statement of the International Diabetes Federation Task Force on Epidemiology and Prevention; National Heart, Lung, and Blood Institute; American Heart Association; World Heart Federation; International Atherosclerosis Society; and International Association for the Study of Obesity. Circulation 2009;120:1640–5. doi:10.1161/CIRCULATIONAHA.109.192644
      OpenUrlAbstract/FREE Full Text
    25. ↵
      1. Lee SY,
      2. Park HS,
      3. Kim DJ, et al
      . Appropriate waist circumference cutoff points for central obesity in Korean adults. Diabetes Res Clin Pract 2007;75:72–80. doi:10.1016/j.diabres.2006.04.013
      OpenUrlCrossRefPubMed
    26. ↵
      1. Caporaso JG,
      2. Lauber CL,
      3. Walters WA, et al
      . Global patterns of 16S rRNA diversity at a depth of millions of sequences per sample. Proc Natl Acad Sci USA 2011;108:4516–22. doi:10.1073/pnas.1000080107
      OpenUrlAbstract/FREE Full Text
    27. ↵
      1. Caporaso JG,
      2. Kuczynski J,
      3. Stombaugh J, et al
      . QIIME allows analysis of high-throughput community sequencing data. Nat Methods 2010;7:335–6. doi:10.1038/nmeth.f.303
      OpenUrlCrossRefPubMedWeb of Science
    28. ↵
      1. Ramette A
      . Multivariate analyses in microbial ecology. FEMS Microbiol Ecol 2007;62:142–60. doi:10.1111/j.1574-6941.2007.00375.x
      OpenUrlAbstract/FREE Full Text
    29. ↵
      1. Morgan XC,
      2. Tickle TL,
      3. Sokol H, et al
      . Dysfunction of the intestinal microbiome in inflammatory bowel disease and treatment. Genome Biol 2012;13:R79. doi:10.1186/gb-2012-13-9-r79
      OpenUrlCrossRefPubMed
    30. ↵
      1. Segata N,
      2. Izard J,
      3. Waldron L, et al
      . Metagenomic biomarker discovery and explanation. Genome Biol 2011;12:R60. doi:10.1186/gb-2011-12-6-r60
      OpenUrlCrossRefPubMed
    31. ↵
      1. Sung J,
      2. Lee K,
      3. Song YM
      . Heritabilities of the metabolic syndrome phenotypes and related factors in Korean twins. J Clin Endocrinol Metab 2009;94:4946–52. doi:10.1210/jc.2009-1268
      OpenUrlCrossRefPubMed
    32. ↵
      1. Lin HF,
      2. Boden-Albala B,
      3. Juo SH, et al
      . Heritabilities of the metabolic syndrome and its components in the Northern Manhattan Family Study. Diabetologia 2005;48:2006–12. doi:10.1007/s00125-005-1892-2
      OpenUrlCrossRefPubMedWeb of Science
    33. ↵
      1. Almasy L,
      2. Blangero J
      . Multipoint quantitative-trait linkage analysis in general pedigrees. Am J Hum Genet 1998;62:1198–211. doi:10.1086/301844
      OpenUrlCrossRefPubMedWeb of Science
    34. ↵
      1. Purcell S,
      2. Neale B,
      3. Todd-Brown K, et al
      . PLINK: a tool set for whole-genome association and population-based linkage analyses. Am J Hum Genet 2007;81:559–75. doi:10.1086/519795
      OpenUrlCrossRefPubMed
    35. ↵
      1. Knights D,
      2. Silverberg MS,
      3. Weersma RK, et al
      . Complex host genetics influence the microbiome in inflammatory bowel disease. Genome Med 2014;6:107. doi:10.1186/s13073-014-0107-1
      OpenUrlCrossRefPubMed
    36. ↵
      1. Le Chatelier E,
      2. Nielsen T,
      3. Qin J, et al
      . Richness of human gut microbiome correlates with metabolic markers. Nature 2013;500:541–6. doi:10.1038/nature12506
      OpenUrlCrossRefPubMedWeb of Science
    37. ↵
      1. Sanz Y,
      2. Olivares M,
      3. Moya-Pérez Á, et al
      . Understanding the role of gut microbiome in metabolic disease risk. Pediatr Res 2014;77:236–44. doi:10.1038/pr.2014.170
      OpenUrlPubMed
    38. ↵
      1. Tagliabue A,
      2. Elli M
      . The role of gut microbiota in human obesity: recent findings and future perspectives. Nutr Metab Cardiovasc Dis 2013;23:160–8. doi:10.1016/j.numecd.2012.09.002
      OpenUrlCrossRefPubMed
    39. ↵
      1. Wu X,
      2. Ma C,
      3. Han L, et al
      . Molecular characterisation of the faecal microbiota in patients with type II diabetes. Curr Microbiol 2010;61:69–78. doi:10.1007/s00284-010-9582-9
      OpenUrlCrossRefPubMed
    40. ↵
      1. Larsen N,
      2. Vogensen FK,
      3. van den Berg FW, et al
      . Gut microbiota in human adults with type 2 diabetes differs from non-diabetic adults. PLoS ONE 2010;5:e9085. doi:10.1371/journal.pone.0009085
      OpenUrlCrossRefPubMed
    41. ↵
      1. Zhang H,
      2. DiBaise JK,
      3. Zuccolo A, et al
      . Human gut microbiota in obesity and after gastric bypass. Proc Natl Acad Sci USA 2009;106:2365–70. doi:10.1073/pnas.0812600106
      OpenUrlAbstract/FREE Full Text
    42. ↵
      1. Angelakis E,
      2. Armougom F,
      3. Million M, et al
      . The relationship between gut microbiota and weight gain in humans. Future Microbiol 2012;7:91–109. doi:10.2217/fmb.11.142
      OpenUrlCrossRefPubMed
    43. ↵
      1. Everard A,
      2. Belzer C,
      3. Geurts L, et al
      . Cross-talk between Akkermansia muciniphila and intestinal epithelium controls diet-induced obesity. Proc Natl Acad Sci USA 2013;110:9066–71. doi:10.1073/pnas.1219451110
      OpenUrlAbstract/FREE Full Text
    44. ↵
      1. Armougom F,
      2. Henry M,
      3. Vialettes B, et al
      . Monitoring bacterial community of human gut microbiota reveals an increase in Lactobacillus in obese patients and Methanogens in anorexic patients. PLoS ONE 2009;4:e7125. doi:10.1371/journal.pone.0007125
      OpenUrlCrossRefPubMed
    45. ↵
      1. Million M,
      2. Maraninchi M,
      3. Henry M, et al
      . Obesity-associated gut microbiota is enriched in Lactobacillus reuteri and depleted in Bifidobacterium animalis and Methanobrevibacter smithii. Int J Obes 2011;36:817–25. doi:10.1038/ijo.2011.153
      OpenUrl
    46. ↵
      1. Karlsson FH,
      2. Tremaroli V,
      3. Nookaew I, et al
      . Gut metagenome in European women with normal, impaired and diabetic glucose control. Nature 2013;498:99–103. doi:10.1038/nature12198
      OpenUrlCrossRefPubMedWeb of Science
    47. ↵
      1. Walter J
      . Ecological role of lactobacilli in the gastrointestinal tract: implications for fundamental and biomedical research. Appl Environ Microbiol 2008;74:4985–96. doi:10.1128/AEM.00753-08
      OpenUrlFREE Full Text
    48. ↵
      1. Wang J,
      2. Tang H,
      3. Zhang C, et al
      . Modulation of gut microbiota during probiotic-mediated attenuation of metabolic syndrome in high fat diet-fed mice. ISME J 2015;9:1–15. doi:10.1038/ismej.2014.99
      OpenUrlCrossRefPubMed
    49. ↵
      1. Yatsunenko T,
      2. Rey FE,
      3. Manary MJ, et al
      . Human gut microbiome viewed across age and geography. Nature 2012;486:222–7. doi:10.1038/nature11053
      OpenUrlCrossRefPubMedWeb of Science
    50. ↵
      1. Schaap FG,
      2. Rensen PC,
      3. Voshol PJ, et al
      . ApoAV reduces plasma triglycerides by inhibiting very low density lipoprotein-triglyceride (VLDL-TG) production and stimulating lipoprotein lipase-mediated VLDL-TG hydrolysis. J Biol Chem 2004;279:27941–7. doi:10.1074/jbc.M403240200
      OpenUrlAbstract/FREE Full Text
    51. ↵
      1. Zhang LS,
      2. Xu M,
      3. Yang Q, et al
      . Apolipoprotein AV deficiency enhances chylomicron production in lymph fistula mice. Am J Physiol Gastrointest Liver Physiol 2015;308:G634–42. doi:10.1152/ajpgi.00339.2014
      OpenUrlAbstract/FREE Full Text
    52. ↵
      1. Cani PD,
      2. Neyrinck AM,
      3. Fava F, et al
      . Selective increases of bifidobacteria in gut microflora improve high-fat-diet-induced diabetes in mice through a mechanism associated with endotoxaemia. Diabetologia 2007;50:2374–83. doi:10.1007/s00125-007-0791-0
      OpenUrlCrossRefPubMedWeb of Science
    53. ↵
      1. Russell DA,
      2. Ross RP,
      3. Fitzgerald GF, et al
      . Metabolic activities and probiotic potential of bifidobacteria. Int J Food Microbiol 2011;149:88–105. doi:10.1016/j.ijfoodmicro.2011.06.003
      OpenUrlCrossRefPubMed
    54. ↵
      1. Griffiths EA,
      2. Duffy LC,
      3. Schanbacher FL, et al
      . In vivo effects of bifidobacteria and lactoferrin on gut endotoxin concentration and mucosal immunity in Balb/c mice. Dig Dis Sci 2004;49:579–89.
      OpenUrlCrossRefPubMedWeb of Science
    55. ↵
      1. Haberman Y,
      2. Tickle TL,
      3. Dexheimer PJ, et al
      . Pediatric Crohn disease patients exhibit specific ileal transcriptome and microbiome signature. J Clin Invest 2014;124:3617. doi:10.1172/JCI75436
      OpenUrlCrossRefPubMed
    View Abstract

    Footnotes

    • Contributors MYL and GPK conceived and designed the study. JS, Y-MS and KL performed sample collection. MYL, HJY, HSY, BK, JYL and SL performed the experiments. MYL analysed the data, and MYL and GPK wrote the manuscript. All authors reviewed the manuscript.

    • Funding This work was supported by the National Research Foundation of Korea (NRF) (NRF-2010-0029113 and NRF-2015R1A2A1A10054078). MYL was supported by the Global PhD Fellowship Program (NRF-2011-0007454).

    • Competing interests None declared.

    • Ethics approval The study protocol was approved by the Institutional Review Boards of Samsung Medical Center (IRB file No. 2005-08-113), Busan Paik Hospital (IRB file No. 05-037) and Seoul National University (IRB file No. 144-2011-07-11). Written informed consent was obtained from all participants.

    • Provenance and peer review Not commissioned; externally peer reviewed.

    • Data sharing statement The sequences from this study were deposited in the European Nucleotide Archive under the study accession number ERP010289. Additional metadata used in this study are available in online supplementary table S7. The operational taxonomic unit (OTU) table containing all OTUs with more than 10 total observations is available as online supplementary table S8.