The authors have declared that no competing interests exist.
Conceived and designed the experiments: CMF ARS NS RH SS. Performed the experiments: KAR SC CJ EM MLZ. Analyzed the data: BLC ZL EFD RK WAW BM DK. Wrote the paper: CMF ARS BLC EFD.
Obesity has been linked to the human gut microbiota; however, the contribution of gut bacterial species to the obese phenotype remains controversial because of conflicting results from studies in different populations. To explore the possible dysbiosis of gut microbiota in obesity and its metabolic complications, we studied men and women over a range of body mass indices from the Old Order Amish sect, a culturally homogeneous Caucasian population of Central European ancestry. We characterized the gut microbiota in 310 subjects by deep pyrosequencing of bar-coded PCR amplicons from the V1–V3 region of the 16S rRNA gene. Three communities of interacting bacteria were identified in the gut microbiota, analogous to previously identified gut enterotypes. Neither BMI nor any metabolic syndrome trait was associated with a particular gut community. Network analysis identified twenty-two bacterial species and four OTUs that were either positively or inversely correlated with metabolic syndrome traits, suggesting that certain members of the gut microbiota may play a role in these metabolic derangements.
Obesity, the accumulation of excess body fat has a negative impact on morbidity, mortality, and quality of life through its complications, which include cardiovascular disease, type 2 diabetes, osteoarthritis, and certain cancers
A 2005 publication by Ley et al.
To explore the possible dysbiosis of gut microbiota in obesity and its metabolic complications in humans, we studied Old Order Amish subjects from Lancaster County, Pennsylvania. The Amish are a genetically closed homogeneous Caucasian population of Central European ancestry ideal for such a study because of their high degree of social cohesiveness and common lifestyle
We enrolled a total of 310 adult subjects, of whom 112 were male and 198 were female. In this cohort, mean age and BMI was higher, and manifestations of the metabolic syndrome
Men | Women | |
|
112 | 198 |
|
46.0±12.7 | 49.5±13.4 |
|
27.2±4.0 (19.3–42.3) | 30.3±5.9 (16.7–51.1) |
|
95.6±11.2 | 90.3±12.3 |
|
117.3±12.2 | 118.2±15.9 |
|
71.0±8.0 | 70.8±9.1 |
|
208.7±42.5 | 214.4±50.2 |
|
54.8±13.1 | 62.0±14.1 |
|
138.6±39.0 | 134.7±45.6 |
|
76.4±45.5 | 88.5±54.0 |
|
87.4±8.0 | 87.3±11.2 |
|
26.8 | 38.9 |
|
14.3 | 21.2 |
|
8.9 | 18.7 |
|
3.6 | 11.6 |
|
7.1 | 10.6 |
Metabolic syndrome traits were defined by NHLBI criteria: (1) fasting triglycerides >150 mg/dl (or on triglyceride lowering medication prescribed by a physician); (2) fasting HDL≤50 mg/dl for women or <40 mg/dl for men (or on HDL raising medication prescribed by a physician); (3) either or both systolic or diastolic blood pressure >130/85 mm Hg (or on anti-hypertension medication prescribed by a physician); (4) fasting glucose ≥100 mg/dl (or on anti-diabetes medication prescribed by a physician). Waist circumference was not included in our definitions because of the high correlation between waist circumference and BMI.
All but one of the twenty-five most abundant genera were present in at least 75% of all subjects, and these include members of the Firmicutes, Bacteroidetes, Tenericutes, Actinobacteria, and Proteobacteria (
Firmicutes | Bacteroidetes | Tenericutes |
|
|
|
|
|
|
|
||
|
||
|
||
|
Despite the substantial overlap in genera present in the gut microbiota of the 310 subjects, a significant amount of inter-individual variation was observed with respect to the relative abundance of both the predominant and rare genera (
Using a random matrix theory-based framework
Bacterial networks were identified based on statistically significant correlations among genera using the Louvain algorithm. Network I:
These are determined by the classical multidimensional scaling algorithm so that the Euclidean distance between points approximates the unweighted UniFrac distance between the OTU profiles of the corresponding samples, colored by Abundance of (A) Prevotella, (B) Bacteroides, and (C) Firmicutes.
The three networks appear to reflect the main large-scale trends in gut microbial populations and suggest that gut bacterial community structures are driven by high abundance and high variability in populations of
Our cohort of 310 subjects included 113 nuclear families having two or more members phenotyped. These nuclear families contributed a total of 54 spouse pairs, 76 sibling pairs, and 42 parent-offspring pairs. Community structure concordance rates tended to be lower for the spouse pairs (38.9%) than for the sibling (46.1%) and parent-offspring (52.4%) pairs (p = 0.32 for comparison of spouse pairs with the combined set of sibling and parent-offspring pairs). Larger sample sizes will be required to get better estimates of the relative contributions of relatedness and household effects to community structure.
We tested the association of groups (coded as a class variable with 3 levels) with each available clinical factor, while adjusting for age and sex (a 2 df test). Neither BMI nor any metabolic syndrome trait was associated with a particular group (
Variable | Network 1 | Network 2 | Network 3 | p-value forNetwork Effect | Contrast | p-value |
Age (yrs) | 47.5±1.1 | 49.6±2.4 | 48.6±1.2 | 0.82 | ||
Sex (% male) | 44.5 | 23.3 | 30.6 | 0.01 | Network 2 vs. Network 3Network 2 vs. Network 1Network 3 vs. Network 1 | 0.370.020.02 |
BMI (kg/m2) | 29.1±0.5 | 29.3±0.8 | 29.4±0.5 | 0.79 | ||
Waist (cm) | 92.8±1.0 | 91.8±29.3 | 91.8±1.1 | 0.95 | ||
HDL-cholesterol (mg/dl) | 59.4±1.3 | 59.3±2.1 | 59.5±1.2 | 0.76 | ||
Triglycerides (mg/dl) | 82.5±4.3 | 96.5±9.8 | 81.8±4.1 | 0.39 | ||
Glucose (mg/dl) | 87.0±0.8 | 89.4±2.1 | 87.1±0.8 | 0.42 | ||
Systolic BP (mm Hg) | 117.6±1.1 | 117.5±2.2 | 118.3±1.4 | 0.88 | ||
Diastolic BP (mm Hg) | 71.1±0.7 | 71.3±1.3 | 70.4±0.9 | 0.69 | ||
Reached Menopause (%) | 39.8 | 19.3 | 41.0 | 0.72 | ||
Has one or more metabolic syndrome traits (Yes/No) (%) | 29.5 | 46.5 | 31.4 | 0.18 |
All analyses adjusted for age and sex except analyses of age and sex, which were unadjusted. See
To determine if gut community type was associated with occupation, we classified study subjects into occupational classes (farmers, tradesmen, farmer’s wives, teachers/shopkeepers, and unknown/retired) and tested the association of phenotype with each occupational class. In men, the occupation of farming was over-represented among those with the
Men | Women | |||||||
Network 1 |
Network 2 |
Network 3Firmicutes(n = 37) | Age adj.p-value |
Network 1 |
Network 2 |
Network 3Firmicutes(n = 84) | Age adj.p-value | |
Farmers | 29 (44.6) | 0 (0.0) | 15 (40.5) | 0.78 | 0 (0.0) | 0 (0.0) | 0 (0.0) | – |
Tradesmen | 22 (33.8) | 5 (50.0) | 16 (43.2) | 0.51 | 0 (0.0) | 0 (0.0) | 0 (0.0) | – |
Farmer’s wives | 0 (0.0) | 0 (0.0) | 0 (0.0) | – | 11 (13.6) | 1 (3.0) | 8 (9.5) | 0.25 |
Teachers/shopkeepers | 12 (18.4) | 3 (30.0) | 4 (10.8) | 0.32 | 69 (85.2) | 29 (87.9) | 73 (86.9) | 0.92 |
Unknown/retired | 2 (3.1) | 2 (20.0) | 2 (5.4) | 0.09 | 1 (1.2) | 3 (9.1) | 3 (3.6) | 0.13 |
p-value for test of association of occupational class with microbiota network.
The availability of extensive clinical data from the OOA cohort allowed us to evaluate the potential role of the microbiota (both species and OTUs) in obesity and its metabolic derangements using network analysis, independent of enterotype. Twenty-two species of bacteria from the phyla Bacteroidetes, Firmicutes, and Actinobacteria, and four OTUs from the order Clostridiales, displayed both positive and inverse correlations with BMI, serum triglycerides, HDL cholesterol, total cholesterol, fasting glucose levels and C-reactive protein (
(A) Known species and Operational Taxonomic Units (OTUs) (green circles) linked to metabolic syndrome traits (yellow diamonds), illustrating statistically significant correlation coefficients using the Louvain algorithm. The size of the circles represents the mean relative abundance in the Amish cohort studied. Numbers connecting microbes are the correlation coefficient (p<0.001 for all). Solid lines represent positive correlations and dashed lines represent inverse correlations. (B) The same network as shown in panel A, but also including the statistically significant associations between bacterial taxa. (C) Phylogenetic tree of 16S rRNA sequences from the bacterial taxa in this network using the R implementation of DNADIST and FASTME. OTUs and known species that are inversely correlated with metabolic syndrome traits are colored in red and that are positively correlated with metabolic syndrome traits are colored in blue (p<0.001 for all).
In a subset of 32 older (≥59 years of age) obese subjects with more severe manifestations of the metabolic syndrome there was an increase in the relative abundance of three gram-negative bacterial genera,
The prevalence of obesity has dramatically increased around the world over the last twenty years, in large part linked to the Western life-style. Many lines of evidence point to a complex etiology for obesity that includes both genetic and environmental factors. Obesity is associated with a panoply of co-morbidities including hypertension, dyslipidemia, insulin resistance, and diabetes, collectively known as the metabolic syndrome, that increase the risk of cardiovascular disease
The present study used sequencing of 16S rRNA amplicons to characterize the gut microbiota in a metabolically well-characterized cohort of 310 Amish subjects over a range of BMI. The Amish are an excellent population to study the relationship between gut microbiota and metabolic traits because of their relatively homogeneous lifestyle. Furthermore, we believe that knowledge gained from the Amish is relevant to the general population because the clinical characteristics of obesity and its related complications in the Amish are indistinguishable from that in the general Caucasian population
Our analyses revealed several novel insights into the structure and role of the gut microbiota in obesity and the metabolic syndrome. First, we identified three groups of bacterial species, each of which include dominant organisms in the three enterotypes previously identified in a small set of European subjects
We hypothesized that we would identify differences in the gut microbiota that would be associated with body weight and/or features of the metabolic syndrome in this cohort. Neither B/F ratio nor community structure was associated with BMI or metabolic syndrome traits. Further investigation using network analysis identified 22 bacterial species and 4 OTUs that represent between 0.03 and 31% of the total sequence reads that were statistically significantly correlated with BMI and several features of the metabolic syndrome. Approximately half of these species are members of the core gut microbiota in the Amish and members of the Bacteroidetes and Firmicutes phyla.
In summary, we have carried out a survey of the gut microbiota and its relationship to obesity and metabolic syndrome in the OOA, a population of common genetic background and similar lifestyle. Our results have identified a subset of bacterial taxa that are linked to metabolic syndrome traits; although the cross-sectional nature of this study makes it difficult to infer cause and effect with these data alone. Follow-on longitudinal studies can begin to address whether specific gut bacterial taxa a play a causal role in the predisposition to or development of the metabolic syndrome, as well as the utility of interventions that modulate the composition of the gut microbiota to mitigate the risk of cardiovascular complications associated with metabolic syndrome.
Our study population consisted of Old Order Amish adults from Lancaster County, Pennsylvania. The Amish sect originated in Berne, Switzerland in 1693 as an ultraconservative wing of the Mennonite movement
All recruitment was performed between April 2008 and September 2010. Eligibility criteria included the following: of Amish descent; age between 20 and 80 years. We recruited subjects over a wide range of BMIs. Exclusion criteria included the following: currently pregnant or have been pregnant in the last 6 months; antibiotic treatment within the prior 6 months; currently taking a medication (e.g., antibiotic, anti-inflammatory agents, glucocorticoids or other immune modulating medications), unwilling to discontinue vitamin or supplements, including probiotics, potentially affecting the gut microbiome (vitamins/supplements and medications that were judged to possibly affect the gut microbiome were discontinued for at least 14 days prior to stool collection); renal insufficiency (serum creatinine >2 mg/dl); hematocrit <32%; uncontrolled thyroid disease (TSH <0.4 or >5.5 mIU/; co-existing malignancy; history of intestinal surgery (except appendectomy or cholecystectomy); history of inflammatory bowel disease, celiac disease, lactose intolerance, chronic pancreatitis or other malabsorption disorder.
Recruitment was performed during an initial home visit by a field team consisting of a nurse and an Amish Liaison. A screening questionnaire, height, weight, and hip measurement, and blood tests (comprehensive metabolic panel, complete blood count, thyroid stimulating hormone (TSH), celiac screen; Quest Diagnostics, Inc, Horsham, PA) were obtained to rule out exclusions (see below). Eligible and consenting volunteers were provided with a stool sample collection kit and instructions for collection (see below). A follow-up home visit was conducted after an overnight fast to obtain additional information through questionnaires (such as medical and family history, food frequency), blood pressure measurements, and collection of blood, urine, and fecal samples.
All procedures were performed by trained personnel using standard operating procedures following the guidelines of the University of Maryland and the Amish Research Clinic
For collection of feces, subjects were instructed to collect a stool sample within 1 day of the scheduled home visit. The sample was collected in a disposable “nuns” hat, and a portion of the sample (approximately 1–2 g) was immediately dispersed in RNALater (QIAgen) and refrigerated overnight. RNALater-stabilized samples were then frozen at −80°C and transported on dry ice to the Institute for Genome Sciences for long-term storage. In a subset of 19 subjects, a second fecal sample was obtained 2 to 24 months later through a home visit. An interval history and follow-up anthropometry was also obtained at that home visit to assess any changes in health status or medication usage (including antibiotics). Protocols and procedures for this visit were similar to those in which the first fecal sample was obtained.
The Institutional Review Board at the University of Maryland School of Medicine approved the protocol and informed consent was obtained from all subjects.
For DNA extraction, a 0.3 g stool aliquot was transferred to a DNA/RNA-free sterile tube, and 1 ml of phosphate-buffered saline was added to the sample. Cell lysis was initiated by adding 50 µL of lyzosyme (10 mg/mL) and 6 µL of mutanolysin (25,000 U/mL; Sigma- Aldrich). After a1 hour incubation at 37°C, each sample was further lysed by addition of 10 µl Proteinase K and 50 µl 10% SDS, followed by incubation at 55°C for 45 minutes. The samples were then disrupted by bead beating, which was performed in a FP120 FastPrep at 6.0 m/s for 40 sec using 0.1 mm silica spheres (QBiogen Lysis Matrix B). The resulting crude lysate was processed using the ZYMO Fecal DNA Kit (Zymogen) according to the manufacturer’s recommendations. Negative extraction controls, where stool samples were omitted, were performed to ensure the samples were not contaminated by exogenous bacterial DNA during the extraction process. The DNA concentrations in the samples were measured using the Quant-iT PicoGreen dsDNA assay kit from Molecular Probes (InVitrogen).
Universal primers 27F and 338R were used for PCR amplification of the V1–V3 hypervariable regions of 16S rRNA genes. The 338R primer included a unique sequence tag to barcode each sample. The primers were as follows: 27F-5′-
Sequences were binned and trimmed, using the sample-specific barcode sequences, using mothur and the following criteria: (i) sequence length >199 bases, (ii) sequence length <501 bases, number of ambiguous bases = 0, exact barcode matching, 1 nucleotide mismatch in primer matching, and maximum homopolymer string of 8 bases
For each network, we transformed the 16S rRNA sequence reads into relative abundance, computed the Spearman rank correlation, and then constructed a genera network and identified a sub-network (module) with those genera that have a direct connection. The network was constructed with the cutoff p value of less than 0.001.
To visualize the interconnectivities between species in the human gut microbiota and clinical phenotypes, we transformed the number of 16S rRNA sequence reads from each sample into relative percentages, and then computed the cross-correlation matrix between clinical metadata and sequence data with the Spearman rank correlation. The network was constructed with correlations that have a p-value of 0.005 or less. The sub-networks (modules) are identified with the Louvain algorithm
Representative 16S rRNA sequences assigned to the 13 species and 12 OTUs with correlations to MST, were aligned to the SILVA database of reference 16S rRNA genes
We compared the distribution of the three major groups across a variety of factors including age, sex, occupation, season of feces collection, and metabolic syndrome-related phenotypic characteristics of the study subjects. We tested the association of bacterial groups with demographic and metabolic factors by regressing group, coded as a class variable with 3 levels, against each metabolic variable separately, and adjusting for age and sex. This was a 2 df test. Continuously distributed variables were compared across groups using analysis of variance. Whenever a significant association was detected, we then tested each pairwise contrast separately (e.g., group 1 vs groups 2 and 3; group 2 vs groups 1 and 3; and group 3 vs groups 1 and 2) to determine the relative contributions of each contrast to the difference. These analyses were run in SAS using the GLM procedure.
Because our sample of 310 individuals included 113 nuclear families, we were able to estimate the heritability of enterotype by comparing concordance rates for enterotype class between spouse pairs (who are unrelated) and sib-pairs and parent-offspring pairs (who share 50% of their genes in common). We estimated heritability as twice the difference in concordance rates between the first-degree relative pairs and the spouse pairs.
(EPS)
(EPS)
(EPS)
(EPS)
(EPS)
(EPS)
(DOCX)
(DOCX)
The authors wish to acknowledge the efforts of Luke Tallon, Lisa Sadzewicz, Kristine Jones, Ivette Santana-Cruz, and Cesar Arze from the Genomics Resource Center at the Institute for Genomic Research for 16S rRNA amplicon pyrosequencing. The authors would also like to acknowledge Patrick D. Schloss for advice on the optimization of 16S rRNA sequence analysis using mothur. Data generated in this study are available at dbGAP under study ID phs000258 and at SRA under accession SRP002465.