Abstract
Background
Studies of the fecal microbiome have implicated the gut microbiota in obesity, but few studies examined the microbial diversity at other sites. We explored the association between obesity and the upper gastrointestinal (UGI) microbial diversity.
Design/Methods
The UGI microbiome of 659 healthy Chinese adults with a measured body mass index (BMI) range of 15.0 to 35.7 was characterized using the 16S rRNA gene DNA microarray (HOMIM).
Results
In multivariate-adjusted models, alpha diversity was not associated with BMI. However, beta diversity, assessed by principal coordinate vectors generated from an unweighted unifrac distance matrix of pairwise comparisons, was associated with BMI (third and fourth vectors, p=0.0132 and p=0.0280, respectively). Moreover, beta diversity, assessed by cluster membership (3 clusters), was also associated with BMI; individuals in the first cluster (median BMI 22.35, odds ratio (OR)=0.48, 95% confidence interval (CI)=0.05–4.34) and second cluster (median BMI 22.55, OR=0.26, 95% CI=0.09–0.75) were significantly less likely to be obese (BMI ≥27.5) than those in the third cluster (median BMI 23.59).
Conclusions
A beta-diversity metric of the UGI microbiome is associated with a four-fold difference in obesity risk in this Asian population. Future studies should address whether the UGI microbiome plays a causal role in obesity.
Keywords: beta-diversity, body mass index, Chinese, microbiome, obesity, upper gastrointestinal tract
Introduction
The prevalence of obesity has increased alarmingly in the past several decades, both in developed and developing countries (1). A complex combination of behavioral, genetic, and environmental risk factors likely contribute to obesity, and understanding the molecular mechanisms that lead to obesity is important for developing prevention and therapeutic strategies. Recent research on the enormously complex, diverse, and vast microbial community in the gastrointestinal tract has provided new insights into the mechanisms of obesity and obesity-related diseases (2).
The microbial ecosystem within the digestive tract contributes substantial benefit to the host, such as digestion of otherwise indigestible plant polysaccharides, but not all interactions may be advantageous (3). For example, studies in experimental animal models suggest that energy harvest and adiposity are modified by gut microbes (4, 5, 6), whereas, several human studies suggest that obesity may be associated with different gut microbial profiles (7, 8, 9, 10, 11, 12, 13, 14), including an altered phylum level Firmicutes-to-Bacteroidetes ratio (9, 14) or bacterial diversity (12). However, most previous human studies have been small and produced mixed results.
The majority of studies of the human microbiome and obesity have focused on the distal gut, evaluating fecal microbiome samples (7, 8, 9, 10, 11, 12, 13, 14), with little attention paid to microbial communities at other sites along the digestive tract. The Human Microbiome Project examined microbial profiles from sampling several distinct body habitats of the digestive tract from healthy individuals (15) and found that the stool microbiota was distinguished from oral cavity microbiota, although several genera were detected in both habitats (16). Compared with stool, oral sites (particularly saliva) have a higher Shannon diversity, a diversity index that measures both richness and evenness, but the stool habitat has a higher species-level richness, and this difference may be driven by environmental selection (17). The Human Microbiome Project showed modest correlations between the relatively stable microbiome (16) at oral sites with body mass index (BMI) (18). Obese individuals may have an increased local inflammatory response and an altered periodontal microflora (19, 20). Studies using a variety of methods for assessing the oral microbiota in association with obesity produced mixed results (15, 21, 22). The important role of the upper digestive tract microbiota in the commencement of digestion provides the rationale for inquiry into the relationship between obesity and the upper digestive tract microbiome.In this study, we used upper gastrointestinal samples collected by esophageal balloon cytology from a cohort of over 650 healthy Chinese adults. We found that individuals in the BMI range for optimum health could be differentiated from those in the BMI range for obesity-related diseases in Asian populations using microbial diversity measures.
Methods
Patient population
The participants were recruited from a commune in Linxian, China, in the spring of 2002, as part of a cancer screening study using esophageal cytology compared with endoscopy (23). The study targeted healthy asymptomatic residents 50- to 64-years old, although approximately 10% of the 720 participating individuals fell outside of that age range but were allowed to participate; individuals who had any signs or symptoms of upper gastrointestinal cancer (dysphagia, hematemesis) or other chronic diseases (liver cirrhosis, congestive heart failure, unstable angina) were excluded. Individuals diagnosed with esophageal or gastric cancer were excluded from this study. The subjects fasted overnight prior to completing an informed consent form and an esophageal balloon cytology exam. A week later, after another overnight fast, all subjects completed a short questionnaire, a physical exam, an oral health exam (24), a saliva collection, and an upper gastrointestinal endoscopy exam with Lugol’s iodine staining and biopsy. The study was approved by the Institutional Review Boards of the Cancer Institute of the Chinese Academy of Medical Sciences and the U.S. National Cancer Institute.
Upper digestive tract biological sample collection
The upper aerodigestive tract cell samples were collected from subjects who were randomly assigned one of two esophageal balloon cytology retrieval devices (23). One device was the traditional Chinese inflatable balloon (CICAMS, Beijing, China), which was a rubber balloon covered with a cotton mesh net and attached to a 0.2 cm in diameter single-lumen rubber tube. The other device was a balloon (Cytomesh Esophageal Cytology Device, Wilson-Cook Medical, Inc., Winston-Salem, North Carolina, USA) that had a mechanically expandable plastic mesh covering. The examination techniques using the two balloons were similar. The patient was given 2 ml of a 2% lidocaine slurry by mouth for local anesthesia, and the balloon was inserted into the back of the throat and swallowed. Once in the stomach, the balloon and mesh covering were expanded and gradually withdrawn through the esophagus. The balloon with its collected cells and luminal material was cut using sterile scissors and placed in 40 mL of saline in a 50-mL centrifuge tube, shaken, and transferred on ice to the central processing laboratory. The sample was then vortexed to remove adherent cells from the balloon. After the balloon was removed, the remaining sample was centrifuged; the pellet that formed was resuspended in 1 mL of saline and snap frozen in liquid nitrogen and stored at −80°C until DNA extraction.
DNA extraction
The Gentra Puregene Cell kit (Qiagen, Valencia, CA) was used according to the manufacturer’s instructions to extract the DNA from 300 ul of the cell suspension collected from the upper digestive tract. The DNA quality and quantity was checked using the 260:280 ratio, Nanodrop, and Picogreen. The presence of human and bacterial DNA in the samples was verified by TaqMan assays using species-specific primers.
HOMIM array
The HOMIM array (25), optimized for oral cavity bacterial species, which uses 16S rRNA-based oligonucleotide probes printed on glass slides, was performed as previously described (26). Briefly, the extracted DNA was PCR-amplified using universal forward and reverse primers and then labeled in a second nested PCR. After hybridization of the DNA on the array, the washed slides were scanned using GenePix Pro software. For each individual feature, the background intensity was subtracted from the median intensity to generate the normalized median intensity score. The relative intensity of each probed species/strain was estimated using feature-specific criteria.
HOMIM data quality
Sufficient DNA was available from 704 subjects for the HOMIM array. These samples were run in 8 batches, with 5 samples repeated in each batch to assess the technical replication of the array results. All of the batches except for the last batch, which included 45 samples, had consistent results and consistent mean alpha-diversity distributions. Therefore, we excluded this last batch from further analyses, leaving 659 samples.
Validation of HOMIM data
To assess whether any high abundance species were present in the samples but missing from the HOMIM array, we cloned and sequenced PCR products (range 59–89) from 10 subjects. Only one species (Lachnospiracaeae, OT107) appeared in multiple samples (4 of the 10 samples) but was not on the array. This absence on the array is unlikely to have influenced our results because we focused primarily on community measures, which are unlikely to be influenced by the presence or absence of data on a single taxon.
Microbial diversity metrics
Alpha-diversity metrics were calculated as the number of genera, families, order, classes, and phyla. Each alpha-diversity metric was scaled by its respective interquartile distance. Beta-diversity metrics were created using an unweighted unifrac distance matrix for all pairs of subjects in the cohort using fast unifrac (http://bmf.colorado.edu/fastunifrac/index.psp) (27) and a phylogenetic tree based on the HOMD (28). The unweighted unifrac distance matrix, which measures pair-wise taxonomic dissimilarity between microbial populations, was analyzed with an unsupervised clustering algorithm (unweighted pair group method with arithmetic mean (UPGMA)), and the maximal pseudo F test and minimal pseudo t2 statistics supported three clusters. Principal coordinate vectors were generated from the unweighted distance matrix. Analyses were conducted using the first five vectors, which together explained approximately 60% of the variation in the matrix. Each vector score was scaled by its respective interquartile distance.
BMI and covariates
BMI (kg/m2) was calculated from measured height and weight. We analyzed BMI in Asian population-specific categories (normal weight, 18.5 – <23; overweight, 23 – >27.5; obese, ≥27.5). Because the distribution of BMI in the cohort was skewed, BMI was log-transformed when used as a continuous variable. The characteristics of the subjects were treated as follows: age in years (continuous), tobacco smoking (ever versus never), alcohol drinking (any in the past 12 months versus none), hypertensive (yes if systolic blood pressure over 140 mm Hg or diastolic blood pressure over 90 mm Hg), antibiotic use (current use or use in the last 3 months versus none), and decayed, missing, or filled teeth (DMFT, characterized as having all 28 permanent teeth or less than the 28 permanent teeth). The balloon collection device, the Chinese balloon or the Wilson-Cook balloon, was included in models as a potential confounder, although no differences were noted by balloon type.
Statistical analysis
We tested whether the alpha-diversity metrics and the beta-diversity unifrac vector scores were ordered by the increasing Asian BMI categories by using the one-sided Jonckheere-Terpstra test. The ordering of cluster membership by the Asian BMI categories was examined by Fisher’s exact test. The associations between BMI (on the log scale) and individual microbial diversity metrics were calculated by linear regression models. Multivariate-adjusted regression models included age, sex, tobacco smoking, alcohol drinking, antibiotic use, and balloon device. Because we previously found that unifrac matrix assignment, poor oral health (DMFT=28), and alpha diversity, among other variables were interrelated (Abnet et al., manuscript submitted), we also adjusted for the oral health variable. We examined the crude and multivariate-adjusted association between BMI categories and individual microbial diversity metrics using multinomial logistic regression models. Multiple microbial diversity metrics found to be important were mutually adjusted in the same regression models. As an exploratory analysis, we determined which unique species were significantly associated with cluster membership using logistic regression and adjusted the p-value using Bonferroni correction due to multiple testing (i.e., 158 species). Unless otherwise specified, all tests were two-sided, and p-values <0.05 and confidence intervals that did not overlap with 1.00 were considered statistically significant. All statistical analyses were conducted using SAS 9.2, and figures were made in GraphPad Prism 5.
Results
Characteristics of the healthy Chinese cohort
For this study, we used esophageal balloon cytology samples collected as part of an esophageal cancer screening study that enrolled and intensively studied a cohort of healthy Chinese adults aged 40–65 years. DNA was extracted from these samples, which contained bacteria from the oral cavity, esophagus, and stomach, and was characterized using the HOMIM arrays. BMI was calculated from height and weight measured in a physical exam. Data were available for 659 individuals in the cohort, and their characteristics by Asian BMI categories (29) are shown in Table 1. Most study subjects were considered normal or overweight (43% and 46%, respectively), whereas a small percentage were considered underweight or obese (3% and 8%, respectively). Females were more likely to be overweight and obese. Hypertensive individuals were more likely to be overweight and obese. In this population, those who reported smoking tobacco were almost exclusively male, and they had lower BMI than non-smokers. Alcohol intake was relatively rare in this population.
Table 1.
Characteristics of subjects by Asian BMI categories
| Asian BMI categories | |||||
|---|---|---|---|---|---|
| cohort | <18.5 (underweight) | 18.6–<23 (normal) | 23–<27.5 (overweight) | ≥27.5 (obese) | |
| N (%) | 659 | 19 (3) | 281 (43) | 304 (46) | 55 (8) |
| Age, years, median (IQR) | 54 (51–58) | 61 (51–62) | 54 (51–59) | 54 (53–58) | 54 (51–57) |
| Sex, males, N (%) | 279 (42) | 8 (42) | 138 (49) | 115 (38) | 18 (33) |
| Tobacco smoking, ever, N (%) | 168 (25) | 4 (21) | 88 (31) | 65 (21) | 11 (20) |
| Alcohol drinking, yes, N (%) | 51 (8) | 2 (11) | 24 (9) | 21 (7) | 4 (8) |
| Systolic blood pressure, mm Hg, median (IQR) | 148 (134–166) | 136 (122–147) | 145 (130–160) | 150 (137–170) | 157 (136–171) |
| Diastolic blood pressure, mm Hg, median (IQR) | 92 (84–101) | 83 (73–91) | 90 (82–98) | 93 (85–103) | 95 (87–108) |
| Hypertensive, yes, N (%) | 469 (71) | 8 (42) | 182 (65) | 234 (77) | 45 (82) |
| Antibiotics, yes, N (%) | 76 (12) | 3 (16) | 26 (9) | 37 (12) | 10 (18) |
| DMFT1, median (IQR) | 9 (3–21) | 12 (5–28) | 10 (3–21) | 8 (3–18) | 9 (2–23) |
| Balloon type, CHB, N (%) | 320 (49) | 6 (32) | 133 (47) | 154(51) | 27 (49) |
Count of decayed, missing, and filled teeth among the 28 core permanent teeth
Alpha-diversity and beta-diversity metrics differed by Asian BMI categories
To assess the microbial alpha diversity, which incorporates microbial richness, in the upper gastrointestinal tract, we calculated the numbers of genera, families, orders, classes, and phyla for the study subjects by Asian BMI categories. Table 2 shows that the numbers of genera, classes, and phyla did not significantly differ by Asian BMI categories, whereas the numbers of families and orders significantly decreased with increasing Asian BMI categories. To test beta diversity, we created an unweighted UniFrac distance matrix for all subject pairs in the cohort, and UniFrac vectors were derived from principal coordinate analysis of the unweighted distance matrix. None of the first five principal coordinate vector scores (which together explained approximately 60% of the variation in the matrix), scaled by their individual interquartile ranges, significantly differed by the Asian BMI categories. The distance matrix was then analyzed with an unsupervised clustering algorithm; there was evidence that individuals in the cohort could be grouped into 3 clusters (Figure S1, labeled as A, B, and C based on increasing numbers of member individuals: 11, 129, and 519, respectively). The individuals in Cluster A had a median BMI of 22.4 and interquartile range (IQR) of 19.2–23.3; individuals in Cluster B had a median BMI of 22.6 and IQR of 20.5–24.4; and individuals in Cluster C had a median BMI of 23.6 and IQR of 21.5–25.3. A higher percentage of individuals in Cluster C were considered overweight and obese than individuals in Clusters A and B.
Table 2.
Microbiome diversity measures by Asian BMI categories
| Asian BMI categories | |||||
|---|---|---|---|---|---|
| Variable | <18.5 (underweight) | 18.6–<23 (normal) | 23–<27.5 (overweight) | ≥27.5 (obese) | p-value1 |
| N | 19 | 281 | 304 | 55 | |
| α - diversity | |||||
| Number of genera, mean (std) | 21.74 (6.28) | 22.77 (6.40) | 22.34 (5.93) | 21.84 (5.11) | 0.13 |
| Number of families, mean (std) | 17.11 (3.83) | 16.88 (3.47) | 16.67 (3.18) | 16.16 (2.52) | 0.03 |
| Number of orders, mean (std) | 12.53 (2.04) | 12.09 (1.92) | 11.99 (1.74) | 11.60 (1.58) | 0.04 |
| Number of classes, mean (std) | 10.00 (1.25) | 9.98 (1.26) | 10.02 (1.17) | 9.89 (1.07) | 0.48 |
| Number of phyla, mean (std) | 5.42 (0.61) | 5.32 (0.72) | 5.42 (0.67) | 5.24 (0.67) | 0.24 |
| β - diversity: principal coordinate vectors | |||||
| UniFrac vector 1 scores, mean (std) | 0.03 (0.16) | −0.001 (0.14) | −0.0008 (0.13) | 0.001 (0.11) | 0.30 |
| UniFrac vector 2 scores, mean (std) | 0.01 (0.11) | 0.003 (0.10) | −0.0002 (0.11) | −0.02 (0.09) | 0.18 |
| UniFrac vector 3 scores, mean (std) | −0.02 (0.12) | −0.005 (0.09) | 0.003 (0.08) | 0.02 (0.07) | 0.08 |
| UniFrac vector 4 scores, mean (std) | 0.06 (0.07) | −0.002 (0.08) | −0.001 (0.08) | −0.005 (0.07) | 0.23 |
| UniFrac vector 5 scores, mean (std) | −0.01 (0.08) | 0.002 (0.07) | −0.0007 (0.07) | −0.002 (0.07) | 0.49 |
| β - diversity: clustering | |||||
| Cluster A, N (%) | 2 (18) | 6 (55) | 2 (18) | 1 (9) | 0.02 |
| Cluster B, N (%) | 7 (5) | 62 (48) | 56 (43) | 4 (3) | 0.01 |
| Cluster C, N (%) | 10 (2) | 213 (41) | 246 (47) | 50 (10) | 2.10E-03 |
One-sided Jonckheere-Terpstra test for the continuous variables and Fisher's exact test for the categorical variables
Beta-diversity but not alpha-diversity metrics were associated with continuous BMI
We also assessed the association between alpha-diversity metrics scaled by their individual interquartile ranges and BMI as a log-transformed continuous outcome in unadjusted and adjusted linear regression models (Table 3). Adjusted models included age, sex, tobacco smoking, alcohol drinking, antibiotic use, balloon sample collection device, and a count for decayed, missing, and filled teeth. None of the alpha-diversity metrics were significantly associated with BMI. We also assessed the association between beta-diversity principal coordinate vector scores and log-transformed continuous BMI and found that vector 3 was significantly associated with BMI in the unadjusted models, and both vectors 3 and 4 were significantly associated with BMI in the adjusted models. Figure 1 shows the distribution of BMI by individual vector scores, and the regression lines represent the log-transformed BMI association from the multivariate-adjusted linear regression models. In addition to the beta-diversity vector scores, we also found that BMI was associated with cluster membership in the unadjusted and adjusted models (Table 3). When we combined the beta-diversity principal coordinate vector 3 and vector 4 scores in the same model that included the beta-diversity clusters, we found that the clusters remained significantly associated with BMI, whereas the vector scores were no longer independently associated with BMI, suggesting that the cluster assignment for beta diversity captures the association with BMI more effectively than the principal coordinate vectors. Figure 2 illustrates the relationship between the three clusters and the first three principal coordinate vectors, suggesting that the clusters of individuals remained distinct in their vector score distributions. Multiple unique species were more commonly detected in Cluster B compared with Cluster C (Supplementary Table 1).
Table 3.
Crude and multivariate-adjusted associations between continuous BMI1 and microbiome diversity measures
| Variable | median (lower, upper quartiles) |
Scale3 | Unadjusted | Adjusted2 | ||
|---|---|---|---|---|---|---|
| β | p | β | p | |||
| α - diversity | ||||||
| Number of genera | 22 (18, 27) | 9 | −0.0033 | 0.65 | −0.0025 | 0.73 |
| Number of families | 17 (15, 19) | 4 | −0.0066 | 0.27 | −0.0062 | 0.29 |
| Number of orders | 12 (11, 13) | 2 | −0.0092 | 0.09 | −0.0086 | 0.10 |
| Number of classes | 10 (9, 11) | 2 | 0.0017 | 0.83 | 0.0025 | 0.75 |
| Number of phyla | 5 (5, 6) | 1 | −0.0005 | 0.95 | −0.0031 | 0.66 |
| β - diversity: principal coordinate vectors | ||||||
| UniFrac vector 1 scores | −0.027 (−0.102, 0.089) | 0.191 | −0.0048 | 0.49 | −0.0066 | 0.35 |
| UniFrac vector 2 scores | 0.009 (−0.076, 0.078) | 0.154 | −0.0068 | 0.33 | −0.0071 | 0.31 |
| UniFrac vector 3 scores | 0.014 (−0.037, 0.058) | 0.095 | 0.0113 | 0.04 | 0.0136 | 0.01 |
| UniFrac vector 4 scores | −0.001 (−0.052, 0.053) | 0.104 | −0.0124 | 0.05 | −0.0141 | 0.03 |
| UniFrac vector 5 scores | −0.008 (−0.034, 0.033) | 0.067 | −0.0031 | 0.52 | −0.0055 | 0.25 |
| β - diversity: clustering | ||||||
| Cluster A | −0.0823 | 0.03 | −0.0942 | 0.01 | ||
| Cluster B | −0.0431 | 4.00E-04 | −0.0454 | 2.00E-04 | ||
| Cluster C | ref | ref | ||||
| Multiple variables combined | ||||||
| Unifrac vector 3 scores | 0.0063 | 0.27 | 0.0086 | 0.12 | ||
| Unifrac vector 4 scores | −0.0095 | 0.13 | −0.0112 | 0.08 | ||
| Cluster A | −0.0730 | 0.05 | −0.0825 | 0.03 | ||
| Cluster B | −0.0378 | 2.70E-03 | −0.0388 | 2.10E-03 | ||
| Cluster C | ref | ref | ||||
log(BMI)
Adjusted for age, sex, tobacco smoking, alcohol drinking, antibiotic use, balloon type, and DMFT
Each variable (except clusters) was scaled by its respective interquartile distance
Figure 1.
Scatterplot distribution of BMI in a healthy Chinese cohort by UniFrac vector scores derived from principal coordinate analysis of the unweighted distance matrix from the upper digestive tract microbiome. (A) BMI by UniFrac vector 3 scores and (B) BMI by UniFrac vector 4 scores. The lines represent the adjusted linear regressions of the log-transformed BMI. The adjusted models included age, sex, tobacco smoking, alcohol drinking, antibiotic use, sample collection device, and the count of decayed, missing, and filled teeth.
Figure 2.
Scatterplot distribution by cluster of the first three UniFrac vector scores derived from principal coordinate analysis of the unweighted distance matrix from the upper digestive tract microbiome in a healthy Chinese cohort. The three clusters were derived from a cluster analysis of the unweighted UniFrac distance matrix (blue, Cluster A; green, Cluster B; black, Cluster C).
Beta-diversity but not alpha-diversity metrics were associated with overweight and obese Asian BMI categories
Table 4 shows the association between alpha- and beta-diversity metrics and risks of having overweight or obese BMI compared with normal BMI from multivariate-adjusted multinomial logistic regression models. In this analysis, the underweight subjects were excluded. The alpha-diversity metrics were not associated with the risk of overweight or obese BMI compared with normal BMI. The beta-diversity vector 3 scores were associated with the risk of obese BMI compared with normal BMI (odds ratio (OR)=1.49, 95% confidence interval (CI)=1.02–2.17). Compared with Cluster C membership, Cluster B membership was associated with a significantly decreased risk of obese BMI (OR=0.26, 95% CI=0.09–0.75) as opposed to normal BMI. Cluster B membership remained independently associated with risk of having an obese BMI compared with normal BMI in a model that combined the vector 3 scores and the clusters; because the principal coordinate vectors and clusters were derived from the same UniFrac distance matrix, this result suggests that the cluster assignment for beta diversity captures the association with BMI more effectively than the principal coordinate vectors.
Table 4.
| Adjusted3 | |||||
|---|---|---|---|---|---|
| 18.6–<23 (normal) | 23–<27.5 (overweight) | ≥27.5 (obese) | |||
| Variable4 | OR | OR | 95% CI | OR | 95% CI |
| α - diversity | |||||
| Number of genera | ref | 0.93 | 0.73–1.20 | 0.82 | 0.53–1.28 |
| Number of families | ref | 0.95 | 0.77–1.16 | 0.79 | 0.55–1.12 |
| Number of orders | ref | 0.96 | 0.80–1.15 | 0.76 | 0.55–1.05 |
| Number of classes | ref | 1.08 | 0.82–1.43 | 0.93 | 0.58–1.49 |
| Number of phyla | ref | 1.19 | 0.93–1.50 | 0.80 | 0.52–1.24 |
| β - diversity: principal coordinate vectors | |||||
| UniFrac vector 1 scores | ref | 0.95 | 0.75–1.21 | 0.99 | 0.65–1.51 |
| UniFrac vector 2 scores | ref | 0.96 | 0.75–1.22 | 0.77 | 0.50–1.18 |
| UniFrac vector 3 scores | ref | 1.17 | 0.97–1.42 | 1.49 | 1.02–2.17 |
| UniFrac vector 4 scores | ref | 0.98 | 0.79–1.22 | 0.89 | 0.60–1.32 |
| UniFrac vector 5 scores | ref | 0.93 | 0.79–1.09 | 0.90 | 0.67–1.21 |
| β - diversity: clustering | |||||
| Cluster A | ref | 0.23 | 0.04–1.16 | 0.48 | 0.05–4.34 |
| Cluster B | ref | 0.72 | 0.47–1.10 | 0.26 | 0.09–0.75 |
| Cluster C | ref | ref | ref | ||
| Multiple variables combined | |||||
| Unifrac vector 3 scores | ref | 1.13 | 0.93–1.37 | 1.36 | 0.92–2.01 |
| Cluster A | ref | 0.24 | 0.05–1.23 | 0.56 | 0.06–5.14 |
| Cluster B | ref | 0.77 | 0.50–1.18 | 0.29 | 0.10–0.87 |
| Cluster C | ref | ref | ref | ||
Odds ratios and confidence intervals were calculated using multinomial logistic regression models for the association between the diversity variable and the risk of being overweight or obese, compared with normal weight; confidence intervals that do not include 1.00 are considered statistically significant
Overweight classified as BMI 23–<27.5; Obese classified as BMI ≥27.5; comparison group, normal BMI 18.6–<23; underweight individuals were excluded from this analysis
Adjusted for age, sex, tobacco smoking, alcohol drinking, antibiotic use, balloon type, and DMFT
Each variable (except clusters) was scaled by its respective interquartile distance
Discussion
In this cross-sectional study of BMI and upper digestive tract microbial diversity in a Chinese cohort of 659 healthy individuals, we found that BMI was not associated with the bacterial community diversity as assessed by alpha diversity in our models after adjusting for multiple potential confounders. However, BMI was significantly associated with the variation in the community composition, as assessed by multiple beta-diversity parameters, even after adjusting for multiple confounders; in particular, the community cluster distinction was independently associated with the risk of having an obese BMI based on categories determined by the WHO for obesity-related diseases among Asian populations (29). The statistically significant odds ratio for the subjects in Cluster B compared with those in Cluster C in having an obese BMI was 0.26, meaning that the subjects in Cluster C were nearly four times more likely to be obese than the subjects in Cluster B.
The gut microbiota can be considered a microbial metabolic organ (30) because of its influence on the regulation of energy uptake from the diet, interaction with signaling molecules involved in host metabolism, and release of gut hormones, among other mechanisms (2). Several studies of the gut microbiome have shown that obese individuals harbor less diverse bacterial communities than lean individuals (12, 31). Studies in both mice and humans have suggested that obese individuals have a lower ratio of bacteria from the phylum Bacteroidetes to bacteria from the phylum Firmicutes than lean individuals (6, 9, 32), although results from other studies attempting to replicate these findings have been conflicting (15, 33). Most previous investigations in the role of gut microbes in obesity have examined stool samples, whereas we sampled the upper digestive tract, including the saliva, esophagus, and gastric contents. Because the Human Microbiome Project showed little similarity in the 16S samples by weighted UniFrac beta-diversity between the oral cavity and stool (15), results from previous stool studies may not provide the most appropriate context for our current findings.
The role of the upper digestive tract microbiota in human metabolism or energy homeostasis begins with the digestive functions of saliva. The composition of salivary bacteria has been suggested to be altered in overweight women compared with women of normal weight (34). Many studies have characterized oral cavity bacteria, individually and communally, for their role in periodontal disease etiology (35). Periodontitis is associated with overweight and obese BMI (20), possibly because the pro-inflammatory proteins produced by adipose tissues may be a risk factor for periodontal inflammation and because these pro-inflammatory factors due to periodontal disease may influence insulin sensitivity in obese individuals (36). Other markers of poor oral health, such as tooth loss, may also be associated with obesity (37), although both may be correlated with socioeconomic status. The relationship between poor oral health and obesity has been correlated with systemic inflammation (38), and the potential impact of infectious agents on obesity has been explored (39).
This study has several strengths. It included a large number of subjects, substantially more than other studies that have examined microbial diversity and obesity. The study subjects were asymptomatic healthy individuals who were part of a well-characterized, population-based study, and they were representative of the community from which they were recruited. While this Chinese community had a relatively homogeneous diet (40), the recruited population had a broad range of BMI, which was based on height and weight measured during a physical exam. This study is among the first to examine the microbiome and obesity in an Asian population, and we had detailed clinical and demographic information to assess potential confounding. Moreover, we used an innovative combination of approaches for measuring microbial diversity in multivariate data, one parameter for assessing community diversity (alpha diversity) along a gradient and another for assessing the variation in community composition (beta-diversity) by finding groupings. However, our study also has several limitations. It was conducted in cross-section and could not provide evidence of a causal effect between the upper digestive tract microbiome and obesity. The study population was a community composed mostly of subsistence farmers, so the upper range of BMI was limited, but the obese category (BMI ≥27.5) was appropriately used for this Asian population. The upper digestive tract microbiome was assessed using samples collected by esophageal balloon cytology, which included a mixture of cells and luminal contents, potentially unevenly, from the stomach, the full length of the esophagus, and the oral cavity, but the HOMIM microarray was optimized only for oral caviety bacterial species. Additionally, the microarray was only semi-quantitative and contained a limited number of bacterial species, so we could not generate information such as the ratio of various phyla to compare with other study results. Moreover, as a microarray based on the 16S rRNA gene, this assay did not produce data that could be used to determine categories of bacterial functions.
As proof of concept, we showed using a microarray that the bacterial community composition in the upper gastrointestinal tract is associated with obesity. Additional work with sequencing technologies is needed to explore bacterial communities that play a role in obesity. Future studies should prospectively examine if variation in the community composition of microbial diversity in the upper digestive tract may be predictive of obesity risk.
Supplementary Material
What is already known about this subject?
Studies in experimental animal models suggest that energy harvest and adiposity are modified by gut microbes.
Several human studies suggest that obesity may be associated with differences in gut microbial profiles, including altered bacterial diversity, but most previous human studies have been small, produced mixed results, and only focused on stool samples rather than samples from other digestive tract sites.
What this study adds?
This study is the first to characterize the microbial diversity in the upper gastrointestinal tract of a large well-characterized cohort of healthy Asian adults with a broad BMI range.
BMI was significantly associated with the variation in the community composition of bacteria residing in the upper gastrointestinal tract, as assessed by multiple beta-diversity parameters, even after adjusting for multiple confounders.
In particular, the community cluster distinction was independently associated with the risk of having an obese BMI based on categories determined by the WHO for obesity-related diseases among Asian populations.
Acknowledgements
Supported by the Intramural Research Program, Division of Cancer Epidemiology & Genetics, National Cancer Institute. Author contributions: CCA and NDF designed the study; VKC, BJP, BAD, GQW, WQW, JHF, YLQ, and SMD contributed to the data acquisition; JS, MHG, CCA, NDF, EV, and SWL analyzed and interpreted the data; SWL drafted the manuscript; all authors revised critically and gave final approval.
Footnotes
Conflict of interest: The authors declare no conflicts of interest.
References
- 1.Flegal KM, Troiano RP. Changes in the distribution of body mass index of adults and children in the US population. Int J Obes Relat Metab Disord. 2000;24:807–818. doi: 10.1038/sj.ijo.0801232. [DOI] [PubMed] [Google Scholar]
- 2.Hullar MA, Lampe JW. The gut microbiome and obesity. Nestle Nutr Inst Workshop Ser. 2012;73:67–79. doi: 10.1159/000341288. [DOI] [PubMed] [Google Scholar]
- 3.Tilg H, Kaser A. Gut microbiome, obesity, and metabolic dysfunction. The Journal of Clinical Investigation. 2011;121:2126–2132. doi: 10.1172/JCI58109. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Backhed F, Manchester JK, Semenkovich CF, Gordon JI. Mechanisms underlying the resistance to diet-induced obesity in germ-free mice. Proc Natl Acad Sci U S A. 2007;104:979–984. doi: 10.1073/pnas.0605374104. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Samuel BS, Gordon JI. A humanized gnotobiotic mouse model of host-archaeal-bacterial mutualism. Proc Natl Acad Sci U S A. 2006;103:10011–10016. doi: 10.1073/pnas.0602187103. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Turnbaugh PJ, Ley RE, Mahowald MA, Magrini V, Mardis ER, Gordon JI. An obesity-associated gut microbiome with increased capacity for energy harvest. Nature. 2006;444:1027–1031. doi: 10.1038/nature05414. [DOI] [PubMed] [Google Scholar]
- 7.Armougom F, Henry M, Vialettes B, Raccah D, Raoult D. 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. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Collado MC, Isolauri E, Laitinen K, Salminen S. Distinct composition of gut microbiota during pregnancy in overweight and normal-weight women. Am J Clin Nutr. 2008;88:894–899. doi: 10.1093/ajcn/88.4.894. [DOI] [PubMed] [Google Scholar]
- 9.Ley RE, Turnbaugh PJ, Klein S, Gordon JI. Microbial ecology: human gut microbes associated with obesity. Nature. 2006;444:1022–1023. doi: 10.1038/4441022a. [DOI] [PubMed] [Google Scholar]
- 10.Schwiertz A, Taras D, Schafer K, Beijer S, Bos NA, Donus C, et al. Microbiota and SCFA in lean and overweight healthy subjects. Obesity (Silver Spring) 2010;18:190–195. doi: 10.1038/oby.2009.167. [DOI] [PubMed] [Google Scholar]
- 11.Zhang H, DiBaise JK, Zuccolo A, Kudrna D, Braidotti M, Yu Y, et al. Human gut microbiota in obesity and after gastric bypass. Proc Natl Acad Sci U S A. 2009;106:2365–2370. doi: 10.1073/pnas.0812600106. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Turnbaugh PJ, Hamady M, Yatsunenko T, Cantarel BL, Duncan A, Ley RE, et al. A core gut microbiome in obese and lean twins. Nature. 2009;457:480–484. doi: 10.1038/nature07540. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Million M, Maraninchi M, Henry M, Armougom F, Richet H, Carrieri P, et al. Obesity-associated gut microbiota is enriched in Lactobacillus reuteri and depleted in Bifidobacterium animalis and Methanobrevibacter smithii. Int J Obes (Lond) 2012;36:817–825. doi: 10.1038/ijo.2011.153. [DOI] [PMC free article] [PubMed] [Google Scholar] [Retracted]
- 14.Xu P, Li M, Zhang J, Zhang T. Correlation of intestinal microbiota with overweight and obesity in Kazakh school children. BMC Microbiol. 2012;12:283. doi: 10.1186/1471-2180-12-283. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Zeigler CC, Persson GR, Wondimu B, Marcus C, Sobko T, Modéer T. Microbiota in the Oral Subgingival Biofilm Is Associated With Obesity in Adolescence. Obesity. 2012;20:157–164. doi: 10.1038/oby.2011.305. [DOI] [PubMed] [Google Scholar]
- 16.Segata N, Haake S, Mannon P, Lemon K, Waldron L, Gevers D, et al. Composition of the adult digestive tract bacterial microbiome based on seven mouth surfaces, tonsils, throat and stool samples. Genome Biology. 2012;13:R42. doi: 10.1186/gb-2012-13-6-r42. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Zhou Y, Gao H, Mihindukulasuriya KA, Rosa PS, Wylie KM, Vishnivetskaya T, et al. Biogeography of the ecosystems of the healthy human body. Genome Biol. 2013;14:R1. doi: 10.1186/gb-2013-14-1-r1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.HMP. Structure, function and diversity of the healthy human microbiome. Nature. 2012;486:207–214. doi: 10.1038/nature11234. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Falagas ME, Kompoti M. Obesity and infection. The Lancet Infectious Diseases. 2006;6:438–446. doi: 10.1016/S1473-3099(06)70523-0. [DOI] [PubMed] [Google Scholar]
- 20.Suvan J, D'Aiuto F, Moles DR, Petrie A, Donos N. Association between overweight/obesity and periodontitis in adults. A systematic review. Obesity Reviews. 2011;12:e381–e404. doi: 10.1111/j.1467-789X.2010.00808.x. [DOI] [PubMed] [Google Scholar]
- 21.Desvarieux M, Demmer RT, Rundek T, Boden-Albala B, Jacobs DR, Sacco RL, et al. Periodontal Microbiota and Carotid Intima-Media Thickness: The Oral Infections and Vascular Disease Epidemiology Study (INVEST) Circulation. 2005;111:576–582. doi: 10.1161/01.CIR.0000154582.37101.15. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Michaud D, Izard J, Rubin Z, Johansson I, Weiderpass E, Tjønneland A, et al. Lifestyle, dietary factors, and antibody levels to oral bacteria in cancer-free participants of a European cohort study. Cancer Causes Control. 2013;24:1901–1909. doi: 10.1007/s10552-013-0265-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Pan QJ, Roth MJ, Guo HQ, Kochman ML, Wang GQ, Henry M, et al. Cytologic detection of esophageal squamous cell carcinoma and its precursor lesions using balloon samplers and liquid-based cytology in asymptomatic adults in Llinxian, China. Acta Cytol. 2008;52:14–23. doi: 10.1159/000325430. [DOI] [PubMed] [Google Scholar]
- 24.Dye BA, Wang R, Lashley R, Wei W, Abnet CC, Wang G, et al. Using NHANES oral health examination protocols as part of an esophageal cancer screening study conducted in a high-risk region of China. BMC Oral Health. 2007;7:10. doi: 10.1186/1472-6831-7-10. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Ahn J, Yang L, Paster BJ, Ganly I, Morris L, Pei Z, et al. Oral Microbiome Profiles: 16S rRNA Pyrosequencing and Microarray Assay Comparison. PLoS ONE. 2011;6:e22788. doi: 10.1371/journal.pone.0022788. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Colombo AP, Boches SK, Cotton SL, Goodson JM, Kent R, Haffajee AD, et al. Comparisons of subgingival microbial profiles of refractory periodontitis, severe periodontitis, and periodontal health using the human oral microbe identification microarray. J Periodontol. 2009;80:1421–1432. doi: 10.1902/jop.2009.090185. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Hamady M, Lozupone C, Knight R. Fast UniFrac: facilitating high-throughput phylogenetic analyses of microbial communities including analysis of pyrosequencing and PhyloChip data. ISME J. 2010;4:17–27. doi: 10.1038/ismej.2009.97. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Dewhirst FE, Chen T, Izard J, Paster BJ, Tanner AC, Yu WH, et al. The human oral microbiome. J Bacteriol. 2010;192:5002–5017. doi: 10.1128/JB.00542-10. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Choo V. WHO reassesses appropriate body-mass index for Asian populations. The Lancet. 2002;360:235. doi: 10.1016/S0140-6736(02)09512-0. [DOI] [PubMed] [Google Scholar]
- 30.Flier JS, Mekalanos JJ. Gut Check: Testing a Role for the Intestinal Microbiome in Human Obesity. Science Translational Medicine. 2009;1:6ps7. doi: 10.1126/scitranslmed.3000483. [DOI] [PubMed] [Google Scholar]
- 31.Le Chatelier E, Nielsen T, Qin J, Prifti E, Hildebrand F, Falony G, et al. Richness of human gut microbiome correlates with metabolic markers. Nature. 2013;500:541–546. doi: 10.1038/nature12506. [DOI] [PubMed] [Google Scholar]
- 32.Ley RE, Backhed F, Turnbaugh P, Lozupone CA, Knight RD, Gordon JI. Obesity alters gut microbial ecology. Proceedings of the National Academy of Sciences of the United States of America. 2005;102:11070–11075. doi: 10.1073/pnas.0504978102. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Arumugam M, Raes J, Pelletier E, Le Paslier D, Yamada T, Mende D, et al. Enterotypes of the human gut microbiome. Nature. 2011;473:174–180. doi: 10.1038/nature09944. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Goodson JM, Groppo D, Halem S, Carpino E. Is Obesity an Oral Bacterial Disease? Journal of Dental Research. 2009;88:519–523. doi: 10.1177/0022034509338353. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Socransky SS, Haffajee AD. Periodontal microbial ecology. Periodontology 2000. 2005;38:135–187. doi: 10.1111/j.1600-0757.2005.00107.x. [DOI] [PubMed] [Google Scholar]
- 36.Nishimura F, Iwamoto Y, Mineshiba J, Shimizu A, Soga Y, Murayama Y. Periodontal disease and diabetes mellitus: the role of tumor necrosis factor-alpha in a 2-way relationship. J Periodontol. 2003;74:97–102. doi: 10.1902/jop.2003.74.1.97. [DOI] [PubMed] [Google Scholar]
- 37.Benguigui C, Bongard V, Ruidavets JB, Sixou M, Chamontin B, Ferrières J, et al. Evaluation of oral health related to body mass index. Oral Diseases. 2012;18:748–755. doi: 10.1111/j.1601-0825.2012.01940.x. [DOI] [PubMed] [Google Scholar]
- 38.Meisel P, Wilke P, Biffar R, Holtfreter B, Wallaschofski H, Kocher T. Total tooth loss and systemic correlates of inflammation: role of obesity. Obesity (Silver Spring) 2012;20:644–650. doi: 10.1038/oby.2011.218. [DOI] [PubMed] [Google Scholar]
- 39.Pasarica M, Dhurandhar NV. Infectobesity: obesity of infectious origin. Adv Food Nutr Res. 2007;52:61–102. doi: 10.1016/S1043-4526(06)52002-9. [DOI] [PubMed] [Google Scholar]
- 40.Zou XN, Taylor PR, Mark SD, Chao A, Wang W, Dawsey SM, et al. Seasonal variation of food consumption and selected nutrient intake in Linxian, a high risk area for esophageal cancer in China. International journal for vitamin and nutrition research Internationale Zeitschrift fur Vitamin- und Ernahrungsforschung Journal international de vitaminologie et de nutrition. 2002;72:375–382. doi: 10.1024/0300-9831.72.6.375. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.