Association between breath methane concentration and visceral fat area: a population-based cross-sectional study

High visceral fat area (VFA) is a stronger predictor of cardiovascular disease and overall mortality, compared with body mass index (BMI) and waist circumference (WC). Recent reports demonstrate that obesity is related to breath gas, which is produced by the intestinal microflora. However, these studies define obesity using BMI, not VFA. In this population-based cross-sectional study, we investigated the relationship between breath gases (methane and hydrogen) and both VFA and BMI. A total of 1033 participants (62% women; age [mean ± standard deviation] 54.4 ± 14.9 years) in the 2015 Iwaki Health Promotion Project in Japan were enrolled in the study. Breath samples were collected using a breath bag and analyzed by gas chromatography. VFA was measured using a visceral fat meter. The proportion of methanogenic bacteria to total intestinal microbiota was measured by polymerase chain reaction and 16S rRNA gene sequencing analysis. Our analysis revealed a significant association between high VFA and low breath methane, even after adjusting for confounding factors (B = −0.024 and P = 0.004). To identify the association between breath methane and VFA in participants with methane-producing bacteria in their intestinal microflora, participants were divided into two groups based on the presence or absence of methanogenic bacteria in their stool. The Methanogen + group was further divided into two subgroups with breath methane higher (Methane-UP) or lower (Methane-LO) than the median breath methane concentration. VFA was significantly lower in the Methane-UP group than in the Methane-LO group. In participants with methanogenic bacteria, breath methane concentration might be an independent biomarker of visceral fat accumulation.


Introduction
Visceral fat accumulation is a widely-recognized risk factor for mortality, likely independent of subcutaneous fat and waist circumference (WC) [1][2][3]. Metabolic risk factors, including hypertension, high blood glucose, and triglyceride (TG) concentrations, and low serum high-density lipoprotein (HDL)-cholesterol, are more strongly associated with visceral fat area (VFA) than higher subcutaneous fat, WC, or body mass index (BMI) [4][5][6]. Thus, reducing VFA decreases the risk of metabolic syndrome-induced fatal disease.
The intestinal microbiota is highly involved in host energy regulation and homeostasis, thereby affecting diabetes and/or obesity status [7]. Some studies report that the intestinal microbiota may cause obesity [8,9]. However, the precise mechanisms underlying how the intestinal microbiota affects obesity are unclear. The role of the intestinal microbiota in the development of obesity has been evaluated using a genome-based approach [8,10]. Although VFA is a stronger predictor of cardiovascular disease and overall mortality than BMI is, it is difficult to obtain in a population-based cross-sectional study. The gold standard for measuring VFA is computed tomography (CT), which is expensive and associated with radiation exposure. Therefore, the association between the intestinal microbiota and obesity is usually studied using BMI instead of VFA.
Several studies evaluating the association between BMI and the intestinal microbiota revealed that Ruminococcus [11] and Akkermansia [12] are related to obesity. In mice, increased prevalence of Methanobrevibacter smithii, a methanogenic bacteria species, is associated with increased weight gain [13]. M smithii produces methane as a byproduct of hydrogen-requiring anaerobic metabolism. Methane produced in the intestine is expelled from the body by exhalation and flatus. Therefore, some studies have evaluated the association between BMI and breath methane concentration as a noninvasive measure of methanogenic bacteria activity. Breath methane-positive people with methanogenic bacteria account for approximately 8% to 50% of the population [14][15][16][17]. A human study found that participants positive for breath methane have a significantly higher BMI [18]. Similar results were reported in a larger study [19]. However, other studies demonstrate conflicting findings [16,17]. In one study, the breath methane concentration was inversely associated with BMI [20]. In another study, no significant relationship was detected between BMI and breath methane concentration in participants with methanogenic bacteria [16]. Thus, studies on the relationship between BMI and breath methane concentration report conflicting results.
We measured VFA using an impedance method whose results highly correlate with using CT results (R>0.8) [21] to investigate the association of breath gases (methane and hydrogen) with VFA and BMI in 1033 participants.

Design, participants, and ethics
The Iwaki Health Promotion Project was launched in 2005 as an annual health check-up for local residents, aiming to prolong a healthy lifespan. Participants were recruited from men and women at least 20 years of age living in the Iwaki region of Hirosaki City in the Aomori Prefecture, Japan [22][23][24][25]. VFA was first introduced as a health check-up parameter in 2015. We used data obtained from the 2015 health check-up as a population-based cross-sectional study. The Iwaki district has approximately 11 000 adults, and 1082 participated in this community health check-up in 2015. Of these, 49 participants did not complete the clinical assessment and were excluded from the analyses. Thus, 1033 participants (397 men and 636 women; mean age±standard deviation, 54.4±14.9 years) were enrolled into the analyses. In addition, 326 individuals who participated in the health check conducted in 2016, but not in 2015 (62% female, mean age 50.7±17.5 years), were enrolled to confirm the findings obtained with the 2015 cohort.
The study was approved by the Ethics Committee of Hirosaki University School of Medicine and conducted in accordance with the principles of the Declaration of Helsinki (2014-377). Written informed consent was obtained from all participants prior to the study. This study was registered in the University Hospital Medical Information Network (UMIN-CTR, https://www.umin.ac.jp) prior to the analyses (UMIN ID: UMIN000030351).
Measurements obtained other than the methanogenic bacteria ratio All participants attended a health check-up early in the morning after fasting for at least 9 h. VFA was measured using a EW-FA90 visceral fat meter (Panasonic Corporation, Osaka, Japan). This instrument is an authorized medical device in Japan (No. 22500BZX00522000) based on high correlation with the CT method [21], which is the gold standard for VFA measurement. Breath samples were collected using a breath bag (Otsuka Pharmaceutical Corporation, Tokyo, Japan). All participants underwent training to provide endalveolar breath before sample collection. The breath was immediately transferred to a gas-tight glass syringe and 1 ml breath was injected into a gas chromatograph with a semiconductor detector (TRIlyzer mBA-3000, Taiyo Ltd, Osaka, Japan) to measure breath hydrogen and methane concentrations. The following clinical characteristics were measured as metabolic risk factors: BMI (calculated from height and weight), WC, VFA, fasting serum glucose, glycated hemoglobin (HbA1c), systolic blood pressure (SBP), diastolic blood pressure (DBP), serum TG, HDL-cholesterol, and low-density lipoprotein (LDL)-cholesterol. Blood samples were collected from the peripheral veins of the participants in the morning after at least 9 h fasting. All laboratory tests were outsourced to LSI Medience Co. (Tokyo, Japan).
Smoking (cigarettes/d), sleep time (hours/d), exercise amount (Mets·h/w), and habitual medicine use (e.g. medicine for hypertension, hyperlipidemia, diabetes, rheumatism, dementia, or allergy) were determined from questionnaires or daily journals. Daily carbohydrate, protein, fat, alcohol, and total dietary fiber intake were estimated from the Brief Diet History Questionnaire [26,27].

Measurements of the methanogenic bacteria ratio
Sample collection and DNA extraction Fecal samples from each participant were collected within 3 d prior to the study using a commercial tube kit (TechnoSuruga Laboratory Co., Ltd., Shizuoka, Japan) and cotton swabs. Sample kits were filled with 3 ml GTC solution (100 mM Tris-HCl [pH 8.0], 40 mM Tris-EDTA [pH 8.0], 4 M guanidine thiocyanate, and 0.001% bromothymol blue). Cotton swabs were saturated with 1 ml GTC solution. Fecal samples were stored at 4°C until DNA extraction.
GTC buffer solutions (800 μl feces) were added to tubes filled with zirconium beads. The tubes were then mixed at room temperature for 2 min at 5 m s −1 using a FastPrep 24 Instrument (MP Biomedicals, Santa Ana, CA, USA). After cooling, samples were centrifuged at 5000 rpm for 1 min. DNA was then extracted from the bead-treated suspension using an automatic nucleic acid extractor (Precision System Science, Chiba, Japan). MagDEA DNA 200 (GC) (Precision System Science) was used for automatic nucleic acid extraction. The final DNA concentration in each sample was adjusted to 10 ng μl −1 .

Polymerase chain reaction amplification and sequencing
To amplify the V3-V4 region of the prokaryotic 16S rRNA gene, universal primer sets were used as described previously [28]. Polymerase chain reaction (PCR) assays were performed as described previously [28]. To check amplicon size, 2.0 μl aliquots of the PCR reaction mixtures were electrophoresed on 1.0% agarose gels. The amplified fragments were purified using PCR Cleanup Filter Plates (Merck Millipore, Burlington, MA, USA). The purified PCR fragments were quantified by real-time quantitative PCR as described by Takahashi et al [28]. MiSeq system (Illumina, San Diego, CA, USA) was used for 2×300 cycle paired-end sequencing.

Taxonomic classification and calculation of the methanogenic bacteria ratio
The multiplexed paired-end reads from the Illumina MiSeq system were processed as follows. The adapter sequences and low-quality bases (threshold=20) were trimmed from the 3ʹ-ends using Cutadapt (version: 1.13). Reads containing N bases and shorter than 150 bases were discarded. The filtered reads were merged to form a single read, called a 'merged read'. Merged reads shorter than 370 bp or longer than 470 bp were excluded using the VSEARCH fastq_mergepairs subcommand (version: 2.4.3). Merged reads with more than one expected sequencing error were also excluded. After removing chimeric reads detected using the VSEARCH uchime_denovo subcommand, the remaining merged reads were clustered at a sequence identity97%. The taxa of the identified clusters were predicted using the RDP Classifier (commit hash: 701e229dde7cbe53-d4261301e23459d91615999d). Identified taxa with a confidence value<0.8 were treated as unclassified. Methanogenic bacteria amount was calculated by combining the total Methanomassiliicoccus, Methanobrevibacter, and Methanosphaera levels, which are commonly known human methanogenic bacteria [29][30][31][32]. The proportion of methanogenic bacteria was calculated as the total methanogenic bacteria divided by the total intestinal microbiota.

Statistical analysis
Participant characteristics are reported as mean-s±standard deviation (SD). Wilcoxon rank-sum tests were performed to compare the two groups. Because VFA is an area dimension (square measure), we transformed the VFA values by square root transformation to improve normality. The square root-transformed VFA value normality was confirmed by the Kolmogorov-Smirnov test. VFA values were normally distributed after square root transformation (P=0.143). The square root-transformed VFA values were used for statistical analyses. The associations between VFA and breath gases (hydrogen and methane) were assessed by multiple regression analysis with a stepwise variable selection method using VFA as an objective variable, and breath gases (hydrogen and methane) and covariates as explanatory variables. Visceral fat is affected by a variety of environmental factors, such as age [33], sex [33], BMI [21], WC [21], smoking [34], alcohol consumption [35], exercise [36], and dietary habits [37], including total dietary fiber intake [38,39]. Furthermore, the human gut microbial composition is affected by a variety of environmental factors, including age [40], dietary habits [41,42], and habitual medicine use [43]. Thus, correlations among the explanatory variates, including breath gases (hydrogen and methane), age, sex, carbohydrate intake, fat intake, protein intake, and total dietary fiber intake, smoking, alcohol consumption, exercise, and habitual medicine use, were assessed by Spearman's correlation coefficient. The correlation coefficients were all less than 0.8, confirming that multicollinearity was not a problem. P<0.05 (two-sided) was considered statistically significant. All analyses were performed using R software (version 3.3.4).

Participant characteristics
Study participant characteristics (N=1033, 62% female) are summarized in table 1. The proportion of overweight individuals (BMI25) was 30.2% for men and 18.8% for women. The obesity rate (BMI30) was 4.0% for men and 3.0% for women. The rates are comparable to the 2010 Japanese national survey (overweight and obesity rate in subjects 30-69 years of age: 33.5% for men and 20.5% for women) [44]. Mean VFA was 83.0±42.5 cm 2 , lower than the value defined as visceral obesity (100 cm 2 ) by the Japan Society for the Study of Obesity [45]. The frequency of a person taking one or more daily medications (e.g. medicine for hypertension, hyperlipidemia, diabetes, rheumatism, dementia, or allergy) was 31%. The participants were divided into two groups based on the presence (Methanogen+, n=141, 14% of all participants) or absence (Methanogen−, n=892, 86% of all participants) of methanogenic bacteria in their stool, as determined by 16S rRNA gene sequencing analysis (table 1). A simple comparison analysis with no confounder adjustment revealed that the Methanogen+group exhaled significantly more methane gas, but metabolic risk factors BMI, WC, VFA, SBP, DBP, serum glucose, TG, HDL-cholesterol, and LDL-cholesterol were not significantly different. The HbA1c level was slightly, but significantly, higher in the Methanogen+group (P=0.030). In the Methanogen+group, the proportion of methanogenic bacteria relative to all microbiota was significantly inversely associated with VFA (Spearman rank test, r=−0.21, P=0.012).
Adjusted assessment between breath gases (hydrogen or methane) and VFA Associations between breath gases (hydrogen or methane) and VFA adjusted for confounding factors; age, sex, carbohydrate intake, fat intake, protein intake, and total dietary fiber intake, smoking, alcohol consumption, exercise, and habitual medicine use were assessed by multiple regression analysis with a stepwise variable selection method using all 1033 participants (table 2). In Model 1 (breath methane versus VFA), the objective variable was VFA, while breath methane and covariates such as age, sex, dietary fiber intake, and habitual medicine use were explanatory variables. Five variables (breath methane, age, sex, dietary fiber intake, and habitual medicine use) were significantly associated with VFA. Breath methane, female sex, and dietary fiber intake were significantly inversely related to VFA (P=0.004, <0.001, and 0.018, respectively). Age and habitual medicine use were significantly positively associated with VFA (P0.001 for both). In Model 2 (breath hydrogen versus VFA), the objective variable was VFA, while breath hydrogen and covariates such as age, sex, dietary fiber intake, and habitual medicine use were the explanatory variables. Four variables (age, sex, dietary fiber intake, and habitual medicine use for yes) were significantly related to VFA. Female sex and  dietary fiber intake were significantly inversely related to VFA (P0.001 and 0.006, respectively). Age was significantly positively associated with VFA (P<0.001). Among these factors, breath methane was significantly inversely associated with VFA, but breath hydrogen was not (P=0.004 for breath methane and 0.050 for breath hydrogen).

Subgroup analyses based on the median breath methane concentration
The Methanogen+group (n=141) was further divided into two subgroups based on the median breath methane concentration: upper half (Methane-UP) or lower half (Methane-LO) of the breath methane concentration (table 3). The Methane-UP group had a significantly higher ratio of the Methanogen phyla than the Methane-LO group. Although the mean age of the Methane-UP group was significantly higher than the Methane-LO group, VFA and serum TG concentration, which increase with aging [30], were significantly lower in the Methane-UP group than in the Methane-LO group. Other metabolic risk factors, including BMI, WC, glucose, HbA1c, SBP, DBP, HDL-cholesterol, and LDL-cholesterol, were not significantly different between the Methane-Up and Methane-LO groups. The breath hydrogen concentration was not significantly different between the two breath methane concentration groups (P= 0.677). Lifestyle habits, including exercise and total fiber intake, were also not significantly different between the two breath methane concentration groups. Alcohol consumption and protein intake tended to differ between the Methane-UP and Methane-LO groups (P=0.061 and 0.085, respectively).

Adjusted assessment between breath methane with VFA in participants with methanogenic bacteria
Associations between methane and VFA in participants harboring methanogenic bacteria were assessed by multiple regression analysis with a stepwise variable selection method (table 4). Model 1 was used to determine the association between breath methane and VFA. Three variables (breath methane, sex, and habitual medicine use) were significantly related to VFA. Breath methane and female sex were significantly inversely related to VFA (B=−0.033 and P=0.002 for breath methane, B=−1.979 and P<0.001 for sex). Habitual medicine use was significantly positively associated with VFA (B=1.100 and P=0.011). Among these factors, VFA was significantly inversely associated with breath methane after adjusting for confounding factors; age, sex, carbohydrate intake, fat intake, protein intake, and total dietary fiber intake, smoking, alcohol consumption, exercise, and habitual medicine use.

Confirmation group analysis
The confirmation group characteristics were similar to the main study group (S1

Discussion
Methane is produced from hydrogen by methanogenic bacteria such as M smithii [16]. Therefore, we investigated if VFA and breath gas (hydrogen and methane) are associated. After analysis of all participant data, the association between VFA and breath hydrogen concentration was not significant, but participants with higher VFA had significantly lower breath methane concentrations (table 2). Breath methane is a product of methanogenic bacteria, which are not present in all individuals [16]. Therefore, we divided the study participants into two groups according to the presence or absence of methanogenic bacteria in their stool. Methanogenic bacteria were present in 14% of all participants, a lower percentage than reported in previous studies (25%-50%) [14][15][16]. Asians have a smaller proportion of methanogenic bacteria than other races [16,17], and the rate observed in the Japanese participants in our study was similar to previous reports. There were no significant differences in BMI or VFA between groups with and without methanogenic bacteria. Similar results were previously reported for BMI [16], but VFA was not previously evaluated. After adjusting for confounding factors, we detected a significant association between breath methane concentration and VFA. However, no association was present with HbA1c, suggesting that the results differ depending on whether or not statistical adjustments are performed. The Methano-gen+group was further divided into two subgroups according to the median breath methane concentration: upper half (Methane-UP) or lower half (Methane-LO) of the breath methane concentration. The mean age of the Methane-UP group was significantly higher than that of the Methane-LO group. BMI did not differ significantly between the two breath methane concentration groups. However, VFA significantly differed between the Methane-UP and Methane-LO groups, suggesting that VFA is more closely associated with breath methane concentration than BMI on the basis of a non-adjusted simple comparison. Even after adjusting for confounding factors, VFA remained significantly associated with breath methane, independent of the confounding factors age and sex [21,33]. Breath methane concentration may be an independent biomarker of visceral fat accumulation and a possible indicator of metabolic risks. Further investigations, including analyses of follow-up data, are required to determine the relationship between baseline breath methane concentration and high VFA over time. Breath methane is a product of methanogenic bacteria, which have a syntrophic relationship with hydrogen-producing bacteria, thus converting hydrogen to methane [46]. We hypothesized that breath methane concentration is inversely associated with breath hydrogen concentration. The data may support this hypothesis, but further investigation is needed. Hydrogen is a source of methane production, short chain fatty acids (SCFA), and hydrogen sulfide [47], which might indicate a nonlinear correlation between breath methane and hydrogen. Methanogenic bacteria change SCFA composition, acetate, and propionate, by coexisting with SCFA-producing microflora [48,49]. Furthermore, SCFA suppresses obesity-related factors by regulating G-protein coupled receptors 41 and 43 [50,51]. In addition, our results showed that participants with high breath methane harbored a higher ratio of methanogenic bacteria (table 3). Participants with Low breath methane harbored low methanogenic bacteria and increased VFA by altering SCFA composition. Furthermore, there is a report that methane can directly stimulate the secretion of glucagon-like peptide-1, which is related to obesity [52]. Although it is completely a contrasting idea, there is a research that methanogenic bacteria promote the increase in SCFA production, which leads to fat accumulation as an additional energy source, suggesting that further research is necessary. The reproducibility of our observations in this study was assessed using an independent confirmation group enrolled from the 2016 health check. The confirmation group analyses were similar to the main study group. There seems to be a commonly held belief that methane is linked to obesity and some reports support this hypothesis [18,19]. However, other studies demonstrate conflicting findings [16,17]. Therefore, the relationship between breath methane and BMI is inconclusive in previous reports [16,[18][19][20]. Therefore, we investigated the relationship between breath methane and VFA, which correlates with BMI but varies with race [53]. Thus, VFA may be a confounding factor for the relationship between breath methane and BMI and could explain the previous inconsistencies [16,[18][19][20], although further investigation is required in populations including different races. Our study revealed no significant association between BMI and breath methane concentration, whereas VFA was significantly inversely associated with breath methane concentration.
In some intestinal microflora studies, diversity is related to BMI as an obesity index [54,55]. Therefore, we evaluated the association between methanogenic bacteria and VFA. The proportion of methanogenic bacteria was significantly inversely associated with VFA in the Methanogen+group, suggesting that methanogenic bacteria is related to this diversity. If methanogenic bacteria maintain or improve VFA status, these bacteria might be a new target/index for obesity and diabetes.
Previously, total dietary fiber intake significantly increased breath methane concentrations [16]. However, we did not observe this relationship in the present study (table 4). After adjusting for confounding factors, total dietary fiber intake tended toward a positive relationship with breath methane. Thus, increasing dietary fiber intake may lead to a higher breath methane concentration and thereby suppress visceral fat accumulation.
Our study has several key strengths and some limitations. The large number of participants and the use of an independent confirmation group are advantageous to our study. It has been reported that exposure to CT may be a risk factor for cancer [56,57]. Therefore, in a study for relatively healthy people, using our method is superior from the ethical point of view. Therefore, another strength was VFA measurement using an impedance method, which highly correlates with gold standard methods such as CT [21]. One limitation is that we used a cross-sectional study and not a longitudinal study, nor could we assess whether subjects with low methane currently having low VFA will go on to develop higher VFA in the future. A longitudinal study is necessary to determine causality and the associated mechanisms. Because this study was performed in a single country and a single race, reproducibility should be confirmed in other demographic settings. Another limitation is that, although the correlation between breath methane levels and VFA can be determined in only the methane producer population, which is 8-50% of the population depending on country, this was not done. Furthermore, a recent study reported that methane can be produced directly by cellular processes in the absence of microbes and that this could be further linked to products of oxidative stress [58]; however, we did not consider this viewpoint.
We are the first to report that the human gut microbial composition is associated with VFA and is more strongly associated with metabolic risk factors than higher subcutaneous fat, WC, or BMI in a relatively large population. The association between breath methane and VFA was evaluated, including the effect of methanogenic bacteria. Reproducibility was confirmed in an independent population that participated in the health check conducted in a different year.

Conclusions
Our large study suggests that breath methane is significantly inversely associated with VFA in participants harboring methanogenic bacteria. Further, breath methane is more closely associated with VFA than BMI in these participants. However, a longitudinal study is necessary to determine causality and the associated mechanisms.