Hydrogen Sulfide and Methane on Breath Test Correlate with Human Small Intestinal Hydrogen Sulfide Producers and Methanogens

Abstract

Background

Evidence indicates that 3 gas breath testing (BT) correlates with stool microbial populations. Breath methane (CH₄) levels correlate with stool Methanobrevibacter smithii levels and constipation, while hydrogen sulfide (H₂S) levels correlate with stool H₂S producers and diarrhea. However, their relationships to small bowel microbes are unknown.

Aims

To assess relationships between small bowel microbes and breath gases.

Methods

REIMAGINE study subjects completed a fasting single-sample BT for CH₄ and H₂S. During esophagogastroduodenoscopy without colon preparation, duodenal aspirates were obtained using double-lumen sterile aspiration catheters. Microbial DNAs underwent shotgun sequencing (NovaSeq6000).

Results

Duodenal bacterial profiles differed significantly in subjects with breath H₂S ≥ 1.5 ppm vs. those with < 1.5 ppm, with 2.08-log₂fold greater prevalence of phylum Thermodesulfobacteriota. Higher breath H₂S levels correlated with greater duodenal prevalences of H₂S producers, including Proteus mirabilis (P = 0.002), Desulfosarcina widdelii (P = 0.027), and Desulfobulbus oligotrophicus (P = 0.041); co-occurrence of all 3 species correlated with ~ 50% higher breath H₂S levels (P = 0.0001). Duodenal archaeal profiles differed significantly in subjects with intestinal methanogen overgrowth (IMO, CH₄ ≥ 10 ppm), with 2.94-log₂fold greater prevalence of family Methanobacteriaceae vs. non-IMO subjects. Higher breath CH₄ correlated with greater prevalences of methanogens including M. smithii (P = 0.02), Halarchaeum sp. CBA1220 (P = 0.003), Desulfurococcus mucosus (P = 0.046), and Halobaculum rubrum (P = 0.049). IMO was more common in subjects with co-occurrence of all 4 species (P = 0.04). In IMO-positive subjects, CH₄ levels correlated with greater constipation severity (P = 0.019); P. mirabilis (P = 0.021) and D. oligotrophicus (P = 0.003) correlated with looser stool in IMO-negative subjects. M. smithii prevalence correlated with known hydrogen-producing syntrophs, e.g., Christensenella minuta (P < 0.001).

Conclusion

This study demonstrates that duodenal prevalences of H₂S-producing bacteria and methanogenic archaea contribute to H₂S and CH₄ levels, respectively, on BT.

Supplementary Information

The online version contains supplementary material available at 10.1007/s10620-025-09156-y.

Introduction

Small intestinal bacterial overgrowth (SIBO) has been recognized as a clinical condition for half a century [1]. Early studies identified SIBO by culturing small bowel aspirates [2, 3]. Subsequently, breath testing (BT) became an established indirect technique for diagnosing SIBO based on levels of hydrogen (H₂) on the breath [4]. Originally, BT only measured H₂, but later testing also incorporated measurement of methane (CH₄). This allowed researchers to elucidate the importance of CH₄, which is produced in the gut by methanogenic archaea[5], in delayed gastrointestinal (GI) transit and constipation, including constipation-predominant irritable bowel syndrome (IBS-C) [6–9]; increased intestinal colonization with methanogens is now known as intestinal methanogen overgrowth (IMO) [10, 11]. Neither H₂ nor CH₄ are produced by human cells, so measuring these gases provides exclusive indirect markers of gut microbial composition. Adopting BT for the diagnosis of SIBO has facilitated an ever-increasing understanding of the role of SIBO in conditions such as irritable bowel syndrome (IBS) [12], celiac disease [13], and others.

Recently, a third gas, hydrogen sulfide (H₂S) has been introduced to BT [14]. H₂S is a gasotransmitter that plays important roles in inflammation and mucosal repair in the GI tract [15] and has been linked to diarrhea-predominant irritable bowel syndrome (IBS-D) [16, 17]. In humans, H₂S is mainly produced by gut bacteria, including sulfate reducers which have been linked to colorectal cancer, ulcerative colitis [18], and persistent diarrhea [19].

An ongoing criticism of BT is the lack of studies comparing BT to the small bowel microbiome. The advent of whole-genome sequencing has facilitated research in this area. A recent study using lactulose demonstrated that a rise in H₂ of ≥ 20 ppm by 90 min on BT was the best marker for SIBO as defined by culture and sequencing, suggesting an increased H₂-producing capacity in the small bowel [20] and confirming the validity of diagnosing SIBO using H₂ on BT. Another whole-genome sequencing study confirmed that breath CH₄ levels correlated with constipation and stool levels of Methanobrevibacter smithii, the predominant methanogen in humans, and also showed that breath H₂S levels correlated with stool levels of H₂S producers [11]. Although there is good correlation between breath gases and stool levels of CH₄ and H₂S producers, we hypothesized that breath CH₄ and H₂S levels would also be reflective of, or influenced by, small bowel microbial populations.

In this study, we analyze and compare CH₄ and H₂S levels on fasting single-sample BT to the results of shotgun sequencing analysis of duodenal microbiome composition.

Methods

Subject Recruitment

Samples for this study were obtained through the REIMAGINE study [21]. Consecutive patients aged 18–85 years undergoing standard-of-care esophagogastroduodenoscopy (EGD) or antegrade double-balloon enteroscopy (DBE) without colonoscopy were asked to participate in the REIMAGINE study. Consenting participants provided fasting serum samples and completed a medical history questionnaire, after which samples of duodenal luminal fluid were obtained during endoscopy. In 2024, the REIMAGINE study was amended to include acquiring a single fasting breath sample on the morning of endoscopy. Cedars-Sinai Institutional Review Board approved the study protocol (Pro00035192) and all participants provided informed written consent prior to participation.

Breath Testing

On the day of endoscopy, a single fasting exhaled breath sample was collected in sealed multilayer foil breath bags to measure carbon dioxide (CO₂), CH₄, and H₂S levels by gas chromatography (Gemelli Biotech, Raleigh, NC). It was not possible to perform full 2-h lactulose breath tests in these subjects prior to the scheduled standard-of-care endoscopy, and as changes in breath hydrogen (H₂) levels cannot be determined from single-sample breath tests, H₂ levels were not considered in this study. Previous studies suggest single fasting breath samples are sufficient to assess CH₄ [22] and H₂S production [11]. A positive CH₄ breath test for intestinal methanogen overgrowth (IMO) was defined as ≥ 10 parts per million (ppm) [22]. There is currently no established consensus or guideline regarding a threshold for H₂S levels.

Small Intestinal Sample Collection and Processing

Aspirates were obtained with a dual-lumen protected catheter (Hobbs Medical, Inc.) via sterile aspiration. Up to 2 mL of luminal fluid were collected from the second portion of the duodenum [21]. Care was taken not to add fluid into the duodenum prior to aspiration. Aspirated samples were immediately transferred on ice to the laboratory, where sterile 1 × Dithiothreitol (DTT) (EMD Millipore Corp., Darmstadt, Germany) was added (1:1 ratio). Samples were processed as described previously and microbial pellets were stored at − 80 °C [21].

DNA Isolation

DNAs were isolated from duodenal aspirate microbial pellets using MagAttract PowerMicrobiome DNA KF kits (Qiagen) on a KingFisher Flex System (Thermo Fisher Scientific) as described previously [21], using 1xDTT aliquots as negative controls. DNA purity and concentration were determined using a NanoDrop One spectrophotometer (Thermo Fisher Scientific).

Library Preparation and Shotgun Sequencing

Libraries for whole-genome (shotgun) sequencing were prepared using Illumina-DNA Prep kit and IDT for Illumina-DNA/RNA UD Indexes (Illumina), as described previously [23]. Library qualities were analyzed on an Agilent 2100 Bioanalyzer System (Agilent Technologies, Santa Clara, CA). Sequencing was performed on a NovaSeq platform (Illumina), using NovaSeq 6000 S2 Reagent Kits v1.5 (300 cycles) and analyzed using CLC Genomics Workbench 22.0.2/20.0.3 and Microbial Module (Qiagen). Reads were demultiplexed and then trimmed using the Trim Reads Tool (Qiagen) and following parameters: quality limit of 0.05, maximum number of ambiguities of 2, and a trim adapter list containing the sequences of Illumina adapters. For co-occurrence analysis, host reads were removed using the Homo sapiens Genome Reference Consortium Human Build 38 as the genomic index. After trimming and two-level de-hosting, reads were mapped to the PlusPFP database (10/9/2023 release) with Dragen Metagenomics Pipeline (version 3.5.13) and the built-in kraken2 tool (version 2.0.8) using default parameters. Microbial metabolic pathways were built as previously described [23] and identified using the MetaCyc Pathway Database (2022-05).

For microbiome comparisons, a deeper de-hosting step was performed using k-mer-based read classification using Kraken 2 v. 2.1.3. As a Homo sapiens reference, a collection of 408 unmasked genome assemblies was compiled from two haploid human genome assemblies (GRCh38.p14 released by the Genome Reference Consortium, NCBI RefSeq assembly GCF_000001405.40, NCBI BioProject PRJNA168, and T2T-CHM13v2.0 released by the Telomere-to-Telomere (T2T) Consortium, NCBI RefSeq assembly GCF_009914755.1, NCBI BioProject PRJNA807723) and a collection of 406 haploid human pangenome assemblies released by the Human Pangenome Reference Consortium (NCBI BioProject PRJNA730822). During this step a confidence of 0.0 and a minimum hit group of 2 were used. After this step, unaligned reads were mapped to the PlusPFP (version 9/4/2024) using kraken2, and a confidence score of 0.5 and a minimum hit group of 2 were used. The PlusPFP database (version 9/4/2024) database also contained a human DNA sequence to retrieve any leftovers not aligned in the previous step.

Statistical Analysis

Continuous variables were compared using t-test or Mann–Whitney U test for 2 groups. Comparisons between three or more groups were performed by one-way ANOVA or Kruskal–Wallis. Correlations were analyzed by Spearman rank correlation coefficients. Statistical analysis was performed using SPSS 24.0 (SPSS, Chicago, IL), Graphs were constructed using GraphPad Prism 9.2.0 (GraphPad Software, La Jolla, CA). Significance was set at P < 0.05. Venn diagrams were constructed using Displayr (New South Wales, Australia).

Comparisons and statistical analyses were performed with CLC Genomics Workbench v.25.0.1 and CLC Microbial Genomics Module v.25.0.1 (Qiagen). The number of reads (after de-hosting steps) assigned to bacterial and archaeal taxa were compared, after removing taxa represented by singletons and doubletons per group. Despite similar library sizes between groups, fold changes were calculated from the GLM, which corrects for differences in library size between the samples and the effects of confounding factors. An unbiased analysis was performed to compare differences in relative abundance (RA) between groups at the species level. A negative binomial GLM model was used to obtain maximum likelihood estimates for the FC of a feature between groups, and the Wald test was used for determination of significance. False discovery rate was performed to correct the P value. FDR P value < 0.05 was considered strongly significant for all bacterial analysis, while a FDR P value < 0.1 was considered a trend towards being significant. Considering the very low diversity and abundance in the small bowel, a FDR P value < 0.1 was considered significant for the archaeome studies. Microbial richness and evenness were evaluated using the Shannon index, and Bray–Curtis was used to evaluate duodenal microbial beta-diversity. Partial Least Squares Discriminant Analysis (PLS-DA) studies were conducted using the MetaboAnalyst, version 6.0 (https://www.metaboanalyst.ca/home.xhtml). Data were log₁₀ transformed and Pareto scaling was applied.

Results

Subject Demographics

Duodenal aspirates from 110 subjects (females: 64.4%, age: 56.4 ± 15.8 years, BMI: 26.3 ± 6.5 kg/m²) were analyzed. BT data were collected for all 110 subjects; however, H₂S levels were not recorded for one subject. Of the 110 subjects, 22 (20%) had breath CH₄ levels ≥ 10 ppm and were considered positive for intestinal methanogen overgrowth (IMO). H₂S levels on BT (from 109 subjects) ranged from 0.31 to 3.54 ppm. The top 10 indications for endoscopy in study subjects are shown in Supplemental Table 1.

Duodenal microbial profiles were analyzed in all subjects via whole-genome (shotgun) sequencing of duodenal aspirates. On average, 174,029,497 reads (SD = ± 110,620,821) were generated per sample. 75 to 96% of the initial reads mapped to human DNA, and approximately 44% (SD = ± 18.52%) of the unmapped reads mapped to the PlusPFP database. Of these, 99% (SD = ± 0.08%, range: 17 to 100%) mapped to Bacteria. The remaining reads mapped to Eukaryotes, Fungi, Viruses, and Archaea, in this order. The average number of classified reads was not statistically different between groups (P = 0.23).

H₂S producers in the Duodenum Correlated with H₂S on Breath Test

The duodenal bacterial profile of subjects with ≥ 1.5-ppm H₂S on BT appeared to be different from subjects with < 1.5 ppm H₂S on BT, regardless of concurrent IMO (Supplemental Fig. 1), with increased bacterial alpha-diversity (P = 0.003, Supplemental Fig. 2). In addition, the relative abundance (RA) of phylum Thermodesulfobacteriota, which comprises the majority of sulfate-reducing bacteria (SRB), was significantly increased in the duodenum of subjects with ≥ 1.5 ppm H₂S on BT compared to subjects with < 1.5 ppm H₂S (log₂ fold change (log₂FC): 2.08, FDR P value = 6.61E−4). A Partial Least Squares Discriminant Analysis (PLS-DA) of this phylum revealed that the majority of species leading to the separation of the groups in a principal component plot were Desulfovibrio desulfuricans, D. fairfieldensis, and other unidentified Desulfovibrio species (Supplemental Fig. 3). The RA of these and other species from phylum Thermodesulfobacteriota were also significantly increased in subjects with ≥ 1.5 ppm H₂S on BT compared to subjects with < 1.5 ppm (FDR P value < 0.05, Supplemental Table 2).

Duodenal RA of known H₂S-producing species [24–26] exhibited correlations with breath H₂S levels, including five species from phylum Thermodesulfobacteriota, Desulfovibrio desulfuricans (R = 0.23, P = 0.018), Desulfobulbus oligotrophicus (R = 0.22, P = 0.025), Desulfobulbus oralis (R = 0.20, P = 0.033), Desulfosarcina widdelii (R = 0.24, P = 0.011), and Desulfuromonas sp. DDH964 (R = 0.23, P = 0.017) and one from phylum Chrysiogenota, Desulfurispirillum indicum (R = 0.24, P = 0.011). Desulfoluna limicola trended toward significance (R = 0.17, P = 0.067). The duodenal RA of another well-known H₂S producer, Proteus mirabilis (phylum Proteobacteria), also correlated with breath H₂S levels (R = 0.32, P = 0.001) (Supplemental Fig. 4, Table 1).

Table 1.

Prevalences of duodenal bacteria in the study population, and correlations between their relative abundances (RA) and breath H₂S levels (ppm)

	Bacterial species	Prevalence within the study population (N = 109)	Correlation between duodenal RA and breath H2S levels
			R	P value
1	Desulfobulbus oligotrophicus	51.38%	0.22	0.025
2	Desulfobulbus oralis	71.56%	0.20	0.033
3	Desulfoluna limicola	41.28%	0.17	0.067
4	Desulfosarcina widdelii	55.05%	0.24	0.011
5	Desulfovibrio desulfuricans	64.22%	0.23	0.018
6	Desulfurispirillum indicum	53.21%	0.24	0.011
7	Desulfuromonas sp. DDH964	56.88%	0.23	0.017
8	Proteus mirabilis	77.98%	0.32	0.001

We next examined associations between the presence and co-occurrence of the eight specific H₂S producers described above and breath H₂S levels. H₂S levels were significantly increased in subjects with any three or more of these H₂S producers in their duodenal microbiome (P = 0.0239) (Fig. 1). Individually, there were no differences in breath H₂S levels in subjects with or without duodenal Desulfurispirillum indicum (P = 0.085), Desulfuromonas sp. DDH964 (P = 0.143), Desulfobulbus oralis (P = 0.346), Desulfoluna limicola (P = 0.109), or Desulfovibrio desulfuricans (P = 0.179) (Supplemental Fig. 5). However, subjects with Proteus mirabilis (P = 0.002), Desulfosarcina widdelii (P = 0.027) or Desulfobulbus oligotrophicus (P = 0.041) had higher breath H₂S levels than subjects without these species (Fig. 2). Subjects who had all three of these H₂S producers in their duodenal microbiome had ~ 50% higher breath H₂S levels than subjects with none of these species (P = 0.0001, Fig. 3).

Fig. 1.

Comparison of breath H₂S levels in subjects (N = 109) with zero to three or more of these H₂S producers in their small bowel (duodenal) microbiome: Proteus mirabilis, Desulfovibrio desulfuricans, Desulfobulbus oligotrophicus, Desulfobulbus oralis, Desulfosarcina widdelii, Desulfurispirillum indicum, Desulfuromonas sp. DDH964, and Desulfoluna limicola

Fig. 2.

Comparison of breath H₂S levels in subjects with (N = 85) or without (N = 24) Proteus mirabilis (A), with (N = 60) or without (N = 49) Desulfosarcina widdelii (B), and with (N = 56) or without (N = 53) Desulfobulbus oligotrophicus (C) in their duodenal microbiome

Fig. 3.

Comparison of breath H₂S levels in subjects with Proteus mirabilis, Desulfosarcina widdelii, and Desulfobulbus oligotrophicus (N = 39) and those without (N = 16) in their duodenal microbiome

Other species known to produce H₂S [24–26] were also evaluated, including C. difficile, H. pylori, Fusobacterium spp., Shewanella spp., Citrobacter spp., Salmonella spp., Proteus spp., and all species from the order Desulfovibrionales. The richness and evenness of these species were very significantly different between subjects with ≥ 1.5-ppm H₂S on BT compared to subjects with breath H₂S < 1.5 ppm (Shannon index P = 0.0025, Supplemental Fig. 6), and several H₂S bacterial producers were overrepresented in the duodenum of subjects with ≥ 1.5-ppm H₂S on BT (Supplemental Tables 3 & 4), including several Proteus species. For example, subjects with ≥ 1.5-ppm H₂S on BT had higher duodenal RA of P. vulgaris (log₂FC:2.24, FDR P value = 1.31E−3) and P. mirabilis when compared to subjects with breath H₂S < 1.5 ppm (log₂FC:1.17, FDR P value = 0.10, Supplemental Table 3).

CH₄ Producers in the Duodenum Correlated with CH₄ on Breath Test

The duodenal archaeal profile of subjects with IMO appeared to differ from subjects without IMO (Supplemental Fig. 7) and trended toward clustering together on a principal component analysis of beta-diversity (P = 0.08, Supplemental Fig. 8). The RA of the entire Methanobacteriaceae family was 2.94-log₂ fold increased in the duodenum of IMO subjects when compared to non-IMO subjects (P value = 1.29E-3, FDR P value = 0.07).

The predominant methanogen, M. smithii, was present in the duodenum of 63.64% of subjects and M. smithii RA correlated with breath CH₄ levels (R = 0.221, P = 0.02). RA of eight other archaeal species [27–30] also correlated with breath CH₄ (Table 2, Supplemental Fig. 9). Three of these, Halarchaeum sp. CBA1220, Desulfurococcus mucosus, and Halobaculum rubrum, co-occurred with M. smithii in the duodenum (Supplemental Fig. 10), but there were no direct correlations between the RAs of any of these species. Breath CH₄ levels increased significantly with co-occurrence of two (P = 0.018), three (P = 0.027), or four (P = 0.008) of these methanogens, as compared to none (Fig. 4A). A positive BT for IMO (CH₄ ≥ 10 ppm) was also more common when all four of these species co-occurred (P = 0.04) (Fig. 4B). Furthermore, breath CH₄ levels in IMO-positive subjects correlated with constipation severity (R = 0.577, P = 0.019).

Table 2.

Prevalences of duodenal archaea in the study population and correlations between their relative abundances (RA) and breath CH₄ levels (ppm) and duodenal Methanobrevibacter smithii RA

	Archaeal species	Prevalence within the study population (N = 110)	Correlation between duodenal RA and breath CH4 levels		Correlation between duodenal RA and duodenal RA of M. smithii
			R	P value	R	P value
1	Methanothermococcus okinawensis	40%	0.319	0.001	0.282	0.003
2	Halomicroarcula sp. SHR3	55.45%	0.291	0.002	0.194	0.042
3	Halarchaeum sp. CBA1220	63.64%	0.277	0.003	0.190	0.362
4	Thermococcus argininiproducens	29.09%	0.226	0.017	0.277	0.003
5	Methanobrevibacter smithii	63.64%	0.221	0.02	N/A	N/A
6	Candidatus Nitrosotenuis cloacae	36.36%	0.209	0.028	0.379	< 0.0001
7	Halobacterium litoreum	57.27%	0.209	0.028	0.312	0.025
8	Desulfurococcus mucosus	27.27%	0.19	0.046	0.077	0.713
9	Halobaculum rubrum	61.82%	0.188	0.049	0.079	0.707

Fig. 4.

A Comparison of breath methane levels in subjects with zero to four of these methane producers in their duodenum: M. smithii, Halarchaeum sp. CBA1220, D. mucosus, and H. rubrum. B Comparison of the percentage of IMO-positive subjects (CH₄ ≥ 10 ppm) when zero to four of these methane producers were present

CH₄ Producers Correlated with Known H₂-Producing Bacterial Syntrophs in the Duodenum

The bacterial families Christensenellaceae and Ruminococcaceae include known H₂-producing syntrophs for methanogens [31–34]. Consistent with this, positive correlations were found between RAs of bacterial species from these families and the archaeal species that correlated with breath CH₄. M. smithii RA correlated with RAs of Christensenella minuta (R = 0.313, P < 0.001), Ruminococcus gnavus (R = 0.433, P < 0.0001), Ruminococcus lactaris (R = 0.329, P < 0.001), Ruminococcus albus (R = 0.349, P < 0.001), Ruminococcus bicirculans (R = 0.244, P = 0.01), Ruminococcus bovis (R = 0.335, P < 0.001), Ruminococcus champanellensis (R = 0.286, P = 0.003), and Ruminococcus gauvreauii (R = 0.346, P < 0.001) (Supplemental Table 5). Other methanogens that correlated with breath CH₄ but not with M. smithii RA (Table 2) also correlated with RAs of Christensenellaceae and Ruminococcaceae species (Supplemental Table 5).

M. smithii also correlated positively with another H₂-producing bacterium, Dorea formicigenerans (R = 0.463, P < 0.0001), but correlated negatively with Enterococcus faecalis (R = −0.212, P = 0.027) and Enterococcus faecium (R = −0.231, P = 0.015).

Pathway Analysis Supported the Presence of H₂S Production and Methanogenesis in the Small Bowel

Microbial metabolic pathways were generated for 105 study subjects. An average of 64.9% of cleaned reads mapped to a microbial genome, of which 0.09% mapped to archaeal genomes (range 0.001–1.9%). Proteus mirabilis RA correlated with assimilatory sulfate reduction pathways I (R = 0.280, P = 0.004), II (R = 0.226, P = 0.020), and III (R = 0.324, P = 0.001), and with the assimilatory sulfite reductase enzyme (NADPH) (R = 0.297, P = 0.002) found in assimilatory sulfate reduction pathways I and III, in which sulfite is converted to H₂S. P. mirabilis also correlated with the dissimilatory sulfate reduction I (to hydrogen sulfide) pathway (R = 0.286, P = 0.003), in which sulfate is reduced to H₂S. D. oligotrophicus RA correlated with assimilatory sulfate reduction pathways I (R = 0.262, P = 0.007), II (R = 0.209, P = 0.033), III (R = 0.296, P = 0.002), and IV (R = 0.234, P = 0.016), and the assimilatory sulfite reductase enzyme (NADPH) (R = 0.287, P = 0.003). P. mirabilis (R = 0.211, P = 0.031) and D. oligotrophicus (R = 0.309, P = 0.001) also correlated with the sulfate assimilation and cysteine biosynthesis superpathway.

M. smithii RA correlated with the ‘methanogenesis from H₂ and CO₂’ pathway (R = 0.197, P = 0.044) and importantly, also correlated with the archaeal coenzyme F420 biosynthesis I pathway (R = 0.256, P = 0.008). Coenzyme F420 is involved in key steps during methanogenesis [35]. Halarchaeum sp. CBA1220 RA also correlated with the ‘methanogenesis from H₂ and CO₂’ pathway (R = 0.230, P = 0.018).

CH₄ Producers Correlate with the Presence of H₂S Producers in the Duodenum

Analyses revealed a trend toward correlation between H₂S and CH₄ levels on BT (R = 0.181, P = 0.059). Further, RAs of some of the duodenal H₂S producers identified in this study also correlated with breath CH₄ levels (P. mirabilis [R = 0.191, P = 0.047], D. oligotrophicus [R = 0.207, P = 0.031], and D. widdelii [R = 0.228, P = 0.017]), and with RAs of some of the CH₄ producers identified: P. mirabilis correlated with M. smithii (R = 0.190, P = 0.047), M. okinawensis (R = 0.192, P = 0.045), and Halomicroarcula sp. SHR3 (R = 0.218, P = 0.022); D. oligotrophicus correlated with Halarchaeum sp. CBA1220 (R = 0.351, P < 0.0001), Candidatus Nitrosotenuis cloacae (R = 0.313, P < 0.0001), and Halobaculum rubrum (R = 0.277, P = 0.004); and D. widdelii correlated with M. smithii (R = 0.254, P = 0.008), Halomicroarcula sp. SHR3 (R = 0.316, P = 0.001), Halarchaeum sp. CBA1220 (R = 0.200, P = 0.037), and Candidatus Nitrosotenuis cloacae (R = 0.201, P = 0.036). In IMO-negative subjects, P. mirabilis (R = 0.281, P = 0.021) and D. oligotrophicus (R = 0.353, P = 0.003) RAs correlated with looser stool.

Discussion

Here, we demonstrate that CH₄ levels on breath testing (BT) exhibit correlations with levels of methanogens in the small intestine. Importantly, the predominant methanogen M. smithii and other methanogens are present in the small intestine and appear to play additive roles in contributing to breath CH₄ levels, in addition to contributions from colonic methanogens. We also show that H₂S levels on BT exhibit correlations with levels of H₂S-producing bacteria in the small intestine, including specifically P. mirabilis, D. oligotrophicus, and D. widdelii. These data support the concept that breath gas profiles are influenced, at least in part, by small bowel microbial composition.

Over the last half century, SIBO and BT have been increasingly understood. Despite the long history of BT as an indirect technique for assessing SIBO, and more recently IMO, the methodology to validate BT was not available until recently. An important question was whether BT findings were related to the levels of microorganisms producing these gases. The lack of data supported criticisms of BT, such as suggestions that BT, and particularly lactulose BT, is merely a marker of transit [36, 37]. Moreover, the measurement of breath H₂S, and the understanding that overgrowth of H₂S producers may contribute to patient symptoms [14], is still evolving. Here, we show that breath CH₄ and H₂S levels are in fact influenced by small bowel microbial composition. Enhanced understanding of different types of microbial overgrowth and potential competition for gas utilization will improve how we interpret breath testing.

Methane is an important gas in breath testing. Substantial evidence now suggests that CH₄ produced in the gut alters gut neuromuscular function [6] and is associated with constipation [7]. Further, constipation severity is proportional to breath CH₄ levels [38] and stool M. smithii levels. In a randomized-controlled study, reducing CH₄ with a combination of antibiotics improved constipation [8, 39]. Recently, stool microbiome analysis in IBS-C subjects found direct correlations between breath CH₄, constipation, and M. smithii levels [11]. Moreover, a recent consensus now recognizes CH₄ as a contributor to constipation in IBS-C [40]. However, studies of CH₄ on BT to date, including microbiome studies, have utilized stool. Here, we confirm that, as previously shown [41], M. smithii and other methanogens also colonize the small intestine and that the duodenal archaeal profile of subjects with IMO is significantly different from that in non-IMO subjects, with an almost threefold increase in the relative abundance (RA) of the entire Methanobacteriaceae family in the duodenum of IMO subjects. Further, we demonstrate relationships between breath CH₄ levels and the duodenal RAs of M. smithii and other archaea including Halarchaeum sp. CBA1220, Desulfurococcus mucosus, and Halobaculum rubrum. Lastly, the RAs of both M. smithii and Halarchaeum sp. CBA1220 correlated with the archaeal pathway for ‘methanogenesis from H₂ and CO₂.’ These findings add to stool data on methanogens, and demonstrate that methanogens are found in both the small and the large intestine.

In contrast to CH₄, there are fewer studies on breath H₂S and human health. Studies suggest that H₂S levels are proportional to diarrhea severity [11, 42], but again, are based on stool. In a recent study, subjects with diarrhea-predominant IBS had higher breath H₂S levels than subjects with other IBS subtypes [11], confirming earlier findings by Banik et al. [16]. Further, breath H₂S levels correlated with increases in H₂S-producing bacteria in stool, including Fusobacterium and Desulfovibrio spp., and with microbial metabolomic markers of H₂S production [11]. A subsequent animal study confirmed forced colonization of rats with H₂S producers Desulfovibrio piger or Fusobacterium varium resulted in increased gut H₂S production and a diarrhea-like phenotype [42]. Here, we show that the duodenal microbial profiles of subjects with ≥ 1.5 ppm H₂S on BT are significantly different from those of subjects with breath H₂S < 1.5 ppm, exhibiting increased alpha-diversity and a more than twofold increase in the duodenal RA of the entire phylum Thermodesulfobacteriota, to which most sulfate-reducing bacteria (SRB) belong, and that the difference between groups was driven by Desulfovibrio desulfuricans and other Desulfovibrio species. We also demonstrate that breath H₂S levels correlate with the RAs of H₂S-producing bacteria in the duodenum, including phylum Thermodesulfobacteriota species Desulfovibrio desulfuricans, D. oligotrophicus, and D. widdelii, as well as Desulfurispirillum indicum (phylum Chrysiogenota) and Proteus mirabilis (phylum Proteobacteria). Interestingly, P. mirabilis is a well-known H₂S producer [43] that was considered years ago as a possible contributor to SIBO [44]. Subjects with duodenal P. mirabilis exhibit higher H₂S levels on BT, and duodenal P. mirabilis RA correlates with the dissimilatory sulfate reduction I pathway (in which sulfate is reduced to H₂S), the assimilatory sulfate reduction pathways, and the enzyme responsible for sulfite conversion to H₂S. Consistent with this, a previous study showed that P. mirabilis can use multiple substrates to produce H₂S, including cysteine, cystine, homocystine, mercaptoacetate, sodium sulfite, and sodium thiosulfate [45]. Lastly, subjects with higher breath H₂S levels (≥ 1.5 ppm) had higher duodenal P. mirabilis RAs than those with breath H₂S < 1.5 ppm. While P. mirabilis duodenal RAs may appear low, we previously showed that the duodenum contains 10³–10⁹ 16S bacterial copies/mL [46]. Therefore, this P. mirabilis RA represents 10² 16S copies/mL of aspirate. Elevated breath H₂S may also require separate terminology from SIBO (i.e., individuals with a positive hydrogen breath test). The concept of intestinal sulfide overproduction (ISO) and how it relates to SIBO and IMO will need to be assessed by breath testing in a large clinical cohort.

Interestingly, there appears to be an overlap in colonization of subjects with methanogens and H₂S producers and the manifestation of stool phenotypes. Looseness of stool appears to only occur in the presence of H₂S producers and absence of methanogens. Constipation is the dominant feature any time methanogens result in a CH₄ level ≥ 10 ppm (i.e., IMO). Below this point, H₂S producers correlate with looser stool. It is, however, important to note that in this study only the small bowel is being assessed. There is a likely possibility that colonic production of both CH₄ and H₂S further compound gas production and levels on breath test, as well as phenotypes.

It is notable to find methanogens in the small intestine as the small intestine is often recognized as a higher oxygen portion of the gut that should not allow for the growth of strict anaerobes such as methanogens. The small intestine is home to both aerotolerant microbes (aerobes and facultative aerobes) and anaerobes. Friedman et al. showed that oxygen levels in the duodenum are much greater than in the distal gut but also showed that radial oxygen levels vary from lumen to mucosa [47], presumably creating niche environments that can be colonized by different types of microbes.

This study has some limitations. Here, analyses were based on fasting breath samples and not a full BT profile after carbohydrate ingestion. However, fasting samples have been shown to be sufficient to assess CH₄ [22], since methanogenesis itself is not a fermentative process, and appear also be sufficient to assess H₂S production [11]. We also could not measure breath H₂ levels or assess for overgrowth of H₂ producers. Hydrogen is rapidly diffused or consumed after production. As such, fasting H₂ is typically very low and requires carbohydrate ingestion for its detection. Here, only a fasting single-sample BT could be performed prior to the scheduled standard-of-care endoscopies. Another limitation is a lack of small bowel shotgun sequencing from healthy subjects. This is a challenge and would require consent for a healthy individual to undergo upper endoscopy. Additionally, this study utilized duodenal luminal samples. We have previously shown that the luminal microbial compositions in different small intestinal segments are highly similar [48] although other studies do not fully support this [49, 50]. Lastly, diet can affect microbial composition, fermentation, and gas production in the gut. While all patients in this study were similarly fasted prior to endoscopy, diet in the 24–48 h before the endoscopy procedure was not controlled for.

In conclusion, CH₄ and H₂S levels on breath testing exhibit correlations with levels of methanogenic archaea and H₂S-producing bacteria in the small intestine. This, combined with previous data (using full breath testing) indicating that H₂ correlates with small bowel Enterobacteriaceae levels [20], suggests that breath test results are also influenced by small intestinal microbiome composition, and that the sources of these gases also include the small intestine. These results have important consequences. First, they reinforce the importance of breath testing and its interpretation. Excessive methane and methanogens are associated with constipation and the absence of excessive CH₄ allows the determination that higher H₂S producers mean looser stool. More importantly, identifying the organisms responsible for these gases, their precise gut locations, and their consequences for human health will help develop a new generation of approaches for treating SIBO, IMO and sulfide overproduction.

Supplementary Information

Below is the link to the electronic supplementary material.

Author Contributions

Conceptualization: MP; Resources: AH, MR, and DB; Investigation: MJVM, GL, CMF, JG, WM, MS, IR, GP, SW, and GB; Formal Analysis: MJVM and MP; Project Administration: MP and RM; Supervision: WM, SW, RM, and MP; Writing – Original Draft: MJVM, GL, GB, and MP; Writing – Review & Editing: MJVM, GL, RM, AR, CMF, JG, WM, MS, IR, GP, SW, MR, AH, DB, GM, and MP.

Funding

Open access funding provided by SCELC, Statewide California Electronic Library Consortium. This study was funded in part by funds from the Scott Gray Foundation, Justine Stamen Arrillaga & John Arrillaga, Arrillaga Foundation, the Steve Perry Foundation, the Eli Gottesdiener Foundation, the Nima Taghavi Foundation, the DiCecco Family Foundation, and the National Philanthropic Trust.

Data Availability

It is publishing industry standard that 16S rRNA gene sequencing data be available on request. However, it is also understood that shotgun sequencing poses a risk for deidentification due to the presence of human genetics in the sequencing. In general, availability of shotgun data is not expected.

Declarations

Conflict of interest

Competing Interests: M.P. is a consultant for Ferring Pharmaceuticals Inc., Salvo Health, Dieta Health, Cylinder Health Inc., and Vivante Health Inc. M.P. has received grant support from Bausch Health and Synthetic Biologics. R.M. has received grant support from Bausch Health. A.R. is a consultant/speaker for and has received grant support from Bausch Health. Cedars-Sinai has a licensing agreement with Gemelli Biotech and Hobbs Medical. A.R., M.P., and R.M. have equity in Gemelli Biotech and GoodLFE. M. P. has equity in Cylinder Health and Salvo Health. The remaining authors report no conflicts of interest.

Ethical approval

All procedures performed in studies involving human participants were in accordance with the ethical standards of the Institutional Review Board and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. The REIMAGINE study protocol was approved by the Cedars-Sinai Institutional Review Board (Pro00035192), and all participants provided informed written consent prior to participation.