Models that rely on abiotic factors have increased uncertainty in high methane-emitting wetlands
Estimates of wetland contributions to the global methane (CH4) budget often rely on ecosystem-scale models, which do not represent soil microbial metabolism, but instead use abiotic variables (such as mean annual air temperature or water content) to approximate environmental states conducive for soil carbon decomposition, methanogenesis, and methanotrophy20. A robust meta-analysis from 42 freshwater wetlands showed that air temperature partially accounted for mean annual CH4 fluxes, explaining 51% of the variance across sites25. This discrepancy between CH4 flux predictions and observations for many wetlands hints at a potential role for microbial contributions in explaining these variations, a feature we sought to examine in more detail in this study.
To understand unifying microbial features across wetlands and how microbial and geochemical properties relate to CH4 flux, we conducted a meta-analysis using data from both published and unpublished wetland soil samples. To qualify for inclusion in our study, sites had to have amplicon sequencing data from at least 12 samples obtained from a minimum of 2 sampling depths and have CH4 flux measurements. From the original 42 wetlands25 in the noted earlier study, we identified 16S rRNA gene amplicon microbial data for three of the sites (OWC, TW1, LA2), of which the amplicon data from LA2 is newly released in this study while OWC and TWI utilize previously published data10,27. We also expanded the dataset to include CH4 flux, 16S rRNA gene amplicon, and temperature data from an additional 6 freshwater wetland sites (JLA, PPR7, PPR8, STM-fen, STM-bog, SPRUCE) (Supplemental Data 1). The incorporation of these additional sites reduced the predictive power of mean annual air temperature to explain 37% of the variability across sites (Fig. 1A). Notably, the addition of sites with the highest CH4 fluxes (PPR8, PPR7) (Fig. 1A & 1D) reveals the limitations of mean annual air temperature as a predictor of CH4 flux in high emitting wetlands, such as Old Woman Creek (OWC) and those within the Prairie Pothole Regional complex (PPR).
We collated and analyzed microbial data from 1,112 samples (10% is newly released in this study) from 9 wetlands to demonstrate how incorporating knowledge of CH4-cycling microorganisms can contribute to improved predictive understanding of these ecosystems (Supplementary Tables 1 & 2). Included data was derived from 5 marshes: Old Woman Creek (OWC), Prairie Potholes Region (PPR 7, PPR 8), AmeriFlux site US-LA2 (LA2), and AmeriFlux, site-ID US-Twt (TWI); 1 swamp: Jean Lafitte National Historical Park and Preserve (JLA); 2 bogs: Marcell Experimental Forest (SPRUCE) and Stordalen Mire (STM-bog); and 1 fen: Stordalen Mire (STM-fen). To account for inter-study variability in depth fractions, we binned these samples into three categories: shallow (0–9 cm), mid (10–19 cm), and deep (20–39 cm) (Fig. 1C).
Additionally, we supplemented these data with genomic information creating a cross-wetland genomic catalog, Multi-omics for Understanding Climate Change (MUCC) v2.0.0 database. Here we expanded the original MUCC v1.0.0 genomic catalog, which was composed of 42 metagenome and 133 metatranscriptome samples obtained from a single, high CH4 emitting marsh (OWC) (Fig. 1A)10. The 2,507 medium and high-quality MAGs recovered from this wetland sampling were combined with 1,529 additional MAGS from previously published palsa, bog, and fen metagenomes from a permafrost thaw gradient at Stordalen mire (STM, Fig. 1A)9. Additionally, we added 50 publicly available MAGs derived from the PPR complex28 and 43 publicly available MAGs from TWI27. Finally, we included 20 new metagenomes from the PPR complex, LA2, and JLA (349 Gbp of new sequencing), resulting in 617 MAGs released as new data as part of this study. In total MUCC 2.0 contains 3,634 high and medium quality, dereplicated (99% genome identity) MAGs derived from six wetland complexes totaling 8.9 Tb of sequence data (Supplementary Table 3). MUCCv2.0.0 compiles previous wetland genomic datasets and expands genome representation across wetland soils spanning diverse geographies, ultimately increasing database read recruitment and reducing the computational requirements for translating reads to functional content. This wetland specific genomic resource database was used to connect microbial community profiles with functional potential.
High CH-emitting wetlands share microbial community composition and structure
Analyses across wetland sites revealed that wetland type, not geographical location, corresponded to microbial community composition and diversity. As might be expected by ecological wetland differences, bog samples derived from Sweden (STM) and Minnesota (SPRUCE), were more alike one another than bog and fen samples collected within the same wetland complex (STM). Wetlands categorized as marshes or swamps had higher bacterial and archaeal alpha diversity, higher pH, and higher CH4 flux than bog and fen sites (Supplementary Fig. 1). Additionally, wetland type had a significant impact on community composition, and separation of communities was linked to pH (Fig. 2A & Supplementary Fig. 2, PERMANOVA, p < 0.001). Notably, communities in bogs with the lowest pH and CH4 flux were most distinct from marsh/swamp communities with the highest pH and CH4 flux. Fens, with intermediate characteristics of bogs and marshes/swamps such as pH, vegetation, and nutrient levels, hosted microbial communities that were similarly intermediate of the bog and marsh communities29.
CH4 flux was loosely correlated with temperature across wetland types but this trend was absent at the level of individual wetland types. In marshes and swamps – the highest CH4 emitting wetland types – no correlation to temperature was observed (R2 = 0.17, p = 0.16) (Supplementary Fig. 3A), suggesting that other factors may be important for predicting CH4 flux3,30. We next assessed the relationships between CH4 flux and CH4-cycling microbial community members including methanogens and methanotrophs across sites. Bog and marsh sites hosted different methanogen communities (Supplementary Fig. 4), with bog sites characterized by dominance of a few methanogens and low relative abundances of acetoclastic methanogens 3,31–33. For example, Methanothrix, an obligate acetoclastic methanogen was significantly more enriched in fen, marsh, and swamp samples than in bog samples. Overall, marsh and swamp sites contained a higher diversity and evenness of methanogen taxa and functional types. Collectively, the functional potential to utilize more diverse methanogenic substrates in high CH4 emitting marsh sites could contribute to higher CH4 fluxes.
To fully understand microbial contributions to the methane cycle, we also assessed the distribution of methanotroph communities across wetland types. Across all sites aerobic methanotrophs were dominant, while the anaerobic methanotrophs assigned to the genus Methanoperedens were enriched only in the three highest methane emitting sites (OWC, PP7, PP8) (Supplementary Fig. 4). We found that the diversity of methanogens (R2 = 0.5, p = 0.034), but not methanotrophs (R2 = 0.22, p = 0.2), was significantly correlated to CH4 flux (Supplementary Fig. 3B). Additionally, the ratio of methanogen to methanotroph relative abundances was correlated to flux (R2 = 0.45, p = 0.047) (Supplementary Fig. 3C), but the relative abundance of methanogens and methanotrophs alone was not. This suggests the coupling of methanogens and methanotrophs act as a control over CH4 flux in wetland environments, highlighting how the balance between these microbial groups likely influences net methane emissions.
Identification of a widespread, core group of CH4 cycling organisms
Given more consistent sampling methodology (i.e., similar sequencing protocols), as well as the higher measured CH4 fluxes, we focused on understanding trends in microbial dynamics across 5 marsh and swamp sites (JLA, LA2, OWC, PPR7, and PPR8) (see methods). We first assessed occupancy patterns across sites to identify if there were core methanogens and methanotrophs for these marsh samples, identifying five methanogens and four methanotrophs detected in at least one sample from each site34 (Fig. 2B). Despite wetland differences in site, depth, and time of year sampling (Fig. 1), five core methanogen genera were found in a majority of samples: Methanothrix (79.7%), Fen 33 (order Methanomassiliicoccales) (72.6%), Methanobacterium B (50.9%), Methanolinea (55.5%), and Methanoregula (93.9%). Interestingly, each methanogenic pathway (hydrogenotrophic, acetoclastic, methylotrophic methanogenesis) was represented within the core community, indicating that all three pathways are consistently important and likely utilized for wetland CH4 production in high emitting marsh and swamp ecosystems (Fig. 2B). Three methanotrophs were identified as core but were found in a lower percentage of samples: Methylomonas (60.3%), Methylobacter (39.8%), and Methylomonadaceae KS41 (85.4%). However, because the core methanotrophs require oxygen for methane oxidation, these methanotrophs may not be as detectable in the deeper anoxic samples sampled here. Constraining our analyses to only the top 10 centimeters of sediment where oxygen might be more available, we found Methylomonas present in 75.1%, Methylobacter in 57.1%, and KS41 in 95.2% of samples. Core microbiomes have become increasingly viewed as important because of their assumed role as critical to a given ecosystems’ functioning35,36. Collectively, these discoveries underscore the pivotal role of select organisms in actively shaping the methane cycle within freshwater marsh ecosystems. These insights carry implications for forthcoming research activities, highlighting these organisms as candidates for more thorough physiological validation and study, as well as focus organisms for scaling to modeling endeavors.
MUCC database enables deeper insight into trophic patterns from co-occurrence networks
For each of the 5 marsh sites, we performed network analysis based on co-occurrence patterns to help unravel possible microbial interactions within these complex, methanogen-oriented communities. We hypothesized that methanogen network structure in wetland communities would act as a predictor of CH4 flux. To test this hypothesis, we built 16S rRNA gene positive co-occurrence networks at each site using both the community-wide amplicon data and only the methanogen community data (Supplementary Fig. 5).
Although network structure of the entire community did not relate to CH4 flux (Fig. 3K), a more constrained network comprising the significant co-occurrences that included a methanogen member did uncover important trends (Fig. 3L). These networks revealed a negative correlation between the number of methanogen-related network nodes and CH4 flux, indicating a relationship between less complex methanogen networks and higher annual CH4 emissions. Furthermore, the number of methanotrophs associated with methanogens in these networks was greater in the lower methane emitting sites (JLA, LA2), indicating that lower CH4 fluxes are associated with communities where methanotrophs and methanogens co-occur. In contrast, while high CH4-emitting sites (OWC, PPR7, PPR8) host methanotrophs and methanogens, they were generally linked by fewer connections (Fig. 3M). Methanotrophs can act as a filter, oxidizing anywhere from 20–60% of the CH4 before it is released into the atmosphere3,37,38 and these results indicate that their absence in wetland samples where methanogens are present could contribute to greater CH4 fluxes.
To determine potential metabolic interactions that underpin CH4 production across these sites, we developed metabolic profiles for methanogen-connected taxa in our 16S rRNA gene networks. Utilizing the MUCC 2.0.0 database, we linked microbes present in the networks with MAG representatives and assigned them functional categories: obligate fermenter, homoacetogen, demethylating, or none of these three criteria (Fig. 3A-E, 4 & Supplementary Table 4). We selected these criteria, as they are thought to cross feed methanogens (Fig. 1B) and are traits that can be inferred from genomes clearly. Methanogen networks were composed of 699 unique co-associated genera, of which 131 genera had a genome representative in the MUCC database (Fig. 4). Summarizing these genome representatives within the methanogen networks, 12 were categorized as methanogens, 7 as methanotrophs, 23 as obligate fermenters, 8 as homoacetogens, 1 as both obligate fermenter and homoacetogen, and 75 demethylating (methyl-x), and 4 did not meet these criteria (Rules for assignment are found in Supplementary Table 4). Additionally, 6 methanogens and 10 methanotrophs identified based on 16S rRNA gene taxonomy alone (no matches to MUCC, but metabolism is defined in literature) were included in the networks (Fig. 4, Supplementary 5).
Specifically, obligate fermenters have the potential to produce acetate, formate, and H2, which we hypothesized would directly promote methanogen activity39,40 and thus be positively associated with our methanogen networks. As we expected, obligate fermenters were highly connected to hydrogenotrophic and acetoclastic methanogens, likely supporting cross feeding. In total, obligate fermenters had 99 significant interactions with methanogens of which 73% were to hydrogenotrophic or acetoclastic methanogens (Fig. 3F-J). Additionally, obligate fermenters were found to highly co-occur with certain methylotrophic methanogens such as Methanofastidiosum, which requires H2 to reduce methylated thiol to form methane. Compared to hydrogenotrophic methanogenesis, this form of methanogenesis is more thermodynamically favorable under low H2 conditions and has been proposed to support H2 producing syntrophs and fermenters by preventing accumulation of H212. In summary, anoxic carbon exchanges between obligate fermenters and methanogens appear vital to carbon cycling in wetlands.
Syntrophy denotes a symbiotic interaction among diverse microorganisms, wherein the exchange of metabolic byproducts mutually supports each organism's metabolism. This phenomenon is particularly prominent in methanogenic environments, where methanogens play a crucial role in regulating product concentrations, thereby rendering otherwise endergonic processes thermodynamically favorable41,42. In our study, we investigated obligate fermenters to uncover evidence of secondary fermentative syntrophs, identifying two prevalent syntrophic genera across methanogen networks: Smithella, present in four marshes except PPR8, and Syntrophorhabdus, found across all five marsh networks. Previous research has demonstrated the capacity for acetate and hydrogen production by Syntrophorhabdus43, aligning with our genome-based characterization of these 7 MAGs in MUCC. Notably, in our networks, Syntrophorhabdus exhibited multiple (8) connections to hydrogenotrophs and acetoclasts, further emphasizing its role in metabolic exchanges. These genomic metabolic insights highlight the intricate connections harbored within these co-association networks, exchanges essential for maintaining metabolic efficiency in methanogenic environments.
Homoacetogens are also interacting with methanogens, as these microorganisms grow on H2/CO2/CO and produce acetate as the main metabolic product. We hypothesized that these organisms could cross-feed acetoclastic methanogens15 and or could compete with hydrogenotrophic methanogens for substrates44. The 9 homoacetogen MAGs identified in the methanogen networks comprised 15 nodes and were closely related across sites, belonging to two main phyla, Desulfobacterota and Chloroflexota despite many other acetogens across other phyla existing in the MUCC database. We observed 32 associations between these acetogens and methanogens, with 50% to hydrogenotrophic, 28% to acetoclastic and 22% to methylotrophic methanogens. Additionally, 6 of the 8 acetoclastic methanogens had at least one connection to an acetogen, supporting our hypothesis that acetogens were cross-feeding methanogens. While our finding does not preclude competition between hydrogenotrophs and other acetogens, these identified positive associations may reflect sufficient hydrogen production within the soil profile to support co-existence of both guilds, or the separation of guilds across microsites.
Finally, demethylating microorganisms, whether bacteria or archaea, are capable of removing methyl groups from oxygen, sulfur, and nitrogen (O, S, N) containing compounds. Unlike methylotrophic methanogens, these taxa do not produce methane directly; however, they may engage in cross-feeding or competition dynamics with methylotrophic methanogens. Depending on the enzymatic systems they encode, these microorganisms can lead to several outcomes: (i) production of trimethylamine (TMA), a substrate for certain methanogens; (ii) formation of quaternary amines (QA), which can could be utilized by select methylotrophic methanogens; or (iii) direct utilization of methylated O, N, or S compounds, which may (iiia) compete with methylotrophic methanogens or (iiib) generate acetate and hydrogen to support hydrogenotrophic or acetoclastic methanogens. The methyl-metabolism category exhibited substantial connectivity with methanogens, comprising nearly half of the connections across sites. Notably, 68% of these connections (comprised mostly of type iii demethylating microorganisms) were linked to acetoclastic and hydrogenotrophic methanogens not methylotrophs suggesting that demethylating metabolisms in soils could indirectly bolster non-methylotrophic methane production. These findings underscore the complexity of microbial interactions beyond methane production and oxidation, thereby contributing to a more comprehensive understanding of microbial cross-feeding and its broader implications for methane emissions.
Methanoregula is critical for CH4 production in wetlands
Two core methanogens (Fig. 2), Methanothrix and Methanoregula, were found in networks across every marsh indicating global importance in the wetland CH4 cycle. Methanothrix is an obligate acetoclastic methanogen already shown to be globally distributed and an important contributor to CH4 emissions in wetlands16. Methanoregula has been found in wetlands and other habitats around the world, and like at many of our sites, is a prominent member of methanogenic networks and consistently a dominant methanogen45,46. We found that its dominance (proportion of methanogens that are Methanoregula) was related to CH4 flux, such that percent of methanogens that are Methanoregula significantly correlated to CH4 flux and the residual values that were not well predicted from the temperature- CH4 flux correlation in Fig. 1 (Fig. 5A). Additionally, we tested how well temperature, Methanoregula dominance, and the two combined explained methane flux. When looking at the 9 study sites, CH4 flux was not predicted by temperature alone (R2 = 0.15, p = 0.30,), was predicted by Methanoregula dominance (R2 = 0.54, p = 0.02,), but that temperature combined with Methanoregula dominance was the best predictor (R2 = 0.84, p = 0.02). This is one example of how incorporating biological insights with already existing abiotic data could improve the predictive power of climate models.
To understand potential physiological drivers that link Methanoregula and predications of CH4 flux, we conducted a genomic analysis of 107 dereplicated MUCC-derived and publicly available (i.e., GTDB, JGI) MAGs. Methanoregula encoded diverse metabolic strategies, the capacity for fixing nitrogen (nitrogenase), viral defense (CRISPR-Cas), and mechanisms to respond to fluctuating redox conditions (reactive oxygen species) (Fig. 5B). Methanoregula are classically designated hydrogenotrophic47, which we broadly confirmed here (Fig. 5B). We also report that some Methanoregula genomes encode genes for methylotrophic methanogenesis, specifically for the demethylation of methylated sulfides48 and methoxylated19 compounds, compounds prevalent in wetlands10,15. Although hydrogenotrophic methanogenesis is generally recognized as the dominant CH4 -generating pathway in wetlands, recent studies have indicated that methylotrophic methanogenesis contributes more to CH4 flux than previously realized17,21,22,49. Therefore, the apparent significance of Methanoregula in contributing to CH4 emissions across diverse wetlands and within wetland gradients could partly be explained by a broader than previously understood ecological niche.
To investigate the role of Methanoregula within a high CH4 emitting wetland, we mined a previously undefined role for Methanoregula from 39 paired metatranscriptome and metabolome datasets across spatial and temporal gradients from a single mudflat at OWC10 (Supplementary Fig. 6A). At this mud-type site, a Methanoregula MAG (OWC-0053) was one of the transcriptionally most active methanogens throughout the entire soil column across 3 months of peak CH4 production (Fig. 5C). This genome was also one of the 9 genomes that predicted 78% of soil porewater CH4 concentration (Supplementary Fig. 6B). In summary, our comprehensive analysis reveals Methanoregula's substantial contribution to CH4 dynamics within a high-emission wetland, highlighting its prominent role as a key player in CH4 production across spatial and temporal scales.
These findings help in part explain the significant correlation between Methanoregula abundance and CH4 flux across wetlands, and its role in marsh CH4 networks. Our results suggest that Methanoregula may possess a broader physiological capacity to produce CH4 and adapt to various abiotic and biotic constraints present in marsh soils. By shedding light on the functional significance of Methanoregula, a core taxon across wetlands, our study contributes to advancing our understanding of wetland CH4 emissions. Our findings use a cross-site analysis to identify core lineages, like Methanoregula, warranting further physiological exploration, as the metabolic assumptions may be constrained by prior strict substrate and redox capabilities. Ultimately our results show promise for biological knowledge to enhance predictive models of wetland emissions, ultimately facilitating more effective management and mitigation strategies.