Introduction

Most CO-utilizing microbes are facultative lithotrophs that are capable of using organic substrates as carbon and energy sources [1,2,3,4,5]; therefore, the term carboxydotroph used here and in other studies defines all the microbes capable of using CO as energy source through carbon monoxide dehydrogenase (CODH) [4,5,6]. As per cofacter contents, CODHs can be classified into aerobic (Mo-containing) and anaerobic (Ni-containing) types. CODHs are present in physiologically and phylogenetically diverse bacteria and archaea [2, 7], in which CO oxidation can be coupled to oxygen respiration, proton respiration, sulfate respiration, acetogenesis, and methanogenesis [8]. Carboxydotrophs play essential roles in global carbon flux such as holding the concentration of atmospheric CO below toxic level [9,10,11,12], and carboxydotrophs are of potential biotechnological interest for synthesis gas (syngas) fermentation as a biorefinery factory [13, 14]. For both fossil-based waste and biogenic but biologically recalcitrantly degradable waste, coupling syngas production to biorefinery is a promising approach to produce value-added biochemicals and to improve bioenergy recovery from waste [15,16,17]. Additionally, this combined process can overcome the inherent limitations of chemical catalyzers in syngas conversion [18, 19]. CO is a main component of syngas, and the potential of CO/CO2 couple (E0′ = −520 mV) theoretically is more favorable for higher energy conservation than hydrogen oxidation (E0′ = −414 mV) [20]. Unfortunately, despite the low reduction potential of CO, the microbial tolerance and utilization for CO is a bottleneck for CO bio-utilization. Thus, insights into carboxydotrophs and their interactions with other microbes will yield useful information about CO bioconversion processes.

Previous studies about the carboxydotrophs in CO bioconversion were primarily concerned with model or isolated microbes in pure culture [21, 22]. Briefly, mesophilic carboxydotrophs can exploit CO to predominantly form chemicals such as carboxylates and alcohols [23, 24]. And various studies have determined the physiological characteristics of carboxydotrophic acetogens; for example, Clostridium ljungdahlii is a well-studied model to produce acetate and ethanol through Wood–Ljungdahl pathway [13]. Methanogens are the best-studied carboxydotrophic archaea [4]. Four methanogens have been demonstrated to grow on CO as a sole substrate [20], and acetate formation was observed during the carboxydotrophic growth of Methanosarcina acetivorans C2A and Methanothermobacter marburgensis [21, 25]. However, there is still a tremendous number of under-characterized taxa with carboxydotrophic potential in artificial and natural environments [26,27,28]. For instance, few mesophilic carboxydotrophs perform thermodynamically favorable H2 production, where CO is firstly oxidized by CODH and the derived electrons are used to reduce protons via an energy-converting hydrogenase [29, 30]. Moreover, the anaerobic CO oxidation in Proteobacteria has not been further investigated [30]. These indicate that carboxydotrophs may be phylogenetically more diverse than expected, and a requirement arises for broadening the phylogenetic scope of carboxydotrophs. Until now, most studies about mixed microbial cultures mainly explored the effects of environmental factors on product spectrum and overall microbial community structures using 16S rRNA gene sequencing [31,32,33,34]. Few studies explicitly explored the under-characterized taxa and the genome-centric interactions in the mixed microbial cultures fed with CO, as well as the functional potential in these cultures. Therefore, the genome reconstruction and annotation of microbes, e.g., the non-carboxydotrophs which utilize the CO-derived metabolites of carboxydotrophs (that is, CO2, acetate and others), is necessary to uncover the black-box microbial ecology of a CO-driven microbiome.

Metagenomics can profile the functional potential of microbial communities and reconstruct the genomes of under-characterized taxa [35]. Despite its success, some limitations to this technique also exist. Metagenomic sequencing of bulk DNA extracted directly from environmental samples may overlook the potentially essential roles of rare species [36]. The increased complexity of microbial communities dramatically hinders metagenomic assembly, leading to the limited reconstruction of high-quality genomes [37, 38]. Furthermore, metagenomics does not allow to confirm the potential functions of under-characterized microbes. However, stable isotope probing (SIP) can help to confirm the connection between functions and microbes [39, 40], and DNA-SIP has an enormous advantage in enriching the DNA of functional microbes through a physical filtering to improve metagenomic assembly [41]. DNA-SIP based metagenomics has been applied to environmental samples [42,43,44,45]; thus, this method should be feasible to explore CO-driven microbiomes.

To address the above knowledge gaps, DNA-SIP based metagenomics was applied in this work to explore the CO-driven microbiomes under different initial CO pressures. The dynamic succession of functional microbiotas was described through the time-series analysis of the labeled 16S rRNA genes. Functional enrichment analysis unraveled the enriched functional potential of the microbiomes under high CO pressure. Additionally, genome-centric analysis reconstructed the metagenome-assembled genomes (MAGs) of under-characterized taxa. Based on the metabolic reconstruction of the MAGs, the ecological role of each species in the conversion of CO to end-products was identified, and the putative microbial interactions in the mixed microbial cultures were established. This in-depth and precise characterization of the CO-driven microbiomes is expected to contribute in the understanding of biodiversities and interactions within mixed microbial cultures, as well as the operating conditions of CO bio-utilization.

Materials and methods

Methods are described in detail in the Supplementary Methods.

Anaerobic incubation with 13CO

Inoculum that had acclimatized to a CO atmosphere of 0.6 atm was sampled from a mesophilic batch reactor at 35 °C with pH ranging from 6.8 to 7.1. Microcosms were set up by using 570 ml serum bottles filled with 150 ml basal medium as in our previous study [46], and the initial pH was adjusted to 7.0. The inoculum was seeded at a concentration of 1 g/l volatile solid to enhance isotope labelling. Each bottle was sealed with a screw cap and a rubber septum, and the headspace was vacuumized and purged with high purity N2 (99.999%) three times before injecting CO. Two experiments were carried out and denoted as “high PCO” and “low PCO”, where the initial CO pressures were 0.95 and 0.35 atm, respectively. For each CO pressure, there were two groups of microcosms. The first group of triplicate microcosms contained 13CO (99 atom% in 13C, Cambridge Isotope Laboratory, USA). The second group of triplicate microcosms was fed with 12CO. The total pressure in each microcosm was 1 atm using complementary N2. All microcosms were maintained at 35 °C and pH 6.0–7.0 using 1 M HCl or 1 M NaOH solution, and were stirred at 125 rpm in a shaker. The negative pressure in serum bottles owing to CO consumption was adjusted to 1 atm by injecting high purity N2 (99.999%). Every 4 days, the headspace of each microcosm was vacuumized and readjusted to the initial CO pressures. Gas pressure, gas composition, pH, and carboxylate (C2–C7, referring to total dissociated and undissociated carboxylic acid) concentration before and after vacuuming were measured, and the incorporation of 13C and 12C into carboxylates was monitored.

DNA extraction, isopycnic centrifugation, and fractionation

Sludge fed with 12CO and 13CO was collected for DNA extraction after 12, 24, and 36 days of incubation (labeled the 12th, 24th, and 36th days, respectively). DNA was extracted and subjected to isopycnic centrifugation and fractionation. CsCl density gradient ultracentrifugation was performed at 51,800 rpm (about 194,991 × g) and 20 °C for 40 h in biological and technological triplicate. After centrifugation, 24 gradient fractions of 200 μl were recovered, and the buoyant density and DNA concentration of each fraction were measured. Most DNA from the sludge fed with 12CO was detected at buoyant densities of 1.68–1.72 g/ml, and the DNA concentration was under detection limit at densities higher than 1.72 g/ml. The DNA from 13C-fed sludge showed an increase in densities of 1.73–1.76 g/ml, regardless of CO pressure (Fig. S1). Therefore, in the sludge fed with 13CO, the fractions with densities of 1.73–1.76 g/ml and the remaining fractions with densities <1.73 g/ml were pooled and designated “heavy fractions” and “light fractions”, respectively. The fractions with densities of about 1.68–1.72 g/ml were designated “control fractions” when 12CO was fed as substrate.

Amplicon sequencing and analysis

The DNA in the heavy, light, and control fractions from the triplicate samples of each condition was used for amplicon sequencing, where the samples at three incubation time were considered (Dataset S1). Amplification of V4 region was carried out using primer F515 (5′-GTGCCAGCMGCCGCGGTAA-3′) and R806 (5′-GGACTACVSGGGTATCTAAT-3′) [47]. Briefly, the PCR program was performed: 95 °C for 3 min, then 27 cycles of 95 °C for 30 s, 55 °C for 30 s, and 72 °C for 45 s. A final extension at 72 °C was kept for 10 min and the products were held at 4 °C. The purified PCR products were pooled in equimolar and paired-end sequenced (2 × 300 bp) on a MiSeq platform (Illumina, San Diego, USA). Raw amplicon reads were denoised and dereplicated into amplicon sequence variants (ASVs) using DADA2 v.1.10 [48] before clustering into 97% operational taxonomic units (OTUs) with VSEARCH v.2.7.0 [49]. Taxonomic classification of each ASV or OTU was performed using the VSEARCH-based classifier in QIIME 2 v.2019.4 [50] against the SILVA 16S rRNA gene database (132 release). Diversity measurements were carried out using Vegan v.2.5-3 [51]. Microbial networks were constructed using the OTUs in the heavy fractions under high CO and low CO pressure through MENAP [52], respectively.

Metagenomic sequencing and analysis

The DNA in the heavy fractions of the triplicate samples for the same condition was used for metagenomic sequencing, where the samples were collected at the 12th and 24th days. Briefly, sequencing libraries were prepared using NEBNext® UltraTM DNA Library Prep Kit for Illumina (New England Biolabs, USA). The libraries were analyzed for fragment size distribution by Agilent2100 Bioanalyzer (Agilent Technologies) and were paired-end sequenced (2 × 150 bp) on Illumina platforms according to the standard protocols. Raw metagenomic reads were trimmed with fastp v.0.20.0 to obtain clean reads [53], which were then corrected and assembled into 12 single assemblies and four co-assemblies using metaSPAdes v.3.13.1 (Table S1) [54]. For each assembly, three bin sets were generated using MaxBin2 v.2.2.6 [55], MetaBat2 v.2.12.1 [56], and CONCOCT v.1.1.0 [57]. All the bins in the three bin sets were checked using CheckM v.1.0.13 [58]. All bins of the 16 assemblies were dereplicated to produce MAGs using dRep v.2.6.2 (Table S1) [59]. The abundances of the MAGs were quantified across 12 samples using metaWRAP v.1.2 [60]. Genome Taxonomy Database(GTDB)-Tk v.1.3.0 was used to identify the taxonomic classifications of the MAGs based on the GTDB (05-RS95 release) [61, 62]. The terms discovered and under-characterized defined the MAGs with and without published genomes, respectively. The coding sequences of the MAGs were predicted and annotated using Prodigal v.2.6.3 and eggNOG-mapper v.2.0.1, respectively [63, 64]. In functional enrichment analysis, the genes were annotated and quantified across 12 samples using HUMAnN2 v.2.8.2 [65]. Then, the genes were clustered into functional categories based on the KEGG Orthogroups (KOs). The principal coordinates analysis using the relative abundances of the KOs showed that the 1st biological triplicate under low CO pressure were outliers (Fig. S2); therefore, for increasing sensitivity, only the 2nd and 3rd biological triplicate under low CO pressure were used to make the differential comparison in the relative abundances of a specific KO between high CO and low CO pressure at the same incubation time. The KOs with significantly differential abundances (p < 0.05, Welch’s t test) between high CO and low CO pressure were used to perform functional enrichment analysis based on the KEGG pathway and module.

Results and discussion

Conversion of CO under different CO pressures

There was no significant difference in the cumulative carbon atom quantity of each carbonaceous composition between the microcosms fed with 13CO and 12CO, regardless of CO pressure (p > 0.05, Wilcoxon test; Fig. 1A and Table S2). This result indicated that when compared with the microcosms fed with 12CO, the addition of 13CO did not affect the microbial utilization of CO or the microbial community structures. Therefore, 13CO just labeled the microbes that participated in the utilization of CO or CO-derived metabolites, and the above confirmation of the fractions based on buoyant densities was reliable. Approximately 98 and 41 mmol CO were totally consumed under high CO and low CO pressure, respectively. More than 90 C-atom% of the consumed CO was converted to carboxylates, CH4 and CO2 throughout the experiments (Fig. S3), where CO2 accounted for more than 50 C-atom% regardless of CO pressure, indicating the active enzymatic reaction of CODH. The conversion ratio of CO to CH4 exceeded 1 C-atom% on the 20th day under high CO pressure, and the ratio increased up to ~5 C-atom%. However, CH4 accounted for more than 1 C-atom% of the consumed CO under low CO pressure from the 12th day, and the ratio rapidly increased up to 16 C-atom% (Fig. S3). These results agreed well with the view that methanogens are less resistant to high CO pressure [26, 66]. Under low CO pressure, acetate, propionate, and butyrate contributed to most of the carboxylates, and almost all kinds of the carboxylates began to decrease from the 24th day on (Figs. S4 and S5), indicating that other microbes might be involved in the degradation of carboxylates in addition to methanogens, which cannot use substrates such as butyrate. Approximately 35 C-atom% of the consumed CO was concentrated in carboxylates under high CO pressure (Fig. S3), and the ratio of acetate was less than that under low CO pressure (Fig. S4); therefore, more carbon flowed into longer carbon chain carboxylates, such as butyrate and heptylate. These results showed that high CO pressure might be conducive to the synthesis of longer carbon chain carboxylates, which reduced the carbon flow into CH4. In the microcosms fed with 13CO, the median ratios of fully 13C-labeled carboxylate fragments were almost all above 90% (Fig. S6), indicating the active uptake and turnover of the added 13CO. Previous studies on mixed microbial cultures showed that acetate was a predominant carboxylate, and CH4 was only produced under low CO pressure [6, 26, 31, 67, 68]. Butyrate might be the carboxylate with the longest carbon chain in the previous studies. Although a previous study achieved longer carbon chain carboxylate production, 2-bro-moethanesulfonate was added to inhibit methanogenesis [69]. Valerate, caproate, and heptylate were produced in the present work, suggesting the active carbon chain elongation in the microbial cultures and demonstrating the feasibility of medium-chain carboxylate production without adding methanogenesis inhibitors. Additionally, increased CO tolerance threshold of methanogenesis in this study might be due to the methanogens’ adaptation to CO after a long-term exposure.

Fig. 1: The properties of the microbiotas between different CO pressures or incubation time.
figure 1

A Cumulative carbon atom quantities of the carbonaceous compositions in the microcosms. The blue bars and its left red bars indicated the carbonaceous compositions in the microcosms fed with 12CO and 13CO at the same incubation time, respectively. The value of each bar represented an average value of biological triplicate (n = 3), and the error bars donated the standard deviation. The error bars of methane were shown when its cumulative carbon atom quantities were more than 0.1 mmol. B The β-diversity measurement using unweighted UniFrac distance (UUD). All the heavy fractions, light fractions, and control fractions under high CO and low CO pressure were considered. C The β-diversity measurement using unweighted UniFrac distance (UUD). Only the heavy fractions under high CO and low CO pressure were considered. D The α-diversity measurements in all the heavy fractions and light fractions using the number of the observed individuals and Shannon index based on ASVs. The densities were calculated from the point data mapped onto the x-axis (n = 3), and the higher ridges, the more concentrated the values of biological triplicate were (Color figure online).

Full size image

Core microbes under different 13CO pressures

Amplicon sequencing of the 16S rRNA genes was performed to determine the composition and diversity of each microbiota. An average read count per sample of 48,855 was obtained, which covered the majority of the microbiotas (Dataset S1 and Fig. S7). All microbiotas in the principle coordinates analysis (PCoA) using unweighted UniFrac distance (UUD) could be grouped significantly by the fractions, as could the PCoAs using other distances (Figs. 1B and S8, p = 0.001, Dataset S2). Partial canonical analysis of principal coordinates also revealed the significant effect of the fractions on the microbiotas (Fig. S9, p = 0.001). These findings suggested that isotope labeling isolated the core microbes and cross-feeding was minor. CO pressure significantly contributed to the differences in the microbiotas and explained 14.4%, 25.4%, 15.7%, and 26.3% of the variation when using UUD, Bray–Curtis (BCD), Jaccard (JCD), and weighted UniFrac (WUD) distances, respectively (Figs. 1C and S10, p < 0.05, Dataset S3 and Dataset S4). BCD and WUD take the microbial abundances into consideration, whereas JCD and UUD do not; thus, the higher explanation ratios of CO pressure using BCD and WUD indicated that CO pressure might cause relatively large shifts in microbial abundances. Another factor significantly shaping the microbiotas was incubation time (Figs. 1C and S10A, B, p < 0.05, Dataset S3 and Dataset S4). Furthermore, the interaction between CO pressure and incubation time significantly comprised 14.0% and 16.9% of the variation based on BCD and WUD, respectively (p < 0.05, Dataset S4), but this interaction had no significant impact on the variation based on JCD and UUD (p > 0.05, Dataset S4), highlighting the changes in the microbial abundances.

The α-diversity was measured using the number of observed individuals and Shannon index, considering the richness and evenness of microbiotas. The median α-diversity was significantly higher in the light fractions than that in the heavy fractions under the same CO pressure or incubation time (Fig. 1D, p < 0.05, Dataset S5). This finding showed that the number of the species which participated in the conversion of CO or CO-derived metabolites might not occupy the majority of the total number of species in the microbiotas, and suggested the existence of the predominant microbes with high abundances in the heavy fractions, resulting in the lower evenness. The observations based on OTUs agreed well with the above results (Dataset S5). The mean α-diversity in the heavy fractions based on ASVs was significantly higher than that based on OTUs (p < 0.001, Wilcoxon test), but the mean α-diversity in the light fractions based on ASVs was similar with that based on OTUs (Fig. S11). This result suggested that the microbiotas in the heavy fractions had higher microdiversity, where one OTU might contain multiple functionally equivalent ASVs, that is, several species consisted of many different subpopulations. Microdiversity is essential for microbial community stability [70], and recent studies have demonstrated that different subpopulations of the same species sustain the distribution of this species across broad environmental gradients [71, 72]. Therefore, the microbes in the heavy fractions might be more resistant to variable CO pressure than those in the light fractions.

Proteobacteria and Firmicutes were significantly enriched in the heavy fractions at the 12th day, regardless of CO pressure (Fig. S12). As the incubation time progressed, the abundance of Proteobacteria showed a significant decrease in the heavy fractions under high CO pressure. However, under low CO pressure, Proteobacteria was always significantly enriched in the heavy fractions where its abundance remained stable (Fig. 2A). The predominant genus in Proteobacteria was Rhodoplanes, which was almost all significantly enriched in the heavy fractions under low CO pressure (Figs. 2B and S13). Complete CODH genes have been found in the isolated Rhodoplanes sp. Z2-YC6860, but the CO metabolism of Rhodoplanes has not been previously investigated in the literature. Under high CO pressure, Firmicutes was always significantly enriched in the heavy fractions where its abundance showed a significant increase, but under low CO pressure, a significant drop in its abundance was noted throughout the incubation (Figs. 2A and S12). Although Clostridium is known for carboxylate synthesis, the prevalent genus of Firmicutes in this work was Acetobacterium (Fig. 2B), of which some species are carboxydotrophic acetogens [22, 23, 73]. Under high CO pressure, Acetobacterium was significantly enriched in the heavy fractions where the abundance of Acetobacterium showed a significant increase (Fig. S13), suggesting that it was a keystone taxon in acetate production. Additionally, Oscillibacter was only detected under high CO pressure (Fig. 2B), which has received limited attention in CO utilization. Methanogenesis was an essential metabolic process in this work and was performed by the methanogens of Euryarchaeota, which was found to be mainly composed of Methanobacterium and Methanosaeta (Fig. 2B). Some previous mesophilic studies showed that acetoclastic methanogenesis was dominant when CO pressure was below 0.5 atm [34, 74], where acetoclastic methanogens should be abundant. However, the present work disclosed that Methanobacterium was the predominant archaeal genus, regardless of CO pressure. The increased abundance of Methanobacterium under both CO pressures implied its wide tolerance range to CO. Despite the high acetate concentration, the acetoclastic methanogen Methanosaeta was almost all significantly enriched in the light fractions (Fig. S13), indicating that cross-feeding might be minor. The truncated power laws of the co-occurrence networks under high CO and low CO pressure were both fitted well (Table S3), and these networks displayed small-world characteristics (Table S4 and Supplementary Notes). The networks also showed the predominant roles of Proteobacteria, Firmicutes, and Euryarchaeota, which totally occupied ~64% and 47% of the nodes under high CO and low CO pressure, respectively (Dataset S6).

Fig. 2: The dynamic succession of predominant taxa.
figure 2

A phylum level. B genus level. The phylum and genus with a maximum relative abundance greater than 9% and 5% were selected, respectively. The label “Others” represented the total relative abundance of remining taxa. The three dots of each taxon represented biological triplicate (n = 3). The labels “High” and “Low” indicated CO pressures, and the labels “heavy” and “light” indicated fractions.

Full size image

Functional potential of the microbiomes under high CO pressure

As far as we know, no previous research has investigated the effects of high CO pressure on functional potential; therefore, the significantly differential KOs between two CO pressures were used to perform functional enrichment analyses (Dataset S7). Overall, 43 and 18 pathways had significantly greater potential under high CO pressure than under low CO pressure at the 12th and the 24th days, respectively (Fig. 3 and Dataset S8). At the 12th day, the pathways relating to fatty acids were significantly enriched under high CO pressure, including carbon fixation in prokaryotes, butanoate metabolism, propanoate metabolism, and fatty acid biosynthesis. Some enzymes of these pathways can participate in Wood–Ljungdahl pathway and reverse β-oxidation (RBO), which are essential for synthesizing acetate and longer carbon chain carboxylates, respectively. These results suggested that the microbiomes under high CO pressure had greater potential in carboxylate production, resulting in the accumulation of carboxylates with 3–7 carbons under high CO pressure. At the 12th and 24th days, the potential in citrate cycle (TCA cycle) was significantly greater under high CO pressure than under low CO pressure. In anaerobic microbes, TCA-related genes have potential in synthesizing essential biosynthetic intermediates such as succinate and 2-oxoglutate [75]. These intermediates can be used in amino acid, pyrimidine, and purine metabolism, and the biosynthesis of amino acids was significantly enriched under high CO pressure. The functional enrichment analyses based on the KEGG module also showed that citrate cycle was significantly enriched under high CO pressure (Fig. S14). These results indicated that the microbes under high CO pressure might exhibit greater potential in maintaining fundamental metabolism than those under low CO pressure.

Fig. 3: The functional enrichment analyses using the significantly increased KOs under high CO pressure.
figure 3

Benjamini and Hochberg method was used to verify the statistical significance. The pathways with KO counts <10 were filtered. KO ratio was calculated by dividing k by n, where k (KO count) was the number of the significantly increased KOs in a specific KEGG pathway and n was the number of the significantly increased KOs in all KEGG pathways.

Full size image

Microbial interactions revealed by genome-centric metagenomics

Twenty-eight MAGs from co-assembly and 20 MAGs from single assembly were reconstructed (completeness >70%, contamination <10%), spanning nine bacterial and three archaeal phyla (Fig. 4 and Dataset S9). More than 94% of the metagenomic reads could be mapped to these MAGs; therefore, it was assumed that these MAGs provided a comprehensive representation of the CO-driven microbiomes. These MAGs were further categorized into six groups (group A–F) based on their functional characteristics.

Fig. 4: The phylogenetic tree of the dereplicated MAGs reconstructed using all assemblies.
figure 4

The bacterial and archaeal MAGs with completeness >70% and contamination <10% were considered. For the label of MAG, “H” and “L” represented this MAG has a higher average abundance under high CO and low CO pressure, respectively. The black labels of the MAGs represented that these MAGs were assigned to discovered species or public genomes, and the red labels of the MAGs represented that these MAGs were under-characterized species. The completeness and contamination of each MAG were shown in the blue and red barplot, respectively. The abundances of each MAG under two CO pressures were shown in the heatmap. The abundances were expressed as genome copies per million reads (CPM), which had been standardized to individual genome size, and these abundances were logarithmically processed as Log2(CPM + 1) for visualization (Color figure online).

Full size image

CO oxidation

As the initial carbon source was CO, carboxydotrophs were essential to fix carbon and provide available carbon sources for other microbes. Carbon monoxide dehydrogenase/acetyl-CoA synthase (CODH/ACS) complex α subunit (acsA) and its homolog anaerobic CODH catalytic subunit (cooS) were used to identify anaerobic carboxydotrophs. In total, six under-characterized MAGs and six discovered MAGs made up group A, where CO oxidation might be performed (Figs. 5 and S15, Dataset S10 and Dataset S11). These carboxydotrophs played an essential role in reducing the inhibition level of CO. MAG H3 was identified as an under-characterized species of phototrophic Rhodoplanes, in which no carboxydotrophic growth is known [76]. The complete genes encoding CODH and CODH-induced hydrogenase, which are essential in carboxydotrophic proton respiration [29, 77], were found in Rhodoplanes H3. Additionally, these genes were highly similar in sequence content and gene organization with the well-characterized homologs of Rhodospirillum rubrum, which is a model hydrogenogenic carboxydotroph (Fig. S16) [78, 79], suggesting that Rhodoplanes H3 could perform carboxydotrophic hydrogenesis. Few mesophiles have been identified as anaerobic hydrogenogenic carboxydotrophs, and these mesophiles are all phototrophic Proteobacteria [30, 76]. Nevertheless, the high relative abundances of Proteobacteria have been detected in some studies fed with CO [46, 80], indicating that anaerobic CO oxidation should be widespread in Proteobacteria. The emergence of Rhodoplanes H3 expanded the phylogenetic diversity of mesophilic hydrogenogenic carboxydotrophs and proposed the crucial role of Proteobacteria in anaerobic CO oxidation [76].

Fig. 5: The representative metabolic pathways in each group, and the microbial interactions among groups.
figure 5

Only the metabolic pathways related to the main function of each group were shown. The under-characterized MAGs Nitratidesulfovibrio L3, Desulfitobacterium_A L17, Propionicimonas H18, Oscillibacter H7, Acetonemaceae H17, and Methanobacterium_C H15 were the representative genomes of groups A–F, respectively. The reverse direction of enzyme only indicated the potential of reversible enzymatic reaction. Asterisk indicated these MAGs were under-characterized taxa, and the other MAGs were assigned to discovered species or published genomes. The exact stoichiometries were not considered in this figure. Mtr tetrahydromethanopterin S-methyltransferase, Rnf Rhodobacter nitrogen fixation complex, Eha energy-converting hydrogenase A, Ehb energy-converting hydrogenase B, Fix electron-transfer-flavoprotein (ETF)-oxidizing hydrogenase complex.

Full size image

Carboxylate synthesis

Apart from CO2, the primary metabolites were carboxylates (C2–C7, referring to total dissociated and undissociated carboxylic acid, Fig. 1A), which were mainly acetate (Fig. S4). Wood–Ljungdahl pathway is necessary for acetate production from CO or CO2, and this pathway has four maker genes, acetyl-CoA synthase (acsB), corrinoid iron sulfur proteins (acsC and acsD), and formate-tetrahydrofolate ligase (fhs) [23, 81, 82]. Two under-characterized and a discovered bacterial MAGs could encode the complete maker genes and other genes in Wood–Ljungdahl pathway; therefore, they constituted group B, which provided acetate or ethanol for other microbes (Figs. 5 and S15, Dataset S10 and Dataset S11). The under-characterized MAGs L21 and L17 were assigned to a novel genus of class Anaerolineae and a novel species of genus Desulfitobacterium_A, respectively. Their emergence suggested that the high phylogenetic diversity of the microbes performing Wood–Ljungdahl pathway and further studies are needed to explore this function guild. Desulfitobacterium_A L17 might also perform carboxydotrophic hydrogenesis (Fig. S17). However, carboxydotrophic hydrogenesis and Wood–Ljungdahl pathway were not simultaneously found in other reference genomes of genus Desulfitobacterium (Table S5), which historically included the majority of Desulfitobacterium_A. The presence of Desulfitobacterium_A L17 implied that mesophilic carboxydotrophs could encode carboxydotrophic hydrogenesis and Wood–Ljungdahl pathway simultaneously, which has been observed in thermophilic carboxydotroph Calderihabitans maritimus [83]. The discovered MAG L4 was Acetobacterium wieringae that can grow on CO as sole carbon and energy source [73].

Acetate resulting from Wood–Ljungdahl pathway can be used for propionate production, where incomplete reductive citrate cycle (from acetyl-CoA to succinyl-CoA) and the conversion of succinyl-CoA to propionate are combined. Six MAGs encoding the complete enzymes in propionate production composed group C (Figs. 5 and S15, Dataset S10 and Dataset S11). Three under-characterized MAGs H8, H18, and H20 and a discovered MAG H12 were assigned to genus Propionicimonas, of which propionate production could be stimulated with acetate [84]. Only two species of this genus were isolated, P. paludicola and P. ferrireducens, and some physiological characteristics in this genus showed distinct differences [84, 85]. Additionally, the previous studies on the mixed microbial cultures fed with CO rarely reported the microbes producing propionate. The emergence of Propionicimonas spp. in this study not only broadened the genome information of Propionicimonas, but also implied the crucial role of Propionicimonas in the conversion of CO to propionate, as well as its tolerance to CO. The remaining under-characterized MAG in group C was Rhodoplanes H3 that could encode carboxydotrophic hydrogenesis as discussed above, and the other discovered MAG H10 was assigned to genus Petrimonas of which some species were also detected in previous studies [34, 69].

Carbon chain elongation

Acetate and propionate can be further elongated into longer carbon chain carboxylates via RBO pathway [86, 87]. The first step of RBO elongates an acyl-CoA by two carbon atoms to form an oxoacyl-CoA, which can be catalyzed by two types of thiolase. Compared with type I thiolase, the substrates of type II thiolase are usually two acetyl-CoA but potentially an acetyl-CoA and a different acyl-CoA. Therefore, type II thiolase mainly participates in butyrate metabolism, and type I thiolase can engage in other longer carbon chain carboxylate production in addition to butyrate production. As per thiolase type, ten MAGs were separated into two groups: group D where the complete enzymes in butyrate production were encoded, and group E that could encode the complete enzymes in synthesizing longer carbon chain carboxylates in addition to butyrate (Figs. 5 and S15, Dataset S10 and Dataset S11, Supplementary Notes).

Group D comprised five under-characterized and five discovered MAGs. The complete genes of Wood–Ljungdahl and RBO pathway existed in the genome of Desulfitobacterium_A L17 in group D. This result suggested that Desulfitobacterium_A L17 could perform carboxydotrophic acetogenesis and butyrate production in a single cell, in addition to carboxydotrophic hydrogenesis as discussed above. Carbon chain elongation using syngas fermentation effluent was separated into two sequential reactors [14, 88], which could reduce the inhibition effects of CO on carbon chain elongation microbes and could easily maintain the optimum condition for each stage. Desulfitobacterium_A L17 provided a possibility to integrate syngas fermentation and carboxylate elongation using single species, which could overcome the drawbacks of two-stage reactors; for example, more complicated operation and increased cost [69]. MAG H7 in group D was a novel species of genus Oscillibacter, of which the type species is a strictly anaerobic and valerate-producing microbe [89]. An isolate Oscillibacter sp. C5 can utilize CO and produce valerate [90]. Additionally, Oscillibacter was a dominant genus in a biological CO-converting system [91]. Group E consisted of four discovered and two under-characterized MAGs, of which Rhodoplanes H3 and Acetonemaceae H17 might perform CO oxidation. Other bacterial MAGs in this work also had potential in multiple metabolic processes, such as CO oxidation, acetogenesis, and carbon chain elongation (Fig. 5). These MAGs extended the functional redundancy in the conversion of CO to end-products, increasing the overall stability of the CO-driven microbiomes. Further investigations however should be undertaken to explore which pathway will be principally expressed under different CO pressures.

Methane production

Methanogens are the best-studied carboxydotrophic archaea. Two under-characterized and three discovered MAGs were classified as methanogens, constituting group F (Figs. 5 and S15, Dataset S10 and Dataset S11, Supplementary Notes). Methanobacterium_C H15, Methanobacterium_A L22, and Methanobacterium formicicum L9 not only encoded the complete enzymes in hydrogenotrophic methanogenesis, but also harbored the complete genes of CODH/ACS complex (cdhABCDE, Fig. S18). Therefore, these three MAGs had potential in carboxydotrophic methanogenesis, which appears to proceed via a common CO2 reduction pathway [29, 92]. The CODH/ACS complex in methanogens is suggested to mainly participate in assimilatory metabolism [77], and these three MAGs contained the complete genes in the proposed acetogenesis of methanogens. A proteomic analysis of Methanothermobacter marburgensis observed higher abundances of CODH/ACS-related proteins and acetate formation in the presence of CO [25]. Furthermore, acetate was a major metabolite of CO-fed Methanosarcina acetivorans [21]. Therefore, the Methanobacteriaceae MAGs in group F might also participate in acetate production in addition to methanogenesis. The other archaeal MAGs H22 and L23 were assigned to Methanothrix soehngenii and a novel species of Methanomassiliicoccus, respectively. Acetoclastic methanogen M. soehngenii might contribute to the CH4 production under low CO pressure since acetoclastic methanogenesis could be dominant when CO pressure was below 0.5 atm [34, 74]. The methanogenic activity of Methanomassiliicoccus is limited to methylotrophic methanogenesis with H2 [93, 94]. The existence of Methanomassiliicoccus L23 was somewhat surprising because no methanol or methylamines was detected in this work. The data available for analyzing the function of Methanomassiliicoccus L23 are limited, and future works are necessary to explore the role of Methanomassiliicoccus in a CO-driven microbiome.

As per the potential functions of the MAGs, the carbon flux and putative microbial interactions were constructed (Fig. 5). CO was first captured by group A, which reduced the inhibition effects of CO and provided available carbon sources for other function guilds. In group B, CO2 or CO was further converted into acetate, an important chemical and precursor for the synthesis of longer carbon chain carboxylates in groups C, D, and E. In addition to these valuable carboxylates, the primary biofuel was CH4, which was produced by the methanogens in group F.

Conclusions

The time-series analysis of the isotope-labeled 16S rRNA genes revealed the different dynamic succession of the functional microbiotas under two CO pressures, proposing the essential roles of Acetobacterium and Rhodoplanes under high CO and low CO pressure, respectively. Methanobacterium was the predominant archaeal genus under two CO pressures, suggesting its wide tolerance range to CO. In addition to community structure, this work also shed light on the enriched functional potential of the microbiomes under high CO pressure, presenting greater potential in synthesizing carboxylates and maintaining fundamental metabolism. Genome-centric metagenomics provided genomic insights into the under-characterized taxa, which might play a key role in the conversion of CO to carboxylates or methane. Some under-characterized taxa could encode complete enzymes in multiple pathways; for instance, both under-characterized Rhodoplanes sp. and Desulfitobacterium_A sp. might perform CO oxidation and carboxylate production. The under-characterized Rhodoplanes sp. expanded the phylogenetical diversity of mesophilic hydrogenogenic carboxydotrophs which is far less common. Moreover, the emergence of the under-characterized Desulfitobacterium_A sp. implied that Wood–Ljungdahl and RBO pathway could be encoded in a single cell. These versatile taxa could improve the functional redundancy in CO conversion, increasing the stability of the CO-driven microbiomes under variable CO pressure. These reconstructed genomes provided a comprehensive genomic baseline for further CO-converting studies. Based on the metabolic reconstruction of the MAGs, the microbial interactions and food web in the CO-driven microbiomes were established, improving our understanding in mediating the bioproduct synthesis from CO. Further studies are necessary to validate and quantify the contribution of each species to CO conversion, as well as to investigate which pathway is primarily expressed in versatile species through improved sequencing technology and sensitive metatranscriptome/metaproteome.