Photo-driven CO2 Reduction to Hydrocarbons: Carbon Flux and Energy Transforming by Marine Microbial Community

Background: Microbial activities play a crucial role in the carbon and energy cycle of the ocean. Microbial communities in the MCP has been proposed the function of transforming carbon, yet the potential for direct CO 2 reduction to the organic carbon is rarely considered in the carbon cycle and energy transforming. Results: Here we showed a naturally inherent photo-driven bioprocess of CO 2 reduction producing C 1 -C 6 alkanes/alkenes by the marine microbial community with CO 2 consumption rate of 100.87 uM day -1 . Upon the metabolite pro�les and metagenomic sequencing analysis, we revealed the mechanism of CO 2 conversion to hydrocarbons that CO 2 �xation was dominantly completed with the Calvin-Benson-Bassham cycle, tricarboxylic acid cycle, and Wood-Ljungdahl cycle, while nitrogenases, aldehyde decarbonylase and carboxylic acid reductase played the crucial roles on the hydrocarbon formations. As these insights, the pathway of CO 2 to hydrocarbons was proposed. Isolated microorganisms from the enriched community, including Pseudomonas sp. , Serratia sp. , Candidatus sp. , Clostridium sp. , Enterococcus sp ., Salmonella sp. , Rhodospirillum sp. , Thalassospira sp. , Thioclava sp. , Stenotrophomonas sp. and Desulfovibrio sp., were tentatively integrated for validating the hypothesized process. Results exhibited that the improved performance of CO 2 reduction to hydrocarbons was achieved by this group. Conclusions: This study demonstrated a natural microbial CO 2 reduction process contributing to the carbon and energy cycle in the ocean, and could provide a microbial “CO 2 -hydrogenation process” or “scher-tropsch process” for industrial application.

Biological carbon pump (BCP), largely responsible for the long-term sequestration of CO 2 in the euphotic zone into the mesopelagic and abyssal zones, includes all those processes in the ocean that form the organic carbon photosynthetically by phytoplankton in the euphotic zone, and thus increase the ux of particulate organic carbon to the deep ocean [6, 9,13,14].The phytoplankton systems and its subsequent pathways and mechanisms for carbon cycling and energy dissipation have been paid signi cant efforts.Yet, the contribution of the microbial activity on carbon and energy cycle, especially in the euphotic zone of the ocean, is less well-understood.
Importance of the microbial communities on regulating the ow of carbon in ocean has been increasingly recognized [4,9,[11][12][13]15]. Concepts such as the Microbial Carbon Pump (MCP) have been proposed to preliminarily explain the transformation of bioavailable dissolved organic carbon (DOC) to refractory DOC with MCP and the returning of DOC to the food web as particulate organic carbon [4,9,12,13,15,16].In addition, microbial respiration mobilizes the bioavailable DOC into CO 2 , which makes the carbon available again to be released back to the atmosphere [4,16,17].Within these hypothesized concepts, the carbon pool in the ocean can be uctuated by the diversity and complexity of marine microorganisms.
Particularly, the metabolism networks of the microbial communities are signi cantly complicated and challenging, and are still being characterized.Majority of the mechanism explanations on the carbon cycle are initiated with the photo-driven xation of BCP in the sunlit surface waters of the ocean [4,9,12,13,[16][17][18].Few of investigation on the microbial xation of CO 2 to the organic carbon in the euphotic zone is taken into account.
Here we showed a photo-driven bioprocess that directly convert CO 2 to light hydrocarbons (C 1 -C 6 alkanes and C 2 -C 3 alkenes) with high selectivity by using the microbial communities enriched from marine water samples.The reduction rate of CO 2 was 48.9% during 50 days culturing.To investigated the metabolic pathway of the microbial community, the analyses of metabolite pro les and metagenomic sequencing indicated that this photo-driven process was initiated with the Calvin-Benson-Bassham cycle, reductive tricarboxylic acid cycle, Wood-Ljungdahl cycle for CO 2 xation, and the hydrocarbon synthesis was completed with nitrogenase and aldehyde decarbonylase.Concomitantly to validate the pathway, we reconstructed an arti cial group with speci c microorganisms, which showed the improved performances on bioconversion.These speci c strains were isolated from the original microbial consortium, and selected according to their abundance in consortium and functions on CO 2 xation and hydrocarbon synthesis.This study shows a natural bioprocess in euphotic zone of ocean for carbon and energy cycle.

Kinetics of CO 2 reduction to Hydrocarbon
Gas samples were transferred from headspace of bioreactor at intervals to carry out the gas chromatograph analysis.The original pro les of gas chromatogram shown in Figure S3.Results demonstrated that CO 2 was obviously utilized at the average consumption rate of 100.87 uM day -1   decreasing from 11573.39 uM to 5905.73 uM through 50 days culturing, and converted to light hydrocarbons including C 1 -C 6 alkanes and C 2 -C 3 alkenes, when exposed to the consistent LED light.The maximum concentration of total light hydrocarbons was 169.67 uM observed at 38 th day (Fig. 1a).The reduction rate of CO 2 was 48.9% during 50 days culturing.The selectivity of C 2 -C 5 was more than 80%. 13C-labeled CO 2 was applied to track the carbon ux and con rm the carbon origin.In the light of GC-MS data shown in Fig. S4, the feeding of 13 CO 2 resulted in the incorporation of isotopic labelling into light hydrocarbons as indicated by the molecular ion, which was increased by 1 to 6 for each alkane or alkane.
The fragment ions were also increased by 1 to 5. Methane production as shown in Fig. 1b, was primarily detected at early stage of our bioprocess, and reached the maximum value of 66.22 uM at 31 st day.
During the period from 4 th to 17 th day, the production rate of methane was highest at 3.99 uM day -1 .
Except methane, other light hydrocarbons were negligibly detected in the headspace before 31 st day, but signi cantly observed after longer time cultivation coupling with the decreasing of methane concentration.After that, n-Butane was dominant in the mixture of light hydrocarbons, the maximum value of 44.09 uM was measured by gas chromatograph at 38 th day, which was approximately 42.68-fold of ethane, 2.47-fold of propane, 4.58-fold of i-butane, 3.88-fold of i-pentane, 1.76-fold of n-pentane, 17.72fold of i-hexane, and 53.64-fold of n-hexane (Fig. 1b).In addition, the isomer type of alkane was less than the normal alkane within our bioprocess, even compared with same carbon number.This trend showed the conformity with the theory of classic geochemistry that the ratio of the isomer type against the normal type generated in biological process was generally less than 1 in contrast with the geothermal process [19,20].Such kinetics of hydrocarbons generation from CO 2 demonstrated the marine microorganisms hold the signi cant ability to transform the CO 2 and solar energy to the organic chemicals.
Microbial diversity and statistical analysis of protein category representation 16S rRNA genes and metagenomic sequencing were performed to investigate the microbial structure the function within the CO 2 reduction process.By 16S rRNA gene sequencing, a total of 765,416 sequencing reads were generated and the reads were grouped into 4,240 operational taxonomic units (OTUs).12 OTUs accounted for 98% to 99% of the total community, and prevalently distributed in all samples (>0.5% of the retrieved sequences, Fig. 2): Proteobacteria, Firmicutes, Basidiomycota, Actinobacteria, Ascomycota, Fusobacteria, Chytridiomycota, Tenericutes, Synergistetes, Mucoromycota, Cyanobacteria and Euryarchaeota.Among them, Proteobacteria (>75% of the retrieved sequences) was accounted for dominate abundance in all samples.The subdivision, including alphaproteobacteria, betaproteobacteria, gamaproteobacteria, deltaproteobacteria and epsilonproteobacteria were detected in all samples.Particularly, gamaproteobacteria with obviously higher percentage (39.88% in S1, 85.91% in S2, 93.56% in S3, and 96.35% in S4) was increased along with culturing, which have been extensively reported that most of this subdivision are capable of photoautotrophic pathway or chemolithoheterotrophic pathway for CO 2 xation [21][22][23][24].Cyanobacteria, Ignavibacteriae, Chlorobi, and Bacteroidetes were only detected in sample S1, which have been well-elaborated the capacity of CO 2 xation, nitrogen xation, and hydrogen production via photoautotrophy or chemolithoautotrophy, yet the phylum (like Ignavibacteriae, Chlorobi, etc) were sensitive to oxygen [22][23][24][25][26]. Oxygen, possibly generated from the oxygenic photoautotroph like Cyanobacteria, was detected (data shown in Fig. 1a).Consequently, the abundance of these oxygensensitive and CO 2 -xing phylotypes was lower than other, even cannot be detected along with culturing (Fig. 2).
We annotated the gene catalogues by NR, KEGG and eggNOG according to metagenomic sequencing.The statistics of the annotation was listed in Table S2.Protein sequences within speci c categories were analyzed to statistically identify protein categories found in each sample.According to functional categories, the predicted genes assigned to energy production and conversion had the highest proportion (Fig. 3).At the functional level, it was obviously observed that functional structures of four samples were changed during culturing, as unique KEGG orthologue group (KO) pro le in sample S3 and S4 showed lower diversity than the sample S1 and S2 (Figure S5a).At pathway level, the subculture enriched Kos were distributed in carbon xation and hydrocarbon synthesis.Genes encoding the carbon xation enzyme in sample S1 and S2 was approximately 2 folds in sample S3 and S4 (Figure S5b).

Metabolic pathways and key microorganisms involved in carbon dioxide xation
According to the data of metagenomic sequencing and KEGG database, six natural CO 2 -xation pathways and the relevant key genes, including the Calvin-Benson-Bassham (CBB) cycle, the 3hydroxypropionate cycle, the Wood-Ljungdahl (WL) pathway, the reductive tricarboxylic acid (rTCA) cycle, the dicarboxylate/4-hydroxybutyrate cycle, and the 3-hydroxypropionate-4-hydroxybutyrate cycle, have been analyzed to predict the potential pathway for such bioconversion process.As well known, the CBB cycle, the 3-hydroxypropionate cycle, and 3-hydroxypropionate-4-hydroxybutyrate cycle were operated under aerobic condition, while the others pathways were anaerobic.Given all six pathways involved in our bioconversion process, the transition of pathways must be occurred during culturing, referring to the oxygen kinetics shown in Fig. 1a.Alignment to KEGG database, it presented that CBB cycle, rTCA cycle and WL pathway were completed, while the others were incomplete in all samples (Fig. S6).Key genes of these completed pathways, namely RubisCO for CBB cycle [23,[27][28][29], ATP citrate lyase for rTCA cycle [22,23,25] and CO dehydrogenase/acetyl-CoA synthase for WL pathway [23,30] were detected and showed the signi cant abundance in all samples (Fig. 4a), demonstrating that the microbial communities embodied the ability of autotrophic carbon xation.Represented by the key genes, the abundance of CBB cycle was decreased during culturing, while the abundance of rTCA cycle and WL pathway were increased (Fig. 4b).This phenomenon indicated that CO 2 reduction was possibly initiated with CBB cycle, and switched to rTCA cycle and WL pathway, when oxygen was depleted or lower than the critical sensitive concentration (Fig. 1a).According to the accumulated gene abundance of such three pathways from metagenomic analysis, the structural dynamics of the microbial consortium in phylum level suggested that Proteobacteria (51.5%-75.6%),Firmucutes (9.6%-19.7%),Acinobacteria (6.8%-14.8%)and Euryarchaeota (2.2%-7.8%)accounting for relative larger abundance were the main taxa responsible for CO 2 xation (Figure S7a-S7c).Numerous literatures have demonstrated that CBB cycle was operated in Alphaand Gammaproteobacteria, rTCA cycle had been found in representatives of the Epsilonproteobacteria, and WL pathway was continually reported on Proteobacteria, Firmucutes, Acinobacteria and Euryarchaeota [22-26, 31, 32].
The contigs representing 76.8% of the total sequence data in all samples were binned into the highquality genome bins.23 of the genome bins recovered were identi ed as prevalent OTUs.Phylogenetic tree was constructed and presented the placement of the selected bins in the relevant phylum or subdivision (Figure S8).Among of them, Nostoc (bin 1), Anabaena (bin 2) and Synechocystis (bin 3) within the phylum of Cyanobacteria, belonging to the oxygenic phototrophs encoding RuBisCO, were detected and showed the abundance of less than 2% in all samples, while their abundance were relatively higher in sample S1 than that in other samples.Nostoc and Anabaena were only detected in sample S1, and Synechocystis was not detected in samples S3 and S4.Although the oxygenic phototrophs showed their presence, anoxygenic CO 2 -xation phylotypes were dominated in all samples, which included Brucella (bin 4), Serratia (bin 5), Shigella (bin 6), Candidatus (bin 7), Chlorobium (bin 8), Azospirillum (bin 9), Azotobacter (bin 10), Vibrio (bin 11), and Simiduia (bin 12) (Figure S7c-S7e).Species from Brucella and Vibrio genus were facultative CO 2 -dependence bacteria capable of producing amino acid, alcohol, aldehyde, etc. [33,34], which can produce the intermediates for hydrocarbon synthesis.Facultative Candidatus genus, capable of CO 2 xation via CBB cycle or rTCA cycle respectively with or without oxygen, has recently been described an intra-aerobic function that is assumed to generate internal oxygen to facilitate bioreaction of those aerobic pathways [35].This insight provided the possibility of oxygenic bioreactions of hydrocarbon synthesis by aldehyde decarbonylase (AD) under anaerobic condition.Speices from Serratia genus were well characterized by chemolithotrophic xation of CO 2 which was supported by the presence of RuBisCo, carbonic anhydrase, and carboxylases [36].
As shown in Fig. 4, key gene abundance of WL pathway and rTCA cycle were increased during cultivation, and higher than that in CBB cycle.Chlorobium sp.(bin 8), as anaerobic phototrophs xing CO 2 via a rTCA cycle [25], was detected with increasing abundance in the enrichment starting from 17 th day, however showed the negligible percentage at the beginning of culturing.The archaeon Halobacterium (bin 13) was detected which can capture light energy to assimilate CO 2 [37].Based on KEGG database, the speci c members of Halobacterium entailed enzymes that can proceed the WL pathway to x CO 2 .

Pathways and key microorganisms involved in hydrocarbon synthesis
Recent researches and developments in biofuels have revealed an assortment of enzymes that can catalyze fatty acids, alcohol, or aldehyde to hydrocarbons [21,25,27,29,[43][44][45], and particularly elaborated the signi cant role of carboxylic acid reductase (CARs), aldehyde decarbonylase, and nitrogenase in the production of alkane or alkene [20,29,30,46].Generally, the activity of AD from microbes is oxygen-dependently triggered.CARs can catalyze the reduction of a carboxylic acid substrate to the corresponding aldehyde, which can be substrate to AD.As presented in Fig. 1a, the concentration of oxygen reached the maximum value (20.64 uM) at 4 th day, then declined to 0.28 uM within 13 days culturing.It was therefore possible that hydrocarbon synthesis could be completed with AD.
Metagenomic analysis showed that the abundance of genes for AD were lower in all samples (Fig. 5a).The species from the phylum of Cyanobacteria, which can express AD, had relatively lower abundance when compared with others (Fig. 2).Vibrio genus, belonging to Gammaproteobacteria, was similarly reported about the synthesis of alkanes via the AD catalysis, however only even-carbon-number and longchain (>C 10 ) alkanes were produced [48].Therefore, the synthesis of light hydrocarbons via the catalysis of CAR and AD was minimal.
As one of potential reductant, 41.44 uM of hydrogen was detected at 4 th day, then dropping extensively to 9.51 uM at 17 th day, and eventually to 0.57 uM at 53 rd day (Fig. 1a).Metagenomic analysis for hydrogenase and hydrogenlyase genes showed Clostridium (bin 8) and Vibrio (bin 11) accounted for high abundance in samples (Figure S10).Such two genus are capable of hydrogen generation [18,39,42].
Based on the analysis of metabolite pro les and metagenomic sequencing, the pathway from CO 2 to light hydrocarbons was proposed as shown in Fig. 6 that CO 2 xation was mainly started with CBB cycle, then transferred to rTCA cycle and WL pathway along with oxygen depletion, providing various precursors and energy for the process of hydrocarbons synthesis.Subsequently, the synthesis of C 1 -C 6 hydrocarbons were mainly catalyzed by nitrogenases.Particularly, the synthesis of methane, observed at the beginning of culturing, was aerobically proceeded by nitrogenase.

Biogroup reconstruction for CO 2 conversion to hydrocarbons
In order to validate the proposed pathway, the isolation of the speci c microbes from the enriched communities was conducted under various conditions.56 strains were obtained, and identi ed by 16S rRNA sequencing.Among of them, some facultatively and obligately anaerobic strains (deposited as GeneBank access No. MN538968 to MN538983) shown in Table S3 were purposely selected for the reconstruction of a biogroup to tentatively perform the bioconversion of CO 2 to light hydrocarbons under the original condition, according to the analysis of the 16S rRNA and metagenomic sequencing, as well as the previous reports of their performances.The selected strains, including Pseudomonas sp., Serratia sp., Candidatus sp., Clostridium sp., Sporomusa sp., Salmonella sp., Rhodospirillum sp., Thalassospira sp., Thioclava sp., Stenotrophomonas sp. and Desulfovibrio sp., had relative higher abundance in microbial communities, and have the functions on the proceeding of CBB, rTCA, WL pathway and encoding AD and nitrogenases.Gas chromatography showed that the reconstructed biogroup had ability of CO 2 reduction to the light hydrocarbons (Fig. 7).The maximum concentration of total light hydrocarbons was 42.97 uM observed at 7 th day (Fig. 7a).CO 2 was obviously assimilated at the average catabolic rate of 107.54 uM day -1 by decreasing from 5617.89 uM to 241.01 uM through 50 days culturing, which was higher than the rate of the original communities shown in Fig 1a .Nitrogen was reduced with rapid decreasing rate at 299.01 uM day -1 in rst three days (Fig. 7a), which was higher than 161.93 uM day -1 of the original communities, indicating more intensive activities of nitrogenase by the reconstructed biogroup.12.69 uM of hydrogen was detected at 3 rd day, then dropped extensively to 0.51 uM at 7 th day and eventually to 0.07 uM at 50 rd day.Compared with process by the original communities, obvious difference was observed that all light hydrocarbons were simultaneously produced by this biogroup, and most of them presented the highest production rate in early three days, while were decreased after 30 th day.n-Butane and n-pentane were dominant hydrocarbon in the mixture, the maximum value of 10.62 uM at 18 th day and 12.75 uM at 7 days were respectively measured (Fig. 7b).The original pro les of gas chromatography from the reconstructed biogroup was presented in Figure S11.These difference between the original communities and the reconstructed biogroup indicated some speci c microorganisms play important role on the e ciency and direction of hydrocarbon synthesis.
Therefore, the exploration and identi cation of the speci c microbes, and microbial diversity need more further works.

Discussion
Microbial community plays an important role on the carbon and energy cycle in ocean, and shows the signi cant diverse functions on the carbon ux [3][4][5][6][7][8][9][10][12][13][14][15][16][17].As presented in Fig. 1, we found the marine microbial communities exhibited the signi cantly impressive performance on CO 2 reduction and hydrocarbon synthesis, the total yield, the metabolite diversity, the production rate, and the consumption rate of CO 2 were obviously higher than the literatures described with soil microbial communities, pure microorganisms or enzymes [18,[35][36][37][38][39][43][44][45][46][47][48][49][50].In ocean, the role of phytoplankton on the carbon ux (CO 2 reduction) was well acknowledged.Bacteria in the ocean that mediate the MCP, are reported to substantially in uence the carbon cycle of the earth system.As tremendously described [2][3][4][5][6][7][8][9][10], they take up labile organic carbon produced by phytoplankton, and transfer it into the recalcitrant form that can resist from degradation for thousands of years.However, rare of efforts on microbial reduction of CO 2 to organics and its mechanism was paid to demonstrate the carbon and energy ux in ocean.The quantitative comparison of CO 2 xation between the microbial community and the phytoplankton, which can putatively measure the percentage of carbon sequestration for each, was not carried out for herein.But, the metabolic kinetics of CO 2 conversion to hydrocarbons shown in Fig. 1 indicated the astonishing e ciency and comparable contribution on sequestrating CO 2 by microbial communities in the euphotic zone of the ocean.
Numerous literatures reported the hydrocarbons production via genetic manipulation on speci c bacteria, but generally CO 2 was not involved [10,30,31,49] or other metabolites (like aromatic hydrocarbons, alcohols, acids, aldehydes, etc.) were targeted initiating with CO 2 reduction [9,32,33].Metabolism networks of the microbial communities are signi cantly complicated and challenging, and are still being characterized due to it's the exibility and diversity of functions on bioconversion [11][12][13][14][15][16].Associated with the pro le of the metabolites, the mechanism of CO 2 conversion by marine microorganisms was explored according to the metagenomic sequencing that the important role of CBB cycle, rTCA cycle and WL cycle on CO 2 reduction, and nitrogenases and AD on hydrocarbon synthesis were tentatively validated.Although each of these pathways have been elaborated well respectively, the synergy of them focusing on CO 2 reduction to hydrocarbon as we found is rarely reported [11-16, 18, 35-39, 43-50].Key genes of three CO 2 xing pathways for encoding speci c enzymes, namely RubisCO for CBB cycle, ATP citrate lyase for rTCA cycle and CODH/acetyl-CoA synthase for WL pathway, were detected and showed the signi cant abundance in samples (Fig. 4a).Other enzymes like AD and nitrogenase showed their attractive functions on hydrocarbon synthesis with aliphatic acids or carbon monoxide as substrate, however the total yield, production rate of hydrocarbons, and consumption rate of CO 2 were lower when comparison with the process we discovered [18, 20, 22-24, 27, 51, 52, 60].The function of aldehyde dehydrogenase to convert aliphatic acids to alkane is generally ful lled with the involvement of oxygen [21,25,27,39,46,52].Therefore, hydrocarbon synthesis via aldehyde dehydrogenase was limited due to the minimal concentration of oxygen (Fig. 1) within our process.Nitrogenase, known for its key role in the global nitrogen cycle, catalyzing the ambient reduction of atmospheric N 2 to bioavailable NH 4 + [48], has been increasingly reporting in recent literatures due to its novel function to reduce C 1 substrates, such as CO and CN -to short hydrocarbons with the presence of the reductants [43][44][45]48], while the characteristics of these reported processes was obviously characterized that methane was dominant in nal products.All of these indicated the complexity of this metabolic network and diverse functions within our enriched microbial community.Fortunately, we preliminarily hypothesized the pathway of CO 2 conversion to hydrocarbons of marine microorganisms, based on the obtained metabolic kinetics and metagenomic analysis.
Subsequently, an arti cial biogroup was reconstructed with the speci c microorganisms isolated from the enriched communities showed the higher rate of CO 2 and N 2 xation, and similar activities of light hydrocarbon synthesis, aiming to verify the possible mechanism.These strains inherently hold the functions on CO 2 reduction or hydrocarbon synthesis .Interesting, their synergy made the high exibility of CO 2 conversion.Through this study, an inherent and effective bioprocess in euphotic zone of the ocean was recognized, which exhibited the carbon and energy ux from CO 2 to hydrocarbons by microbial activities.

Conclusion
A photo-driven bioprocess that directly x CO 2 to light hydrocarbons (C 1 -C 6 alkanes and C 2 -C 3 alkenes) with high selectivity by marine microbial communities was discovered.The reduction rate of CO 2 was 48.9% during 50 days culturing.Mechanisms of such bioprocess was preliminarily described according to metabolic kinetics and metagenomic sequencing, and indicated that the CBB cycle, rTCA cycle and WL pathway were responsible for CO 2 xation, and nitrogenase played dominant role on hydrocarbon synthesis.Tentative efforts to arti cially develop a biogroup with using the isolated strains from the enrichment including Pseudomonas sp., Serratia sp., Candidatus sp., Clostridium sp., Enterococcus sp., Salmonella sp., Rhodospirillum sp., Thalassospira sp., Thioclava sp., Stenotrophomonas sp. and Desulfovibrio sp., were successfully made to simulate such bioconversion process.This study presented a natural bioprocess in euphotic zone of ocean for carbon and energy cycle.Additionally, this bioprocess could be further optimized and provide a microbial "CO 2 -hydrogenation process" or " scher-tropsch process" for industrial application.

Methods
Samples, cultures, CO 2 conversion 500 milliliters of water samples fetched from beach (location shown Figure S1) were ltrated for cell collection, then suspended in sterilized water and transferred into 1000 milliliters of pyrex reaction vessel (Figure S2) which supplemented with mineral salts medium (including g L -1 : 0.3 K 2 HPO 4 , 0.3 NaH 2 PO 4 , 0.05 MgSO 4 , 0.01 FeSO 4 , 1.0 CaCO 3 , 0.1 NaNO 3 ), and nally sealed and exposed to white LED light (6 mol photons m -2 s -1 ) for culturing.The vessel was purged with nitrogen gas, and 30 milliliters of CO 2 (Research 5.0 Grade, Airgas, USA) with 29.29 psi (25 o C) was syringed into the glass reactor as carbon source.At intervals, 30 milliliters of the collected samples from the reactor was syringed into to a new autoclaved tube for further analysis.Samples collected at 4 th , 17 th , 31 st , 38 th day during the culturing were labeled as S1, S2, S3, S4 respectively.

Gas chromatograph determination of hydrocarbons
Alkane and alkene analysis were carried out with a gas chromatographs (Capillary FID GC System, SRI Instruments, USA) equipped an 8600-WB5B 30M×0.53mmI.D.1.0uDB-5 Type MXT-5 capillary column [19,20].One microliter of the headspace gas was syringed into 10:1 split mode (1.9 ml min -1 He ow), the .Short-chain hydrocarbon was con rmed by comparison with a commercial standard (Airgas, USA).Quanti cation was completed by relating the GC peak area of the sample to the peak area of the standard alkane and alkene gas mixtures (Praxair, Geismar, LA), which were obtained from the headspace of sealed culture vials.In all cases, three replicates were carried out in parallel.

Gas chromatography-mass spectrometry validation
For the validation of the carbon ux, the produced hydrocarbons with using 13 C labeled CO 2 were identi ed by GC-MS using a Thermo Trace 1300 GC coupled to an ISQ-QD GC-MS (Thermo Fisher Scienti c Inc., USA) [19,20].A total of 1 ml headspace gas was injected, and operated at 125°C in splitless mode.Gas separation was achieved on a Restek (Bellafonte, PA) PLOT-QS capillary column (0.320 mm ID x 30 m length), which was held at 40°C for 2 min, heated to 180°C at 10°C/min, held at 180°C for 1 min, heated to 220°C at 40°C/min, and nally held at 220°C for 2 min.The carrier gas, helium (He), was passed through the column at 1.1 mL/min.The mass spectrometer was operated in electron impact (EI) ionization and selected ion monitoring (SIM) mode.

Total DNA extraction
Total metagenomic DNA was isolated from 30 mL of each sample using the QIAamp DNA Mini Kit (Qiagen, California, USA) according to the manufacturer's protocol.Quality and molecular weight of the DNA extracts were evaluated by agarose gelelectrophoresis.Quantity and 260/280 nm absorbance ratio as a measure of DNA purity was determined with NanoDrop® ND-1000 Spectrophotometer (Thermo Fisher Scienti c, Wilmington, DE, USA) and Qubit uorometer (Life Technologies, Carlsbad, CA, USA).All DNA extracts were stored at -20℃ until further processing.

Taxonomic assignments based on 16S rRNA sequencing
The variable V3-V4 region of 16S rDNA was ampli ed using primer sets Bac336F/Bac806R for bacteria and Arc334F/Arc806R for archaea [53,54].Indexing sequences (10 nt) were added to the 5' end of the forward and reverse primers for differentiating multiple samples.Ampli cation primers were derived from standard Illumina adapters, and PCR reaction was carried out according to the method described previously [55].The concentration of the ampli ed libraries was determined by quanti cation with Real-Time PCR to con rm the presence of suitable primers for Illumina sequencing.Paired-end sequencing with read length of 300 bp was performed on Miseq platform.Raw reads that had a low-quality score (≤20 for more than 10%) or did not match to primer and index sequences were discarded for subsequent analysis.The Quantative Insights Into Microbial Ecology (QIIME, version 1.8.0) software was used to process these high-quality reads and calculate the abundance and composition of microbial communities [56].A similarity threshold of 97% was set for clustered operational taxonomical units (OTUs) according to the Greengenes reference database [57].

Metagenomic sequencing
For each water sample a shotgun library was constructed.The libraries were prepared using the NEBNext® Ultra™ DNA Library Prep Kit for Illumina® (New England Biolabs, UK).The libraries were size selected using Agencourt® AMPure® XP beads (Beckman Coulter, Pasadena, CA, USA) with a bead to DNA ratio of 0.6 to 1 (v v -1 ).Quality and purity of the libraries has been analyzed with the High Sensitivity DNA Analysis Kit (Agilent Technologies, Santa Clara, CA, USA) on a 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA).The libraries were quanti ed using the KAPA Library Quanti cation Kits (Kapa Biosystems, Wilmington, MA, USA).Sequencing was performed on Illumina Hiseq® 2500 sequencing system (Illumina, San Diego, CA, USA) with the HiSeq PE Cluster Kit v4 cBot and HiSeq SBS Kit v4 (Illumina, San Diego, CA, USA) in a paired end mode (2 ×150 bp).

Bioinformatic analysis
Raw reads were quality-ltered and the sequencing adaptors were removed by using Trimmomatic software (V0.35) with default parameters [58].MEGAHIT was used for de novo assembly of clean reads with the "meta-large" preset [59].For each assembled contig, automated gene annotation was performed by using Prokka program.Salmon was utilized to calculate the normalized abundance for each annotated gene.Maxbin and MetaBAT2 were used for metagenomic binning analysis to recover individual genomes [60,61].
The species pro ling analysis was accomplished by self-built databases and customized Perl scripts.In brief, the taxonomy databases were downloaded from NCBI (ftp://ftp.ncbi.nlm.nih.gov/pub/taxonomy/accession2taxid/) and EBI (ftp://ftp.ebi.ac.uk/pub/databases/taxonomy/) for linking the accession ID, taxonomic number and lineage information.And then, the NT database was downloaded and all microbic sequences were extracted, including archaea, bacteria, fungi, viroids and viruses.Subsequently, annotated genes were BLASTN against this micro-NT database, and only the best hit was extracted for species abundance analysis.At last, the species distribution on each taxonomic level could be calculated by using the sum of normalized abundance of best-hit genes for each sample.
For functional analysis, available information of genes, modules and pathways was collected from the kyoto encyclopedia of genes and genomes (KEGG) database as the reference system [62].For each dataset, previous annotated genes were compared with known KEGG genes by using BLASTP (E-value < 1e-5).Only the best-hit pair was extracted for abundance estimation of designated gene families.For speci ed pathways or modules, the sum of abundance of the complete set of genes was deemed as normalized values for comparative analysis between samples.All raw reads of 16S rRNA and metagenomic sequencing had been deposited at Sequence Read Archive under accession number PRJNA577349 and PRJNA512068.
Metagenomic sequencing of all samples generated total 80.66 Gb raw reads.Subsequent data cleaning and Megahit assembly resulted in 327.3 Mbp of sequences in 334039 contigs, with an average size of 980 bp and an N50 of 1669 bp.Approximately 98.85% of clean reads could be mapped back to the assembled contigs.The statistics of sequencing data of each sample was tabulated in Table S1    encoding at least one full enzymatic pathway capable of carrying out that metabolic process were found in the genome.A bioenergetic complex was considered present only if all genes encoding the necessary subunits for a functional protein complex were found in the genome.
Page 22/25 testing program: initial temperature 35 o C, hold 7 min; ramp 15 o C min -1 to 180 o C, hold 7 min at 180 o C; ramp 20 o C min -1 to 245 o C; hold 9 min at 245 o C

Figure 2 DNA
Figure 2

Figure 3 Functional
Figure 3