Salinity-triggered compositional and metabolic responses of autotrophic prokaryotes in Tibetan lacustrine sediments

Autotrophic prokaryotes are crucial participants in the global carbon cycle, and autotrophic carbon fixation contributes approximately 50% of the global net primary production in aquatic ecosystems per year. Salinity is a prominent regulator structuring microbial communities in diverse aquatic ecosystems. However, little information is available regarding the compositional and physiological response of autotrophic microbial communities to salinity change. Here, we used genome-resolved metagenomics to study autotrophic microbial communities in 25 Tibetan lacustrine sediments with a salinity gradient (from 0.54‰ to 82.6‰). Overall, these pathway shifts along the salinity gradient suggest that it provides a selective advantage between different mechanisms for carbon fixation. our the of autotrophic microbial to change and provide potential strategies for genome-enable cultivation of novel autotrophic microorganisms from saline environments. Further study is to characterize ecological functions of carbon-fixing prokaryotes by using “omics” and stable isotope probing (SIP) technologies, to better understand their occupied niches, response to environmental disturbance and contribution to the biogeochemical cycling of life-essential elements.

Changes in the sediment autotrophic community and carbon fixation pathways along a salinity gradient Across the five salinity groups, an abundance-salinity correlation was observed that the average autotrophic abundance decreased from G1 to G5 (one-way ANOVA, P < 0.001, Fig. 1e). This inverse correlation is likely due to two reasons: 1) salinity-derived osmotic stress directly limits metabolic activity of carbon fixers, and consumes their energy to maintain cellular osmotic balance, resulting in a decrease in autotrophic biomass; and 2) high salinity decreases the solubility and diffusion coefficient of CO 2 , thus limiting its availability for carbon fixation, and finally autotrophic biomass decreases [25,35]. On the other hand, the autotrophic community composition was significantly different among salinity groups (Anosim, Adonis and MRPP, all P = 0.001). The community in the freshwater group G1 (with the lowest salinity) was mainly composed of the Betaproteobacteria (average relative abundance, 7.28%) and Deltaproteobacteria (1.23%), while the Deinococcus-Thermus (2.50%) and Deltaproteobacteria (0.95%) were predominant in the second lowest salinity G2 (Additional file 1: Figure S2). In G3-G5 (from moderate to high salinity), the Deltaproteobacteria (0.86-1.57%) dominated, followed by the Gammaproteobacteria and Cyanobacteria. More specifically, at the species level, no MAGs were shared between G1 and G5 (Additional file 1: Figure S3), in agreement with previous findings that the freshwater taxa could rarely survive under hypersaline conditions [36,37], implying that the dominant pathway of carbon fixation may be different between the freshwater and hypersaline conditions. Analyses of the top 3 dominant species per salinity group revealed that a specific dominant species from the corresponding group had a higher relative abundance than that in the other groups (one-way ANOVA, all P < 0.01), except the Desulfobacteraceae S20.Bin068, demonstrating that these dominant species have adapted to specific salinity levels (Fig. 2). Previous studies have provided evidences that most known bacteria can growth well under pH 7.5-8.8 instead of a wide range of salinity (freshwater to hypersaline) [26]. Thus, it is reasonable to conclude that salinity has a greater impact than pH on autotrophic microbial community structure in Tibetan sediments.
However, recent studies reported that forms IV and IV-like RubisCOs might perform other functions instead of carbon fixation [11,12]. Hence, we constructed a phylogenetic tree of the RubisCO large subunit to determine the RubisCO forms present in the MAGs generated here ( Fig. 3a and Additional file 1: Figure S4). This revealed that among 41 MAGs containing the CBB cycle, almost all had the potential to encode forms I or/and II RubisCOs, except Methanothrix sp. S02.Bin032 that possessed forms II/III and III-a RubisCOs-encoding rbcL genes. Forms III and II/III RubisCOs, primarily found in archaea, enable light-independent CO 2 incorporation into sugars derived from nucleotides like adenosine monophosphate (AMP) [11,38]. The findings illustrated that all above 41 MAGs carried the potential to fix atmospheric carbon dioxide via the CBB cycle.
Exploring relationships between taxa and carbon fixation pathways showed that all MAGs affiliated with the Deinococci ( n = 4) and Epsilonproteobacteria ( n = 5) harbored the rTCA cycle, while MAGs belonging to the Cyanobacteria ( n = 7) and Gemmatimonadetes ( n = 4) exclusively coded for CBB cycle and 3-HP bicycle, respectively ( Fig. 2 and Fig. 3c).
Meanwhile, ~ 91% of the MAGs belonging to the Betaproteobacteria ( n = 22) and Gammaproteobacteria ( n = 11) contained the CBB cycle with the WL pathway identified iñ 88% of the 27 deltaproteobacterial MAGs (Fig. 2). There was a significant negative correlation (P < 0.001) between the number of the CBB cycle-containing MAGs and salinity ( Fig. 4), supporting the view that the diversity of the CBB-performing autotrophic prokaryotes decreases with enhancing salinity in aquatic ecosystems [4]. For the main carriers of the CBB cycle, taxonomy shifted from the Betaproteobacteria in G1 to the Cyanobacteria and Gammaproteobacteria in G2 and G3, to the Cyanobacteria in G4, and finally to the Gammaproteobacteria in G5, which is consistent with the conclusion of a recent study that the Betaproteobacteria dominated the CBB-performing autotrophic community in the freshwater lake and the Gammaproteobacteria dominated the hypersaline lake [4]. Moreover, for the rTCA pathway, the dominant carriers belonged to the Epsilonproteobacteria in G1 (the average relative abundance: ~0.47%) and the Deinococci in G2-G5 (0.18-2.50%), respectively. The WL pathway were mainly encoded by the Deltaproteobacteria in all five salinity groups.
Comparison of carbon fixation pathways among the five salinity groups revealed that the CBB cycle was predominant in the freshwater G1 samples, with the rTCA cycle dominating in the second lowest salinity G2 (t test, all P < 0.001). Comparative analysis showed that the CBB cycle-containing MAGs in G1 had a higher abundance in comparison with that in the other groups (one-way ANOVA, P < 0.01). The top 3 dominant MAGs (Rhodoferax sp. S03.Bin061 and Hydrogenophaga sp. S05.Bin024 and S04.Bin021) in G1 encoded the CBB cycle and the top 2 dominant MAGs (Trueperaceae sp. S10.Bin132 and S23.Bin013) in G2 encoded rTCA cycle. The discovery of the potential autotrophic Rhodoferax bacterium was inconsistent with previous results that members of the Rhodoferax could not fix CO 2 [39,40], expanding the autotrophic diversity and the niche of the Rhodoferax. Besides, previous studies also gave evidences that members of the genus Hydrogenophaga (e.g., Hydrogenophaga pseudoflava) could use the CBB cycle for carbon fixation [41]. In saline to hypersaline sediments (G3-G5), autotrophs mainly contained the CBB and WL pathways, and the WL pathway became the most important in the G5 samples (Fig. 3b). Statistical analysis indicated a significantly positive (P < 0.001) correlation between the fraction of MAGs with the WL pathway and salinity (Fig. 4). There may be two explanations for this: 1) autotrophs need to reallocate energy to accumulate inorganic and organic osmoregulators (e.g., K + , sugars, polyols and quaternary amines) to deal with intracellular osmotic stress caused by salinity [42,43]. In this instance perhaps the more energy-efficient (no ATP consumption) WL pathway provides an advantage for survival in comparison to the energy-demanding (9 ATP consuming) CBB cycle [44]. A recent study indicated that the CO 2 -reduction via WL is more ATP-efficient and yields more biomass than CO 2carboxylation dependent pathways (including CCB, rTCA and 3-HP) [7]. 2) Primordial life on earth may have started in hypersaline environments [45], and the WL pathway is considered to be the most ancient pathway for carbon fixation [46,47], so microbes containing the WL pathway may be more suitable for living in high salinity environments.
The inference of an operative WL pathway in the G5 dominant Desulfotignum sp.
S23.Bin123 and S15.Bin083 was supported by previous research [48][49][50]. Overall, these pathway shifts along the salinity gradient suggest that it provides a selective advantage between different mechanisms for carbon fixation.
Metabolic potential of sediment autotrophs in response to salinity We examined the potential of these microbes for driving biogeochemical cycles. To do this, the important pathways and related key genes in carbon, nitrogen, sulfur and iron metabolisms and respiration were searched ( Fig. 2  Hydrogenophaga S07.Bin178 contained a complete PPP. These findings greatly increase our knowledge of catabolism mechanisms of these genera. Further analysis showed a significant positive trend (P < 0.001) between the fraction of MAGs containing the EMP pathway and salinity (Fig. 4). In the highest salinity G5 sites, the top 3 dominants all carried the EMP pathway. This correlation is likely due to that the fermentative EMP pathway yields more energy from C6 and C5 sugars than the ED pathway [53], and the additional energy generated is beneficial for autotrophs to resist environmental stresses, especially salinity stress. Furthermore, roughly 41% of the 117 MAGs might produce energy via the TCA cycle, along with the gluconeogenesis pathway identified in ~ 50%.
Notably, the fraction of MAGs with gluconeogenesis potential showed a significant linear positive relationship (P < 0.001) with TOC, whereas it was significantly negatively correlated with salinity (P < 0.001). It could be explained that TOC and increased salinity inhibited carbon fixation, resulting in a decline in related glucose yield [4,33,34,54], while facultative autotrophs could utilize organic matter to produce glycose [4,33,34] and possibly use gluconeogenesis to maintain energy homeostasis (ATP synthesis) and glucose production. Meanwhile, autotrophs with the EMP pathway would be selected and enriched under high salinity conditions as they had a competitive advantage in glucose utilization over autotrophs with the ED pathway [53]. In addition, although more than a half (~ 58%) of MAGs harbored the oxidative or/and non-oxidative branches of PPP, a major source of NADPH and metabolic intermediates for biosynthetic processes [55], only 14% had the ability to perform the complete PPP.
Then we identified the coxLMS genes encoding aerobic CO dehydrogenase in 11 MAGs (such as the dominant Hydrogenophaga S05.Bin024, Desulfotignum S14.Bin031, S15.Bin083 and S23.Bin123, and Maritimibacter S23.Bin089), suggesting they are capable of CO oxidization. Previous studies also confirmed that members of the Hydrogenophaga and Desulfotignum genera could oxidize CO and generate equivalents for nitrate reduction [49,52]. The 11 species may be oxidizing CO formed from photochemical degradation of dissolved organic carbon in aquatic environments and then use the generated electrons to support ATP generation and CO 2 fixation [56,57]. This CO oxidation potential showed the highest number in the freshwater G1 (t test, all P < 0.05), and the highest fraction in the highest salinity G5 sites (t test, all P < 0.01). Besides, all seven cyanobacterial MAGs code for oxygenic photosynthesis (psa or psb), evidenced by previous findings [58]. The pufABLM genes were detected in several MAGs affiliated with Alphaproteobacteria ( n = 3), Betaproteobacteria ( n = 3) and Gemmatimonadetes ( n = 3), indicating they are potential aerobic anoxygenic photosynthetic (AAP) bacteria [59]. The three betaproteobacterial RuBisCO in AAP bacteria was involved in both CO 2 fixation and the central redox cofactor recycling, because CO 2 fixation could be used to maintain redox balance by recycling reduced redox cofactors for photoheterotrophic metabolism [60, 61].
In assessing the potential for respiration, genes encoding the aa 3 -and cbb 3 -type cytochrome c oxidases (coxABCD and ccoNOQP, respectively) and cytochrome bd ubiquinol oxidase (cydAB) were found in these MAGs (n = 14, 33 and 59, respectively) (Additional file 6: Table S5), indicating that most (~ 68%) putative autotrophs likely utilize oxygen as a terminal electron acceptor in the studied sediments. Note that in comparison with the low-oxygen-affinity aa 3 -type oxidase induced under oxic conditions, the cbb 3 -type and bd oxidases are high-affinity terminal oxygen reductases capable of functioning under microoxic to anoxic conditions [62], and the bd oxidase has a lower energetic efficiency than the heme-copper oxidases (aa 3 -and cbb 3

-type) as it does not pump protons [63].
Given the presence of the ccoNOQP or/and cydAB genes, more than a half (~ 56%) of these potential autotrophs were adapted to microoxic to anoxic conditions. Meanwhile, given the occurrence of genes (coxABCD and ccoNOQP, coxABCD and cydAB, or coxABCD, ccoNOQP and cydAB) encoding the low-and high-affinity oxidases, ~ 11% of the MAGs affiliated with the Betaproteobacteria ( n = 6), Gammaproteobacteria ( n = 5), Alphaproteobacteria ( n = 1) and Epsilonproteobacteria ( n = 1) could survive under fully aerobic to anoxic conditions, according to previous reports [63-65]. Further statistics revealed the highest species number and fraction of putative autotrophic MAGs with the ccoNOQP genes in the freshwater G1 (Fig. 4), while no significant difference was observed in the MAGs with the cydAB genes among the five salinity groups.

Autotrophic sulfur cycling in sediments of different salinities
Recent studies have revealed the important roles of autotrophs in the biogeochemical sulfur cycle [14,66]. We found 49 MAGs with dissimilatory sulfite reductase genes (dsrAB) (Additional file 7: Table S6). Phylogeny of concatenated DsrAB proteins indicates 31 reductive and 18 oxidative bacterial-types ( Fig. 5a and Additional file 1: Figure S5). Deltaproteobacteria S05.Bin005 likely gained the dsrAB genes by lateral gene transfer (LGT), supporting the fact that this genes in organisms from Deltaproteobacteria may be acquired in multiple LGT events [66,67], as shown in Fig. 5a. Besides, we summarized seven types of dsr operons in 48 of the 49 MAGs (Fig. 5b). A recent study reported that the gene composition (dsrAB and dsrD/dsrEFH) of the dsr operon may determine the direction of the dissimilatory pathway between sulfite and sulfide [66,67]. Therefore, 23 deltaproteobacterial MAGs were inferred to reduce sulfate to sulfide via the dissimilatory sulfate reduction (dsr) pathway and 13 MAGs affiliated with the Gammaproteobacteria ( n = 7), Betaproteobacteria ( n = 5) and Deltaproteobacteria ( n = 1) likely oxidize sulfide to sulfate through the reverse dsr (rdsr) pathway. Only Desulfotignum S25.Bin111 had the potential to perform both functions, likely depending on oxygen concentration and/or oxidation reduction potential, that increased the number of this especial genomes to 14 (the Deltaproteobacteria n = 5, Ca. Lambdaproteobacteria n = 4, Actinobacteria n = 4, and Nitrospirae n = 1) [66,67]. Desulfotignum species were previously inferred to be sulfatereducing bacteria [50], but our findings suggested that some specific Desulfotignum species could be both sulfate reducers and sulfide oxidizers, increasing our understanding of ecological roles of these bacteria. In the freshwater G1 samples, MAGs with dsr genes were the least prominent, while rdsr (sulfide oxidation) were the highest (t test, all P < 0.05), in good agreement with the lowest concentration of sulfate and that the most dominant species Rhodoferax S03.Bin061 and Hydrogenophaga S05.Bin024 and S04.Bin021 had the capability to oxidize sulfide via the rdsr pathway. Besides, although the Rhodoferax and Hydrogenophaga species were evidenced to oxidize reduced sulfur compounds in many studies [52,68], our findings provide a new oxidation mechanism in these genera and strengthen their importance in the sulfur cycle.
Interestingly, more than a half (~ 52%) of the 117 MAGs contained sqr or/and fccAB genes (encoding sulfide:quinone oxidoreductase and sulfide dehydrogenase, respectively) and thus might be able to directly oxidize sulfide to sulfur. Additionally, around 38% of the MAGs, mainly in G1 sites (t test, all P < 0.05), likely utilize S(0) as an electron source and produce SO 3 2− by sulfur dioxygenase (sdo). In brief, these findings indicated a coordination in sulfur cycling among sediment autotrophs.
Autotrophic nitrogen-cycling capabilities along a salinity gradient To better assess the role of sediment autotrophic microbes in nutrient cycling, we reconstructed key nitrogen utilization pathways (Additional file 8: with the Betaproteobacteria ( n = 7), Deltaproteobacteria ( n = 2) and Gammaproteobacteria (n = 1), with the highest species number and fraction in the freshwater G1 sites (Fig. 4).
More than a half of the individuals (n = 59) identified here involved in denitrification, but only Rhodocyclaceae S06.Bin133 is capable of complete denitrification (to N 2 ) and considered as an autotrophic sulfide-oxidizing denitrifier, and four species affiliated with the Alphaproteobacteria ( Rhodobacteraceae S24.Bin137) and Gammaproteobacteria (S13.Bin138, S22.Bin169 and Thiotrichaceae S06.Bin195) code a partial pathway for nitrite to nitrogen. This suggests a coordination in denitrification among sediment autotrophs.
Autotrophic nitrate-reducing and denitrifying bacteria are commonly found in sediments [75,76], and the coupling of carbon fixation, nitrate reduction and oxidation of reduced sulfur compounds performed by such microorganisms is successfully applied to wastewater treatment [77,78]. Urease genes (ureDABCEFG) genes were identified in 14 MAGs (Fig. 2), suggesting these species could acquire C and N by hydrolyzing urea to ammonia and carbon dioxide [79], which then could be fixed to yield glucose via a carbon fixation pathway. The number of species and the fraction of these potential urea-utilizers declined with increased salinity (both P < 0.001, Fig. 4). Two abundant species Hydrogenophaga S05.Bin024 and S04.Bin021 in the freshwater G1 sites possess the capability of urea utilization, evidenced by previous research [80]. To our surprise, genes encoding ammonia monooxygenase, which is responsible for the transformation from ammonia to hydroxylamine [81], were not detected in any of the MAGs. However, 42 MAGs likely participated in a partial process of nitrification (converting hydroxylamine to nitrate). In general, autotrophic microorganisms in Xiaochaidan Lake sediments likely use multiple strategies to gain nitrogen resources for growth, illustrating that they are important participants in the nitrogen cycle.

Iron metabolism by the sediment autotrophs
Iron is a ubiquitously abundant redox active transition metal in sedimentary systems [82], serving as an essential nutrient and an electron donor or acceptor to many prokaryotes [83]. More than a half (~ 56%) of the MAGs encode proteins with similarity to iron oxidase (iro) or/and sulfocyanin and the cbb 3 -type oxidase (ccoNOQP) (Additional file 9: Table S8), suggesting they are capable of transferring electrons to oxidases during iron oxidation [84]. This suggests that many autotrophic microbes in Xiaochaidan Lake sediments are also oxidizing iron, consistent with that abundant bacteria could oxidize iron in lake sediments [85]. Among them, eight were inferred to catalyze nitrate-dependent Fe(II) oxidation under anaerobic conditions, which have been observed in several different sediments [86,87], suggesting a tight coupling of Fe and N redox cycles in anaerobic sedimentary environments that has significant implications for mechanisms of NO 3 − removal and the regeneration of reactive Fe(III) oxides in hydromorphic sediments, as well as the transformation of various natural and contaminant organic and inorganic compounds [83]. It was notable that the species number and fraction of the MAGs with the sulfocyanin-encoding gene were significantly negatively associated with salinity (both P < 0.001). In the freshwater G1 sites, the most dominant species Rhodoferax S03.Bin061 codes for sulfocyanin to oxidize iron, inconsistent with the fact that the Rhodoferax species are Fe(III)-reducers [88,89]. In addition to iron oxidation, 10 MAGs possessed ferric-chelate reductase (feR-like) gene, potentially catalyzing ferric iron reduction with NAD(P)H as the electron donor [90]. FeR was first characterized in a strictly anaerobic sulfate-reducing archaeon, Archaeoglobus fulgidus [91]. Intriguingly, among the abovementioned 10 ferR-like gene-containing MAGs, four deltaproteobacterial species (Desulfotignum S09.Bin188 and S14.Bin031, and Desulfobacteraceae S24.Bin046 and S24.Bin080) also had the potential to perform sulfate reduction. In addition, the mtrB/pioB-like genes encoding decaheme-associated outer membrane proteins of MtrB/PioB family were identified in 19 MAGs, which are common among Fe(II) oxidizers and Fe(III) reducers [92]. These findings implied that autotrophic microorganisms likely be important players in the biogeochemical cycle of iron (especially iron oxidation) in sedimentary systems.

Increased physiological efficiency of sediment autotrophs in response to increased salinity
This study revealed that autotrophic prokaryotes in lacustrine sediments were taxonomically diverse with metabolic versatility, suggesting that autotrophs have wider biogeochemical niches. The comparison of the freshwater G1 and hypersaline G5 sediments illustrated that carbon-fixing microorganisms with higher physiological efficiency were selected by high salinity. The conclusion was supported by the following three reasons: 1) Different electron donors could affect the carbon fixation efficiency [93] and the changes in electron donors and acceptors resulted in the decline of energy generated with increasing salinity, because the main reactions changed from highly exergonic (such as the oxidation of sulfide and ferrous iron and dissimilatory nitrate reduction) to low exergonic (such as CO oxidation and dissimilatory sulfate reduction) [

Methods
Site description, sampling, geochemical measurements, and DNA extraction and sequencing Xiaochaidan Lake (37°27'-37°31' N, 95°25'-95°35'E) is a salty lake near the northeastern edge of Qaidam Basin (Fig. 1a). In the past forty years, the surface area has experienced expansion, reduction, and expansion again to its current size of 103.94 km 2 [95]. This lake has no outlet but is fed by a freshwater river, Tataling River, at its northwestern margin, and groundwater. In November 2016, 25 surface sediment samples were collected at ~ 1.0 m water depth, along the shoreline, from Tataling River in northwestern corner to southeastern margin of the lake. A sterile spade, washed with ethanol and in situ lake water before use, was used to collect sediments into 50 ml sterile centrifuge tubes. The samples for DNA extraction were immediately frozen in dry ice until return to the laboratory, and then stored at -80 ℃ for further analysis.
Sediment pore water was collected by centrifugation (5,000 x g, 10 min, 4 ℃), and was used to measure pH and salinity with a meter PP-20 (Sartorius, Germany). The concentrations of major ions (e.g., Na + , K + , Ca 2+ , Mg 2+ , Cl − and SO 4 2− ) were determined by using a Dionex DX 600 ion chromatography (Dionex, USA). Additionally, total organic carbon (TOC) and dissolved organic carbon (DOC) were measured with a Multi N/C 2100S analyzer (Analytik Jena, Germany). Physicochemical characteristics were summarized in Additional file 10: Table S9.
For the 25 sediment samples, genomic DNA was extracted with our modified phenol/chloroform method [96]. The quality and quantity of extracted DNA were checked by using agarose gel electrophoresis and a Nano-Drop 2000 spectrophotometer, respectively. Standard shotgun libraries of 300-bp in insert size were prepared at the Guangdong Magigene Company, and were sequenced on an Illumina HiSeq 4000 platform (paired-end 150-bp mode).

Genome-resolved metagenomic analysis
All raw reads were dereplicated using an in-house perl script, and the resulting unique reads were trimmed based on quality scores using Sickle (version 1.33) with the parameters "-q 20 -l 50" [97]. The 25 individual samples were each assembled de novo to obtain 25 separate assemblies using SPAdes (version 3.11.0) with the parameters "-k 21, 33, 55, 77 --meta" [98]. For each sample, the scaffold coverage was calculated by mapping the qualified reads to the assembled scaffolds (length ≥ 2000 bp) with BBMap.
These scaffolds were clustered into genome bins based on their tetranucleotide frequencies and coverages using MetaBAT 2 with the parameters "-m 2000 --unbinned" [99].
These retrieved genomes were further used to call genes with the single mode of Prodigal [100], and these predicted genes were annotated against the Pfam, KEGG, eggNOG, and NCBI-nr databases. According to the above results, genomes containing all key genes of any carbon fixation pathway were aggregated, dereplicated, and optimized using dRep [101] and ESOM [102]. For obtaining high-quality genomes, the best representative from each cluster was further optimized and re-assembled as previously described [68].
Next, the optimized genome bins were evaluated for taxonomic assignment, genome completeness, potential contamination, and strain heterogeneity using CheckM [103]. The predicted gene functions were manually curated and modified by comparing to the NCBInr, KEGG, eggNOG, Pfam, and in-house databases. Subsequently, metabolic pathways of each bin were constructed based on gene annotation. The 16S rRNA gene sequences identified by CheckM were used to help ensure taxonomy by searching for the closely related sequences in NCBI GenBank with BLASTn, whilst taxonomic assignment of these genomes was confirmed by phylogenomic analyses. Lastly, calculation of relative abundances of these genomes in communities was described in the Supplementary Information.

Phylogenetic analyses
The 117 genomes retrieved in this study and 544 reference genomes available in the NCBI prokaryote genome database and the Integrated Microbial Genomes and Microbiomes (IMG/M) online database were used to construct a phylogenomic tree based on a concatenated alignment of 32 marker genes (Additional file 11: Table S10) as previously suggested [104]. Specifically, marker proteins were identified in these genomes with hmmsearch (version 3.1b1) using specific hidden Markov models. Each of the 32 marker protein sequences was aligned using MAFFT (version 7.313) with the parameters "-localpair --maxiterate 1000" [105], and subsequently filtered with trimAl 1.4 to remove columns comprised of ≥ 95% gaps [106]. Then the curated alignments were concatenated, and a phylogenetic tree was built using IQ-TREE (version 1.6.10) with the parameters "LG + I + G4 -alrt 1000 -bb 1000" [107]. For the phylogenetic analysis of functional marker proteins (RbcL and DsrAB), the respective protein sequences were collected from the previous studies [12, 67,108]. Alignment and filtering were performed with the same programs as described above. Phylogenetic trees were constructed using IQ-TREE with the parameters "-m TEST -alrt 1000 -bb 1000". The newick files with the best tree topology were uploaded to iTOL for visualization and formatting [109].

Statistical analyses
The fraction of potential carbon fixers with a given function in the potential carbon-fixing community was determined by:   The retrieved MAGs with carbon fixation potential in lacustrine sediments. a Geographic location of 25 sediments collected from Xiaochaidan Lake and its inflow Tataling River. b Phylogenetic placement of the 117 retrieved MAGs. The phylogenomic tree was constructed using the concatenated alignment of 32 marker genes listed in Additional file 11: Table S10. A detailed tree is shown in    The species number and fraction of potential carbon-fixing MAGs with metabolic capabilities along the salinity gradient.

Figure 5
The dsr operons of MAGs with carbon fixation potential. a Phylogenic analysis of the concatenated DsrAB proteins. The detailed tree was provided in Additional file 1: Figure S5. Bootstrap values were based on 100 replicates, and only bootstrap values higher than 50% are shown with red (between 50% and 75%) and black (≥ 75%) circles. b The dsr operon structure in 48 putative autotrophic MAGs. The number of MAGs containing a specific dsr operon structure was showed within parentheses.

Supplementary Files
This is a list of supplementary files associated with the primary manuscript. Click to 45 download.