Composition and key-influencing factors of bacterial communities active in sulfur cycling of soda lake sediments

Bacteria are important participants in sulfur cycle of the extremely haloalkaline environment, e.g. soda lake. The effects of physicochemical factors on the composition of sulfide-oxidizing bacteria (SOB) and sulfate-reducing bacteria (SRB) in soda lake have remained elusive. Here, we surveyed the community structure of total bacteria, SOB and SRB based on 16S rRNA, soxB and dsrB gene sequencing, respectively, in five soda lakes with different physicochemical factors. The results showed that the dominant bacteria belonged to the phyla Proteobacteria, Bacteroidetes, Halanaerobiaeota, Firmicutes and Actinobacteria. SOB and SRB were widely distributed in lakes with different physicochemical characteristics, and the community composition were different. In general, salinity and inorganic nitrogen sources (NH4+-N, NO3−-N) were the most significant factors. Specifically, the communities of SOB, mainly including Thioalkalivibrio, Burkholderia, Paracoccus, Bradyrhizobium, and Hydrogenophaga genera, were remarkably influenced by the levels of NH4+-N and salinity. Yet, for SRB communities, including Desulfurivibrio, Candidatus Electrothrix, Desulfonatronospira, Desulfonatronum, Desulfonatronovibrio, Desulfonatronobacter and so on, the most significant determinants were salinity and NO3−-N. Besides, Rhodoplanes played a significant role in the interaction between SOB and SRB. From our results, the knowledge regarding the community structures of SOB and SRB in extremely haloalkaline environment was extended.


Introduction
Soda lake is a unique ecosystem with both high salinity and high pH, in which the total salt concentration exceeds 50 g/L, and the pH value can reach 10 (Schagerl and Renaut 2016;Furian et al. 2013). Despite these double extreme conditions, most lakes have high productivity and contain fully functional microbial systems. Due to the presence of high concentration of inorganic sulfur compounds, microbial sulfur cycle is one of the most active cycles in soda lake, which provides sufficient energy for microorganisms to cope with costly life at double extreme conditions (Sorokin 2017;Sorokin et al. 2015).
Haloalkalophilic sulfur oxidizing bacteria (SOB) participate in the oxidation part of the sulfur cycle in soda lake. The majority of detected SOB in soda lakes belong to Gammaproteobacteria and Alphaproteobacteria classes, which can be divided into the following three categories: chemoautotrophic (colorless) sulfur bacteria, purple sulfur bacteria, and purple non-sulfur bacteria (Berben 2019). Up to now, the following four genera of chemolithotrophic SOB have been found in soda lakes: Thioalkalimicrobium, Thioalkalispira, Thioalkalivibrio and Thioalkalibacter. They can utilize reduced sulfur compounds, such as sulfide, polysulfide, thiosulfate, polythionates, and elemental sulfur as e-donor (Sorokin 2017;Sorokin et al. 2015). The purple sulfur bacteria mainly include canoxygenic phototrophs in soda lakes, Communicated by Erko Stackebrandt. such as the genera Ecthothiorhodospira, Halorhodospira, Thiorhodospira, Thioalkalicoccus, and Ectothiorhodosinus (Gorlenko 2007). The sulfate-reducing bacteria (SRB) found in the anoxic layer of soda lake sediments belong to the Deltaproteobacteria class (Sorokin 2017;Sorokin et al. 2015). Members of the deltaproteobacterial haloalkaliphilic genera Desulfonatronum, Desulfonatronovibrio, Desulfonatronospira and Desulfohalophilus have been detected in soda lakes. They can grow either by oxidizing hydrogen, formate or small organic molecules using sulfate, thiosulfate or sulfite as electron acceptor, or by thiosulfate or sulfite disproportionation (Sorokin et al. 2011a, b).
Previous studies have shown that the structure of SOB and SRB communities could be affected by the environmental factors. For instance, Edwardson et al. compared the composition of microbial communities along the redox gradient of Mono Lake, California, USA at five depths. They found that the prokaryotic community was dominated by bacteria and the community diversity increased with depth (Edwardson and Hollibaugh 2018). In petroleum reservoirs, the SRB and SOB communities were closely associated with temperature of a reservoir, pH of formation brine, and sulfate concentration (Tian et al. 2017). And several scholars demonstrated that environmental parameters, such as depth, dissolved inorganic nitrogen concentrations, and the sedimentary dynamic environment, can significantly influence the community structures of total bacteria, SOB and SRB in sea (Zhang et al. 2017). However, the effects of physicochemical characteristics on the composition of SOB and SRB in haloalkaline environment have remained elusive. Therefore, the present study aimed to assess the diversity and structures of bacteria associated with sulfur cycle in soda lakes with different physicochemical characteristics in Inner Mongolia, China (Fig. 1).
Due to the high phylogenetic diversity of SOB and SRB, we used high-throughput sequencing of 16 s rRNA gene coupled to clone libraries of the functional genes to determine the community structures of SOB and SRB. The soxB gene, encoding sulfate thiohydrolase, a key enzyme of the Sox pathway, had been detected in sulfur oxidizers irrespective of the pathway, and thus it was selected as the biomarker of SOB in our study (Zhang et al. 2017;Meyer et al. 2007). The dsrB gene is the main subunit of isomerization (bi) sulfite reductase (DsrB) which is highly essential for the reduction of sulfite to sulfide. This gene had been used as a phylogenetic marker for SRB identification (Giloteaux et al. 2013;Gao et al. 2015). In this study, the combination of 16S rRNA gene and functional genes sequencing would improve our ability to analyze these microbial populations in detail.

Chemical analyses of soda lake sediments
To measure the sediment pH and conductivity, the sediment was suspended in ultrapure water at a ratio of 1:5 (w/w) and then the suspension was vortexed. Following the centrifugation at 21,500 ×g for 1 min at 4 °C, pH and electrical conductivity of the supernatant were measured with a multiparameter probe (Mettler Toledo Delta 320, Switzerland). The concentrations of NO 3 − and NH 4 + were analyzed by spectrophotometry (Perkin-Elmer Lambda 35, USA) (Sheibley et al. 2003). The concentrations of SO 4 2− , Cl − , and Na + were determined by ion chromatography (DIONEX ICS-1000) using a cation-exchange column (DIONEX ICs-1000) and an anion-exchange column (Shim-Pack IC-C3). The sediment was dried at 105 °C until the weight became constant to calculate the moisture concentration (MC). After that, the dried sediment was heated in a muffle furnace at 550 °C for 5 h to measure the total organic matter (TOM) (Santisteban et al. 2004). All analyses were performed in triplicate, and the mean values with standard deviations (SD) were given.

Genome extraction and high-throughput sequencing of 16S rRNA, soxB, and dsrB genes
Total DNA was extracted from sediment samples using FastDNA ® SPIN Kit for Soil (MP Biomedicals, Santa Ana, CA, USA) according to manufacturer's instructions. The DNA extract was checked on 1% agarose gel, while concentration and purity of DNA were detected with NanoDrop 2000 UV-vis spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). V3-V4 hypervariable region of 16S rRNA, soxB gene and dsrB gene were amplified by an ABI GeneAmp ® 9700 PCR thermocycler (ABI, CA, USA). The primer sequences are shown in Table 1 (Tian et al. 2017). The PCR product was extracted from 2% agarose gel and then purified using the AxyPrep DNA Gel Extraction Kit (Axygen Biosciences, Union City, CA, USA) according to the manufacturer's instructions and quantified by using Quantus™ Fluorometer (Promega Corp., Madison, WI, USA). Purified amplicons were pooled in an equimolar and paired-end sequenced on an Illumina MiSeq PE300 platform (Illumina, San Diego, CA, USA) according to the standard protocols provided by Majorbio Bio-Pharm Technology Co. Ltd. (Shanghai, China). The raw reads were deposited into the NCBI Sequence Read Archive (SRA) database (Accession Number: PRJNA678019).

Sequence analysis of the 16S rRNA, soxB, and dsrB genes
Microbiome bioinformatics were performed with QIIME2 2019.4C (Bokulich et al. 2018) with slight modification according to the official tutorials (https:// docs. qiime2. org/ 2019.4/ tutor ials/). Briefly, raw sequence data were demultiplexed using the demux plugin following by primers cutting with cutadapt plugin (Martin 2011). Sequences were then merged, quality filtered and dereplicated using functions of fastq_mergepairs, fastq_filter and derep fullength in Vsearch plugen. All the unique sequences were then clustered at 98% (via cluster_size) followed by chimera removing (via uchime_denovo). At last, the non_chimera sequences were re_clustered at 97% to generate OTU representative sequences and OTU table. Non-singleton amplicon sequence variants (OTUs) were aligned with mafft (Katoh et al. 2002) and used to construct a phylogeny with fasttree2 (Price et al. 2009). Alpha-diversity metrics Chao1, Shannon, Simpson, Pielou's evenness and Good's coverage were estimated using the diversity plugin. SOB and SRB samples were rarefied to 2293 and 707 sequences per sample, respectively. Taxonomy was assigned to OTUs using the classify-sklearn naïve Bayes taxonomy classifier in feature-classifier plugin (Bokulich et al. 2018) against the SILVA Release 132 and NCBI Database (Koljalg et al. 2013).

Bioinformatics and statistical analysis
Sequence data analyses were mainly performed using QIIME2 and R packages (v3.2.0). The taxonomy  compositions and abundances were visualized using MEGAN (Koljalg et al. 2013) and GraPhlAn software (Asnicar et al. 2015). The abundance comparison among groups at the phylum and genus levels was performed by Metastats analysis (White et al. 2009) and then visualized as a histogram. R programming language (Ihaka and Gentleman 1996) was used to analyze and draw Heatmap graphs, and Gephi 0.9.2 (Bastian et al. 2009) was employed to analyze and draw microbial interaction networks. SPSS 21.0 software (IBM, Armonk, NY, USA) was used for correlation analysis via Spearman's correlation analysis. The redundancy analysis (RDA) was carried out by Canoco 5 software.

Physicochemical characteristics of five soda lakes
The physicochemical parameters of the samples collected from five soda lakes were significantly different ( Table 2). The salinity (Na + ) of the samples ranged from 7.99 g/L to 68.45 g/L. The salinity (Na + ) was as high as 60.00 g/L and 68.45 g/L in samples of lakes B and D, respectively, while it was medium in samples of lakes C and E and very low in lake A (7.99 g/L), which were consistent with the conductivity data of water samples. In the samples, CO 3 2− , Cl − and SO 4 2− were the main anions, and their concentrations were positively correlated with salinity (Na + ), which were significantly higher than those of freshwater lakes. The color of water samples of lakes B and D was pink, and other lakes were colorless. The pink lake could be explained by dense blooms of halophilic microorganism, which synthesized lots of carotenoid to adapt to the environment of low dissolved oxygen and high light intensity (Grant and Sorokin 2011).
Nitrogen content and total organic matter were also considered ( Table 2). There were significant differences in concentrations of NH 4 + -N and NO 3 − -N (P < 0.05, n = 3), in which the NH 4 + -N concentration in the samples of lake A was obviously higher than that in other lakes, and the NO 3 − -N concentration in the samples of lake B was markedly higher. The total organic matter (TOM) of the three sampling points in the same lake was significantly different, but no obvious regularity was found. The pH of sediment samples was between 9.0 and 10.0. There was no significant difference in moisture content (MC) of sediment samples in the five lakes.

Bacteria diversity in sediment samples
High-throughput sequencing sample description and alphadiversity index of total bacteria (16S rRNA), SOB (soxB gene), and SRB (dsrB gene) in sediment samples were shown in Supplementary Table 1. The coverage of the sequencing libraries of all samples is over 96%, most of which reached over 99%, which could reflect the microbial community in each sample. Shannon index and Chao index were calculated. The results showed that the diversity of bacteria in most samples had little difference. However, there were some exceptions. The diversity of SOB in A1 and D3 samples and SRB in A3 sample was significantly lower than that in other samples. Table 3 summarized the correlation between diversity index and physicochemical characteristics of the samples. In all the samples, the α-diversity index of total bacteria, SOB and SRB was highly correlated with the salinity (Na + /SO 4 2− /Cl − ) and concentration of NO 3 − -N, while it had a low correlation with other parameters. Shannon and Chao indexs of total bacteria and SOB were negatively correlated with the salinity and concentration of NO 3 − -N. The Shannon and Chao indexs of SRB were positively correlated with salinity, while were negatively correlated with NO 3 − -N concentration. These results indicated that the diversity of total bacteria and SOB decreased with the increase of salinity, especially the total bacteria. With the increase of NO 3 − -N concentration, the diversity of total bacteria, SOB and SRB all decreased.

Total bacterial community in sediment samples
We compared and annotated sequencing results of 16 s rRNA and performed taxonomic analysis on phylum level ( Fig. 2A). Total bacterial community mainly included phyla Proteobacteria (11.34-46.00%), Bacteroidetes (10.10-45.24%), Halanaerobiaeota (0.03-53.53%), Firmicutes (0.73-21.95%), Actinobacteria (0.75-16.31%) and Gemmatimonadetes (0.1-17.59%). The results showed that there were extreme differences in bacterial composition among the samples from different lakes. However, three samples from the same lake, with similar physicochemical parameters, had similar bacterial composition, which indicated that the physicochemical parameters were important factors affecting the bacterial composition. Meanwhile, it should be noted that the bacterial diversity of samples B and D with high salinity was lower than that of other samples, and the content of phyla Haloanaerobiaeota, which was famous for its salt tolerance and anaerobic characteristics, was significantly higher than that of other samples.

SOB community in sediment samples
Many microorganisms, including obligate anaerobes, facultative anaerobes and aerobic bacteria, play roles in sulfur oxidation (Ghosh and Dam 2009;Jung et al. 2010). Based on the results of high-throughput sequencing of soxB gene, the taxonomic composition of SOB was analyzed on genus level ( Fig. 2B, Supplementary Fig. 1A), in which 26 SOB genera were detected, including Thioalkalivibrio (Gammaproteobacteria), Burkholderia, Hydrogenophaga (Betaproteobacteria), Paracoccus, Bradyrhizobium (Alphaproteobacteria) and so on. The compositions of SOB were different in the sediment samples of five soda lakes. However, the cluster of bacteria was basically consistent with the sampling sites, i.e., the samples from the same lake had similar composition and proportion of SOB. On the whole, based on the annotated results, the diversity of SOB was relatively low. Thioalkalivibrio was found dominant in the majority of samples with relative abundances ranging from 0.9 to 81.7%. Burkholderia was only detected in the sediment samples of lakes A and E with a low salinity, and the relative abundance was the highest in the sediment samples of lake E3, which was as high as 10.3%. Paracoccus was found in a small amount of the majority of samples, while it was relatively high in lakes A and E with a low salinity, which reached 5.92%, and the abundance of Hydrogenophaga was 4.51% only in lake C. Purple non-sulfur bacteria and purple sulfur bacteria, including Rhodoplanes (0.94%) and Halorhodospira (0.58%), were detected in lake D with a high salinity.

SRB community in sediment samples
Based on the results of high-throughput sequencing of dsrB gene, the taxonomic composition of SRB was analyzed at the genus level (Fig. 2C, Supplementary Fig. 1B), and it was revealed that 39 SRB genera, including Desulfurivibrio, Candidatus Electrothrix, Desulfonatronum, Desulfonatronovibrio, Desulfonatronobacter, Desulfohalophilus, Desulfonatronospira and so on, all belong to the deltaproteobacteria class. It was observed that the community structure of the SRB varied across the different soda lakes. Similar to SOB, the cluster of SRB was consistent with the sampling sites. However, compared with SOB, the diversity of SRB was higher in all samples. Desulfunvibno was detected in the majority of the samples, while it was dominant only in lake D with a high-salinity, Desulfonatronospira and Desulfohalophilus were found to have higher abundance only in lakes B and D with a high salinity, which reached 18.69 and 8.79%, respectively.
Desulfonatronobacter was detected to have a higher abundance in lake B, which was as high as 22.70%.

Effects of physicochemical factors on the SOB and SRB community
To illustrate the relationship between the microbial communities and physicochemical characteristics in soda lakes, Spearman's correlation analysis and Redundant analysis (RDA) were conducted. Spearman correlation analysis showed that most of the dominant SOB were highly correlated with the concentrations of NH 4 + -N and salinity (Na + ) (Table 4). Specifically, Bradyrhizobium, stappia and rhodoplanes were positively correlated with salinity. Burkholderia and Paracoccus were positively correlated with NH 4 + -N concentration, while Bradyrhizobium and Hydrogenophaga were the opposite. In addition, Bradyrhizobium and Hydrogenophaga were significantly affected by MC. As the main SOB in the soda lake, Thioalkalivibrio was not significantly affected by the factors investigated, which indicated that Thioalkalivibrio had a wide range of adaptability. Overall, as shown in Fig. 3A, NH 4 + -N was the most  important factor influencing the composition of SOB in the soda lake, followed by salinity. As shown in Table 4 and Fig. 3B, the influence of physicochemical factors on SRB composition was analyzed. Similar to SOB, salinity and NH 4 + -N had significant effects on SRB. Candidatus electrothrix and Desulfonatronum were negatively correlated with salinity, while Desulfonatronospira and Desulfonalophilus were on the contrary. For the effect of NH 4 + -N, it was found that the significant effect of NH 4 + -N on Candidatus electrotherix was positively correlated, while the effect on Desulfohalophilus was negatively correlated. Different from SOB analysis, the effect of NO 3 − -N concentration on SRB was more obvious, especially on Desulfonatronospira and Desulfonatronum. Desulfurivibrio was the main SRB bacteria in soda lakes, which had strong adaptability and was not affected by environmental factors, which was similar to Thioalkalivibrio. In short, salinity was noted as the most important factor that influenced the composition of SRB, followed by the NO 3 − -N and NH 4 + -N. In addition, it can be seen from the above data that TOM, pH and MC had no significant effect on the composition of SOB and SRB.

Co-occurences of SOB and SRB
In order to study the interaction between SOB and SRB, we used R programming language and Gephi 0.9.2 software to analyze and draw a network diagram. As shown in Fig. 4, there were 65 nodes (bacteria) and 309 edges in the graph. A group of closely associated microorganisms in the microbial network were divided into the same module. The bacterium in the same module had similar niches. The colors of different nodes denoted different modules. As shown in Fig. 4, there were five modules (I-V), and the proportion of each module was 29. 23, 18.46, 18.46, 15.38 and 6.15%. The network diagram showed the complex relationship between sulfur metabolizing bacteria in soda lake sediments. Most SRB were distributed in the same module (module IV), which indicated that their functions were similar and had relatively close niche. Similar to SRB, photophilic sulfur oxidizing bacteria were mostly found in module III. However, as the highest abundance of SOB, Thioalkalivibrio occupied a less important position, which implied the diversity of sulfur oxidizers and their unimportance in determining the niche.
It can be seen that Rhodoplanes was highly connected with other bacteria, which was called kinless hubs, although its relative abundance was low in the samples. Previous research had shown that the relative abundance of taxa classified as kinless hubs within the ecological network were positively and significantly correlated with the abundance of functional genes (Shi et al. 2020). The demise of kinless hubs would bring huge changes to the community structure and its functions (Banerjee et al. 2018). Rhodoplanes had a unique position in the microbial community of this study. Rhodoplanes was a kind of purple non-sulfur bacteria, which belonged to Alphaproteobacteria (Okamura et al. 2009). Rhodoplanes could grow aerobically in the atmosphere or grow anaerobically through denitrification in the dark, but the preferred growth way was to use simple organic acids, such as pyruvate, for anaerobic organic growth (Chakravarthy et al. 2012). Rhodoplanes had been reported to have an optimal pH of 7.0 (Chakravarthy et al. 2012). However, it was found that Rhodoplanes play an important role in the community of soda lake sediments, indicating the existence of some haloalkalophilic Rhodoplanes species.

Diversity of the sulfur bacteria community in soda lakes
For the first time, we depended on high-throughput sequencing of 16S rRNA, soxB and dsrB genes to accurately estimate phylogenetic richness and composition in five soda lakes with different physicochemical factors, located in Ordos, Inner Mongolia Autonomous Region, China. A surprisingly diverse bacterial community was discovered in five lake sediments. The bacterial community was dominated by the phyla Proteobacteria, Bacteroidetes, Halanaerobiaeota, Firmicutes, Actinobacteria and Gemmatimonadetes. From the results, it was found that chemolithoautotrophic genus Thioalkalivibrio covered the whole spectrum of salinity, NO 3 − , NH 4 + conditions present in soda lakes. Thioalkalivibrio presented the obvious physiological diversity, the most prominent of which was that, the dominant subgroup of this genus was able to grow in saturated soda brines containing 4 M total Na + (Sorokin et al. 2011a, b). Due to its strong contribution in biological desulfurization process, the growth characteristics of Thioalkalivibrio have been studied in depth (Mu et al. 2016;Grant and Sorokin 2011), and the genome editing methods for Thioalkalivibrio have been established (Sharshar et al. 2020). Important genes related to sulfur oxidation, such as fcc, sox, and sat/apr, have been detected in most of Thioalkalivibrio genus that have been sequenced, indicating that the genus has rich sulfur oxidation pathways, showing that it can adapt to more complex environments and contribute to the sulfur cycling process, which was consistent with the results of the present study (Foti et al. 2006;Liu et al. 2021).
Although our data highlighted the importance of Thioalkalivibrio to sulfur biogeochemistry in soda lakes, purple non-sulfur bacteria and purple sulfur bacteria, including Rhodoplanes (0.94%) and Halorhodospira (0.58%), were detected in lake D with a high salinity. Photoheterotrophic bacteria could oxidize sulfide, thiosulfate and sulfur to sulfate during heterotrophic growth with additional energy benefit and occupy the niche rapidly, which placed them in a favorable position in the rapidly changing environment of soda lakes (Tourova et al. 2013). In addition, heterotrophic sulfur oxidizing bacteria Halomonas was also detected by soxB gene. The most remarkable feature of the bacteria was their ability to grow and concomitantly oxidize sulfide and polysulfide to elemental sulfur anaerobically under denitrifying conditions (Sorokin and Kuenen 2005).
The reductive part of the sulfur cycle was active in the anoxic layers of the sediments of soda lakes. Microbiological analysis indicated a domination of haloalkaliphilic SRB, including Desulfurivibrio, Candidatus Electrothrix, Desulfonatronovibrio, Desulfonatronum and Desulfonatronospira. At present, 15 genera of haloalkaliphilic SRB had been isolated, of which 11 strains have completed genome sequencing. Most of these bacteria were isolated from soda lakes in Russia, the United States and Kenya (Sorokin et al. 2015). Desulfurivibrio had the characteristics of elemental sulfur disproportionation to adapt to the environment and became the dominant bacteria (Poser et al. 2013). Recent studies suggested that Desulfurivibrio might have the potential of extracellular electron transfer, because similar conductive pili synthesis genes responsible for electron transfer had been detected in the genome of D. alkaliphiles (Melton et al. 2016;Ni et al. 2019). Although as to whether Desulfurivibrio could carry out extracellular electron transfer still needed further verification, it was speculated that Desulfurivibrio played an important role in the electron sharing of anaerobic bacterial community.
In addition, a great number of Candidatus Electrothrix were firstly detected in the sediment samples of soda lakes with a low salinity. Candidatus Electrothrix was a kind of filamentous multicellular microorganism, which can electrically couple the oxygen reduction on the surface of the sediment with the oxidation of sulfide in the deep hypoxic layer through long-distance electron transmission (Trojan et al. 2016). Candidatus Electrothrix can also use nitrate or nitrite as an electron acceptor to couple nitrate reduction and sulfide oxidation to transfer electrons over a long distance (Bjerg et al. 2018). Uniquely, it can perform sulfide oxidation and sulfate reduction at the same time (Nielsen et al. 2010), but in this study, Candidatus Electrothrix was not detected with soxB as the marker gene. It should be because most strains of this genus use opposite pathway of classical sulfur reduction to complete sulfur oxidation (Müller et al. 2020).

− -N were the significant factors
We compared the effects of salinity, nitrogen and other physicochemical parameters on the diversity and composition of total bacteria, SOB and SRB. In general, salinity, NH 4 + and NO 3 − concentration were the most important factors. The diversity of total bacteria and SOB was negatively correlated with salinity. It has been proved that salinity can cause 317 Page 10 of 12 changes in the sulfur cycle. There were two sulfur cycles in soda lakes mainly depending on the salinity of the lake water. At medium salinity, a complete sulfur cycle, namely "long" sulfur cycle, is more likely between HS − and SO 4 2− , where HS − is completely oxidized to SO 4 2− , and SO 4 2− is reduced to HS − by SRB. Under saturation salinity, the cycle may be shortened due to the presence of sulfur intermediate products such as S 0 , Sn 2− , S 2 O 3 2− , namely "short" sulfur cycle. In this cycle, HS − is not completely oxidized to S 2 O 3 2− and then formed HS − by sulfur reduction reaction (Sorokin et al. 2011a, b).
Additionally, it was showed that the diversity of total bacteria and SRB was significantly correlated with NO 3 − , and the RDA analysis revealed that NO 3 − was the second important factor influencing the composition of SRB. This could be related to the high NO 3 − concentration contributing to the growth of denitrifying bacteria (DNB), because the intermediate products of denitrifying, e.g. NO and N 2 O, could inhibit the growth of SRB (Zumft 1993). Moreover, studies demonstrated that DNB could preferentially use matrix and thus, it possessed an advantage compared with SRB (Chidthaisong and Conrad 2000). In addition, NO 3 − is an important electron acceptor for SOB, especially in oxygen deficient environment. Now, three out of nine Thioalkalivibrio species had been confirmed to be able to grow anaerobically with NO 3 − as electron acceptor.

Rhodoplanes was the core bacterium
In the process of sulfur cycle in soda lakes, there were complex relationships between sulfur-related bacteria. Many species of bacteria were located in the same niche, which indicated the complexity of sulfur cycle in soda lakes and the redundancy of the same functional bacteria, which ensured the stability of the system to the maximum extent. According to the co-occurences of SOB and SRB, it was found that the bacteria with high abundance was not the kinless hubs. The most representative example was Thioalkalivibrio, which had little contact with other bacteria. The results showed that Rhodoplanes was the core bacterium, which may be related to its metabolic diversity. This genus comprised purple non-sulfur bacteria expressing preferably photoheterotrophic growth in the light under anoxic conditions, whereas chemoorganotrophic growth was possible in the dark under both oxic and anoxic conditions. Due to its metabolic versatility, Rhodoplanes has been isolated from very diverse aquatic environments, ranging from freshwater through to activated sludge in wastewater treatment plants (Rojas et al. 2018).

Conclusions
In this paper, the high-throughput sequencing of 16 s rRNA, dsrB, and soxB genes was carried out, and the microbial communities in different soda lakes were analyzed. The results showed that most of the total bacteria in soda lakes belong to phyla Proteobacteria, Bacteroidetes, Halanaerobiaeota, Firmicutes, and Actinobacteria. SOB and SRB were broadly distributed in soda lakes. The detected SOB was affiliated to Alpha-, Beta-and Gamma-proteobacteria populations, while most of SRB was affiliated to Deltaproteobacteria. The distribution of SOB and SRB was significantly affected by levels of salinity and inorganic nitrogen. The communities of SOB, mainly including Thioalkalivibrio, Burkholderia, Paracoccus, Bradyrhizobium, and Hydrogenophaga genera, were remarkably influenced by the levels of NH 4 + -N and salinity. For SRB communities, including Desulfurivibrio, Candidatus Electrothrix, Desulfonatronospira, Desulfonatronum, Desulfonatronovibrio, Desulfonatronobacter and so on, the most significant determinants were salinity and NO 3 − -N. Besides, Rhodoplanes played a significant role in the interaction between SOB and SRB. In summary, this study extended our knowledge regarding the distribution of community structures of SOB and SRB under haloalkaline conditions.