Metagenomic insights into mixotrophic denitrification facilitated nitrogen removal in a full-scale A2/O wastewater treatment plant

Wastewater treatment plants (WWTPs) are important for pollutant removal from wastewater, elimination of point discharges of nutrients into the environment and water resource protection. The anaerobic/anoxic/oxic (A2/O) process is widely used in WWTPs for nitrogen removal, but the requirement for additional organics to ensure a suitable nitrogen removal efficiency makes this process costly and energy consuming. In this study, we report mixotrophic denitrification at a low COD (chemical oxygen demand)/TN (total nitrogen) ratio in a full-scale A2/O WWTP with relatively high sulfate in the inlet. Nitrogen and sulfur species analysis in different units of this A2/O WWTP showed that the internal sulfur cycle of sulfate reduction and reoxidation occurred and that the reduced sulfur species might contribute to denitrification. Microbial community analysis revealed that Thiobacillus, an autotrophic sulfur-oxidizing denitrifier, dominated the activated sludge bacterial community. Metagenomics data also supported the potential of sulfur-based denitrification when high levels of denitrification occurred, and sulfur oxidation and sulfate reduction genes coexisted in the activated sludge. Although most of the denitrification genes were affiliated with heterotrophic denitrifiers with high abundance, the narG and napA genes were mainly associated with autotrophic sulfur-oxidizing denitrifiers. The functional genes related to nitrogen removal were actively expressed even in the unit containing relatively highly reduced sulfur species, indicating that the mixotrophic denitrification process in A2/O could overcome not only a shortage of carbon sources but also the inhibition by reduced sulfur of nitrification and denitrification. Our results indicate that a mixotrophic denitrification process could be developed in full-scale WWTPs and reduce the requirement for additional carbon sources, which could endow WWTPs with more flexible and adaptable nitrogen removal.


Introduction
With the increasing realization of the impacts of excess nitrogen (N) discharge on the environment and human health, N effluent regulations have become increasingly stringent worldwide. Until now, the biological process of nitrification/denitrification has been the most prevalent wastewater treatment plant (WWTP) used to remove N from wastewater. Normally, the ratios of COD (chemical oxygen demand)/TN (total nitrogen) and BOD (biological oxygen demand)/TN are required to be higher than 15 and 8, respectively, to supply a sufficient carbon source for the traditional nitrification/denitrification process [1,2]. To ensure N removal, external organics are often required, and the additional consumption of carbon and energy has become a challenge in some WWTPs. As sustainable wastewater treatment becomes a priority, autotrophic denitrification processes based on sulfur oxidation have become increasingly popular [3][4][5]. Sulfur-based autotrophic denitrification, such as the sulfate reduction autotrophic denitrification nitrification integrated (SANI 1 ) process and sulfur-limestone autotrophic denitrification (SLAD), has been comprehensively studied [3,[5][6][7][8]. In these processes, chemolithotrophic sulfide-oxidizing denitrifying bacteria (SONB) (Thiobacillus sp., Sulfurimonas denitrificans, Beggiatoa sp., and Thiothrix sp.) and heterotrophic sulfide-oxidizing denitrifiers (Thauera-like taxa, Azoarcus, Pseudomonas, and Dechloromonas) cooperate in the processes of N removal [3,5,[9][10][11][12][13][14]. Moreover, sulfate reducing bacteria (SRB), such as Desulfobacteraceae, Desulfonema, and Thermotogaceae, in activated sludge (AS), which convert sulfate to sulfur, sulfide, and poly-S, could cooperate with SONB and enhance N removal by providing electron donors [3,8,15].
Normally, studies of sulfur-based denitrification have focused on wastewater containing high concentration of sulfate, such as saline sewage [3,16]. In fact, some inland WWTPs, especially WWTPs located in industrial parks, also treat influent containing high sulfate concentrations due to the high concentration of sulfate in industrial wastewater [17,18], which might boost the development of sulfur-based denitrification. In our previous study, we analyzed the performance of a full-scale WWTP (referred to hereafter as the YXM WWTP) based on an anaerobic/anoxic/oxic (A2/O) process in an industrial park in Yixing, China, and found that the WWTP could achieve efficient N removal under a low ratio of COD/TN [19]. In the YXM WWTP, approximately 40% of the influent was collected from an industrial park, and the average concentration of SO 4 2--S in the inlet was over 100 mg L -1 , which could provide the sulfate needed in mixotrophic denitrification. Moreover, organic carbon, N and sulfate coexist in actual wastewater. Mixotrophic denitrification, which integrates the advantages of heterotrophic and autotrophic denitrification, is promising in removing N, and it can match the actual conditions of wastewater and reduce the requirement for carbon sources [6,11,20]. Furthermore, the A2/O process has temporal changes in redox conditions that create an environment favorable for sustaining sulfide for extended time periods, and previous researchers have reported sulfide-related corrosion in the A2/O process [14]. Accordingly, we hypothesized that the N removal under a low COD/TN ratio in the YXM WWTP was due to mixotrophic denitrification facilitated by the sulfur cycle.
In this study, we sought to reveal i) whether mixotrophic denitrification facilitated by sulfur-based autotrophic denitrification could be developed in full-scale WWTPs using A2/O through acclimation to low COD/TN and high sulfate concentrations and ii) the underlying microbial and genetic mechanisms of mixotrophic denitrification in full-scale A2/O WWTPs. Until now, A2/O process, i.e., anoxic denitrification followed by aerobic oxidation of organic and N, and then recycling of nitrate back to the anoxic reactor for denitrification to N 2 is still the most common mainstream biological N and phosphorus removal process used in full-scale WWTPs [2]. Despite innovations leading to energy-efficient N management, the improvement of existing WWTPs to fulfil stringent N regulation will be increasingly valuable [14]. Understanding the underlying mechanisms of mixotrophic denitrification in full-scale A2/O WWTPs will help to design and reform existing full-scale A2/O WWTPs to improve the N removal efficiency, cost savings and sludge minimization.

Description of the YXM WWTP
The WWTP built in Yixing, Jiangsu, China (YXM WWTP), was designed to receive 60% local urban sewage and 40% industrial wastewater. The process flow of the YXM WWTP, consisting of pre-anoxic (PRAN), anaerobic (ANA), anoxic (AN), and aerobic (AE) units, followed by post-anoxic (POAN) and post-aerobic (POA) units (S1 Fig and S1 Table), was described in Chen et al. [19]. During the operation, the external cycling liquid was returned to PRAN, and the internal cycling liquid was returned to the AN unit [19]. The influent was distributed to PRAN and ANA at a ratio of 3:7. Stable operation of the YXM WWTP began after nearly one year of operation [19]. After one year of stable operation, mixed liquid samples were collected from PRAN, ANA, AN, and POAN units.

Analytical methods
The mixed liquid was sampled from the PRAN, ANA, AN, and POAN units and filtered through a millipore filter (0.45 μm), and the filtrates were analyzed for TN, NH 4 + -N, Bureau [21]. Sulfate (SO 4 2--S) was analyzed using an ion chromatography method with a conductivity detector (DIONEX-100, IonPac AS9-HC analytical column). The TDS (total dissolved sulfide, including H 2 S, HS -, and S 2-) was measured using an iodometric method [22]. The subsamples used for TDS analysis were supplemented with 0.1% v/v of 10 mol/L NaOH to prevent H 2 S from escaping. All samples were collected and run in triplicate.
The TN removal efficiency, ΔCOD and ΔTN were calculated as follows: The relationships of COD/TN and ΔCOD/ΔTN with TN removal efficiency were analyzed with linear regression in GraphPad Prism 7.0.

Nucleic acid extraction and cDNA synthesis
Mixed active sludge (AS) liquid (100 mL) was collected from the PRAN, ANA, AN, and POAN units and then centrifuged at 10,000 g for 10 min to collect the AS pellet. Total DNA was extracted from 100 mg of the AS pellet using the FastDNA spin kit for soil and the Fas-tPrep instrument (both from MP Biomedicals, Santa Ana, CA) following the manufacturer's instructions. The concentration and purity of the extracted DNA were quantified microspectrophotometrically (NanoDrop1 ND-1000, NanoDrop Technologies, Wilmington, DE, USA).
Total RNA was extracted from 100 mg of the AS pellet using the FastRNA Pro Soil-Direct kit (MP Biomedicals) according to the manufacturer's instructions. The extracted RNA was immediately reverse transcribed into cDNA using the PrimeScript TM RT reagent kit (Takara, Dalian, China) following the manufacturer's protocol. DNA and cDNA were stored at -80 o C before further analysis.

DNA library construction and metagenomic sequencing
The DNA library was constructed following the manufacturer's instructions (Illumina). Briefly, the DNA was sheared using an M220 Focused-ultrasonicator TM (Covaris Inc., Woburn, MA, USA), and fragments of~300 bp were extracted for paired-end library construction. The DNA fragments were processed by end repair, A-tailing, adapter ligation, DNA size selection, PCR, and PCR-product purification according to the Illumina TruSeq DNA sample prep v2 guide. The average insert size of the library was 350 bp. Paired-end sequencing (2 × 100 bp) was performed on an Illumina genome analyzer (HiSeq2000, Illumina) at Majorbio Bio-Pharm Technology Co., Ltd. (Shanghai, China) using the TruSeq PE cluster kit v3-cBot-HS and the TruSeq SBS kit v3-HS according to the manufacturer's instructions (Illumina). All original metagenomic sequences were archived at the NCBI Sequence Read Archive (SRA) under accession SRP140747.

Taxonomic and functional annotation analyses
Merged long-read sequence sets were blasted for sequence matching using the NR (nonredundant) database. The BLASTX alignments were further processed using MEGAN5 [23] to statistically analyze the abundance of each taxon. MEGAN (MEta Genome ANalyzer) software uses a homology-matching algorithm to generate a phylogenetic tree based on the GenBank taxonomic database.
The protein-coding reads were annotated in two steps: 1) the predicted open reading frame (ORF) was subjected to Rap-search (version 2.0) [24] against the entries in the NCBI NR database, with an E value cutoff of 1e-5; 2) the reads were mapped to the ORFs using the bow-tie program [25], allowing two mismatches. The number of reads mapped to each ORF was counted using the SAM tools package [26] and used to quantify ORF abundance within the metagenome. MEGAN5 [23] was used to parse the tabular outputs of BLASTN and Rap-search into various taxonomic and functional (SEED) levels. For comparison purposes, all distributions were normalized as a function of the number of annotated sequences/genes. KEGG (Kyoto Encyclopedia of Genes and Genomes) Orthology (KO) annotation was conducted in a BLAST search (v2.2.25) against the KEGG database (http://www.genome.jp/keeg/ ). Hierarchical clustering analysis of the KO annotation results obtained for the different samples were carried out using Hierarchical Clustering Explorer.

Taxonomic analysis of the denitrification gene sequences
MG-RAST was used to assign the narG, napA, nirS, nirK, norB, and nosZ genes to specific bacterial genera by BLASTX against the entries in the NCBI-NR database, with an E-value cut off of 1e-5. The BLASTX results were visualized using MEGAN (http://ab.inf.uni-tuebingen.de/ software/megan/) at a threshold of a bit score > 50 [27].

Gene expression activity analysis using quantitative PCR (qPCR) and reverse transcription qPCR (RT-qPCR)
SYBR Green I qPCR was performed to estimate the abundance and expression of the Ncycling functional genes (amoA, narG, napA, nirS, nirK, norB, and nosZ) by using DNA and cDNA as templates. The primer pairs and programs used in the qPCR are listed in S2 Table. Standard plasmids of the genes of interest were prepared as described in a previous study [28]. Each reaction was performed in triplicate. SPSS 19.0 was used to calculate the Spearman index and to perform the correlation analysis. A p-value < 0.05 was considered to indicate statistical significance. Data were expressed as the mean ± SD.

Sulfur-cycling facilitated nitrogen removal
The influent of the YXM WWTP was characterized by a low ratio of COD/TN (< 6) [19] and a high sulfate concentration (> 100 mg SO 4 2--S L -1 ). Additionally, the ratio of BOD/COD in the YXM WWTP was only approximately 0.35. However, the effluent quality could meet the first class A criteria of effluent discharge, and the average removal efficiency of COD, TN and NH 4 + -N could reach 83%, 72.4%, and 98.6%, respectively [19]. Both COD/TN and ΔCOD/ ΔTN showed no significant relationship with TN removal efficiency (p > 0.05, Fig 1), which indicated that the ratio of COD/TN did not significantly influence the N removal processes in the YXM WWTP.
In the PRAN unit, NO 3 --N decreased from 7.41 ± 0.44 mg L -1 in the influent to 1.03 ± 0.25 mg L -1 in the effluent, and the concentration of NO 3 --N continued to decrease to 0.41 ± 0.08 mg L -1 in the ANA unit effluent (Table 1). Meanwhile, approximately 12 mg L -1 SO 4 2--S was reduced, and 10 mg L -1 TDS was produced in the ANA unit. In AN, the main denitrification unit in the A2/O process received 400% nitrate-rich internal influx from the AO unit (S1 Fig), and approximately 17 mg L -1 COD was removed, which at most corresponded to

PLOS ONE
approximately 7 mg L -1 TN removal. Previous studies reported that 2.86 mg L -1 BOD was needed for complete denitrification of 1 mg L -1 nitrate [29]. However, approximately 8 mg L -1 TN was removed in AN, which indicated that heterotrophic denitrification alone could not meet the need for N removal even though all of the consumed COD was used for denitrification. In addition to COD reduction, we also found that approximately 9 mg L -1 TDS was oxidized to SO 4 2--S in AN, which stoichiometrically corresponded to approximately 6 mg L -1 TN removal following the equation of 5HS -+ 8NO 3 -+ 3H + ! 5SO 4 2-+ 4N 2 + 4H 2 O [30]. Thus, the combined data showed that the internal sulfur cycle of sulfate reduction and re-oxidation did occur in the A2/O system, and sulfur-based autotrophic denitrification could be a supplement to the integrated denitrification capacity. Simultaneous sulfur-based autotrophic and heterotrophic denitrification has been reported to be achieved in pilot and full-scale bioreactors [6,31,32]. Wu et al. [16] reported that a high sulfate-to-COD ratio (> 1.25 mg SO 4 2-/mg COD) facilitated the development of sulfur-based autotrophic denitrification. Accordingly, the average concentration of SO 4 2--S was over 100 mg L -1 in the inlet of the YXM WWTP, which was high enough to provide sufficient sulfur through sulfate reduction. In addition, the A2/O process has varying redox environments, including anaerobic and anoxic zones, which are favorable for sulfate reduction and sulfur oxidation. Normally, the redox potential needed for sulfate reduction is lower than that needed for denitrification, and NO 3 --N can interfere with sulfate reduction. In the YXM WWTP, NO 3 --N and DO were consumed in PRAN before entering the ANA unit (Table 1), which facilitated the establishment of an anaerobic environment suitable for sulfate reduction in ANA [33]. Furthermore, the floc micro-environment might also play an important role in sulfur-based denitrification in the A2/O process. AS flocs are partially penetrated by oxygen, and the outer portion of the flocs is aerobic, while the inner portion of the flocs will be anoxic and/ or anaerobic [34], which could provide a critical micro-environment for the harmonious coexistence of SRB, heterotrophic denitrifiers and sulfur-based autotrophic denitrifiers.
At the genus level, the top 20 abundant genera collectively accounted for 42.1-44.1% of the total bacterial reads (Fig 3). Many members among the top 20 genera were associated with nitrification and sulfur-based denitrification. Nitrospira (4.0%-6.3%) and Nitrosomonas (0.8%-1.1%), typical nitrifiers, also dominated in other AS communities [19,35]. Ignavibacterium (7.3%-8.4%) was the most dominant genus in the AS community, which was associated with sulfur-based autotrophic denitrifying processes, converting NO 2 and NO to N 2 [36,37]. Thiobacillus (4.9%-5.6%) was the third most abundant taxon, which was previously reported to be the dominant taxon in sulfur-based autotrophic denitrification bioreactors and could also cooperate with heterotrophic denitrifiers in heterotrophic environments [8,11,15,32]. Sulfuritalea (0.80%-0.91%) and Thauera (0.65%-0.76%) were also found to act as sulfur-driven autotrophic denitrifiers [13]. Dechloromonas (1.40%-2.30%), a heterotrophic denitrifier, was also found to contain sulfur oxidation genes [38]. Sulfate reduction by SRB provides the reduced S species for sulfur-assisted denitrification. The relative abundance of SRB (e.g., Geobacter, Desulfobulbaceae, Pelobacteraceae, and Desulfobacteraceae) was 1.74%-1.88% of the total bacterial reads. Unlike the mainstream sulfur-driven  PLOS ONE denitrification process [3,8,15], SRB were not the dominant taxa in the A2/O process. However, the guild of SRB showed high diversity in our study (40 genera) (S3 Table). The high diversity of SRB could facilitate the adaption and activity of SRB guilds when they went through units with different environments in A2/O facilities. In addition, Ignavibacterium might also act as the potential SRB because it was reported to have polysulfide, thiosulfate reductases, and tetrathionate reductases [36]. Meanwhile, Ignavibacterium, together with other fermenters, could provide organic acids (butyrate, lactate, fumarate) or alcohols (ethanol), which could be used by SRBs and heterotrophic denitrifiers [33,36,39]. Generally, the metagenomics data showed the co-existence of nitrifiers, anaerobic SRB, autotrophic SOB denitrifiers and heterotrophic denitrifiers in AS, which could provide a substantial microbial basis for sulfur-based mixotrophic N removal.

Functional genes related to sulfur reduction and oxidation
Sulfate reduction mediated by SRB was anticipated to provide reduced S species for electron donors in sulfur-based autotrophic denitrification, and the genes involved in dissimilatory sulfate reduction, including sulfate adenylyltransferase (sat), adenylylsulfate reductase (aprAB) and dissimilatory sulfite reductase (dsrAB), were found in high abundance in the AS communities of the YXM WWTP (Fig 4 and Table 2). In addition, a high abundance of the poly-S formation gene coding sulfide quinone oxidoreductase (sqr) was found in the AS community (Fig 4 and Table 2). Poly-S could also serve as electron and energy storage material for denitrification [8,16,40].
Sulfur oxidation genes, including those for sulfur oxidation multienzyme complex (sox) and heterodisulfide reductase (hdr), were detected in the AS communities (Fig 4 and Table 2). The Sox enzyme system oxidizes thiosulfate, which is produced by SRB or chemical oxidation of H 2 S [33,41,42]. A previous study showed that the sox and hdr genes were simultaneously expressed with denitrification genes in Thiobacillus and Thauera [13], indicating the important role of sox and hdr in sulfur-based denitrification. Therefore, the metagenomic analysis of internal sulfur cycling-related genes showed that the AS of the YXM WWTP could proceed with the internal cycling of sulfur through the cooperation of enzymes encoded by sulfate reduction and sulfur oxidation genes, which consolidated the possibility of sulfur-based autotrophic denitrification without additional TDS supplementation (Fig 4).

Functional genes related to nitrogen transformation and taxonomic classification of denitrification genes
Of the detected functional genes related to N cycling, those related to denitrification had the highest number of hits, followed by ammonification and nitrification genes. The functional
The taxonomic origins of the denitrification genes were also analyzed. The results showed that denitrification genes were unevenly distributed in eighty genera, whereas most of them were affiliated with β-proteobacteria (Fig 5), which was consistent with the dominance of βproteobacteria in the AS communities. The narG gene was mostly associated with Thauera, Pseudomonas, Thermus, Azoarcus, Stenotrophomonas, Acidovorax, Rhodanobacter, and Thiobacillus. The napA gene was mainly associated with Thauera, Thioalkalivibrio, Sulfuritalea, Dechloromonas, Cupriavidus, and Leptothrix (Fig 5). Among the genes required for the transformation of NO 2 --N to NO, nirS was mainly associated with Labrenzia, Pseudomonas, Alicycliphilus, Thauera, Sulfuritalea, and Dechloromonas, while nirK was mainly affiliated with Azospirillum, Rhodanobacter, and Polaromonas. The norB gene was mostly associated with Thauera, Rubrivivax, Hyphomicrobium, Hydrogenophaga, and Variovorax. In our metagenomics data, 23 genera were found to harbor nosZ, predominately Dechloromonas, Sulfuritalea, and Candidatus Accumulibacter phosphatis. The diversity of denitrifying gene associated taxa showed that every step of denitrification was metabolized by a combination of gene products that originated from different bacterial species. Complete denitrification is achieved through the combined activity of taxonomically diverse co-occurring bacteria performing successive metabolic steps. The diverse denitrifiers catalyze N transformation at very different rates, however, the diverse denitrification consortia were suitable to adapt to varying environments in different units in the A2/O process.
The genes encoding enzymes initiating nitrate reduction, narG and napA, were mainly associated with autotrophic Thiobacillus, Thioalkalivibrio, and Sulfuritalea (Fig 5). Consistent with the dominance of Thiobacillus in the microbial communities, the results showed the important roles of autotrophic sulfur-based denitrifiers at the functional gene level. Thauera

PLOS ONE
was associated with a high relative abundance of the narG, napA, nirS, norB, and nosZ genes (Fig 5). Both Thauera and Thiobacillus can use reduced sulfur as the electron donor and contribute to sulfur-driven autotrophic denitrification [12,13]. The analysis provided insight into microbial and functional gene foundations for the occurrence of sulfur-based mixotrophic denitrification in the A2/O process. It must be noted that most of the denitrifier taxa and high abundance of the denitrification genes narG, napA, nirS, norB, and nosZ were associated with heterotrophic denitrifiers, such as Dechloromonas, Pseudomonas, and Rhodanobacter. Dechloromonas, which carries the napA, nirS, norB, and nosZ genes, accounted for a relatively high percentage of the denitrifiers ( Fig  5). Dechloromonas, which was found to act as a heterotrophic sulfur oxidizing denitrifier [43], might also contain sulfur oxidation genes [38]. The results were not surprising because heterotrophic denitrification was still the main N removal pathway in the A2/O process. Notwithstanding, the coexistence of autotrophic and heterotrophic denitrifiers provided evidence that denitrification in the YXM WWTP could be a mixotrophic process.
Mixotrophic denitrification involves the use of fewer carbon sources and less sulfate production than heterotrophic or autotrophic denitrification and thus allows for a relatively good rate of nitrate removal over a long period of time while achieving efficient COD and N removal [44]. Notably, the mixotrophic denitrification process could achieve high efficiency of NH 4 + -N removal, as indicated in our study and others [43]. We should keep in mind that electron donors significantly impact the microbial community structure and composition [32]. Our previous study also showed that the addition of external carbon source could improve the efficiency of TN removal, however, the external carbon source elevated the proportion of heterotrophic denitrifiers in the AS communities and incurred feedback for more carbon sources [45]. Additional exploration of the communities in sulfur-based mixotrophic denitrification, with clearer variations in performance parameters, as well as the interactions between

PLOS ONE
"Thiobacillus" and heterotrophic denitrifiers in A2/O, will facilitate the development of lowcost and more flexible denitrification processes.

Activities of nitrification and denitrification genes
Although SONB can tolerate a certain concentration of sulfide, a high concentration of sulfide can inhibit their denitrification activity [12]. A previous study also showed that the activities of nitrification were inhibited by sulfide because ammonia oxidizers were sensitive to sulfide [14]. To evaluate the influence of TDS on mixotrophic denitrification in the YXM WWTP, the expression activities of functional genes related to N removal were analyzed using RT-qPCR. In our study, a high amoA mRNA/DNA ratio (S4 Table) was found in all four units (from 0.43 in POAN to 2.88 in AN), and the expression of amoA had no significant relationship with the content of reduced sulfur, which indicated that amoA expression was not inhibited by sulfide, as reported in a previous study [14]. The high-level expression of amoA ensured a high rate of NH 4 + -N transformation and subsequent denitrification in the A2/O process despite the low abundance of nitrifiers (Fig 6) [46].
In agreement with the metagenomics data, the abundance and expression of the narG gene were higher than those of napA (Fig 6). Moreover, the expression of narG and napA in ANA and AN was significantly higher than that in PRAN and POAN. The high expression of narG confirmed the important role of Thiobacillus and Thauera in denitrification, as Thiobacillus

PLOS ONE
and Thauera were the main taxa carrying narG (Fig 5). We also found a higher mRNA/DNA ratio of narG and napA in the ANA chamber than in other units (S4 Table), indicating that the anaerobic environment was favorable to the expression of narG and napA, while narG and napA were not sensitive to TDS (Fig 6). However, the low content of nitrate and short hydraulic retention time (HRT) implied no significant denitrification in the ANA chamber.
The abundance of nirS was higher than that of nirK, and the mRNA abundance of nirS was 3.1-to 93.3-fold higher than that of nirK, indicating that nirS was a key player in NO 2 --N reduction (Fig 6). Although among the denitrification genes, the abundance of norB was relatively high, its expression was low, as evidenced by its mRNA/DNA ratio, which only varied from 0.05 to 0.21. The nosZ gene is often used as a marker of complete denitrification [47].

PLOS ONE
The activity of nosZ is critical for the elimination of N 2 O in the denitrification process. Some recent studies have shown that sulfur-based autotrophic denitrification could promote the production and accumulation of N 2 O as a significant intermediate product [48]. In this study, the abundances of the nosZ gene and mRNA were the lowest among the denitrification genes (Fig 6). Therefore, the bioaugmentation of nosZ-type denitrifiers, including Thauera and Candidatus Accumulibacter phosphatis, might facilitate the enhancement of the TN removal performance of WWTPs.
The correlation of the expression of amoA and denitrification genes and the content of TDS was analyzed, and no significant relationship was found between them (p > 0.05). Our results suggested that reduced sulfur did not inhibit or stimulate the potential of N removal in the AS of the YXM WWTP. The expression pattern of functional genes might be attributed to mixotrophic denitrification using both reduced sulfur and organic compounds as electron donors, which compensate for the influence of high reduced sulfur levels on functional gene activity. In the cooperation of SRB with SONB and other heterotrophic bacteria, metabolic diversity helps to relieve the inhibition of sulfide. As discussed above, HSproduced by SRB might be transformed to poly-S and thiosulfide, and these reduced S species could also be oxidized by Thiobacillus in denitrification.

Conclusion
Mixotrophic denitrification was reported in the A2/O process with low COD/TN and high sulfate in the influent. Metagenomic analyses revealed the mixotrophic denitrification potential based on sulfur reduction and oxidation cycles at the level of microbial communities and functional genes. High abundances of sulfur oxidation, sulfate reduction and denitrification genes were found in our study. Thiobacillus dominated the microbial communities, and the narG gene was mainly harbored by Thiobacillus and Thauera, however, denitrification genes were enriched in heterotrophic denitrifiers with high abundance. Both autotrophic and heterotrophic denitrifiers contributed to nitrogen removal. Reduced sulfur showed no inhibition of the expression of nitrifying and denitrifying genes. Further investigation aimed at manipulating the denitrification process will help to further identify the key autotrophic denitrification metabolism pathways involved in nitrogen removal and the important factors related to denitrification under different conditions.