1. Introduction
Advanced denitrification biofilters, serving as a tertiary treatment process, have been widely implemented in wastewater treatment plants (WWTPs) [[
1], [
2], [
3]]. However, achieving optimal biological nitrogen reduction and phosphorus removal often necessitates supplementation with external electron donors and chemical agents owing to the low C/N ratio in the secondary effluent of WWTPs [
2,
4], which increases operating costs, sludge yield, and clogging risk and creates excessive chemical oxygen demand in the effluent [
5]. The main form of nitrogen in secondary effluent is nitrate nitrogen (NO
3–-N), followed by dissolved organic nitrogen, particulate nitrogen, and ammonia nitrogen (NH
4+-N) [
6,
7]. Accordingly, there is an imperative need for technologies that can effectively remove nitrogen, especially NO
3–-N and phosphorus from the secondary effluent of WWTPs.
Recently, the advent of pyrite-based mixotrophic denitrification biofilters packed with a mixture of pyrite and solid-phase carbon sources has offered considerable advantages, including saving external organic carbon sources, simultaneous removal of NO
3–-N and phosphorus, and reducing sludge production and sulfate (SO
42–) and Fe
2+/Fe
3+ emissions [[
8], [
9], [
10]]. These attributes render such biofilters an economical, efficient, and environmentally friendly tertiary treatment technology [
11,
12]. Moreover, tight microbial interactions between sulfur-based autotrophic and heterotrophic denitrifying bacteria are conducive to the stability of the wastewater treatment process [[
13], [
14], [
15]]. In our previous study [
16], constructed pyrite-sawdust composite-based biofilters achieved average removal efficiencies of total nitrogen (TN) and total phosphorus (TP) of 85% and 62%, respectively, with low N
2O, CH
4, and CO
2 emissions when treating simulated secondary effluent for 304 days. Preliminarily results using 16S ribosomal RNA (rRNA) gene amplicon sequencing revealed that the main autotrophic and heterotrophic denitrifying bacteria were
Thiothrix and
unclassified_f_Rhodocyclaceae, respectively. However, the deeper intrinsic drivers of the functional microorganisms, their metabolism, and electron transfer patterns at the molecular-gene level remain unknown in pyrite-sawdust composite-based biofilters.
Although attention is beginning to be paid to microbial nitrogen removal in denitrification systems, most previous studies have provided basic microbial information, such as the abundance or composition of dominant nitrogen-removing functional bacteria (e.g., denitrifiers) [
14,
17,
18]. In fact, in addition to the dominant nitrogen-removing functional bacteria, other nonfunctional bacteria, such as cellulose-degrading bacteria, methanogenic archaea, and sulfate-reducing bacteria (SRB), may not be directly involved in nitrogen removal processes but cooperate with functional microorganisms related to nitrogen transformation, significantly affecting the nitrate (NO
3−) reduction process [
17]. Thus, comprehensive and deep data mining is required to elucidate the species involved using high-resolution methods (e.g., whole-metagenome sequencing) since 16S rRNA gene amplicon sequencing suffers from variable 16S rRNA gene copy numbers and primer efficiencies across microbes [
19]. Additionally, the functions of functional and nonfunctional bacteria involved in C, N, S, and Fe metabolism are unknown in pyrite-based mixotrophic denitrification systems. Consequently, gaining an in-depth understanding of the comprehensive species involved and pyrite-driven C, N, S, and Fe functional metabolism is imperative for maintaining the stability of mixotrophic systems [
20].
Moreover, microorganisms within mixotrophic systems can simultaneously employ multiple metabolic strategies or alternate between strategies as needed [
21]. Metabolite cross-feeding is widely distributed across mixotrophic systems [
11]. Beyond enhancing process performance, metabolites play a pivotal ecological role. For instance, in mixotrophic systems that combine anammox and heterotrophic denitrification, anammox bacteria have been identified to possess the capability to synthesize B vitamins, which they supply to the heterotrophic bacterial community, fostering a symbiotic relationship [
22,
23]. However, studies exploring metabolites and metabolic exchange between autotrophic and heterotrophic denitrifying bacteria in mixotrophic denitrification systems remain limited. Therefore, placing a greater emphasis on understanding the microbial interactions mediated by cellular metabolites and signaling molecules within these systems is imperative. Such research will be instrumental in unraveling the complex ecological dynamics at play and optimizing the performance of mixotrophic denitrification processes.
Hence, this study explored the microbial spatial stratification, multiple nitrogen removal pathways, and metabolites in the long-term operation of pyrite-sawdust composite-based biofilters via metagenomics and untargeted metabolomics. The objectives of the study were: ①to examine pollutant variations at different heights within the biofilters and decipher the hot zones; ②to explore the spatial dynamics of the microbial community composition at different heights within the biofilters; ③to elucidate the microbial contributions of functional genes and perform a differential analysis of the metabolic pathways involved in the N, S, C and Fe cycles via metagenomics; and ④to reveal the differential metabolites via untargeted metabolomics and elaborate the coupling mechanisms of N, S, C, and Fe transformations and their electron transfer modes. This study comprehensively evaluates the microbial functions and metabolites in pyrite-sawdust composite-based biofilters and innovatively highlights the deeper intrinsic drivers of nitrogen removal, including microbial spatial stratification, multiple nitrogen removal pathways, and differential metabolites.
2. Materials and methods
2.1. Biofilter set-up and operation
Four up-flow biofilters (internal diameter, 100 mm; height, 900 mm) were packed with pyrite-sawdust composite as the experimental groups (bioreactors 1, 2, 3, and 5), and one up-flow biofilter was packed with poly(3-hydroxybutyrate-
co-3-hydroxyvalerate (PHBV)/sawdust composite as the control group (bioreactor 4) inoculated with heterotrophic sludge (Fig. S1). Therein, bioreactors 1, 2, and 3 were inoculated with autotrophic sludge, corresponding to HRT of 2.5, 3.5, and 1 h, respectively, while bioreactor 5 was inoculated with mixed sludge (autotrophic and heterotrophic sludge) with HRT of 2.5 h to compare with bioreactor 1. The pyrite/sawdust and PHBV/sawdust composites were prepared by blending pyrite or PHBV with sawdust in a rubber internal mixer, as described in our previous study [
14]. The structure, seeding sludge, and operational mode of each bioreactor were described in our previous study [
16]. The influent was synthetic wastewater prepared using tap water, with 0.5–1.4 mg·L
–1 NH
4+-N, 15–19 mg·L
–1 NO
3–-N, 0.9–1.5 mg·L
–1 PO
43–-P, 180 mg·L
–1 NaHCO
3, 80–90 mg·L
–1 SO
42–, and dissolved oxygen (DO) content of 2.1–4.8 mg·L
−1 at pH 7.2 ± 0.1. To reduce the negative impact of changing the hydraulic retention time (HRT), bioreactors 1–5 were only set at HRTs of 2.5, 3.5, 1, 2.5, and 2.5 h, respectively. The trial lasted 304 days at room temperature (27–30 °C).
2.2. Physicochemical analytical methods
Water samples were obtained every three days from the inlet and outlet of each biofilter until the end of the experiment (day 304), and the corresponding nitrogen concentration changes were published in our previous study [
16]. Especially, to reveal the changes of pollutants along the path of bioreactors, the concentrations of NO
3–-N, NO
2–-N, NH
4+-N, TN, TP, SO
42–, and dissolved organic carbon (DOC) were measured at different heights within the biofilters under stable operating conditions (day 158) [
13,
24]. The elemental distribution and valence states on the biofilms were characterized using an EscaLab 250Xi X-ray photoelectron spectrometer (XPS; Thermo Fisher Scientific, USA).
2.3. Biofilm sample extraction methods
Biofilm samples were collected at the same height of the biofilm sampling port of 60 cm in each biofilter on days 36, 109, 149, and 215 for high-throughput sequencing analysis to reveal the microbial community, whose results were published in our previous study [
25]. Also, at the end of the bioreactor operation (day 304), 3–4 pieces of the pyrite–sawdust and PHBV–sawdust composites were collected at biofilm sampling ports at different heights within the biofilters to perform metagenomics sequencing and untargeted metabolomic analysis. For instance, biofilm samples were collected from the bottom (20 cm) of bioreactor 1 (EG_1_B), the bottom of bioreactor 2 (EG_2_B), the middle (40 cm) of bioreactor 2 (EG_2_M), the middle of bioreactor 4 (CG_4_M), and the bottom of bioreactor 5 (EG_5_B); EG is experimental group, CG is control group, B is bottom, and M is middle. Biofilm samples were extracted from the 3–4 pieces of pyrite–sawdust and PHBV–sawdust composites in the biofilters using a whirlpool mixer and oscillated three times (5 min·time
–1) to ensure that the biofilms were fully detached. Then, the detached biofilms were centrifuged at 10 000 ×
g for 10 min at 4 °C to obtain cell pellets. The biofilms were extracted in triplicate. The 15 obtained biofilm samples were cryopreserved at −20 °C until DNA and metabolite extraction, and performed downstream analysis after the inspection passes of DNA.
2.4. Metagenomics and untargeted metabolomics analysis
As stated above, the 15 qualified biofilm samples were used for metagenomic sequencing by the Illumina HiSeq4000 platform in Majorbio laboratory (China). Briefly, the obtained high-quality reads were annotated using BLASTP software according to the non-redundant protein database and Kyoto Encyclopedia of Genes and Genomes (KEGG) database to identify species information and functional genes. Detailed data processing information is presented in Text S1 in Appendix A.
Further, genome binning of each metagenome was performed using MetaWRAP‡ to obtain the high-quality metagenomic assembled genomes (MAGs) (completeness > 50%, contamination < 10%). According to sequence mapping, the assembly and binning process recovered 407 high-quality MAGs with completeness > 50% and contamination < 10%. 11 MAGs with completeness > 75% and contamination < 5% were chosen for further analysis. The detailed description was also found in Text S1in Appendix A.
Additionally, the extracted biofilm samples underwent untargeted metabolomics analysis by Guangdong Magigene Biotechnology Co., Ltd. (China). Briefly, the spectra obtained via LC-MS/MS analysis were preprocessed using Compound Discoverer 2.1 software (Thermo Fisher Scientific), including peak extraction, alignment, correction, and standardization. Then, metabolite annotation was performed using the self-built secondary mass spectrum database, BiotreeDB [
22]. Therein, the quality control procedures for LC-MS/MS analysis mainly include the following steps: ①Filter individual peaks to remove noise. ②Filter a single Peak. ③Simulate the missing value recoding in the original data. ④Normalization of data. Detailed data processing and statistical analysis information is presented in Text S2 in Appendix A. Therein, a microbe–metabolite co-occurrence network was constructed by utilizing the Pearson correlation matrix, which was calculated based on the correlation coefficient (
r > 0.7,
P < 0.05) between the detected microbe and metabolites using the R igraph package.
3. Results and discussion
3.1. Spatial variation of pollutants within the biofilters
In a previous study, pyrite-based mixotrophic denitrifying biofilters achieved good removal performance during operation for 304 days, with average TN and TP removal rates of 85% and 62%, respectively [
16]. To gain a better understanding of the spatial operation characteristics of the bioreactors, we examined the variations in nitrogen (NO
3–-N, NO
2–-N, and NH
4+-N), sulfur (SO
42–), and carbon (DOC, TC, inorganic carbon (IC)) concentrations on day 158 at different heights within the biofilters, including 0–10 cm (bottom); 10–20, 20–30, and 30–40 cm (middle); and 40–50, 50–60, and 60–70 cm (upper). Additionally, the reaction hot zones for various pollutants were analyzed.
3.1.1. Variations in nitrogen concentrations and hot zones for nitrogen removal within the biofilters
The variations in nitrogen concentrations (NO
3–-N, NO
2–-N, and NH
4+-N) within the bioreactors are presented in
Fig. 1. The NO
3–-N concentrations were continuously reduced and utilized by a microbe, resulting in a gradual concentration decrease from the bottom to the top in each bioreactor. However, the corresponding intermediate products, such as NO
2–-N and NH
4+-N, exhibited different accumulation concentrations within the bioreactors. A comparison of bioreactors 1, 2, 3, and 5 revealed that only bioreactor 2 exhibited a slight accumulation of NH
4+-N in the upper 50 cm, while ammonium accumulation was not shown significantly (
P > 0.05) in other bioreactors. The lack of NH
4+-N accumulation from the bottom to the top in bioreactors 1 and 3 with short HRTs suggested that prolongation of HRT led to the occurrence of the dissimilatory nitrate reduction to ammonia (DNRA) process in the pyrite-sawdust composite-based biofilters, which mainly occurred in the upper part of the biofilters. Similarly, the lack of NH
4+-N accumulation in bioreactor 5 also indicated that inoculated mixed sludge was conducive to NH
4+-N removal. A recent study demonstrated that NO
3–-N reduction is preferred in the DNRA pathway when the nitrogen source (electron acceptor) is relatively insufficient [
26]. Corresponding with the highly efficient denitrification in bioreactor 2, nitrate concentrations in the upper part of the bioreactor were low, which led to a severe shortage of electron acceptors, and thus the DNRA process mainly occurred in the upper part. The characteristics of NO
2–-N accumulation are described in detail in Text S3 in Appendix A. Further, the nitrogen removal in each section of the bioreactors and its hot zones are presented in Fig. S2 in Appendix A and described in detail in Text S4 in Appendix A.
3.1.2. Variations in SO42− concentrations within the biofilters and hot zones of its generation and reduction
The variations in SO
42− concentrations within each bioreactor are also presented in
Fig. 1. From the bottom to the top, compared with that in the influent (80.1 mg·L
–1), the SO
42− concentrations in the experimental groups (bioreactors 1, 2, 3, and 5) first increased to a maximum and then decreased rapidly, demonstrating two phases of SO
42− generation and reduction. By contrast, the SO
42− concentration in the control group (bioreactor 4) decreased from the bottom to the top, indicating the occurrence of only the sulfate reduction process. As an important reaction product of sulfur-based autotrophic denitrification (SAD), the amount of generated SO
42− characterizes the activity of the SAD process.
During the sulfate generation stage, peak sulfate generation in bioreactors 2, 1, and 3 occurred at heights of 50, 60, and 60 cm within the biofilters, respectively, with peak accumulated concentrations as high as 201.8, 166.4, and 111.2 mg·L–1, respectively. This finding indicated that maximum SO42− generation gradually decreased with decreasing HRT and that HRT affected the SAD process. Notably, maximum SO42− generation was significantly and positively correlated (P < 0.05) with TN removal, again indicating the dominance of the SAD process in the pyrite-sawdust composite-based biofilters.
In the SO
42− reduction stage, the SO
42− reduction process in bioreactors 2, 1, and 3 occurred at heights of 50–70, 60–70, and 60–70 cm within the biofilters, respectively, with accumulated SO
42− reduction amounts of 90.9, 39.3, and 16.3 mg·L
–1, corresponding to effluent NO
3–-N concentrations of 0.89, 5.7, and 6.3 mg·L
–1, respectively. Generally, SO
42− reduction occurs only when levels of electron acceptors O
2/NO
3− are very low or even depleted due to SO
42− having a higher electron potential than O
2/NO
3− on the energetic electron tower [
27]. However, the results indicated that the SO
42− reduction process continued to occur in the presence of NO
3–. Therefore, in actual bioreactor microenvironments, the SO
42− reduction process can break through the theoretical limitations within a certain NO
3–-N concentration range due to the multiple metabolic pathways of microorganisms and can be successfully carried out. A similar phenomenon was observed in the control group (bioreactor 4), in which the sulfate reduction process continued to occur even though the influent NO
3–-N concentrations were higher. Overall, the results indicated that as long as sufficient organic matter, low DO, and high SO
42− concentrations were present in the influent, the SO
42− reduction process could take place even in the presence of NO
3– [
28]. Notably, the accumulated SO
42− reduction amount was significantly and positively correlated (
p < 0.05) with TN removal, which also suggested that the SO
42− reduction process could greatly enhance the nitrogen removal performance [
13,
14]. The possible reason could be that the regenerative Sn
2− can also be used as a new electron donor for denitrification, which realizes multi-stage utilization of pyrite. Thus, the denitrification efficiency was enhanced by improving sulfur bioavailability, and the service life of pyrite was also prolonged by the S cycling process in these biofilters [
13]. The SO
42− generation in each section within the bioreactors is presented in Fig. S3 in Appendix A and the hot zones of its generation and reduction are described in detail in Text S5. Most importantly, according to electron balance analysis (Text S5 in Appendix A), and assuming that 10% sulfide produced from the sulfate reduction process would participate in the autotrophic denitrification process [
29], the electron contribution of sulfate reduction on denitrification could be quantified as 6.59%, 11.35%, 3.20%, and 7.00% in bioreactors 1, 2, 3, and 5, respectively. Additionally, the variations in DOC concentrations within the biofilters and their hot zones of accumulation are presented in Figs. S4 and S5 in Appendix A and described in detail in Texts S6 and S7 in Appendix A.
In summary, the hot zones of the heterotrophic denitrification, SAD, and sulfate reduction processes were in the bottom, middle-lower, and upper parts of the biofilters, respectively. These results indicated that the main reaction hot zones of C, N, and S metabolism formed a certain spatial distribution without overlap, collaborating to ensure the high efficiency and long-term stability of mixotrophic denitrification processes.
3.2. Spatial dynamics of microbial community composition within the biofilters
As mentioned above, the main reaction hot zones of C, N, and S metabolism formed a certain spatial distribution along the water flow direction. A heatmap of the top 50 species at the phylum and genus levels for each bioreactor at different heights is presented in
Fig. 2(a), and a detailed discussion at the phylum level is provided in Text S8 in Appendix A. Briefly, metagenomics sequencing revealed new phyla in the pyrite-sawdust composite-based biofilters compared to 16S rRNA sequencing, such as those involved in anaerobic fermentation for methane production and lignocellulose degradation of sawdust. The species composition and their relative abundance at different heights are presented in Fig. S6 in Appendix A. Comparing biofilm samples from the bottom and middle of bioreactor 2 revealed that the relative abundances of methanogenic (
Methanosarcina,
Methanothrix) and cellulose-degrading (
unclassified_f_Chitinispirillaceae) bacteria were higher in the bottom of the bioreactor, while those of sulfur-based denitrifying bacteria (
Thiothrix,
Thiobacillus, and
Sulfuritalea), iron-based autotrophic denitrifying bacteria (
Gemmatimonas), and SRB (
Desulfobacter) were higher in the middle of the bioreactor. The results demonstrated the spatial dynamics of the microbial community composition in the pyrite-sawdust composite-based biofilters. For example, methanogenic archaea and cellulose-degrading bacteria were mainly concentrated in the bottom of the bioreactor, while sulfur/ferric-based autotrophic/mixotrophic denitrifying bacteria and SRB were mainly concentrated in the middle of the bioreactor. Notably, the spatial distribution characteristics of the core functional bacteria were consistent with those of the reaction hot zones of C, N, and S metabolism, highlighting that microbial composition and functions were intrinsic drivers of pollutant removal.
By combining the 16S rRNA sequencing results in our previous study and those of the macro-genomic sequencing performed herein, we reconstructed the two-dimensional ecological amplitude of the core functional microorganisms from the direction of operation time and reactor height in the pyrite-sawdust composite-based biofilters based on the one-dimensional ecological amplitude of the core functional bacteria (Fig. S7 in Appendix A). As shown in
Fig. 2(b), the microbial community succession with time was mainly from anaerobic fermenting to mixotrophic denitrifying bacteria. The SAD process coupled with the sulfate reduction process was mainly concentrated in the middle of the bioreactor, while heterotrophic denitrification coupled with cellulose degradation was mainly concentrated in the bottom of the reactor, which was also accompanied by anaerobic methanogenesis.
3.3. Microbial contributions of functional genes involved in N, S, C, and Fe cycles
3.3.1. Microbial contributions of functional genes involved in the N cycle
The microbial contribution analysis (genus level) of genes related to nitrogen metabolism is displayed in Fig. S10(a) in Appendix A. The dominant genera involved in nitrogen metabolism for each sample included
unclassified_f_Rhodocyclaceae,
Methanosarcina,
Thiobacillus,
Thiothrix,
Dechloromonas,
unclassified_f_Chitinispirillaceae,
Desulfobacter, and
Sulfuritalea. Among them,
unclassified_f_Rhodocyclaceae,
Thiobacillus, and
Thiothrix were mainly involved in denitrification and DNRA processes;
Dechloromonas and
Sulfuritalea were mainly involved in the denitrification process;
Methanosarcina and
unclassified_f_Chitinispirillaceae were mainly involved in nitrogen fixation; and
Desulfobacter was mainly involved in nitrogen fixation and complete nitrification. Thus, in addition to
Dechloromonas, which is a heterotrophic denitrifying bacteria [
30], other genera involved in the denitrification process were sulfur-based denitrifiers, which once again demonstrated the dominance of sulfur-based autotrophic/mixotrophic denitrification in the pyrite-sawdust composite-based biofilters. Although previous studies have only focused on the denitrification capability of
Thiobacillus and
Thiothrix [
13,
31,
32], the macrogenomic sequencing results indicated that
Thiobacillus and
Thiothrix were also involved in the DNRA process. Furthermore, methanogenic archaea and cellulose-degrading bacteria also participated in the nitrogen fixation process, and SRB participated in the complete nitrification process. The above results provide new information about the diversity of microbial functions related to nitrogen metabolism.
3.3.2. Microbial contributions of functional genes involved in the S cycle
The microbial contribution analysis of genes related to sulfur metabolism is displayed in Fig. S10(b) in Appendix A. The dominant genera involved in sulfur metabolism for each sample included
unclassified_f_Rhodocyclaceae,
Methanosarcina,
Thiobacillus,
Sulfuritalea,
Thiothrix,
Chlorobaculum, and
Dechloromonas. Among them,
unclassified_f_Rhodocyclaceae,
Thiobacillus,
Sulfuritalea, and
Thiothrix were mainly involved in sulfur oxidation and dissimilatory sulfate reduction, and
Dechloromonas was mainly involved in sox sulfur oxidation. Notably,
Dechloromonas, a heterotrophic denitrifying bacteria, can also be involved in the sulfur oxidation process. Thus, the results indicated that
Dechloromonas might be a heterotrophic sulfur-oxidizing denitrifying bacteria (h-soNRB). In recent years, h-soNRB, representing a novel pathway for sulfur oxidation, has been extensively reported in mixotrophic denitrification systems. For instance, Chen et al. [
33] isolated and identified the heterotrophic sulfur-oxidizing bacterium
Pseudomonas C27 from mixed-culture denitrification sludge, which demonstrated the capacity for sulfide oxidation while utilizing organic matter for denitrification metabolism. Moreover, to verify that
Dechloromonas was a heterotrophic sulfur-oxidizing denitrifying bacteria, we have also performed genome binning of each metagenome to obtain the high-quality MAGs and annotate these MAGs (
Figs. 3(e) and
3(f)). The detailed analysis was described in Text S9 in Appendix A. Briefly, we demonstrated that MAG356 (
Dechloromonas genus) simultaneously possessed the function of sulfur oxidation, denitrification, and organic carbon degradation (Figs. S8 and S9 in Appendix A), which could speculate that MAG356 (
Dechloromonas genus) might belong to a heterotrophic sulfur-oxidizing denitrifying bacteria.
Herein, the presence of h-soNRB (Dechloromonas) detected by macrogenomic sequencing provides insight into the core functional bacteria in pyrite-sawdust composite-based biofilters. In addition, the dominant sulfur-based autotrophic denitrifying bacterial genera (Thiobacillus, Thiothrix, and Sulfuritalea) have the capacity for sulfate reduction, which may explain the bioreactors’ active sulfate-reducing process. Additionally, the microbial contribution of functional genes related to C degradation, under the annotation of fatty acid degradation metabolism genes, is displayed in Fig. S10(d) and is described in detail in Text S10 in Appendix A.
3.3.3. Microbial contributions of functional genes involved in the Fe cycle
The microbial contribution analysis of genes related to Fe metabolism is displayed in Fig. S10(c) in Appendix A. The dominant genera involved in the Fe cycle for each sample included
Methanosarcina,
unclassified_f_Rhodocyclaceae,
Thiobacillus,
Sulfuritalea, and
Gemmatimonas. Among them,
Methanosarcina was mainly involved in Fe
2+ transport and pumping (
feoA/
feoB), while
unclassified_f_Rhodocyclaceae,
Thiobacillus, Sulfuritalea, and
Gemmatimonas were mainly involved in Fe
2+/Fe
3+ transport permeation (
afuA/
B/
C). These results suggested that methanogenic archaea and sulfur-based autotrophic denitrifying bacteria also have Fe redox potential. Specifically, to verify that
Gemmatimonas genus was an autotrophic denitrifying bacterium, we have also further analyzed these MAGs (
Figs. 3(a) and
3(b)). The detailed analysis was described in Text S9 in Appendix A. Briefly, we demonstrated that MAG400 (
Gemmatimonas genus) simultaneously possessed the function of Fe metabolism, denitrification, and carbon fixation (Fig. S11 in Appendix A), which could speculate that MAG400 (
Gemmatimonas genus) might belong to an autotrophic denitrifying bacterium. Therefore, the results directly demonstrated that
Gemmatimonas might be a Fe-based autotrophic denitrifying bacteria, thereby verifying the existence of the Fe-based autotrophic denitrification pathway in the pyrite-sawdust composite-based biofilters.
In summary, this study deciphered the dominant Fe-related functional genes and Fe-metabolizing bacteria in the pyrite-sawdust composite-based biofilters via metagenomics sequencing. In addition, we constructed the Fe migration transformation pathway coupled with nitrogen and phosphorus removal processes in the pyrite-based mixotrophic systems according to the dominant Fe-related functional genes (
Fig. 3(c)). Briefly, Fe
2+ released from pyrite first diffuses into the periplasmic layer of the cell with the assistance of
troA iron transporter protein, and then transfers to the cytoplasmic membrane via the
feoA/
feoB Fe
2+ transporter protein system, and the Fe
2+ on the cytoplasmic membrane is oxidized to Fe
3+ by the relevant enzymes [
34]. The detailed process was described in Text S11 in Appendix A.
3.4. Differential analysis of metabolic pathways involved in N, S, and C cycles
As mentioned above, the dominant genera and functions in the pyrite-sawdust composite-based biofilters were resolved via macrogenomic sequencing. Significant differences in functional genes between the experimental and control groups were marked in the pathway map to excavate significantly upregulated genes in the pyrite-based mixotrophic systems. Subsequently, we reconstructed the metabolic network of the N, S, and C cycles in the pyrite-sawdust composite-based biofilters.
3.4.1. Differential analysis of N metabolism pathways
The differences in nitrogen-based gene abundance in the experimental (EG_2_M) and control (CG_4_M) groups are presented in
Fig. 4(a), and the nitrogen cycle was reconstructed based on the differential genes. As shown in
Fig. 4(a), the relative abundances of nitrogen fixation functional (
nifH), denitrification functional (
napA,
norB, and
nosZ), DNRA functional (
nirB), and anammox functional (
hzs and
hdh) genes were higher in the experimental group than in the control group, indicating the occurrence of active denitrification, DNRA, and anammox processes, which was attributed to the addition of pyrite. Therein, DNRA and anammox functional genes were active but present in low abundance, suggesting that partial nitrate removal may have also been achieved by coupling the DNRA and anammox processes, in addition to the mainstream denitrification process [
35].
Overall, the addition of pyrite not only promoted the mainstream denitrification process but also enhanced the coupling of DNRA and anammox processes as an auxiliary nitrogen removal pathway. The rare but critical anammox process was revealed in pyrite-based mixotrophic systems.
3.4.2. Differential analysis of S metabolism pathways
The differences in sulfur-based gene abundance in the experimental (EG_2_M) and control (CG_4_M) groups are presented in
Fig. 4(b), and the sulfur cycle was reconstructed based on the differential genes. As shown in
Fig. 4(b), S oxidation and reduction genes were dominant in the experimental group, which was consistent with the SO
42− concentrations increasing and then decreasing in the experimental groups. For the S oxidation process, the dominant functional genes were
sor,
dsrA/
B,
sat, and
aprA/
B, which belong to the sulfide quinone oxidoreductase (SQR) pathway [
36]. Therein, sulfide is oxidized to intermediate S
0 by the
fccA/B , then oxidized to SO
32− by the
sor, and finally oxidized to SO
42− by the
cysB/
sat [
14].
As for the S reduction process, the dominant sulfate-reducing genes in the experimental group (
sat,
aprA/B, and
dsrA/B) were mainly involved in the dissimilated sulfate reduction process, while those in the control group (
cysHJI and
sir) were mainly involved in the assimilated sulfate reduction process. This finding suggested that only the dissimilatory sulfate reduction process can continuously provide S
2− as a new electron donor, thus strengthening the nitrogen removal capacity of the experimental groups. Further, the formation of the end product derived from SO
42− reduction, elemental sulfur (S
0), was mainly driven by Fcc enzymes, encoded by the
fccA/B functional gene, in the experimental groups. However, the stability of S
0 as a new denitrification electron donor is questionable as it can only maintain thermodynamic stability under very limited environmental conditions (−0.05 V < Eh < 1.5 V) [
37]. A recent study reported that biological sulfur (Bio-S
0) can be encapsulated by an extracellular polymeric substance (EPS), making its surface hydrophilic and slowing down its conversion to the crystalline state [
37]. Therefore, it was speculated that EPS could inhibit the aging and crystallization of Bio-S
0, ensuring its potential as a new electron donor for sulfur-based denitrification.
To test the above conjecture, we performed XPS to characterize the elemental distribution and valence state of sulfur on the biofilms of the pyrite-based composites (Fig. S12 in Appendix A), The presence of S0 was detected on the biofilms, which confirmed the ability of EPS to protect the stability of S0. In summary, the differential analysis of functional genes indicated that the end product of the dissimilatory sulfate reduction process was Bio-S0 in the pyrite-sawdust composite-based biofilters, which was secreted into the extracellular layer and encapsulated with EPS, making its surface hydrophilic and slowing its transformation to the crystalline state. Thus, the potential of Bio-S0 as a new electron donor was guaranteed, thereby strengthening the system denitrification process.
3.4.3. Differential analysis of C metabolism pathways
The glycolytic process and tricarboxylic acid (TCA) cycle are two important C metabolism pathways during sawdust degradation. The differences in the abundance of genes related to the glycolytic process and TCA cycle in the experimental (EG_2_M) and control (CG_4_M) groups are shown in
Fig. 4(c) and are described in detail in Fig. S13 and Text S12 in Appendix A. In summary, the pyrite-sawdust composite-based biofilters positively stimulated the activities of key enzymes involved in the glycolytic process and TCA cycle, producing abundant ATP and NADH, which not only enhanced the metabolic activities of the core functional bacteria but also indirectly facilitated nitrogen removal. Thus, high ATP and NADH production was one of the internal driving forces for efficient and stable nitrogen removal performance by the pyrite-sawdust composite-based biofilters.
3.5. Untargeted metabolomic analysis and microbe-metabolite network
3.5.1. Differential analysis of metabolites and metabolic pathways
The metabolic activities of microorganisms play a key role in maintaining the mass transfer efficiency and nitrogen removal performance of mixotrophic systems [
38]. Metabolite information was obtained through metabolomic analysis of the experimental and control groups. In total, 25 292 and 31 876 differential metabolites were found in negative and positive modes, respectively. Among them, 252 and 889 differential metabolites were annotated in negative and positive modes, respectively (Figs. S14(a) and (b) in Appendix A). In addition, the top 20 differential metabolites with the largest fold change (FC) and variable importance on projection values in negative and positive modes are presented in Figs. S14(c) and (d) in Appendix A, respectively. Among upregulated differential metabolites, 1-(5'-phosphoribosyl)-5-amino-4-imidazolecarboxamide (AICAR), picrotin, alpha-bixin, 9-
cis-retinoic acid, 5,6-dimethyl benzimidazole, and hexacosanoyl carnitine exhibited the most positive significant differences (
P < 0.05), which could be utilized as potential biomarkers in pyrite-based mixotrophic denitrification systems to explore their biological metabolic mechanisms. Further, the metabolite pathway enrichment analysis (FC > 2) of the experimental and control groups based on the KEGG database is presented in Figs. S14(e) and (f) in Appendix A, respectively. Most of the metabolites were classified as being involved in amino acid (tyrosine, tryptophan, histidine, and phenylalanine metabolism), nucleotide (riboflavin and purine metabolism), energy (retinol metabolism and cofactor biosynthesis), and carbohydrate (fatty acid degradation) metabolism. The results implied that higher metabolic activity was induced in the pyrite-sawdust composite-based biofilters.
To investigate the key metabolic pathways in more detail, the upregulated differential metabolites were linked to metabolic pathways by searching against the KEGG database. As shown in
Fig. 5(a), the abundance of indole-3-acetamide, imidazole-4-acetaldehyde, and 3-(3-hydroxyphenyl) propanoic acid, which were involved in tryptophan, histidine, and phenylalanine metabolism, respectively, were significantly changed. Active metabolism of these three amino acids could contribute to protein synthesis for bacterial growth and reproduction. The product of active phenylalanine metabolism demonstrated the unique phenylpropane structure of lignin, derived from sawdust in the biofilters [
39].
Another key metabolite, 5,6-dimethyl benzimidazole, involved in purine, riboflavin, and porphyrin metabolism, was also significantly upregulated. Specifically, bacteria can utilize 5,6-dimethylbenzimidazole as a precursor in cobalamin (vitamin B
12) biosynthesis and secrete vitamin B
12 to promote central microbial metabolic processes, such as C and N compound catabolism, nucleotide metabolism, and biosynthesis of products like methionine [
18]. As a typical example of cofactors, vitamin B
12 is the most well-known member and is important in the metabolic exchanges of microbial communities [
40]. Further, to verify vitamin B
12’s involvement in sulfur or nitrogen metabolism, the major biosynthesis intermediates (uroporphyrinogen III, Succinyl-CoA, and 5,6-dimethyl benzimidazole) [
41] and metabolic intermediates (
S-adenosyl-
L-methionine and
S-adenosyl-
L-homocysteine)[
42] of vitamin B
12 in these biofilters have also been detected via LC-MS/MS analysis and their relative abundance was shown in Fig. S15 in Appendix A. As shown in Fig. S15, the uroporphyrinogen III, succinyl-CoA, 5,6-dimethyl benzimidazole and
S-adenosyl-
L-methionine in bioreactor 2 had higher abundance compared with the control groups (bioreactor 4), it was directly demonstrated that active biosynthesis and utilization of vitamin B
12, which could enhance the communication of nutrients or metabolic products via vitamin B
12 in the pyrite-sawdust composite-based biofilters. Wang et al. [
18] also suggested that microbes with functional genes involved in the S cycling process may benefit from an increase in the abundance of vitamin B
12 producers.
The third key mechanism involved in the adipocytokine signaling pathway revolved around the utilization of 9-cis-retinoic acid, indicating that 9-cis-retinoic acid may play a vital role in fatty acid metabolism, further suggesting the involvement of the fatty acid metabolism pathway as mentioned above. In summary, vitamin B12 and tryptophan might be key metabolites in realizing synergistic promotion of autotrophic and heterotrophic denitrification with metabolic change.
3.5.2. Correlation and network analysis between metabolites and microbes
To identify the relationships between differential genera and metabolites, we used Spearman’s rank correlations and hierarchical clustering of differentially abundant genera (Fig. S16 in Appendix A) and metabolites, as described in detail in Fig. S17 and Text S13 in Appendix A. After confirming the mutualistic relationship between microbes and metabolites, we constructed the microbe–metabolite co-occurrence network, a direct and powerful method for revealing the underlying cross-feeding relationships between the different microbes and metabolites. As shown in
Fig. 5(b), 40 genera were connected to 40 metabolites via 441 links, including 245 positive and 196 negative links, with an average of ten links per node. It was observed that three metabolites (picrotin, 9-
cis-retinoic acid, and 5,6-dimethyl benzimidazole) dominated by positive correlations exhibited positive links (78%) with a wide range of bacteria, and positive links primarily with Bacteroidetes, Euryarchaeota, Proteobacteria, and Chloroflexi. Two metabolites (AICAR and hexacosanoyl carnitine) dominated by negative correlations exhibited negative links (94%) with a wide range of bacteria, and negative links primarily with Proteobacteria and Chloroflexi. Four genera (
g_unclassified_o_Bacteroidales,
g_unclassified_o_Anaerolineales,
Thiobacillus, and
Desulfobacter) dominated by positive correlations exhibited positive links (98%) with a wide range of metabolites, while two genera (
Geobacter and
Gemmatimonas) dominated by negative correlations exhibited positive links (95%) with a wide range of metabolites. Combined with the relative abundances of these genera, the results indicated that cross-feeding occurred between these species with low abundance via amino acid, nucleotide, energy, and carbohydrate metabolism during mixotrophic denitrification processes [
43]. Interestingly, single links between microbes and metabolites, such as between
Thiothrix and phaseollidin and between
unclassified_f_Rhodocyclaceae and L-aspartic acid, were observed in the network. On the one hand, the results provided direct evidence of microbial metabolites through simple biological pathways, such as,
Thiothrix is only responsible for the formation of phaseollidin, and
unclassified_f_Rhodocyclaceae is only responsible for L-aspartic acid production. On the other hand, combined with the high relative abundances of
Thiothrix and
unclassified_f_Rhodocyclaceae in the mixotrophic denitrification systems, the results indicated that no cross-feeding occurred between dominant mixotrophic denitrifiers, and it was speculated that the autotrophic and mixotrophic denitrification process is relatively independent and less affected. Generally, generalist microbes are not masters of their habitats and are present in low abundance, whereas specialists adapt to become dominant within their habitats under stable conditions, thus reflecting a trade-off wherein specialists gain local dominance at the expense of ecological versatility [
44]. Thus, the results indicated that dominant bacteria (
Thiothrix and
unclassified_f_Rhodocyclaceae) were specialists with low cross-feeding metabolism, whereas rare species in the pyrite-based mixotrophic denitrification systems (
Thiobacillus and
Desulfobacter) were generalists with complex cross-feeding metabolism. Wang et al. [
45] suggested that generalist species can tolerate a wider range of environments and have a more flexible metabolism than specialist species . The constructed microbe–metabolite network directly provides new insight into the metabolic exchanges between microbes and metabolism in mixotrophic denitrification systems. Further proof may come through in-depth techniques such as microbial pure culture feeding with these special metabolites in the future.
3.6. Coupling mechanism of C, N, S, and Fe metabolism and electron transfer pathway
Combined with the species, function, and differential gene annotations of the metabolic pathways by macrogenomic sequencing and untargeted metabolomics analysis, the biological mechanisms of N, S, C, and Fe transformations and their electron transfer modes in the pyrite-sawdust composite-based biofilters were further elaborated, as shown in
Fig. 6.
In terms of C metabolism, the three main pathways for small-molecule organic matter (CH2O) released from fatty acid metabolism were driven by the metabolite 9-cis-retinoic acid after the hydrolysis of sawdust by cellulose-degrading bacteria in the pyrite-sawdust composite-based biofilters, as follows: ①Most CH2O enters the glycolytic and TCA cycling processes, which produces large amounts of ATP and NADH for microbial growth and metabolism. Additionally, the electron produced from NADH is utilized to reduce NO3–-N to N2 for heterotrophic denitrification; ②some CH2O enters into the sulfur cycling process, providing an organic carbon source for h-soNRB (Dechloromonas) and SRB (Desulfobacter); and ③the remaining CH2O is utilized by methanogenic archaea (Methanosarcina and Methanothrix) to produce CH4 via anaerobic methanogenesis in the bottom of the biofilters.
In terms of S metabolism, the chemical source S2− is released from the dissolution of pyrite, which is utilized by sulfur-based autotrophic/mixotrophic denitrifying bacteria (Thiothrix and Thiobacillus) and h-soNRB (Dechloromonas) to SO42− via the SQR pathway. Therein, the electrons released from S2− oxidation are utilized to reduce NO3–-N to N2 for SAD. Further, SO42− is captured by SRB, which reduces SO42− to biogenic S0 via the Fcc enzymes and secretes it to the EPS of the biofilms. Thus, an S cycling process is formed, which continues to provide new electron donors for NO3–-N reduction, enhancing mixotrophic denitrification processes.
In terms of Fe metabolism, Fe2+ released from the dissolution of pyrite is utilized by Fe-based autotrophic denitrifying bacteria (Gemmatimonas) to form Fe3+ in the periplasmic layer, and the released electrons are also utilized to reduce NO3–-N for Fe-based autotrophic denitrification. The formed Fe3+ is pumped out of the cell to react with PO43–, thereby achieving phosphorus removal.
Finally, the electrons released from C, S, and Fe oxidation are transferred to NO
3–-N reduction, and most are mainly utilized in denitrification processes as the dominant denitrification pathway, including S
2–/S
0-based autotrophic, fermentation acetic acid production–heterotrophic, and Fe(II)-based autotrophic denitrification. Additionally, some of the electrons are utilized in the coupling of DNRA and anammox processes as an auxiliary pathway for systemic nitrogen removal. Especially, combined with the results of spatial dynamics of microbial community composition within the biofilters (
Fig. 2(b)), it was speculated that synergy between heterotrophic denitrification and cellulose degradation mainly occurred under the high nitrate and DOC concentrations, while the synergy between S
2–/S
0–based autotrophic and Fe(II)-based autotrophic denitrification was mainly occurred under the moderate nitrate and DOC concentrations for extreme nitrogen removal.
4. Environmental significance and future perspective
Pyrite-driven mixotrophic denitrification is considered a promising technology for wastewater treatment [
11,
15,[
46], [
47], [
48]]. However, the complex core bacteria and multiple nitrogen removal pathways have yet to be clarified. Herein, the constructed two-dimensional ecological amplitude of the core functional microorganisms from the direction of operation time and reactor height in pyrite-sawdust composite-based biofilters revealed that the core functional bacteria in the bottom, middle-lower, and upper parts of the bioreactors were cellulose-degrading bacteria combined with heterotrophic denitrifying bacteria, S-based mixotrophic denitrifying bacteria, and SRB, respectively. The dominant nitrogen removal pathway included S
2–/S
0-based autotrophic, fermentation acetic acid production-heterotrophic, and Fe(II)-based autotrophic denitrification, while an auxiliary pathway comprised the coupling of DNRA and anammox processes. Additionally, the untargeted metabolomic analysis indicated that vitamin B
12 and tryptophan might be key metabolites for realizing the synergistic promotion of autotrophic and heterotrophic denitrification with metabolic change. In summary, the introduction of pyrite–sawdust composites led to heightened metabolic intensity, fostering enzyme synthesis and bolstering electron transfer processes. This, in turn, optimized the microbial community structure and reinforced the dominance of the SAD process. Collectively, these effects enhanced the material’s metabolic capacity and its ability to change, while concurrently reducing its reliance on external carbon sources. Future research will apply pyrite-sawdust composites to treat actual wastewater, delving deeper into the parameters and intrinsic mechanisms using metatranscriptomics and metaproteomics to study the activities of important genes and enzymes at the expression level, thereby revealing their contributions to nitrogen species removal.
CRediT authorship contribution statement
Qi Zhou: Writing – original draft, Visualization, Software, Data curation, Conceptualization. Weizhong Wu: Writing – review & editing, Supervision, Resources, Funding acquisition, Conceptualization. Jianlong Wang: Writing – review & editing, Supervision, Funding acquisition, Conceptualization.
Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgments
This work was supported by the National Natural Science Foundation of China (42407077 and 52070001), and the Postdoctoral Fellowship Program of CPSF (GZC20240815).
PDF
