Anammox Bacteria Are Potentially Involved in Anaerobic Ammonium Oxidation Coupled to Iron(III) Reduction in the Wastewater Treatment System

Anaerobic ammonium oxidation coupled to nitrite reduction (termed as Anammox) was demonstrated as an efficient pathway to remove nitrogen from a wastewater treatment system. Recently, anaerobic ammonium oxidation was also identified to be linked to iron(III) reduction (termed Feammox) with dinitrogen, nitrite, or nitrate as end-product, reporting to enhance nitrogen removal from the wastewater treatment system. However, little is known about the role of Anammox bacteria in the Feammox process. Here, slurry from wastewater reactor amended with ferrihydrite was employed to investigate activity of Anammox bacteria in the Feammox process using the 15N isotopic tracing technique combined with 16S rRNA gene amplicon sequencing. A significantly positive relationship between rates of 15N2 production and iron(III) reduction indicated the occurrence of Feammox during incubation. Relative abundances of Anammox bacteria including Brocadia, Kuenenia, Jettenia, and unclassified Brocadiaceae were detected with low relative abundances, whereas Geobacteraceae dominated in the treatment throughout the incubation. 15N2 production rates significantly positively correlated with relative abundances of Geobacter, unclassified Geobacteraceae, and Anammox bacteria, revealing their contribution to nitrogen generation via Feammox. Overall, these findings suggested Anammox bacteria or cooperation between Anammox bacteria and iron(III) reducers serves a potential role in Feammox process.


INTRODUCTION
Feammox [anaerobic ammonium oxidation coupled to iron(III) reduction] is a pathway of nitrogen cycling identified recently and makes a contribution to nitrogen loss in various environments, such as terrestrial (e.g., wetland, tropical rainforest, and paddy soils) and aquatic ecosystem (e.g., freshwater and marine) in addition to denitrification, co-denitrification, and anaerobic ammonium oxidation (Clement et al., 2005;Yang et al., 2012;Ding et al., 2014Ding et al., , 2017Ding et al., , 2019Huang and Jaffé, 2014;Zhou et al., 2016). Previous studies showed that iron(III)-reducing bacteria such as Anaeromyxobacter, Pseudomonas, Geobacter, Desulfosporosinus, Dechloromonas, and Geothrix always dominated in the Feammox "pool" (Zhou et al., 2016;Li et al., 2019); however, these iron(III) reducers prefer to utilize organic carbon for iron(III) reduction. Only a minor part of iron(III) reduction (0.4-6.1%) by these bacteria is estimated to be associated with Feammox in the natural or artificial environments (Yang et al., 2012;Ding et al., 2014Ding et al., , 2017Ding et al., , 2019Li et al., 2015;Zhou et al., 2016). As a replacement, 15 NH 4 + is generally employed to trace the occurrence of Feammox; however, it is hard to directly identify the Feammox microbes through DNA or RNA stable isotope probing because the Feammox process results in conversion of 15 NH 4 + to 15 N 2 / 15 NO 3 − / 15 NO 2 − but not assimilation of 15 NH 4 + into DNA and RNA. As a result, "Feammox microbes" are difficult to be directly captured from the incubation. Sawayama (2006) has reported that the possibility of Feammox happened in a wastewater treatment system, which is recently expected to become another important way to remove nitrogen from the sludge of wastewater in addition to Anammox. Anammox is another anaerobic ammonium oxidation pathway that used nitrite as an electron acceptor with forming dinitrogen (Rikmann et al., 2014). Furthermore, the addition of iron(III) oxides is indicated to increase the efficiency of nitrogen removal in the system (Chen et al., 2014;Wang et al., 2016;Yin et al., 2019). These discoveries further suggest the occurrence of Feammox in the Anammox-based nitrogen removal system of a wastewater treatment reactor. However, the Feammoxinvolving microorganisms in the wastewater treatment system is still under-characterized. Anammox bacteria, including members of the Planctomycetales such as Brocadia, Kuenenia, Anammoxoglobus, Jettenia, Anammoximicrobium moscowii, and Scalindua (Rikmann et al., 2014), possesses a versatile metabolism involved in the utilization of diverse electron donors (e.g., NH 4 + and propionate) and acceptors (e.g., NO 2 − , NO 3 − , and SO 4 2− ) (Strous et al., 2002;Cervantes et al., 2009;Rikmann et al., 2014;Rios-Del Toro and Cervantes, 2016;Rios-Del Toro et al., 2018). Inspired by these recent findings, we hypothesized that Fe(III) can play as terminal electron acceptors in anaerobic ammonium oxidation mediated by Anammox bacteria in the sludge from a wastewater treatment reactor.
In order to verify our hypothesis, sludge from a wastewater treatment reactor, which has been demonstrated to employ the Anammox process to remove nitrogen (Zhang et al., 2012), was used in this study. Through the 15 NH 4 + -based isotopic tracing technique with 16S rRNA gene Illumina sequencing, we amended the sludge with ferrihydrite to investigate whether (1) Feammox occurred in the Anammox enrichment and (2) the Anammox organisms were involved in Feammox process.

Experimental Procedures
The sludge was obtained from a wastewater treatment reactor (Zhang et al., 2012). The dominant electron donor and acceptor were NH 4 + and NO 2 − in the wastewater treatment reactor, respectively, which has been established as a stable, completely autotrophic nitrogen removal process over nitrite (Canon). Theoretically, the Canon is a process combining partial nitrification with Anammox within a single reactor, which has been shown to be a cost-efficient autotrophic process for nitrogen removal, as it has no need for a carbon source and has low requirement for oxygen (Zhang et al., 2012). The characteristic of the sludge is detailed in Supplementary Table 1. The Feammox experiment was initiated by inoculating 10% (v/v) sludge into 20 ml basal medium and incubated at 30 • C in the dark. The basal medium (pH 6.8-7.2) consists of MgCl 2 ·6H 2 O (0.4 g L −1 ), CaCl 2 ·H 2 O (0.1 g L −1 ), 14 NH 4 Cl (0.027 g L −1 ), KH 2 PO 4 (0.6 g L −1 ), 1 ml L −1 vitamin solution (Lovley and Phillips, 1988), 1 ml L −1 trace element solution (Lovley and Phillips, 1988), 30 mmol L −1 bicarbonate buffer, and 4 mmol L −1 Fe(III). The headspace of the serum vials was flushed with ultrapure helium. Ferrihydrite was synthesized as previously described (Kappler et al., 2014) and used as the Fe(III) source. The basal medium and ferrihydrite were autoclaved (120 • C for 20 min) before use, and the vitamin solution and trace element solution were filtrated with a 0.22-µm filter from the stock solutions. In order to enrich the Feammox-associated microbial population, the cultures were anaerobically transferred (10%, v/v) to fresh medium for three generations once the Fe(III) was used up.
For the labeled experiment, 14 NH 4 Cl was replaced by 15 NH 4 Cl ( 15 N, 99.14%; Cambridge Isotope Laboratories, Andover, MA, United States) to prepare the fresh medium. In brief, aliquots (2 ml) of the Feammox enrichment were centrifuged and washed three times with sterile deionized water before inoculating into the 20 ml of fresh 15 NH 4 Cl-labeled medium. Three treatments were set up: (1) NH 4 + : Feammox enrichment was inoculated in the 20 ml of 15 NH 4 Cl-added basal medium amended without ferrihydrite; (2) Fe(III): Feammox enrichment was inoculated in the 20 ml of ferrihydritecontaining basal medium amended without 15 NH 4 Cl; and (3) Fe(III) + NH 4 + : Feammox enrichment was inoculated in the 20 ml of basal medium amended with both ferrihydrite and 15 NH 4 Cl. The final concentrations of 15 NH 4 Cl and ferrihydrite were 0.5 and 4 mmol L −1 in the treatments, respectively. The number of serum vials for each treatment was 4, 4, and 24, respectively. All the treatments were incubated at 30 • C in a dark under anaerobic condition.

Chemical Analysis
Ferrous iron and total iron were determined as described previously (Kappler et al., 2005). Briefly, Fe(II) was determined by anaerobically transferring 100 µl of culture suspension with a syringe into 900 µl of 40 mmol L −1 sulfamic acid and incubating for 1 h at room temperature. Total Fe was extracted using a mixture of 20 mmol L −1 hydroxylamine hydrochloride and 20 mmol L −1 sulfamic acid (v:v = 1:1) (Klueglein et al., 2015). A 100-µl extract was then added with 1 ml ferrozine solution (1 g ferrozine in 50 mmol L −1 HEPES buffer, pH 7) to generate the ferrous complex, which was quantified at 562 nm UV/Vis spectrometer. Change in Fe(II) concentrations between two given time points (23 days) were used to calculate iron(III) reduction rates.
For the 15 N-N 2 analysis, vials were shaken vigorously to equilibrate the dissolved phase with gaseous phase, and 1 ml of gas samples was collected from the headspace using gastight syringes and then injected into 12-ml glass vials (Exetainer; Labco, Lampeter, United Kingdom). The gas samples were taken on days 1, 4, 8, 10, 12, 14, 18, and 22, respectively. 30 N 2 and 29 N 2 concentrations were calculated by multiplying the moles of total N 2 in the headspace by the 30 N 2 and 29 N 2 mole fractions (Zhou et al., 2016). The total N 2 and N 2 O concentration in the headspace was measured using a robotized system coupled to a gas chromatograph (Agilent Technologies, Santa Clara, CA, United States) as previously described (Zhou et al., 2016). The mole fractions of 30 N 2 and 29 N 2 were determined by isotope ratio mass spectrometry (IRMS; Thermo Finnigan Delta V Advantage, Bremen, Germany) coupled with Gasbench II, respectively (Zhou et al., 2016). After gas collection, the remaining cultures were immediately centrifuged at 14,000 × g for 15 min and the pellets were used for DNA extraction. The resulting supernatant was filtered through 0.22-µm filters and then subjected to measurement of NH 4 + , NO 2 − , and NO 3 − concentrations by ion chromatography (Dionex ICS-3000 system; Diones, Sunnyvale, CA, United States). All the liquid was sampled in the anaerobic glovebox (Shel Lab Bactron IV; Shel Lab, Cornelius, OR, United States) to avoid chemical oxidation. pH was analyzed using a dual-channel pH-ion-conductivitydissolved oxygenmeter (X60; Thermo Fisher Scientific, Carlsbad, CA, United States) in the anaerobic glovebox.

DNA Extraction and Illumina Sequencing
DNA was extracted using FastDNA Spin Kit (MP Biomedical, Illkirch-Graffenstaden, France) according to the manufacturer's protocol and stored at −20 • C for the molecular analyses. Since the biomass in the treatments amended with only Fe(III) or NH 4 + was extremely low, there was enough DNA extracted from these treatments.
To investigate the bacterial community structures and compositions, the V4-V5 region of bacterial was amplified using the DNA extracted from the samples in the treatment of Fe(III) + NH 4 + as template. The amplicons were purified, quantified, pooled, and then sequenced on an Illumina Miseq PE 250 platform (Novogene, Beijing, China) (Zhou et al., 2016). The forward primer was 515F (5 -GTGCCAGCMGCCGCGG-3 ), and the reverse primer consisted of a 6-bp barcode and 907R (5 -CCGTCAATTCMTTTRAGTTT-3 ) (Ren et al., 2014). Quantitative Insights into Microbial Ecology toolkit-version 1.9.0 (QIIME) was used to process and analyze sequences as previously described (Su et al., 2015). After removal of low-quality or ambiguous reads, operational taxonomic units (OTU) were determined at 97% similarity level using UCLUST clustering in accordance with the online instruction of QIIME for open-reference OTU pick, definition, and determination (Wang et al., 2007). The representative sequences of each OTU were assigned to taxonomy using an RDP classifier (Version 11). 1 1 http://rdp.cme.msu.edu

Quantitative PCR
The abundance of relevant genes and microbial organisms, including bacterial 16S rRNA gene, Geobacteraceae spp., Acidimicrobiaceae spp. (the reported potential microbe responsible for Feammox) (Huang and Jaffé, 2014), hzsB (hydrazine synthase), nirS (nitrite reductase), and nosZ (nitrous oxide reductase) was analyzed with a real-time PCR Detection System (Roche 480; Roche, Indianapolis, IN, United States). The primer sets and thermal cycles were detailed in Supplementary Table 2. The 20-µl qPCR reaction contained 10 µl 2 × TransStart R Top Green qPCR SuperMix (AQ131; Transgen Biotech, Beijing, China), 0.25 µM each primer, 0.8 µl bovine serum albumin (BSA, 20 mg ml −1 ), and 2 µl of fivefold diluted DNA as a template. The standard curve was obtained using 10-fold serial dilutions of plasmid DNA with targetgenes. Three non-template controls were carried out for each quantitative assay. A melting curve for each reaction showed that only one special peak was detected. Only the reactions with efficiencies between 90 and 110%, and standard curves with correlation coefficient above 0.99 were employed in this study.

Statistical Analyses
Analysis of variance (ANOVA) and Pearson correlation analysis were conducted by SPSS 18.0 (SPSS Inc., Chicago, IL, United States) and Origin 9.0 (OriginLab, Northampton, MA, United States). Statistical significance was performed using Duncan's multiple range test and denoted at p < 0.05. The differences of the bacterial communities were analyzed by nonmetric multidimensional scaling (NMDS) based on weighted UniFrac dissimilarity among samples, which was represented by the ordination axes (Tunney et al., 2013).

Data Accessibility
The 16S rRNA gene sequences have been deposited in GenBank with accession number SRP116169.

Iron(III) Reduction and Changes of N Species in the Enrichment
In the Fe(III) + NH 4 + treatment, Fe(II) increased up to 1.83 ± 0.010 mmol L −1 after 22-day anaerobic incubation (Figure 1A). In comparison, iron(III) reduction was not detected in the treatments only amended with Fe(III) or NH 4 + after the 22-day incubation ( Figure 1A).
Significant (p < 0.05) accumulation of 30 N 2 was detected in the Fe(III) + NH 4 + treatment (0.36 µmol L −1 ) compared to that in the treatment of Fe(III) (0.064 µmol L −1 ) or NH 4 + (0.080 µmol L −1 ) during the incubation ( Figure 1B   , and N 2 (C) production in treatments of NH 4 + , Fe(III), and Fe(III) + NH 4 + during 22-day incubation. Error bars represent standard deviations of three replications. The different lowercase letters above the error bar denote statistically significant (p < 0.05) differences among different treatments. (D,E) The correlation between rates of iron(III) reduction and 15 N 2 production in treatments of Fe(III) + NH 4 + during 22-day incubation. "*" and "**" represent a statistically significant relationship between rates of iron(III) reduction and 15 N 2 production, which is denoted at p < 0.05 and p < 0.01, respectively.

Changes in Abundances of Bacteria and the N Cycling-Relevant Genes
16S rRNA gene copy number increased up to 2.36 × 10 10 copies L −1 medium after incubation in the Fe(III) + NH 4 + treatment (Figure 2A). Also, the abundances of hzsB, nirS, and nosZ rapidly elevated in the treatment of Fe(III) + NH 4 + after 12 days (Figure 2 and Supplementary  Figure 2). Especially, the gene copy numbers were higher for the genes nirS (1.26 × 10 7 ) and nosZ (2.62 × 10 7 ) than that for the gene hzsB (7.36 × 10 6 copies L −1 medium) in the treatment of Fe(III) + NH 4 + (Figure 2A). The abundance of Geobacteraceae spp. was about 2.03 × 10 6 copies L −1 medium after 22-day incubation in the treatment of Fe(III) + NH 4 + (Figure 2B). By contrast, Acidimicrobiaceae spp. kept constantly low abundance throughout the incubation ( Figure 2B).

Shift of Bacterial Community Composition
In the initial inoculant, 42.94% of the total bacterial community were affiliated to the family of Ignavibacteriaceae, followed by unclassified Anaerolineae (23.68%) and unclassified Chlorobi (8.02%) (Figure 3). While in the treatment of Fe(III) + NH 4 + , Geobacteraceae was the family with the highest relative abundance, followed by Comamonadaceae, Pelobacteraceae, and Pseudomonadaceae (Figure 3). These four families occupied up to 96.13% of the total microbial community (Figure 3).

Shift in the Relative Abundances of Iron(III)-Reducers and Anammox Bacteria
The detected iron(III) reducers included Geobacter, Pseudomonas, Clostridium, Bacillus, Thiobacillus, unclassified Geobacteraceae, Desulfotomaculum, Desulfovibrio, Desulfobulbus, and Pelobacter in the treatment of Fe(III) + NH 4 + (Figure 4A). In comparison with the initial inoculant, the genera of Geobacter, Frontiers in Microbiology | www.frontiersin.org  Pseudomonas, Clostridium, Desulfovibrio, Desulfotomaculum, and Pelobacter were significantly enriched, while the relative abundance of Bacillus and Thiobacillus decreased in the treatment of Fe(III) + NH 4 + during the incubation (Figure 4A). Of all the iron(III) reducers, the change in relative abundances of Geobacter and unclassified Geobacteraceae were significantly (p < 0.05) correlated with the 30 N 2 and 29 N 2 production rates in the treatment of Fe(III) + NH 4 + (Figure 4B). The Anammox-relevant Planctomycetes remained with low relative abundances in the treatment of Fe(III) + NH 4 + after 22-day incubation. Dynamics of relative abundances of Anammox-relevant taxa, including unclassified Brocadiaceae, Brocadia, Kuenenia, and Jettenia, displayed a similar pattern in the treatment of Fe(III) + NH 4 + . These relative abundances of Anammox bacteria reached a peak on days 10-12 and then decreased during the incubation (Figure 4C). The relative abundances of these Anammox bacteria significantly (p < 0.05) correlated with the rates of 30 N 2 and 29 N 2 production ( Figure 4D).

DISCUSSION
Feammox is a recently identified pathway of dinitrogen generation, expecting to be applied to remove nitrogen from a wastewater treatment system. In this study, we aimed to verify the occurrence of Feammox and investigate the Feammoxassociated microbes in the sludge of a wastewater treatment system. Detection of 15 N 2 production from sludge amended with iron(III) indicated the existence of Feammox in the incubation. Anammox bacteria such as Brocadiaceae, Kuenenia, and Jettenia and iron(III) reducers including Geobacter and unclassified Geobacteraceae were found potentially involved in the Feammox process.
The Feammox incubation was established using the sludge as inoculant, which was subjected to three generations with freshly prepared medium under anaerobic condition. As a result, the major electron acceptor for anaerobic oxidation of 15 N-NH 4 + was ferrihydrite during the incubation. Previous reports suggested that Anammox can be linked to the microbial reduction of natural organic matters (Rios-Del Toro et al., 2018), likely originating from the breakdown of dead biomass, for example, carbohydrate residues including cellulose and lignin (Siemann et al., 2012). The transformation between oxidized and reduced state enables the quinone group-abundant organic matters serving as electron shuttles to mediate anaerobic oxidation of ammonium to N 2 (Westereng et al., 2015;Rios-Del Toro et al., 2018). However, the standard Gibbs free energy released from this process (NH 4 + + 1.5 quinone-NOM ox → 0.5N 2 + 1.5 quinoneH 2 -NOM red + 4H + ) ranged from 5.8 kJ mol −1 to −124.6 kJ mol −1 (Rios-Del Toro et al., 2018), greatly lower than that produced from the Feammox process [3Fe(OH) 3 + 5H + + NH 4 (Yang et al., 2012). Therefore, Anammox was probably coupled to ferrihydrite reduction during the incubation. The significant accumulation of 30 N 2 provided a solid evidence for the occurrence of Feammox in the treatment Frontiers in Microbiology | www.frontiersin.org FIGURE 4 | The relative abundances of iron(III) reducers and Anammox bacteria during incubation. The relative abundances of iron(III) reducers (A) and the correlation between the relative abundances of Geobacter/unclassified Geobacteraceae and 15 N 2 production rates (B) in the inoculant-based treatment of Fe(III) + NH 4 + during 22-day incubation. Error bars represent standard deviations of three replications. The relative abundances of Brocadiaceae (unclassified), Brocadia, Kuenenia, and Jettenia (C), and the correlation between the relative abundance of Geobacter and 15 N 2 production rates (D) in the inoculant-based treatment of Fe(III) + NH 4 + . Error bars represent standard deviations of three replication. "*" and "**" represent a statistically significant relationship between relative abundances of iron(III) reducers, Anammox bacteria, and 15 N 2 production, which is denoted at p < 0.05 and p < 0.01, respectively.
of Fe(III) + NH 4 + . Codenitrification is another potential source of 30 N 2 (Laughlin and Stevens, 2002). However, it can be ruled out in this study because other 15 N-labeled nitrogen compounds (e.g., hydrazine and amino compounds) that could reduce 15 NO 2 − and 15 NO 3 − to N 2 were not available in the culture. Under these conditions, direct N 2 production from Feammox, or Feammox-produced NO 2 − or NO 3 − followed by denitrification or Anammox are the possible pathways for 30 N 2 generation, supporting the occurrence of Feammox in the treatment of Fe(III) + NH 4 + . The positive correlation (p < 0.0001; Figures 1D,E) between Fe(III) reduction and 30 N 2 / 29 N 2 production rates further verified the existence of Feammox in the treatment of Fe(III) + NH 4 + during the anoxic incubation.  Table 3) produced from the treatment of Fe(III) + NH 4 + was far less than that of NH 4 + depleted during the incubation, indicating that the majority of NH 4 + was assimilated into microbial biomass (Tupas and Koike, 1990). The abundant genera such as Pseudomonas and Bacillus detected in this study, which are able to assimilate ammonia during their growth (Kim and Hollocher, 1982;Kanamori et al., 1989), may be the major NH 4 + consumers in the treatment of Fe(III) + NH 4 + . The molar ratio of reduced Fe (III) to the total NH 4 + oxidation was about 15.41 in the Fe(III) + NH 4 + treatment (Figure 1), which did not match the stoichiometry (ranging from 3 to 8) in the three Feammox equations (Yang et al., 2012;Ding et al., 2014;Zhou et al., 2016). This suggested that only a minor fraction of reduced Fe(III) was linked to the NH 4 + oxidation in the Fe(III) + NH 4 + treatment. According to the thermodynamic calculations, the amount of iron reduction associated with Feammox was 0.36-0.96 mmol L −1 , accounting for 19.67-52.46% of total Fe(III) reduction in the treatment of Fe(III) + NH 4 + . Thus, a majority of the reduced Fe(III) was linked to the oxidation of other substrates mediated by microorganisms. Because exogenous organic matter was not added in the treatment of Fe(III) + NH 4 + , the substrates might be organic compounds sourced from the dead biomass or extracellular secretion from the microbes. Among all the iron(III)-reducing bacteria, the family of Geobacteraceae was abundant in the treatment of Fe(III) + NH 4 + (Figure 3). The genus of Geobacter is able to reduce Fe(III) associated with organic substrate oxidation to support its growth (Lovley, 1991). The relative abundance of Geobacteraceae showed an increase on days 10 and 12, which was in agreement with the rapidly increasing copy number of Geobacter after 12-day incubation in the treatment of Fe(III) + NH 4 + (Figures 2B, 3, 4A), indicating Geobacter was the dominant genus in Geobacteraceae. The positive correlation between the abundance of Geobacter and accumulation of 15 N 2 in the treatment of Fe(III) + NH 4 + ( Figure 4B) suggested that Geobacter may exert a potential role in Feammox.
The abundance of Anammox bacteria, including Brocadiaceae, Kuenenia, and Jettenia, showed significant correlation with 15 N 2 production in the treatment of Fe(III) + NH 4 + (Figure 4D), suggesting Anammox bacteria were linked with Feammox during the incubation. Although the mechanism about the role of Anammox bacteria as Feammox players was still unknown, a variety of reports have shown their versatile metabolism. Firstly, Anammox bacteria are capable of anaerobically oxidizing ammonium via anammoxosome coupled with other electron acceptors such as sulfate in addition to nitrite (Liu et al., 2008;Rikmann et al., 2014). The G o of sulfate-reducing anaerobic ammonium oxidation proceeded by Anammox species Anammoxoglobus sulfate (2NH 4 is obviously lower than that of Feammox (Liu et al., 2008;Rikmann et al., 2014), suggesting that Anammox bacteria in the enrichment should favor Feammox. Secondly, genera of Brocadia and Kuenenia have iron(III)-reducing ability using organic matter (e.g., formate, acetate, and propionate) as electron donor (Graaf et al., 1996;Zhao et al., 2014); 80% of the ferric iron reductase in these Anammox bacteria locates in the membrane fraction and part of them termed as dissimilatory ferric iron reductases are the essential terminal reductase of the Fe(III) respiratory pathway in iron(III)-reducing bacteria (Schröder et al., 2003;Zhao et al., 2014). It provided a cue that the Anammox bacteria detected in our study, including Brocadiaceae, Kuenenia, and Jettenia, were capable of reducing iron(III) linked to oxidizing ammonium anaerobically at same time. Moreover, several publications demonstrated the potential role of Anammox bacteria in Feammox based on the increase in the N 2 production after amendment with iron(III) oxides in the Anammox sludge (Chen et al., 2014;Wang et al., 2016;Li X. et al., 2018;Yin et al., 2019). The Feammox bacteria Acidimicrobiaceae sp. were previously identified in acid soil with pH between 3.5 and 4.5 (Huang and Jaffé, 2014); however, the gene copy of the Acidimicrobiaceae spp. was extremely low (Figure 2B), and members of this family were not detected via Illumina sequencing in this study, which might indicate that this reported family made very little contribution to Feammox-linked N 2 production under neutral condition. In addition, the aerobic ammonium-oxidizing bacteria Nitrosomonas spp. were found to be highly enriched in the ammonium-containing anoxic condition (Lek Noophan et al., 2009), whereas they were not detected through Illumina sequencing and amoA-based qPCR (data not shown) in the treatment of Fe(III) + NH 4 + in this study. All of these disclosed that the Anammox bacteria Brocadiaceae, Kuenenia, and Jettenia had potential for Feammox-associated anaerobic ammonium oxidation ( Figure 5A).
Cooperation between Anammox bacteria and iron(III)reducers such as Geobacter could also complete the Feammox process ( Figure 5B). The conductive pili of Geobacter provide the chance to extend electron transfer ability beyond the outer surface of their cells (Reguera et al., 2005). These pili offer the possibility for Geobacter to accept electrons from anaerobic ammonium oxidation by Anammox bacteria through periplasmic or outer membrane electron transfer protein (Reguera et al., 2005). The increase in the gene copy numbers of nirS, nosZ, hzsB, and relative abundances in nitrate reducers/denitrifiers (including Geobacter, Pseudomonas, Bacillus, Clostridium, and Desulfotomaculum)  further indicated the contribution of denitrification, nitrate reduction, Anammox and nitrate reduction dependent iron(II) oxidation to N turnover in the Fe(III) + NH 4 + treatment during the incubation (Figures 5A,B).
Potential Feammox rate was estimated with a value of 0.49 µg N kg −1 d −1 based on the 30 N 2 production rates, which was comparable to that reported in paddy soil (0.17-0.59 µg N kg −1 d −1 ), intertidal wetland (0.24-0.36 µg N kg −1 d −1 ), and tropical forest soil (about 0.32 mg N kg −1 d −1 ) (Clement et al., 2005;Yang et al., 2012;Ding et al., 2014;Li et al., 2019). However, the contribution of Feammox to N loss was much lower than that of Anammox enrichment from the sludge (4.1 × 10 6 µg N kg −1 d −1 ) (Chen et al., 2014), suggesting a substantially low N removal efficiency via Feammox. Furthermore, the minor contribution ratio of Feammox to N loss in this study was inconsistent with the previous reports in the Anammox sludge, which showed that Fe(III) addition increased the N removal up to 0.8 × 10 6 µg N kg −1 d −1 (Chen et al., 2014;Wang et al., 2016). The significantly lower relative abundance of Anammox bacteria in the treatment of Fe(III) + NH 4 + (Figures 2A, 3, 4C) than the those reported in the previous wastewater treatment system (Chen et al., 2014;Wang et al., 2016) may be an important reason for low amount of N 2 via Feammox. Firstly, limited capability of substrate utilization by Anammox bacteria led to slow growth rate and long doubling time (Jetten et al., 2005). As a result, these Anammox bacteria were overwhelmed by iron(III) reducers that could efficiently obtain energy from dissimilatory iron(III) reduction. Secondly, the relative abundances of the Brocadiaceae, Kuenenia, and Jettenia decreased in the treatment of Fe(III) + NH 4 + (Figure 4C), which was likely due to the accumulation of NO 2 − /NO 3 − after 12-day incubation (Supplementary Table 3). Iron(II) coexistence with NO 2 − (Supplementary Figure 2), which might lead to the NO x − dependent Fe(II) oxidation, can severely inhibit the activity of Anammox bacteria (Zhao et al., 2014). Thirdly, dilution of Anammox bacteria via three generations of subculture may be another important reason for the low activity of Feammox-associated N 2 production decreased during the incubation. Besides, Fe 3 O 4 can form from rapid iron(III) reduction with the production of Fe(II) absorbed on the ferric oxides surface and then improve N removal during Feammox process (Kappler et al., 2014;Li X. et al., 2018). Hence, it can be inferred that Fe(II) was accumulated with relatively low extent after subculturing in the treatment of Fe(III) + NH 4 + ; therefore, the amount of Fe 3 O 4 produced in the treatment was lower compared to the sludge reactor in continuous operation.

CONCLUSION
This study demonstrated the occurrence of Feammox in Anammox inoculant-based enrichment. The relative abundances of Geobacter and Anammox bacteria such as Brocadiaceae, Kuenenia, and Jettenia were significantly correlated with 15 N 2 production rates, indicating their potential role in Feammoxinvolved N removal. We proposed that sole Anammox bacteria or cooperation between Anammox bacteria and Geobacter or unclassified Geobacteraceae could complete the Feammox process during the incubation. Our results suggested the potential role of Anammox bacteria in the nitrogen removal via the Feammox process.

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/ Supplementary Material. in lab work and laboratory analyses. X-RY wrote the manuscript. X-RY, HL, J-QS, and G-WZ revised the manuscript. All authors read and approved the final manuscript.