Newly established process combining partial hydrogenotrophic denitri fi cation and anammox for nitrogen removal

The anaerobic ammonium oxidation (anammox) process holds great promise for treating nitrogencontaminated water; stable nitrite-nitrogen (NO 2 -N) production is significant to anammox performance. In this study, partial hydrogenotrophic denitrification (PHD) was used to stably and efficiently produce NO 2 -N from nitrate-nitrogen (NO 3 -N). An investigation of the effects of initial pH on the PHD process revealed that a high NO 2 -N production efficiency (77.9%) could be ensured by setting an initial pH of 10.5. A combined PHD-anammox process was run for more than three months with maximal ammonium-nitrogen (NH4 -N), NO 3 -N, and total dissolved inorganic nitrogen removal efficiencies of 93.4, 98.0, and 86.9%, respectively. The NO 2 -N to NH þ 4 -N and NO 3 -N to NH þ 4 -N ratios indicated that various bioprocesses were involved in nitrogen removal during the anammox stage, and a 16S rRNA gene amplicon sequencing was performed to further clarify the composition of microbial communities and mechanisms involved in the nitrogen removal process.


GRAPHICAL ABSTRACT INTRODUCTION
The anaerobic ammonium oxidation (anammox) process, which simultaneously converts ammonium-nitrogen (NH þ 4 -N) and nitrite-nitrogen (NO À 2 -N) to nitrogen gas (Van de Graaf et al. ), has been studied intensively as a remarkably effective biological nitrogen removal process. The anammox process is superior, in terms of both energy efficiency and cost-effectiveness, to conventional processes such as nitrification and denitrification (Siegrist et al. ). However, the application of anammox-based technology has been limited by its high NO À 2 -N requirement. The NO À 2 -N used in the anammox process is generally produced from NH þ 4 -N or nitrate-nitrogen (NO À 3 -N). Although most research has focused on the production of NO À 2 -N from NH þ 4 -N sources through partial nitrification (Li et al. ; Laureni et al. ), the long-term stability of NO À 2 -N production, especially when low-strength water is used, remains a significant concern (Shi et al. ).
Recently, a considerable amount of attention has been given to NO À 2 -N production by partial denitrification (PD), in which NO À 3 -N is partially denitrified to NO À 2 -N under anoxic conditions (Du et al. , b; Cui et al. ; Cao et al. ; Shi et al. ). Previous studies primarily addressed heterotrophic denitrification processes in which organic carbon was used as an electron donor. However, anammox activity is heavily inhibited by the presence of organic carbon ( Jin et al. ), making it preferable to minimize the addition of organic carbon.
To address this issue, a novel process combining partial hydrogenotrophic denitrification (PHD) and anammox has been promoted as a promising method for removing NH þ 4 -N and NO À 3 -N simultaneously (Kamei et al. ). The advantages of this combination arise from the high compatibility of the two bioprocesses, as the groups driving them are both categorized as autotrophic bacteria and the processes both take place under a conditions. Additionally, hydrogen (H 2 ) has no harmful environmental effects, while the use of organic matter can elevate the risk of secondary contamination (Khanitchaidecha & Kazama ; Eamrat et al. ). However, the stability of NO À 2 -N production remains a problem, and the NO À 2 -N production through PHD is of considerable importance in ensuring a robust combination process. High NO À 2 -N accumulation at high pH has been frequently observed during hydrogenotrophic denitrification ( Nguyen et al. ), and it is known that pH is one of the critical factors for NO À 2 -N accumulation. However, there is no current understanding of how to efficiently produce NO À 2 -N via hydrogenotrophic denitrification by controlling the pH. In previously used single reactor processes (Kamei et al. , ), there has been competition for NO À 2 -N acquisition between the anammox and denitrifying bacteria, which can result in undesirable deterioration in the performance of the process (Du et al. a). Moreover, anammox bacteria are extremely slow growing (Van de Graaf et al. ) and can go into a gradual decline in the system. Furthermore, high pH conditions severely inhibit the anammox process (Jin et al. ). These factors suggest that a double-stage process would be better than a single-stage process for carrying out the combined reaction. Currently, there is extremely limited information regarding the combination of the two bioprocesses, and no double-stage system for carrying out the combined process has been reported to date.
The objective of this research was to achieve stable NO À 2 -N production for the anammox process. As a first step, the effects of initial pH on the hydrogenotrophic denitrification process were investigated. Based on the findings of the assessment, a double-stage combination process was carried out for more than three months. Finally, 16S rRNA gene amplicon sequencing was performed to further clarify the makeup of the microbial communities in both stages and the overall mechanisms of nitrogen removal. Our results show the significant potential of this combined process for nitrogen removal.

Batch test
All batch tests were conducted in vial containers with 14 ml of synthetic NO À 3 -N water and 1 ml of mixed sludge with 12.3 mg mixed liquor suspended solid (MLSS) content. After covering the vial containers with aluminum caps, the headspaces and liquid phases were purged with argon gas to eliminate oxygen and the headspaces were then further purged with H 2 from a gas cylinder. The containers were placed in a thermostatic shaker (Bioshaker, BR-43FL) at a temperature of 35 C and rotated at 100 rpm. Samples were removed from the shaker every hour over a period of five hours. Each batch test was repeated three times in parallel. The synthetic NO À 3 -N water was prepared from 0.24 g/L of NaNO 3 , 0.50 g/L of NaHCO 3 , 0.3 g/L of MgSO 4 , 0.027 g/L of KH 2 PO 4 , 0.180 g/L of CaCl 2 ·H 2 O, and trace elements I and II (Eamrat et al. ). The biomass was taken from an ongoing hydrogenotrophic denitrification reactor, as described in our previous paper (Eamrat et al. ), and washed three times with deionized water before the inoculations. The initial pH of the synthetic medium was adjusted to 8.5, 9.5, 10.5, and 11.5 with 1 N NaOH to investigate the effects of pH on NO À 2 -N accumulation in the PHD process.

Reactor setup
The combination PHD-anammox process was established in two individual reactors, as shown in Figure 1. In the first stage of the process (PHD stage), NO À 3 -N was converted to NO À 2 -N via the PHD process and, in the second stage (anammox stage), the resulting NO À 2 -N was removed along with NH þ 4 -N via an anammox process. The combined system was continuously operated for more than 110 d under varying operating conditions (Table 1).
Lab-scale attached growth reactors with a working volume of 1 L were used to carry out the reactions. In each reactor, a hanging non-woven fabric (7 cm × 6.3 cm × 1 cm) was used as a platform for the microbial attachment of the enriched hydrogenotrophic denitrification (HD) and anammox sludges. The biomass used for the batch tests was used as seeding sludge for the PHD stage, while that used for the anammox stage was extracted from an anammox reactor, in which it had been cultivated from activated sludge, as reported in previous paper (Kamei et al. ), and then washed three times with deionized water before the inoculations into the reactors.
The synthetic inorganic water used in the PHD stage was prepared from NaNO 3 , 0.50 g/L of NaHCO 3 , 0.3 g/L of MgSO 4 , 0.027 g/L of KH 2 PO 4 , 0.180 g/L of CaCl 2 ·H 2 O, and trace elements I and II. During the course of the experiment, the concentration of NO À 3 -N was increased in a stepwise manner at a variable H 2 flow rate, as shown in Table 1. The synthetic water was alkalified (pH of approximately 10.5) using 1 N NaOH to accumulate NO À 2 -N and then deoxygenated using argon gas to reduce the concentration of dissolved oxygen to less than 0.3 mg-O 2 /L, with the hydraulic retention time (HRT) maintained at 4.2 h. The reactor was operated at a temperature of 25 ± 2 C, with H 2 gas (>99.99% purity) supplied from a water electrolytic H 2 generator (HG-26; GL Science, Tokyo, Japan) via a commercial air stone diffuser with a diameter of 15 mm and length of 30 mm. A magnetic stirrer was used to ensure that the liquid, gas, and sludge were completely mixed inside the reactor.
Synthetic deoxygenated water containing NH þ 4 -N and the effluent of the PHD stage were continuously and simultaneously fed into the anammox stage at a feeding ratio of 1:1 with the HRT maintained at 2.1 h. As the optimum pH for the anammox process is understood to be lower than that for the PHD process (Jin et al. ), the pH in the PHD process effluent should be adjusted before the anammox process or within the anammox reactor. In this case, the pH of the reactor was maintained below 8.3 by using 1 N HCl to prevent the adverse effects of high pH from the effluent of the first stage on the anammox process. The reactor was operated at a temperature of 35 ± 1 C and completely mixed using a magnetic stirrer.

Analytical methods
Samples were collected from the inlet and outlet and immediately filtered using a 0.45-μm pore-size membrane filter. The samples were then stored in a freezer (-18 C) until water quality analysis could be performed. The concentrations of NH þ 4 -N, NO À 3 -N, and NO À 2 -N were analyzed in accordance with standard methods (Standard Methods for the Examination of Water and Wastewater ), and the pH, temperature, and dissolved hydrogen (DH) concentrations were measured using a pH meter (Horiba-B712, Japan), a digital thermometer (WT-6, China), and a DH meter (ENH-1000, Japan), respectively. The performance of the PHD and anammox reactors was evaluated on the basis of the water quality analysis as follows.

PHD reactor performance
To evaluate the performance of the PHD process, the following equations were used: where ΔNO À 3 -N represents the difference in NO À 3 -N concentration between the influent and effluent, NO À 3 -N in represents the NO À 3 -N concentration in the influent, and NO À 2 -N eff represents the NO À 2 -N concentration in the effluent.

Anammox reactor performance
The percentage contributions of the anammox and denitrification processes to the total amount of nitrogen removal were evaluated using the following equations (Du et al. b): where ΔNH 4 þ -N is the difference in NH þ 4 -N concentration between the influent and effluent, ΔDIN is the difference in the concentration of dissolved inorganic nitrogen (DIN)-given as the sum of NH þ 4 -N, NO À 2 -N, and NO À 3 -N between the influent and effluent, and DIN in is the DIN concentration in the influent. 1.32 in the Equation (3) is based on the coefficient of the stoichiometric equation of the anammox process.

Microbial community analysis
Six samples were collected from the PHD and anammox reactors on the first (0th), 45th, and 64th days. DNA extractions were carried out from 0.1 g of wet solid samples using the FastDNA ® SPIN Kit for Soil (MP-Biomedicals, Santa Ana, CA, USA) in accordance with the kit protocol. Amplification of the V4 region of the 16S rRNA gene was conducted via polymerase chain reaction using the universal primer set 515F (5 0 -GTGCCAGCMGCCGCGGTAA-3 0 ) and 806R (5 0 -GGAC-TACHVGGGTWTCTAAT-3 0 ) (Eamrat et al. )under the following conditions: 94 C for 3 min, followed by 30 cycles at 94 C for 15 s, 55 C for 30 s, 72 C for 30 s, and a final elongation at 72 C for 5 min.
Using an Illumina MiSeq platform provided by a commercial sequencing service (FASMAC Co., Ltd Atsugi, Japan), amplified metagenomic sequencing was performed to carry out read preprocessing, operational taxonomic unit (OTU) generation, and identification. As a result, 62 250 sequence reads were obtained, with the resulting OTUs generated at an identity threshold of 97%.

Statistical analysis
The mean and standard deviation (SD) of concentrations of NO À 3 -N and NO À 2 -N during the batch experiments, the NO À 3 -N reduction rates (ΥNO 3 -N), and NO À 2 -N production efficiencies at different initial pH values were calculated. A one-way analysis of variance (ANOVA) was used to confirm significant differences among different initial pH values (p < 0.05). The data were processed using the statistical analysis software package SPSS v. 22 (IBM Corp., Armonk, NY, USA).

RESULTS AND DISCUSSION
Effect of pH on the PHD process Figure 2 shows the variations in pH and the concentrations of NO À 3 -N and NO À 2 -N at each initial pH value. At initial pH values of 8.5, 9.5, and 10.5, the NO À 3 -N concentrations decrease linearly and rapidly, although at a rate that reduces slightly as the initial pH increases from 8.5 to 10.5. By contrast, the NO À 3 -N reduction is strongly inhibited at an initial pH value of 11.5. Contrary to the NO À 3 -N results, there are significant discrepancies in the NO À 2 -N accumulation patterns at initial pH values of 8.5, 9.5, and 10.5. NO À 2 -N accumulation increases as the initial pH value increases from 8.5 to 10.5, reaching a maximum of 9.3 mg-N/L at an initial pH of 10.5, before sharply decreasing as the initial pH value increases from 10.5 to 11.5. At an initial pH of 8.5, the pH moderately increases to 9.7 after 5 h. At an initial pH of 9.5, the pH also gently increases (to 10.0), but no pH increases are observed at initial values pH of 10.5 and 11.5. Figure 3 shows a comparison of the NO À 3 -N reduction rates (ϒ NO3-N ) and NO À 2 -N production efficiencies at different initial pH values. The value of ϒ NO3-N does not significantly differ at initial pH values of 8.5, 9.5, and 10.5 but is significantly reduced at an initial pH of 11.5 (p < 0.05). The highest ϒ NO3-N of 3.36 ± 0.25 mg-N/(g-MLSS·h) is achieved at an initial pH value of 8.5, followed by 3.17 ± 0.019 mg-N/(g-MLSS·h), 2.96 ± 0.061 mg-N/ (g-MLSS·h), and 0.73 ± 0.064 mg-N/(g-MLSS·h) at initial pH values of 9.5, 10.5, and 11.5, respectively, indicating that the NO À 3 -N reduction process of the hydrogenotrophic denitrification is strongly inhibited at a pH of 11.5. The trend in maximum NO À 2 -N production efficiency is contrary to that of ϒ NO3-N , peaking at 82.4 ± 0.4% at an initial pH value of 11.5. The maximum NO À 2 -N production efficiencies, achieved at initial pH values of 10.5 (77.9 ± 4.8%) and 11.5, are significantly higher than those at initial pH values of 8.5 (10.7 ± 0.6%) and 9.5 (26.8 ± 1.2%) (p < 0.05). The ϒ NO3-N results obtained in this study, in which NO À 3 -N reduction dropped significantly as the pH increased above 8.6, were not consistent with those of previous studies (Lee & Rittmann ). Nguyen et al. () obtained different results, with decreases in NO À 3 -N reduction rate occurring when the pH increased from 6.5 to 9.5 and enhanced NO À 2 -N was observed at a pH 8.5 rather than that at a pH 9.5 (Ghafari et al. ; Nguyen et al. ), whereas Lee & Rittmann () found significant NO À 2 -N accumulations at pH values above 9.5.
The complete hydrogenotrophic denitrification process can be roughly divided into NO À 3 -N reduction and NO À 2 -N reduction processes resulting from the presence of nitrate and nitrite reductases. In particular, given that NO À 2 -N accumulation occurs as a result of imbalances in enzyme activity, the results clearly indicate that nitrite reductase is more susceptible to pH variation than nitrate reductase, which is consistent with previous results (Li et al. ). As can be seen from Equations (6) and (7) (Lee & Rittmann ), the second process requires protons to reduce the nitrate; a potential triggering mechanism for NO À 2 -N accumulation during hydrogenotrophic denitrification under high pH conditions (Nguyen et al. ): There are considerable discrepancies in the existing literature on hydrogenotrophic denitrification with respect to the optimal pH and the effect of pH on NO À 3 -N reduction and NO À 2 -N accumulation; these arise from the fact that the cultivation conditions make a definitive difference in the composition of the microbial community, leading to diverse denitrification characteristics (Li et al. ).
Given the features of the annamox process, the prime requirements for anammox pretreatment include (1) a high capability for reducing NO À 3 -N and (2) a stable and high NO À 2 -N production rate. Accordingly, it is not unreasonable to expect that the optimal pH conditionin terms of maintaining a good ϒ NO3-N and a high NO À 2 -N production efficiencyfor using this biomass for NO À 2 -N production in the PHD process is 10.5. NO À 3 -N removal and NO À 2 -N production during the PHD stage The long-term effect of the high pH condition on NO À 2 -N production under hydrogenotrophic denitrification was assessed by monitoring the stability and continuity of NO À 3 -N removal and NO À 2 -N production.

Figure 3 | Comparison of NO À
The PHD reactor was initially operated at a NO À 3 -N concentration of approximately 40 mg-N/L with a H 2 gas flow rate of 20 ml/min for 45 d (Phase I). As soon as the operation started, the NO À 3 -N concentration gradually decreased, reaching 0 mg/L, with a high NO À 2 -N concentration of 34.1 mg-N/L, on day 10. Average NO À 3 -N removal and NO À 2 -N production remained as high as 86.2 and 81.9%, respectively, throughout Phase I. By contrast, a deterioration in NO À 3 -N removal efficiency to a mean value of 63.7 mg-N/L was observed during Phase II, during which the influent NO À 3 -N concentration was approximately 60 mg-N/L. This deterioration was potentially attributable to the insufficient H 2 gas supply, which can directly impact the activity of hydrogenotrophic denitrification (Nguyen et al. ). To maintain the stability of the PHD process, the H 2 gas flow rate was increased to 30 ml/min during Phase III, resulting in increases in both NO À 3 -N removal and NO À 2 -N production efficiencies to 85.9 and 95.7%, respectively, on average. Increasing the H 2 gas flow rate to 40 ml/min during Phase IV, however, led to slight decreases in the NO À 2 -N production efficiency, although the average NO À 3 -N removal efficiency remained as high as 92.2%. These results indicate that NO À 2 -N production might be susceptible to high availabilities of H 2 gas. During Phase V, the influent NO À 3 -N concentration was further increased to 80 mg-N/L and the H 2 gas flow rate reduced to 30 ml/min. This reduced the NO À 3 -N removal efficiency, probably as a result of the insufficient Figure 4 | Profiles of (a) concentration of nitrogen compounds; (b) NO À H 2 gas supply; restoring the H 2 gas flow rate to 40 ml/min during phase VI resulted in high NO À 3 -N removal and NO À 2 -N production efficiencies with mean values of 88.1 and 96.7%, respectively.
These results show that high NO À 2 -N production can be stably and efficiently achieved over long intervals of time under alkalified conditions without a deterioration in NO À 3 -N removal. The NO À 2 -N concentrations produced in this study reached as high as 36.7, 56.1, and 74.6 mg-N/L at influent concentrations of 40, 60, and 80 mg-N/L, respectively. High NO À 3 -N removal and NO À 2 -N production efficiencies were reliably maintained even under increased NO À 3 -N influx. It should be noted that the H 2 gas supply should be appropriately controlled to avoid impacting NO À 3 -N removal and NO À 2 -N production.

Performance of the anammox stage
With the effluent of the PHD stage combined with the synthetic NH þ 4 -N water and supplied to the anammox stage at a constant rate, the stability of the combined process was explored for more than three months.
During Phase I, low NH þ 4 -N concentrations (average: 5.3 mg-N/L) were found with nearly depleted NO À 2 -N (average: 3.1 mg-N/L), although there were slight increases in NO À 3 -N (average: 7.6 mg-N/L) in the effluent. Nevertheless, the decreased NO À 3 -N removal and NO À 2 -N production during the PHD stage led to a high concentration of leftover NH þ 4 -N with a mean value of 14.2 mg-N/L in the effluent during Phase II, presumably because there was a lack in NO À 2 -N for the anammox process, which resulted in insufficient NH þ 4 -N removal. The increased NO À 2 -N production in the PHD stage during Phase III resulted in a recovery of the NH þ 4 -N removal, indicating that NO À 2 -N production during the PHD stage was intensely intertwined with NH þ 4 -N removal during the anammox stage. Furthermore, high levels of NO À 3 -N remnant in the effluent during the PHD stage negatively affected nitrogen removal efficiency because of the further NO À 3 -N production by the anammox process. Thus, the performance of the PHD stage was found to be of significant importance to the overall nitrogen removal capability of the combined system. NH þ 4 -N removal was maintained even when the concentration of NH þ 4 -N in the influent was increased to 60 mg-N/L during Phase V. NO À 2 -N concentrations in the effluent remained low and were relatively stable compared to those of NH þ 4 -N, even when the concentrations of nitrogen compounds in the influent fluctuated. This suggests a high demand for NO À 2 -N by the bacteria in the anammox stage, as NO À 2 -N is a substrate that can be used by other bioprocesses such as nitrification and denitrification. NO À 3 -N concentrations in the effluent were slightly higher than in the influent (after mixing), probably because of the anammox process during Phases I to IV. However, lower NO À 3 -N concentrations in the effluent were found during phase V, indicating strengthened activities by other bioprocesses during the anammox stage.
The removal efficiencies of NH þ 4 -N, NO À 3 -N, and DIN for the overall system are listed in Table 2. The performance (d) nitrogen removal efficiency over the entire process. Before mixing; original N-species concentrations of synthetic deoxygenated water containing NH þ 4 -N and the effluent of the PHD stage containing NO À 3 -N and NO À 2 -N. After mixing; N-species concentrations of the influent for the anammox stage after mixing the synthetic deoxygenated water and the effluent of the PHD stage at a feeding ratio of 1:1. Effluent; N-species concentrations of the effluent of the anammox stage.
of the system was stable throughout the experiment, with the maximal NH þ 4 -N, NO À 3 -N, and DIN removal efficiencies reaching 93.4, 98.0, and 86.9%, respectively, higher values than have been reported for a combined, single-reactor hydrogenotrophic denitrification and anammox process (Kamei et al. ; Kamei et al. ). These results indicate the superiority of the combination system proposed in this study over previously reported systems. The anammox process is typically applied to high-strength water treatments and, although the initial NH þ 4 -N concentrations in this study were at most 60 mg-N/L and were halved following the mixing of the two influents, the anammox activity was maintained at high levels even for low-strength water treatment.

Key factors for the anammox stage
To achieve a high level of anammox bacteria performance, NO À 2 -N to NH þ 4 -N and NO À 3 -N to NH þ 4 -N ratios should be monitored, as anammox bacteria require large amounts of NO À 2 -N and produce NO À 3 -N. Throughout the experiments carried out in this study, the average ratios of influent NO À 2 -N to influent NH þ 4 -N and removed NO À 2 -N to removed NH þ 4 -N were 0.83 and 1.13, respectively. Those values were lower than the theoretical value of 1.32 even though the performance of the combined system was maintained at a suitable level. The average ratios of produced NO À 3 -N to removed NH þ 4 -N during Phases I-IV were 0.44, 0.33, 0.28, and 0.35, respectively, which are all slightly higher than the theoretical value of 0.26. These results indicate the presence of nitrifying bacteria in the anammox stage and suggest the possibility of nitrifying bacteria surviving the anammox stage because the system does not enter a completely anoxic condition during that stage. In fact, nitrifying bacteria were observed during the anammox stage at day 64 (Figure 7), which probably resulted in the nitrification of NH þ 4 -N to NO À 3 -N. Interestingly, however, the average ratio of produced NO À 3 -N to removed NH þ 4 -N suddenly dropped to À0.20 during Phase V, indicating increased NO À 3 -N removal by denitrifying bacteria. We assume that either some denitrifying bacteria were washed out of the PHD stage and flowed into the anammox stage or that denitrifying bacteria originally coexisting with the anammox bacteria contributed to the nitrogen removal. Although there was no dissolved hydrogen remaining, or addition of organic carbon sources during the anammox stage, some biodegradable organic carbon was still available, including soluble microbial products from biomass growth or decay, volatile fatty acids, and extracellular polymeric substances (Khanitchaidecha & Kazama ; Nguyen et al. ). It has been reported that some hydrogenotrophic denitrifiers can utilize organic carbon in addition to H 2 and can behave both autotrophically and heterotrophically (Xing et al. ). In these experiments, these bacteria might have used organic carbon as a carbon source, thus NO À 3 -N removal was enhanced as the operation proceeded.
To gain a better understanding of the mechanism of nitrogen removal during the anammox stage, the contribution of each bioprocess to nitrogen removal was calculated (Figure 6(c)). It was found that the anammox process made consistently high contributions to the nitrogen removal, with average contributions of 84.0, 90.6, 75.9, and 70.7% during Phases I, II, III, and IV, respectively. The anammox contribution to nitrogen removal decreased significantly to 53.9% as the denitrification contribution increased to 46.1% during Phase V. Although the production of NO À 3 -N by anammox is a bottleneck in processes such as these, the combination used in these experiments was capable of compensating for it.
These results clearly show that the coexistence of different bioprocesses, including anammox, nitrification, and denitrification, performed a significant role in a nitrogen cycle that enhanced the performance of the system.

Microbial community characterization in both stages
As the makeup of a system's bacterial community has a vital impact on system performance, it was necessary to characterize the community present in the reactors used in this study. Therefore, the variations in microbial communities during both stages were analyzed to better understand their distributions and evolution. To carry out next-generation sequencing analysis, six samples were collected. The first two samples, which were used as seeding sludge, were collected from the parent reactors on day zero. The The microbial communities at the phylum level for the six samples are shown in Figure 7(a) and 7(b), in which Proteobacteria abundances are represented using subclass classifications. Among the three PHD stage samples, β-Proteobacteria and γ-Proteobacteria were the two most dominant phyla, accounting for 73.9-79.2% and 18.2-20.5% of the total bacteria, respectively. The major taxa in the three anammox stage samples were Planctomycetes (30.5-46.5%), β-Proteobacteria (16.7-26.4%), Chloroflexi (10.3-13.2%), Armatimonadetes (8.4-16.2%), and Chlorobi (3.5-3.7%). As the amount of substrate increased, the relative abundance of Planctomycetes increased, making it the most dominant phylum, while the relative abundance of β-Proteobacteria decreased. Proteobacteria are found in bioreactors related to partial denitrification processes (Du et   the seven genera was Hydrogenophaga spp., which increased in relative abundance during this stage (from 44.5 to 61.7%). Hydrogenophaga spp. has been identified as a hydrogenotrophic denitrifier, and it is known that most hydrogenotrophic denitrifiers are facultative autotrophic bacteria (Xing et al. ); Hydrogenophaga spp. has been found in bioreactors with organic (Li et al. ) or organic-free conditions (Eamrat et al. ; Kamei et al. ). Rhodocyclaceae had the second highest abundance at the beginning of the stage but decreased to third (from 23.0 to 10.3%) after 45 d of operation. The abundance of Xanthomonadaceae, by contrast, did not change significantly and increased to second highest on day 45. The major genera at the beginning of the anammox stage were Candidatus Jettenia spp., followed by [Fimbriimonadaceae] and Dok59 spp. Candidatus Brocadia spp. appeared on day 45 and were the most dominant genus on days 45 and 64, representing a significant increase from the beginning of the operation. Candidatus Jettenia spp. and Candidatus Brocadia spp. are recognized as anammox bacteria (Pereira et al. ). Interestingly, the overall relative abundance of anammox bacteria has been reported to drop over the course of a single-reactor experiment (Kamei et al. ); by contrast, the overall abundances of anammox bacteria increased during our experiments. This increased dominance of anammox bacteria can lead to superior system performance based on the contribution of the bacteria to nitrogen removal in our experiments ( Figure 6). In addition to anammox bacteria, Nitrosomonadaceae, a group of lithoautotrophic ammonia oxidizers (Prosser et al. ), appeared on day 45 and increased in relative abundance up to a value of 3.3% on day 64. This change in the Nitrosomonadaceae abundance might have contributed to the enhancement of NH þ 4 -N removal. The above results indicate that Hydrogenophaga spp. played a significant role in both NO À 3 -N removal and NO À 2 -N production during the PHD stage in conjunction with the exclusive microbial community present in this stage. The calculated contribution ratio of anammox bacteria to nitrogen removal and the relative abundance of anammox groups such as Candidatus Jettenia spp. and Candidatus Brocadia spp. suggest that they played a significant role in the anammox stage. Furthermore, our results corroborated the coexistence of anammox bacteria, ammonia oxidizers, and denitrifying bacteria in the anammox stage.
The significance and applications of the newly established system As an improvement to processes combining heterotrophic denitrification and anammox reactions, the PHD-anammox process has several advantages relative to the partial denitrification and anammox process. First, it requires no external organic carbon source dosage, thereby reducing the risk of