Investigation of nitrite accumulation by hydrogenotrophic denitrification in a moving bed biofilm reactor for partial denitrification and anammox process Open Access

Anaerobic ammonium oxidation (anammox) is a process in which -N and -N are converted to nitrogen gas and a small amount of -N by chemoautotrophic bacteria (anammox bacteria) (Strous et al. 1997). The application of the annamox process in low carbon to nitrogen (C/N) ratio and ammonium-rich wastewater treatment is more cost-efficient than conventional techniques based on nitrification and denitrification (Siegrist et al. 2008). -N production is necessary for practical use of the anammox process, as -N is not generally present in wastewater. Thus, the partial nitrification and subsequent anammox (PNA) process has been implemented in many full-scale wastewater treatment plants, in which the influent -N is partially oxidized to -N by nitrification and -N is then removed with producing -N by the anammox process. However, Lackner et al. (2014) reported that -N was generated after the PNA process in 50% of full-scale plants, which lasted from a few days to several weeks. Furthermore, the PNA process has rarely been implemented in the treatment of wastewater that contains low -N (e.g., domestic wastewater). This is due to the lack of inhibition factors for nitrite-oxidizing bacteria to achieve substantial -N accumulation (Miao et al. 2016). Thus, it is crucial to minimize -N generation while the wastewater is treated by the anammox process and to develop an alternative process for stable -N accumulation in wastewater containing low -N for further improvement.

The partial denitrification and anammox (PDA) process is a combination of partial denitrification (⁠-N is reduced to -N) and subsequent anammox (⁠-N is reduced by -N to N₂) processes, and is also expected to achieve stable -N production prior to the anammox process (Sumino et al. 2006). The influent used for the PDA process should contain equal concentrations of -N and -N. This can be achieved stably through the complete oxidation of -N in wastewater to -N by nitrification and subsequent mixing with the influent. The -N generated by the anammox process is ideally reduced to -N; thus, the PDA process is a promising option for applying anammox process into treatment for wastewater containing low -N concentration. Heterotrophic denitrification is commonly utilized in the PDA process (Sumino et al. 2006). However, when the C/N ratio exceeds the optimal ratio (0.5–1.0), nitrogen removal by the annamox reaction decreases during the PDA process (Takekawa et al. 2014), indicating that controlling the influent water quality is necessary. Another option is the PDA process that is based on sulfur autotrophic denitrification (Chen et al. 2018). In this type of PDA process, the -N accumulation efficiency is also influenced by the sulfide to nitrogen (S/N) ratio (Liu et al. 2017). Extensive research about the PDA process has been conducted to further optimize the anammox process. However, a buffering unit is required to maintain a suitable balance between electron donors and -N, which makes the treatment process more complex and requires sensitive system control. These complexities are major drawbacks for facilitating the implementation of the PDA process.

Stoichiometrically, the denitrification efficiency of HD is influenced by the H₂ dose. Thus, a HD-based PDA process is expected to be controlled by managing the H₂ flow rate immediately following water quality fluctuations, thereby eliminating the need for a buffering unit, which is a considerable advantage. In addition, H₂ is naturally removed from the water, which eliminates the need for secondary wastewater treatment for electron donor removal. Our previous research has demonstrated the feasibility of simultaneous -N and -N removal using the HD-based PDA process in a fixed bed anammox reactor by supplying H₂ (Kamei et al. 2018). Major concern for the HD-based PDA is H₂ source. But, H₂ can be produced by electrolysis of water. Furthermore, H₂ production from wastewater sludge by biological processes has attracted attention as a cost beneficial process and because of its efficiency (Kapdan & Kargi 2006; Wang et al. 2019), which may also be used as H₂ source. Thus, H₂ for the HD-based PDA process can be provided from multiple sources.

Although the HD-based PDA process may be effective, suitable conditions remain unclear for maximizing -N accumulation by HD, which is necessary for stable operation of the PDA process. Lee & Rittmann (2002) reported a sharp increase in the -N concentration when a autohydrogenotrophic hollow-fiber membrane biofilm reactor was operated with increased nitrate loading or lower H₂ pressure. Similarly, Lee et al. (2010) also observed a slight increase in the -N concentration (reaching ∼1 mg-N/L) in a packed bed HD reactor with a shorter hydraulic retention time (HRT) or insufficient H₂ supply. As reported in these previous studies, -N accumulation has occasionally been observed in HD reactors due to a short reaction time or a low H₂ supply. As HD has been applied to water treatment (Vasiliadou et al. 2009), -N should not be present in the treated effluent from the reactor. Therefore, the relationship between H₂ input and -N to maximize -N accumulation in HD reactors is poorly understood and requires further study. In this study, we investigated optimal conditions required for the maximization of -N accumulation by HD. A moving bed biofilm reactor (MBBR), which has an efficient volumetric treatment capability compared to other biofilm reactors (di Biase et al. 2019), was used. To evaluate the optimal conditions, the influences of H₂ flow rate and operational conditions on the -N accumulation efficiency during the synthetic wastewater treatment experiment were monitored using a laboratory-scale MBBR.

The overall experimental setup is shown in Figure 1. The MBBR was prepared in a cylindrical polymethyl pentene beaker (φ14 cm × 20 cm; 2 L working volume). As anammox plant is usually operated at 30–35 °C to enhance its nitrogen removal performance (Lackner et al. 2014), the MBBR was placed in a water bath, maintaining temperature at 35±1 °C using a thermostat. H₂ (99.99% purity) was supplied from a H₂ generator (HG-260, GL-science, Tokyo, Japan) through a commercial air stone diffuser (φ15 mm × 30 mm). To minimize gas release, the water surface in the MBBR was covered with polypropylene beads. An internal reactor pH of 6.5–7.5 was maintained using a periodic supply of 0.1 N HCl that was managed by a pH controller. In this study, the MBBR was stirred continuously at 100–150 rpm by a magnetic stirrer to move the sponge carrier constantly. The MBBR was operated under continuous flow conditions.

Polyolefin sponges (1.0 cm × 1.0 cm × 1.0 cm; 30 cm⁻¹ effective surface area; 95% porosity; Aqua cube, Sekisui aqua system, Japan) were utilized as the bacterial carrier. The sponges were first immobilized with 1.28 g of volatile suspended sludge (VSS) per liter of HD sludge that exhibited HD activity, as in a previous study (Rujakom et al. 2020). This immobilization was conducted for 2 months in the same scale MBBR (hereafter referred to as the parent reactor) under the following operational conditions: H₂ flow rate of 40 mL/min, supplemented with synthetic wastewater containing 40 mg-N/L of -N, and a 7.0 h HRT. These operational conditions were determined following our previous study, in which we achieved the HD-based PDA process (Kamei et al. 2018). However, to enhance the denitrification activity of HD, H₂ flow rate was increased four-fold and set to 40 mL/min. After immobilization, 200 sponges were withdrawn from the parent reactor and inoculated into newly prepared MBBRs for each experimental run. The initial HD sludge concentration attached to the biomass on the sponge carrier in each MBBR was set to 0.12–0.14 g-VSS/L.

Synthetic wastewater was prepared from tap water following the methods described in Mulder et al. (1995). The components of the synthetic wastewater are listed in Table 1. In each experimental run, the -N was a sole nitrogen source in the synthetic wastewater. The -N concentration was fixed at 40 mg-N/L by adding NaNO₃.

The initial VSS concentration of the HD sludge immobilized on the sponge carrier was determined by the weight difference between the cultivated sponge and a new sponge. The sponge carrier was placed in an incubator (FLT-30, Tokyo Garasu Kikai, Japan) at 60 °C until it was completely dried, and then incinerated to measure the VSS following the standard method (APHA/AWWA/WEF 2012). The measurements were conducted in triplicate.

The microbial community structure was analyzed in run 1 to identify the bacteria related to HD and -N accumulation. When experimental run 1 was started, suspended sludge was collected from the parent reactor and analyzed as the initial microbial community (hereafter referred to as the inoculum). In addition, the sponge carrier was collected after the operation of 15 mL/min of H₂ flow rate (maximized -N accumulation) and analyzed to evaluate the changes in the microbial community. The total DNA from the samples was extracted following the protocols in the FastDNA SPIN Kit for Soil (MP-Biomedicals, USA). The extracted DNA was then submitted to a next-generation sequencing (NGS) analysis service (Food analysis and biotechnology company, FASMAC, Atsugi, Japan). All microbial community analyses mentioned herein were conducted by FASMAC. Analysis of the V4 hypervariable region on the 16S rRNA gene was conducted using the Illumina Miseq gene sequencer using the specific primers 515F (ACACTCTTTCCCTACACGACGCTCTTCCGATCT-GTGCCAGCMGCCGCGGTAA) and 806R (GTGACTGGAGTTCAGACGTGTGCTCTTCCGATCT-GGACTACHVGGGTWTCTAAT). Based on a 3% divergence, the read sequences were grouped into operational taxonomical units (OTUs), and their homology was compared to the SILVA database in the QIIME platform.

Quantitative PCR (qPCR) was also performed to determine the abundances of total bacteria and functional genes for denitrification. Details of the PCR primers and conditions are summarized in Table S1. Specific primers for amplifying partial regions of bacterial 16S rRNA: 515F and 806R (Caporaso et al. 2011) were used to evaluate the total bacterial abundances. Four functional genes related to the denitrification pathway, including the narG gene (⁠-N reduction to -N) (López-Gutiérrez et al. 2004), nirK and nirS genes (⁠-N reduction to NO) (Yan et al. 2003; Throbäck et al. 2004), and nosZ gene (N₂O reduction to N₂) (Scala & Kerkhof 1998), were evaluated. All of the qPCR assays were conducted in a 25 μL reaction mixture containing 12.5 μL of SYBR^® Premix Ex TaqTM II (Takara, Bio, Shiga, Japan), 2 μL of template, 9.5 μL of deionized H₂O, and 0.5 μL of each specific primer (50 nM).

The means and standard deviations (SD) of the data were calculated for each experimental run. To evaluate the relationship between -N accumulation performance and the MBBR operational conditions, a correlation analysis was conducted using the R software package (version 3.6.2; available at http://www.r-project.org).

The NAR and NRR variations in the MBBR at different H₂ flow rates are shown in Figure 2. The MBBR suppressed -N removal and induced -N accumulation when the H₂ flow rate decreased from 40 mL/min to 15 mL/min (Figure 2(a)). Although -N accumulated, a H₂ flow rate of less than 10 mL/min suppressed the -N removal process. During reactor operation, the NAR was maximized at H₂ flow rates of 10 mL/min (0.06±0.01 kg-N/m³/d; mean±SD) and 15 mL/min (0.06±0.02 kg-N/m³/d). The NRR at H₂ flow rates of 10 and 15 mL/min were 0.10±0.01 kg-N/m³/d and 0.13±0.01 kg-N/m³/d, respectively. The NAEs at H₂ flow rates of 10 and 15 mL/min were calculated as 60.5±13.1% and 48±14.5%, respectively. Under these conditions, the higher H₂ flow rate yielded a higher NRE (90.0±5.4%), whereas the lower H₂ flow rate yielded a lower NRE (72.7±5.3%) (Figure 2(b)). The lower H₂ flow rate also yielded a higher NAE, but the NRE indicated the presence of -N residue in the effluent. -N accumulation should be maximized when -N removal is also maximized. Moreover, H₂ flow rates greater than 20 mL/min induced increase in NRR (reaching more than 0.14 kg-N/m³/d) and NRE (reaching more than 95%) and suppressed the NAR (less than 0.01 kg-N/m³/d). In addition, the MBBR experienced deteriorating NARs and NRRs at a H₂ flow rate of 5 mL/min. A H₂/N ratio greater than 8.1 exhibited a sufficient electron donor supply and induced enhanced denitrification activity (Figure 2(c)). Sharp decreases in NRR were observed when the H₂/N ratio was less than 4.1, indicating a lack of electron donors. A H₂/N ratio of 6.1 (at 15 mL/min H₂ flow rate) provided a suitable balance between H₂ and -N loading to maintain -N removal and induce -N accumulation. These results suggest that using a H₂ flow rate of 15 mL/min for further experiments can achieve higher NARs and NREs.

To monitor the stability of the -N accumulation, the MBBR was operated for 60 d in run 2. The -N accumulation occurred immediately and was stable until the end of the experimental run (Figure 3). The -N concentration increased with a decrease in influent -N concentration (Figure 3(a)). The mean -N and -N concentrations in the effluent were 17.1±3.1 mg-N/L and 5.9±4.0 mg-N/L, respectively. The mean NAE during run 2 was 53.0±11.5%. Although the NRE fluctuated during the first 10 days of operation, it was generally above 75% (Figure 3(b)). Similarly, the NAR also fluctuated during this initial period, but then remained stable until the end of run 2 with NAR of 0.06±0.01 kg-N/m³/d, which was similar to the results obtained from run 1 for the entire operational period.

Chang et al. (1999) reported that the -N reduction process of the HD was inhibited more dramatically by an insufficient H₂ supply than -N reduction. The influence of a lower H₂ flow rate on -N accumulation in the MBBR, while maintaining the NRR and NRE, was in accordance with the findings of Chang et al. (1999). Maintaining -N accumulation through HD is crucial for stabilizing the PDA process, as -N is a common nitrogen source for both HD and anammox processes. The MBBR continuously exhibited stable -N accumulation under the optimal H₂ flow rate for 60 d, indicating the potential to achieve stabilization of the HD-based PDA process. However, 0.14 g of H₂ is required to reduce 1 g of -N according to Equation (1). The H₂/N ratio observed in run 1 was 6.1, which was ∼44 times larger than the above ideal ratio. Furthermore, the NAEs obtained from each run did not reach 100%, indicating that excessive H₂ was supplied to the MBBR and was used for -N reduction. To maximize NAR and minimize H₂, the effects of HRT and an intermittent H₂ supply on NAR and NRR were further evaluated in runs 3 and 4.

The qPCR analyses indicate that the abundances of the 16S rRNA and functional genes for denitrification were similar for the inoculum and the sample collected from the MBBR after operation at a H₂ flow rate of 15 mL/min (Fig. S1). Total bacterial abundances in the inoculum and sludge from run 1 were 4.5×10¹⁰ and 5.3×10¹⁰ copies/g-VSS, respectively. Functional genes for all of the denitrification genes were observed at similar levels in each sample. The abundances of the narG, nirS, nirK, and nosZ functional genes were 6.6–8.7×10⁶, 0.8–3.1×10⁷, 2.3–3.3×10⁶, and 0.7–5.0×10⁶ copies/g-VSS, respectively. The abundances of nirS and nirK were equivalent to that of narG, even after achieving -N accumulation, which suggests that functional genes related to -N reduction might be not functioning due to the insufficient H₂ supply. Conversely, the microbial community structure shifted after achieving -N accumulation (Figure 4). The major genera in the inoculum were unidentified genus of Hyphomicrobiaceae, Flavobacterium, and unidentified genus of Ignavibacteriaceae, with relative abundances of 14.7%, 12.3%, and 11.1%, respectively. The most dominant bacterium, with 39.2% of the total bacterial abundance, after achieving -N accumulation was an unidentified genus of the family Comamonadaceae. Flavobacterium was the second-most dominant bacterium, comprising 12.8% of the total bacteria, which was similar to the inoculum.

The results indicate that the unidentified genus of Comamonadaceae was dominantly enriched in the sludge after achieving -N accumulation. The sequence data of this taxon (253 bp length) was further analyzed using the EzBioCloud analytical service (https://www.ezbiocloud.net/). This OTU exhibited 100.00% similarity to partial 16S rRNA sequences of some species in the genus Hydrogenophaga, which is a H₂ oxidizing bacterial group (Willems et al. 1989). In addition, this genus is often found in denitrification reactor with H₂ supply (Zhang et al. 2009). Although the microbial community structure shifted drastically after achieving -N accumulation in this study, the abundances of the functional genes related to denitrification were similar in all samples. Therefore, these results indicate that bacteria in the genus Hydrogenophaga contributed to -N accumulation in the MBBR.

The MBBR was operated using different HRTs and an intermittent H₂ supply to enhance the -N accumulation performance (Figure 5). With a decreasing HRT, the NAR increased slightly in the range of 0.06–0.10 kg-N/m³/d (Figure 5(a)). The HRT and NAR were found to be inversely correlated by a 3.5 h HRT (r=–0.74, p<0.05). A weak inverse correlation was also observed between HRT and NAE (r=–0.51, p<0.05) (Fig. S2). In contrast, the NRR was not suppressed significantly when the HRT was reduced to 3.5 h (r=–0.22, p>0.05), and the NRR ranged from 0.13 to 0.15 kg-N/m³/d. Lee et al. (2010) also reported that a reduced HRT caused an increase in the -N concentration in the effluent of a packed bed HD reactor. The decrease in HRT led to a shorter contact time between the HD bacteria and H₂, which might have suppressed -N reduction instead of -N reduction. However, a linear decrease in NRE from 96.3% to 45.7% in the MBBR, which resulted from an increased NLR due to a shorter HRT, resulted in the presence of -N residue in the effluent. Thus, we determined that a 7.0 h HRT was the optimal in this set of experiments.

Our previous study regarding the HD-based PDA process indicated that the nitrogen removal performance was stable after applying an intermittent H₂ supply (Kamei et al. 2018). Therefore, we analyzed the performances at different H₂ supply rates to determine the optimal H₂ supply for higher -N accumulation (Figure 5(b)). The NAR was maintained at 0.07 kg-N/m³/d by maintaining the H₂ supply rate at 0.7 (e.g., ON: 7 min, OFF: 3 min, 15 mL/min flow rate), with the NRE at ∼89–94% and NRR at 0.10–0.13 kg-N/m³/d. Conversely, the NAR, NRE, and NRR all decreased when the H₂ supply ratio was set at or below 0.6, indicating that the -N reduction in the MBBR deteriorated. A H₂ supply ratio of 0.7 was the optimal operational condition, in which NAR, NRR and NRE reached 0.07±0.01 kg-N/m³/d, 0.12±0.02 kg-N/m³/d, and 89.2±8.9%, respectively. Compared with the results of the other experimental runs, MBBR operation with a H₂ supply ratio of 0.7 exhibited similar NARs and NREs, but had a higher NAE (64.4±14.5%) with a 30% savings of H₂ input (Table 3). Furthermore, the H₂/N ratio of this optimal condition was calculated as 4.3, which was 1.4 times lower than those in runs 1 and 2, and was assumed to be the optimal balance between H₂ and -N loading.

In this study, we determined and demonstrated the suitable operational conditions for -N accumulation using the HD process. The -N accumulation efficiency was affected by the H₂ flow rate. Moreover, an intermittent H₂ supply could be an effective operational factor for achieving -N accumulation with higher NAEs, NREs, and H₂ savings. In ideal operating conditions, the MBBR had an NRE of 89.2±8.9%, with -N accumulation at an NAE of 64.4±14.5%, indicating that -N residue was minimized and -N accumulation was simultaneously sustained for the subsequent anammox reaction in the PDA process. Shinoda et al. (2020) reported NRE and NAE values of 88.1% and 96.7%, respectively, in a HD reactor operated at a highly alkaline pH (10.5). This NAE is 1.5 times higher than the value obtained herein. Since it was necessary to strictly control the reactor pH at 10.5, double stage reactor operations with respective pH controls (i.e., alkaline pH in the HD reactor and a neutral pH in the anammox reactor) was required for operation. The MBBR used in this study required only a single pH control, which was controlled to not exceed the optimal pH for the anammox process. Therefore, the -N accumulation induced by controlling the H₂ flow rate would be superior for a single stage PDA system, and may yield operational cost savings and simplification of the overall treatment process.

The performance of the MBBR was compared to other partial denitrification processes. Cao et al. (2013) reported that 80% of the reduced -N was accumulated as -N, with complete reduction of the -N by HD using supplemental supernatant from fermented sludge (C/N ratio of 3) in a laboratory-scale sequencing batch reactor (SBR). Similarly, a 71.7% NAE with complete -N removal was reported for an SBR with a limited supply of acetate and a readily biodegradable chemical oxygen demand to -N ratio of 2.5 (Gong et al. 2013). Moreover, Liu et al. (2017) reported a 70% -N removal and a 70% nitrite conversion rate at an S/N ratio of 0.76 in an up-flow anaerobic sludge blanket reactor using sulfide as an electron donor. Huang et al. (2021) recently reported a maximum -N accumulation rate of 55.3% with 95.3% -N removal from an SBR reactor with supplemental sulfide at an S/N ratio of 0.8. Although the efficiency of -N removal was similar, the NAE obtained herein was slightly lower than in the other denitrification processes.

According to the experimental results, the H₂/N ratio could show the efficiency for -N accumulation and -N reduction by HD. But, the study was limited by the fact that the optimal H₂/N ratio (4.3) is 30 times larger than the ideal ratio of HD-based -N reduction. Furthermore, the NAR at this optimal H₂/N ratio was 64.4±14.5%, indicating that the H₂ supply was excessive and the surplus H₂ reduced some of the -N to gaseous nitrogen, thereby limiting the -N accumulation. In this study, we evaluated -N accumulation conditions based on a simple relationship between the H₂ flow rate and -N. Eamrat et al. (2020) reported that small bubbles with low rising velocities and a gas diffuser with a higher gas transfer coefficient could enhance HD activity by increasing the H₂ utilization for nitrogen removal by up to six times. Considering such physicochemical effects, -N accumulation via HD may increase the NAE as a result of suppressing -N reduction and may serve to minimize the H₂ input.

The intermittent H₂ supply minimizes operational cost, but may also require sensitive system controls that increase the total cost. Maintaining the water temperature at 35 °C during the PDA process also requires additional operational costs. An economic feasibility study should be conducted to facilitate the implementation of the HD-based PDA process, including an evaluation of the applicability of the H₂ supplied from multiple sources (i.e. electrolysis of water, biohydrogen production process). Further, the MBBR was operated in conditions without organic carbon that is usually contained in wastewater. We recommend verifying the -N accumulation performance in the presence of organic carbon by evaluating any changes in the microbial community structure and its long-term stability for implementing the HD-based PDA process.

This study evaluated the conditions required for -N accumulation through HD in a laboratory-scale MBBR. The results indicate that -N accumulated significantly under suitable H₂ supply and operational conditions (e.g., 15 mL/min H₂ flow rate, 40 mg-N/L influent -N, 7.0 h HRT, and 2 L working volume). The H₂ oxidizing bacterial genus Hydrogenophaga was detected and was the most dominant bacterium (39.2% relative abundance) after achieving -N accumulation, indicating its contributions to -N accumulation and HD. Furthermore, an intermittent H₂ supply ratio of 0.7 (e.g., ON: 7 min, OFF: 3 min) maintained the -N accumulation and -N removal, while enhancing the NAE and saving 30% of the H₂ supply. This also increased NAR, NRR, NRE, and NAE, which reached 0.07±0.01 kg-N/m³/d, 0.12±0.02 kg-N/m³/d, 89.2±8.9%, and 64.4±14.5%, respectively. The H₂/N ratio in this set of experiments was calculated as 4.3, which was assumed to be a suitable input ratio for maximizing the -N accumulation. The findings of this study can be used to improve the anammox process for treatment of wastewater containing low -N. Further studies should be conducted to maximize NAE and to evaluate the effect of organic carbon on the -N accumulation efficiency by HD

The authors wish to acknowledge the financial support of the Steel Foundation for Environmental Protection Technology (Grant ID:C-41-14). The authors also would like to thank Editage (www.editage.com) for English language editing.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.