Introduction

In recent years, dissimilatory nitrate reduction to ammonium (DNRA), which could convert nitrate to ammonium by two steps, has received increasing attention because of its ability to recover ammonium from wastewater with a high nitrate concentration1. In contrast to denitrification, DNRA converts nitrate into ammonium with nitrite as the sole intermediate. Both denitrification and DNRA involve similar nitrate reductase enzymes (narGHI, napAB), but DNRA utilizes distinct nitrite reductase enzymes (nirBD, nrfAH) specific to the DNRA pathway2. The traditional denitrification step wisely reduces nitrate to N2 via NO2−, NO, and N2O as intermediates3. Among the intermediates, N2O is a potent greenhouse gas, with a greenhouse effect approximately 265 times greater than that of CO2. DNRA not only retains nitrogen but reduces N2O production, thus it deserves further investigation. Previous reports highlighting the occurrence and significance of DNRA in diverse natural and human-impacted ecosystems have significantly advanced our comprehension of the global nitrogen cycle4,5. In the field of sewage treatment, Wang et al. suggested that DNRA could couple with anaerobic ammonium oxidization (anammox) to realize complete nitrogen removal6. In another innovative study, DNRA activities were utilized to convert excess nitrite to ammonium, thereby minimizing nitrite toxicity to anammox bacteria7. Moreover, DNRA could conserve bioavailable nitrogen in soil ecosystems to reduce nitrogen loss via denitrification and leaching8. Generally, in low nitrogen-fertilized paddies, DNRA could convert about 60% of the soil’s available nitrate into ammonium, thus facilitating nitrogen retention9. Thus, DNRA could assume a more significant role by participating in nutrient cycling. However, the intricate competition for carbon sources between denitrification and DNRA, coupled with the slow growth rate of DNRA bacteria, hindered the investigation and application of DNRA10. Hence, it is very necessary to discover efficient promotion strategies in bioreactors.

Electromagnetic field (EMF) is a new environmental factor, the level of which is increased as technology advances11. The ability of bacteria and other cells to communicate through EMF is a captivating phenomenon, shedding light on the complex mechanisms underlying intercellular interactions12. It has been reported that EMF could mimic the control signals produced by cells, driving a cellular response by correcting metabolic processes, particularly in the plasma membranes13. According to Maxwell’s equations, EMF is generated by charged particle moving with accelerated velocity, thus EMF of different frequency or intensity possess varying energies, which can exert either positive or negative effects on cells14,15. On the one hand, EMF of high frequency or intensity could inhibit cell growth and biofilm formation by changing water cluster structuring, membrane protein properties, DNA or prophage conformations, and other cellular structures16,17. On the other hand, low-frequency EMF has been applied to a number of bacteria, and interesting effects have been observed11,18,19. In the field of wastewater treatment, the utilization of low-frequency infrared electromagnetic fields (IR-EMF) has the potential to enhance partial nitrification and anammox activity20,21. Therefore, it is reasonable to assume that the assorted IR-EMF could promote DNRA process.

Here, the use of IR-EMF to stimulate fermentative DNRA was explored for the first time. The primary objective of this study was to assess the long-term effects of IR-EMF on DNRA. Quantitative real-time polymerase chain (qPCR) assay was used to estimate the DNRA and denitrification functional gene content and Illumina MiSeq assays were used to reveal the composition and changes in the microbial community. In addition, Bacteroides fragilis, Thauera sp. RT1901 and Shewanella oneidensis MR-1 were selected to validate the improvement of IR-EMF on DNRA. Overall, this study aimed to propose an effective and environmental-friendly biotechnology method for recovering ammonium from high nitrate concentration wastewater.

Results and discussion

Nitrogen conversion under different IR-EMF

The nitrogen transformation in the three reactors was monitored (Fig. 1A–C). The average effluent NO2−-N concentration was in the range of 0.1–0.2 mg/L and NO3−-N was nearly undetectable during the first 38 days, suggesting that denitrification was the dominant process, where NO3−-N was almost converted to N2 and expelled. The appearance of NH4+-N in the effluent at the early stage may be attributed to sludge autolysis22. NH4+-N concentrations in RCK, R1, and R2 started to increase from 38, 40, and 60 days and reached 50.8, 73.0, and 67.9 mg/L on 66, 66, and 69 days, respectively, representing the successful start-up of the DNRA process. Therefore, based on the start-up time, two distinct phases were identified: the acclimation phase (day 0–80) and the maturation phase (day 80–180). During the long-term domestication process, the average nitrate to ammonium conversion efficiency (hereafter referred to as “ammonium conversion efficiency”) of RCK, R1, and R2 were 33.3%, 72.5%, and 54.0%, respectively, indicating a clear difference among various IR-EMF intensities (Fig. 1D). At the early stage, RCK started to produce NH4+-N, earlier than R1 (42 d) and R2 (60 d). After the start of NH4+–N production, the time required to achieve an NH3 conversion efficiency of more than 60% was greatly shortened with an increase in IR-EMF intensity. In particular, the NH3 conversion efficiency reached 64.12% from 6.21% in 36 d for RCK, 73.23% from 6.04% in 24 d for R1, and 67.14% from 3.81% in 12 d for R2. Compared with those of RCK, the start-up times of R1 and R2 from the beginning of NH4+-N production were reduced by 33.33% and 66.67%, respectively. A previous study reported that magnetic field could stimulate the abundance of bacteria responsible for different simultaneous functions, such as nitrogen removal, organic matter removal, and saline tolerance simultaneously23. The effect of EMF at improving the denitrifying flora was also validated24. Thus, IR-EMF may not specifically target DNRA-functioning bacteria for promotion, but had promotion effect on both DNRA and denitrification bacteria. Denitrification was dominant during the initial domestication (0–38 d), whereas IR-EMF drove the DNRA to outcompete the denitrification process under more suitable conditions for DNRA.

Fig. 1: Nitrogen conversion performance of different reactors in long-term operation.
figure 1

Control (A), 0.04 μT IR-EMF irradiated reactor (B), and 0.06 μT IR-EMF irradiated reactor (C). D Average ammonia concentration in three reactors. E Free ammonia concentration in long-term experiments. Within the box, horizontal line denotes median value; box extend from the 25th to the 75th percentile of the group’s distribution of values; vertical extending lines denote the most extreme values within 1.5 interquartile range of the 25th and 75th percentile of the group; dots denote observations. The significance of differences between the treatments and the control was indicated by asterisks: p < 0.05 (*), p < 0.01 (**), p < 0.001 (***).

Full size image

The concentration of free ammonium (FA) in the reactor was analyzed (Fig. 1E). The variation of FA was calculated according to Eq. (2). To date, the mechanism of DNRA inhibition by FA is yet to be elucidated. In this study, R1 and R2 showed higher tolerance to FA compared to RCK. However, more information is needed to understand the mechanism underlying the inhibition of DNRA bacteria by FA.

$$\text{FA}=\frac{17}{14}\frac{[\text{N}{\text{H}}_{4}^{+}-\text{N}]\times {10}^{\text{pH}}}{\exp \left(\frac{6344}{273+\text{T}}\right)+{10}^{\text{pH}}}$$
(1)

Potential DNRA rate and functional gene abundance

The potential DNRA rates on day 120 of RCK, R1, and R2 were 8.49, 26.43, and 10.81 μmol/(kg·h), respectively (Fig. 2A). R1 achieved the highest potential DNRA rate, corresponding to the water quality. DNRA and denitrification functional genes were tested by qPCR technology (Fig. 2B–D). In all three reactors, the abundance of nrfA continuously increased during the acclimation phase, peaked, and then slightly declined during the maturation phase. The increase rate in nrfA gene abundance was highest in RCK (50.82%) on day 40, followed by R1 (19.41%) and R2 (21.79%). However, on day 80, nrfA abundance in RCK only increased by 9.24%, whereas that in R1 and R2 increased by 101.51% and 157.30%, respectively. Additionally, the abundance of nirK in all three reactors was two to three orders of magnitude lower than nirS, indicating that denitrifying bacteria were mainly of the nirS-type. The abundances of both nirK and nirS initially increased and then decreased with notable differences in timing. The abundances of nirS and nirK in R1 first increased (0–40 d) and reached the highest abundances of 1.82 × 1010 and 3.51 × 107 copies/g sludge, respectively, on day 40, followed by a rapid decline (40–180 days). In contrast, the abundance of nirS gene in RCK and R2 reached its peak on day 80, lagging behind R1, with values of 1.83 × 1010 copies/g sludge and 2.34 × 1010 copies/g sludge, respectively. After day 80, the abundance of nirS decreased by 8.18% and 34.65% in RCK and R1, respectively, on day 120 and further declined by 3.38% and 33.48% in RCK and R1, respectively, on day 180. R2 exhibited a faster decline than RCK, which can be ascribed to the IR-EMP irradiation. As reported, nirS, nosZ, and nrfA genes are all inducible genes25. Temporary changes in environmental conditions cannot induce immediate higher reaction rates even under environmental conditions favoring DNRA or denitrification26. They could only be induced in a continuous and stable environment and the corresponding responses could be observed. Based on this, IR-EMF accelerated the speed of microbial succession and turnover in the reactor, leading to a competition in which the DNRA bacteria dominated over denitrifying bacteria.

Fig. 2: Potential DNRA rates and functional gene abundances during long-term operation.
figure 2

Potential DNRA rates during stable operation on day 120 (A) and results of q-PCR for functional genes of nrfA (B), nirS (C), nirK (D). Error bars represent the standard deviation.

Full size image

Effect of IR-EMF on the microbial community structure

Ten major phyla were detected in four samples from the three reactors (>0.5% abundance in at least one sample), with the dominant phyla Proteobacteria, Bacteroidetes, and Chloroflexi (Fig. 3A). Proteobacteria were the main phylum, widely detected in activated sludge from different municipal wastewater treatment plants. The abundance of Bacteroides increased during the incubation. Bacteroides could be the main DNRA functional phylum in this study, which contained some DNRA functional bacteria27. On day 40, the growth of Bacteroides was inhibited by 0.04 μT IR-EMF, probably because the environment inside the reactor was more suitable for denitrifying bacteria in the early stage. On days 80 and 120, the abundance of Bacteroides in R1 increased substantially, exceeding that of RCK and R2. Chloroflexi was also one of the three dominant phyla in the domestication process of the three reactors. Some bacteria in Chloroflexi can also perform DNRA function28. However, in this study, Chloroflexi did not exhibit a clear pattern of variation. The dominant genera were Geobacter, Lentimicrobiaceae, Bacteroidetes_vadinHA17, Lentimicrobium, and Aegiribacteria (Fig. 3B). Geobacter, which has been proven to be a fermentative DNRA functional bacteria, was enriched in three reactors29. Lentimicrobiaceae, an autotrophic denitrifying bacterium with only 0.02% abundance in the initial sludge, increased to 2.34%, 1.25%, and 2.36% in RCK, R1, and R2, respectively, on day 40. Lentimicrobium belongs to the phylum Bacteroidetes, a class of fermentative bacteria that lives in strictly anaerobic environments. The removal of COD in the reactor depended mainly on fermentative bacteria. The abundance of Lentimicrobium in the initial sludge was 0.03%, and then increased to 1.40%, 2.05%, and 1.24% in RCK, R1, and R2 on day 40, and 3.35%, 6.97%, and 8.11% on day 80, respectively.

Fig. 3: Relative abundance of 16S rRNA dominant bacteria (>0.5% in at least one sample) in the three reactors.
figure 3

A 16S rRNA bacterial dominant phylum; (B) 16S rRNA bacterial dominant genus; R0 was seed sludge sample. C Microbial association network analysis. Each node represents a species, and species that may perform the same function are divided into a module, and the size of the node represents the role of the species in the module; the larger the node, the greater the role of the species in the module, and the connection line between different nodes represents the correlation between species. The higher the absolute value of the number, the stronger the correlation. Negative values are negative correlations, positive values are positive correlations. D PCA cluster analysis. Groups A, B, C, D, and E represent sludge samples taken on day 0, 40, 80, 120, and 180. The numbers of 1, 2, and 3 after the letters represent the RCK, R1, and R2, respectively.

Full size image

The relative abundance of DNRA functional communities was shown in Supplementary Fig. 1. The dominant phyla were mainly Bacteroidetes, Proteobacteria, Chloroflexi, Acidobacteria, and Planctomycetes (Supplementary Fig. 1A). The dominant genera were Luteitalea, Desulfomicrobium, and Ignavibacterium (Supplementary Fig. 1B). However, Geobacter, which was enriched in this study, was not detected. The missing detection of Geobacter could be due to the limitation of the sequencing methods and primer selection of nrfA gene. Thus, only some of the bacteria with the nrfA gene could be monitored because of the limitations of the methods and primers. The discussion related to the detected bacteria which contained nrfA gene was placed in Supplementary Note 1.

Network analysis and PCA analysis

Microbial network analysis was performed on the 16 most dominant genera in the reactor (Fig. 3C). Module 1 included only two genera: Sulfuritalea and Ellin6067. Sulfuritalea existed in anoxic water with abundant sulfides and low concentration of nitrate30. Ellin6067 has also been noted autotrophic denitrification systems31. Thus, these two bacteria were considered to be associated with sulfur-driven denitrification processes, and their nodes were similar in size and played the same role in module 1. The correlation between these genera was negative, as shown by their connecting line, indicating their competitive relationship. In Module 2, genera, such as Bacteroidetes vadinHA17, Lentimicrobium, and Geobacter, played a dominant role with their variations influencing other bacteria32. As Lentimicrobium is a denitrifying genus33, this module may be related to NO3− reduction. The correlations between Geobacter and other genera associated with this module were all negative, which can be ascribed to the DNRA function of Geobacter, whereas other genera performed denitrification. These two bacteria with different NO3− reduction functions competed for the substrate; therefore, their correlation was negative, which was consistent with the competition between DNRA and denitrification in the reactor.

In Fig. 3D, principal components analysis (PCA) has been used to show the similarities and differences in the community composition of different samples. The largest discrepancy between the three samples on day 40 can be ascribed to the different electromagnetic radiation conditions causing different rates of microbial community succession within the reactor, resulting in variability between samples. As the incubation proceeded, the variability between the samples at the same sampling time continued to decrease, indicating the enriched DNRA bacteria in the system and evolution of the community structure for DNRA dominance. B2 is extremely close to E1, E2, and E3, implying their similar community structures, indicating the possible acceleration of the community succession with 0.04 μT IR-EMF.

Predictive relative abundances of genes encoding key enzymes and metabolic pathway

The influence of 0.04 μT IR-EMF on the relative abundance of functional genes involved in tricarboxylic acid (TCA) cycle, ethanol degradation, and nitrogen metabolism was analyzed (Supplementary Fig. 2), and the relative pathway is proposed (Fig. 4). Descriptions of genes and enzymes were listed in Supplementary Table 3. The TCA cycle is a central carbon metabolism pathway that plays a vital role in generating the reducing power required for various cellular activities. After exposure to 0.04 μT IR-EMF, related genes, such as K00174, K00175, K00658, K18118, K01902, and K01903, involved in succinyl-CoA metabolism in the TCA cycle were upregulated (Supplementary Fig. 2A). Succinyl-CoA synthesis can stimulate amino acid and protein synthesis, thereby improving bacterial activity34. Succinyl-CoA for succinate conversion results in adenosine triphosphate (ATP) production, providing cells with more energy for cellular metabolic activities. Additionally, the anaerobic metabolism of ethanol is conducted by syntrophic bacteria for alcohol formation and consumption35. Compared to those in RCK, most ethanol degradation-related genes in R1 were not upregulated and were even expressed at lesser levels than those in RCK (e.g., K01895, K13788, K00128, and K00114) (Supplementary Fig. 2B). In terms of N2 metabolism, all functional genes related to denitrification (K00368, K00376, and K04561) were downregulated, whereas the DNRA functional gene nrfA (K03385) was upregulated (Supplementary Fig. 2C).

Fig. 4: Metabolic pathway of ethanol degradation, TCA cycle, and nitrogen metabolism.
figure 4

Predictive relative abundances of genes encoding key enzymes involved in ethanol degradation (A), TCA cycle (B), and nitrogen metabolism (denitrification and dissimilatory nitrate reduction) (C) under 0.04 μT IR-EMF influence in reactors.

Full size image

The main metabolic pathways in the reactor, i.e., ethanol degradation, TCA cycle, and N2 metabolism are illustrated at the enzyme level (Fig. 4). In ethanol metabolism, IR-EMF resulted in slight upregulation of the enzyme (EC:1.1.1.1) involved in alcohol decomposition. Enzymes related to succinyl-CoA metabolism (EC:1.2.7.11, EC:1.2.7.3, EC:6.2.1.5, EC:2.8.3.18, and EC:2.3.1.61) in the TCA cycle were also upregulated, which could provide more energy for cellular metabolism. In nitrogen metabolism, enzymes related to NO3− transport proteins (COG0715, COD0600, and COG1116) were upregulated, allowing more NO3− to enter the cytoplasm. Enzymes related to NO3− reduction (COG1140, COG3043, and COG5013) were also downregulated. The enzymes related to the reduction of nitrite to ammonium during the DNRA process (COG3005 and COG3303) were upregulated under the influence of IR-EMF. In contrast, the enzymes involved in nitrite (EC:1.7.2.1), NO (EC:1.7.2.5), and N2O (1.7.2.4) reduction during denitrification were downregulated. These findings are consistent with the qPCR results, demonstrating an upregulation of the DNRA functional gene nrfA and a downregulation of the denitrification functional genes nirS and nirK under conditions of 0.04 μT IR-EMF exposure.

Effect of IR-EMF on the DNRA and denitrifying bacteria, and the underlying mechanism

Bacteroides fragilis was used to verify whether IR-EMF promoted the DNRA process. The fastest NH4+-N production was observed in A2, therefore, 50 mg/L NO2−-N was selected for following experiments (Fig. 5A). Compared with control and 0.06 μT IR-EMF, 0.04 μT IR-EMF promoted the production of ammonium nitrogen, and this effect became increasingly apparent over time (Fig. 5B). The improvement of DNRA with the use of 0.04 μT IR-EMF was verified by collecting the samples and performing transcriptome sequencing (Fig. 5C). Gene Ontology (GO) analysis was performed (Fig. 5D) and possible mechanism (Fig. 5E) was proposed based on transcriptome results. Functions related to ATP synthesis (GO: 0042773, 0006119, 0008137, 0003955, 0050136, 0003954, 0016655, 0016651, 0048038, 0015980), transmembrane transport (GO: 0015453, 0022804), and electron transport (GO: 0022904, 0022900, 0009055) were upregulated. During fermentative DNRA, bacteria uptake fermentable organic substrates such as glucose through the phosphotransferase system. These substrates are utilized for growth, leading to ATP production via the glycolysis of glucose to lactate under anaerobic conditions10. Anaerobic glucose metabolism occurs in the cytoplasm. In this work, transmembrane transporter activity was enhanced, indicating that extracellular glucose could be transported and utilized in cytoplasm more efficiently. Additionally, NADH dehydrogenase is the key cellular mechanism for replenishing NAD+, crucial for sustaining glycolysis36. Membrane-bound quinone molecules are essential for redox reactions that drive cellular bioenergetics in bacterial organisms. NADH dehydrogenase, the initial enzyme in the respiratory chain, facilitates the transfer of electrons from NADH to ubiquinone, a process that is coupled with the pumping of protons out of the matrix37. Based on GO analysis, NADH dehydrogenase activity was enhanced which contributed to the energy production and then accelerated glycolysis. Subsequently, more ATP and formate were produced and prepared for nitrite reduction. Formate can reduce nitrite via nitrite reductase (NrfA), the terminal enzyme in dissimilatory reduction of nitrite to ammonium, even at low nitrite concentrations (<3 mM). NrfA reduces periplasmic nitrite with formate-derived electrons via formate dehydrogenase, potentially linking to the quinone pool through cytochrome c nitrite reductase small subunit (NrfH) proteins38. Numerous bacteria use the NrfHA system for converting nitrite into ammonium39. NrfH contains a hydrophilic domain binding four c hemes and is membrane-anchored by a transmembrane helix and a perpendicular helix. Heme 1 in NrfH, located closest to the membrane and likely to interact with menaquinol, exhibits a unique coordination involving methionine, resulting in a high-spin heme. Similarly, heme 4, situated near NrfA, displays an uncommon coordination involving a histidine/lysine combination, with the lysine originating from the NrfA chain40. NrfH is vital for electron transfer between the quinone pool and NrfA, facilitating the transfer of electrons from the quinone pool to NrfA41. Based on GO analysis, the function related to electron transport was improved. IR-EMF could strengthen the electron transport from Fdh and quinone pool to NrfH and then to NrfA. When more hydrogen or electrons were transferred to the final electron acceptor, the NrfA activity would be higher and the reaction of nitrite reduced to ammonium could be more efficient.

Fig. 5: The effect of IR-EMF on Bacteroides fragilis.
figure 5

A NH4+–N production by Bacteroides fragilis at different nitrite nitrogen concentrations. The NO2−–N concentration was 0, 25, 50, and 100 mg/L in A0, A1, A2, and A3, respectively. NO2−–N addition started at 0 h. and the reaction was completed at 24 h. B NH4+–N production by Bacteroides fragilis under exposure to different IR-EMF intensities. B0 is the control group without IR-EMF. B1 and B2 are the experimental group with the use of 0.04 and 0.06 μT IR-EMF, respectively. C Ammonium production by Bacteroides fragilis under conditions of exposure to 0.04 μT IR-EMF. C0 is the control group without IR-EMF, whereas C1 is the experimental group with 0.04 μT IR-EMF. D Genes upregulated between C0 and C1 based on GO analysis. E Possible mechanism underlying IR-EMF promotion of DNRA. The red line is the possible upregulated pathway. Error bars represent the standard deviation.

Full size image

Based on the discussion above, the dynamic response of Thauera sp. RT1901 and Shewanella oneidensis MR-1 to IR-EMF were further investigated. Thauera sp. RT1901 could perform denitrification42 and Shewanella oneidensis MR-1 was not only DNRA but typical electroactive bacteria43. During the experiments, IR-EMF of 0.04 μT or 0.06 μT promoted the nitrogen transformation of Thauera and Shewanella. IR-EMF of 0.04 μT and 0.06 μT enhanced the ammonium transformation of Shewanella by 17.3% and 10.7%, respectively (Fig. 6A). IR-EMF of 0.04 μT and 0.06 μT also improved the nitrite removal of Thauera by 11.9% and 16.8% within 180 min, respectively (Fig. 6B). Nevertheless, the optimal IR-EMF intensities for Thauera and Shewanella were 0.04 μT and 0.06 μT, respectively.

Fig. 6: Dynamic response of Shewanelle oneidensis MR-1 and Thauera sp. RT1901 to IR-EMF.
figure 6

The nitrogen performance of Shewanelle oneidensis MR-1 (A) and Thauera sp. RT1901 (B) under different IR-EMFs. The metabolic schematic of cell energy, electron transfer, purine metabolism, tricarboxylic acid cycle, and nitrogen metabolism (C). The RT-qPCR results related to functional genes of Shewanelle oneidensis MR-1 (D) and Thauera sp. RT1901 (E) under different IR-EMFs. Error bars represent the standard deviation.

Full size image

To further investigate the dynamic response of DNRA and denitrifying bacteria to different IR-EMF intensities, the transcriptional levels of functional genes related to energy metabolism were analyzed. The metabolic schematic of cell energy, electron transfer, purine metabolism, TCA cycle, and nitrogen metabolism were provided in Fig. 6C. The expression of functional genes of Thauera and Shewanella were significantly affected by IR-EMF (Fig. 6D, E). NrfA was the marked gene for DNRA44. Both 0.04 and 0.06 μT IR-EMF promoted the nrfA expression of Shewanella and enhanced the synthesis of related DNRA enzymes, which matched the nitrogen transformation result. NirS, nirK, norB, and nosZ were related to denitrification45. IR-EMF improved the expression of nirS and norB to enhance denitrification activity of Thauera. Moreover, microorganisms obtained energy (NAD(+/H) and ATP) to maintain activity. As discussed above, the genes involved in NAD(+/H) and ATP synthesis might be the key genes contributing to improvement by IR-EMF. Reverse transcription qPCR (RT-qPCR) results revealed a trend whereby the expression of nadD, nadE, nadR, nadV, and pncC genes, which related to the salvage and universal biosynthesis of NAD+46, were upregulated in both Shewanella and Thauera (nadV and pncC were undetected). Furthermore, PetA, petB, and petC genes encoded Fe-S protein, cytochrome b, and cytochrome c1 subunits, respectively, playing a key role in both aerobic and anaerobic respiration for ATP production47. As expected, the transcript levels of petA, petB, and petC genes were also improved in both Shewanella and Thauera. Thus, the findings further confirmed that IR-EMF could promote DNRA or denitrification process by increasing energy production. Discrepancies in the genes involved in the TCA cycle were also detected. A complete TCA cycle produces 3 mol NADH per mol acetate (Fig. 6C). These NADH molecules subsequently enter the electron transport chain for oxidative phosphorylation, generating substantial ATP to drive microbial metabolism48. Remarkably, IR-EMF positively influenced expression of the genes encoding isocitrate dehydrogenase (icd) and malate dehydrogenase (mdh and mqo), which were directly related to NADH production49. This indicated that IR-EMF supported more efficient TCA cycle, leading to increased utilization of acetate and NADH production. Thus, nitrogen metabolism could access additional energy resources to advance further. In the purine metabolism, the purA/B/M/N/Q expression of Thauera was upregulated and the purA/C/D/K/M/N/Q expression of Shewanella was upregulated. These genes were involved in AMP synthesis. AMP was not only the precursor of RNA, but also the important participant of the transformation with ADP and ATP50. It could be inferred that the IR-EMF improved RNA metabolic activity to further enhance nitrogen transformation50. IR-EMF also upregulated the expression of genes related to electron transfer (ribA/B/C/D, mtrA/D/E/F, cymA, omcA, cctA of Shewanella, and mtrA/D/E/F, sirA, cctA of Thauera). Furthermore, Shewanella oneidensis MR-1 seemed to be more sensitive to higher IR-EMF intensity than Thauera sp. RT1901. First, of all the functional genes detected, 34 genes were upregulated in Shewanella, compared to only 22 genes upregulated in Thauera. Second, among all 34 upregulated genes of Shewanella, 24 genes gained higher expression under 0.04 μT IR-EMF. But for Thauera, most genes were more abundant under 0.06 μT IR-EMF. It could be due to its outstanding electroactive characteristic of Shewanella, where electroactive bacteria have been reported to exhibit significant adaptability in reaction to magnetic fields51 and Wang et al. found that electromagnetic field could boost the extracellular electron transfer of Shewanella52.

Application of IR-EMF for recovery of ammonium by DNRA process

Widespread nitrate wastewater could have dual impacts, causing ecological damage while also serving as a potential source for ammonium recovery. Compared to conventional nitrate removal strategies, the application of DNRA offers the advantages of nitrogen recovery from nitrate wastewater while simultaneously mitigating the production of the greenhouse gas N2O1. Ammonium produced by DNRA has two main destinations: it can either be concentrated and recovered through physicochemical methods or utilized in situ by other bacteria. Several techniques have been documented for ammonium recovery from aqueous solutions, such as stripping53, ion exchange, and forward osmosis54, enabling recovery of ammonium generated by DNRA process. In one reported ammonium recovery system, nitrate was first converted to ammonium by enriched DNRA bacteria and subsequently, the system was connected to a circuit to force electrons to the counter electrode. Then, an electric field was generated and aided the migration and concentration of the generated ammonium to the catholyte for recovery55. The sketch of this recovery system was shown in Supplementary Fig. 3A. Moreover, DNRA has been deemed a feasible mechanism for producing nitrite and ammonium from nitrate in anammox bioprocess and has been employed in diverse bioreactor setups to improve anammox nitrogen removal56. The cooperative reaction mechanism was shown in Supplementary Fig. 3B.

In this study, a high ammonium conversion efficiency in DNRA systems improved by IR-EMF was realized. In addition, Vialkova et al. reported that the cost of treating and utilizing wastewater sludge decreased by an average of 1.3 times when EMF was applied11. In this work, electricity consumption is minimized, as an intensity of 0.04 ~ 0.06 μT IR-EMF can notably enhance the efficiency of the DNRA process. In addition, IR-EMF could be evenly transferred from the surface to the inside of the reactor by using a tube made of methyl methacrylate material20, so the sludge can be evenly and effectively exposed to EMF. Thus, applying IR-EMF could be an effective method to shorten start-up time, enhance ammonium conversion efficiency, and reduce operation costs.

In summary, the effect of IR-EMF on the enrichment of DNRA bacteria and its impact on microbial community structure and nitrogen transformation were evaluated using mixed bacteria culture and pure culture. With the increase of IR-EMF intensity, the start-up process of NH4+-N production by DNRA bacteria was prolonged for a while, but the time from initial to 60% ammonium nitrogen conversion was greatly accelerated by IR-EMF. R1 and R2 showed higher average ammonium nitrogen conversion than RCK. The abundance of DNRA functional gene nrfA was increased, and the nrfA abundance was maintained at a high level under 0.04 μT IR-EMF. The investigation of dynamic response of DNRA and denitrifying bacteria to IR-EMF suggested that the expression of genes related to nitrogen metabolism, TCA cycle, cell energy, purine metabolism, and electron transfer were improved by both 0.04 and 0.06 μT IR-EMF. This study demonstrated the viability of enhancing ammonium recovery efficiency through DNRA by applying IR-EMF and the underlying mechanism.

Methods

Reactor set-up, synthetic medium, and inoculation

To enrich DNRA bacteria, three non-woven fabric membrane bioreactors (nMBRs) of the same configuration were designed. Each reactor featured a double-layer cylindrical structure with an effective volume of 4 L. The reactors were encased by a water bath, which could keep the temperature at approximately 22 ± 1 °C using a cooling-water machine. Two rectangular non-woven fabric membrane modules with a pore size of 0.1 mm were securely installed inside each reactor. The membrane material was polymethyl methacrylate. The reactor configuration was provided in Supplementary Fig. 4. A mechanical agitator was installed and worked continuously. The hydraulic retention time (HRT) was 38 h. pH was controlled at 7.1 ± 0.1 by a pH Controller (Bluelab, New Zealand). The influent and effluent were regulated by peristaltic pumps (BT100-2J, Longer Pump, China). The low-frequency IR-EMF generation device used in this experiment was a modification version of a specific IR-EMF therapy instrument. The specific IR-EMF generator can produce an IR-EMF intensity of 0.06–0.25 μT and wavelength of 2–25 μm, which has bio-affinity. The distribution of IR-EMF inside the reactor was provided in Supplementary Fig. 5. The composition of the synthetic wastewater was determined as the previous study22, with slight modification (Supplementary Table 1 and Supplementary Table 2.). The ratio of C/NO3−-N in the influent was set to 7.7. The synthetic wastewater buckets were purged with high-purity N2 gas for 20 min/d to maintain an anaerobic environment. The concentrations of nitrogen compounds (NH4+-N, NO2−-N, and NO3−-N) were determined using the standard methods1.

Calculation of NH4 +-N transformation efficiency in the reactor

The conversion efficiency of NH4+-N can be calculated using Eq. (1). (this formula ignores adsorption and assimilation).

$$\begin{array}{ll}{{{\rm{NH}}}_{4}}^{+}-{\rm{N}}\; {\rm{transformation}}\; {\rm{efficiency}}\left( \% \right)\\=\frac{{c\left({{{\rm{NH}}}_{4}}^{+}\right)}_{{\rm{Eff}}}}{{c\left({{{\rm{NO}}}_{3}}^{-}\right)}_{{\rm{Inf}}}-{c\left({{{\rm{NO}}}_{3}}^{-}\right)}_{{\rm{Eff}}}-{c\left({{{\rm{NO}}}_{2}}^{-}\right)}_{{\rm{Eff}}}}\times 100 \%\end{array}$$
(2)

where c(NH4+) Eff represents the effluent NH4+-N concentration, mg/L; c(NO3−)Inf and c(NO3−)Eff represent the NO3−-N concentration in the influent and effluent, respectively, mg/L; c(NO2−) Eff represents the concentration of NO2−-N concentration in the effluent, mg/L.

15N tracer incubation, DNA extraction, and real-time qPCR analysis

The potential DNRA rate was assessed by conducting slurry incubation experiments using15N isotope tracing technology for the precise measurement and tracking of nitrogen transformations as previous work57. Genomic DNA was extracted from the sludge samples was performed using the DNeasy Power Soil DNA Kit (Qiagen, Germany) following the manufacturer’s instructions. The DNA concentration and purity were measured using an ultraviolet microspectrophotometer (K5500; Kaiao, China). DNA samples were stored at −20 °C for preservation and subsequent experiments. Quantitative polymerase chain reaction (qPCR) was employed to assess the abundance of target genes in the three reactors. The nrfA gene, responsible for encoding nitrite reductase enzymes, was used as a molecular marker to quantify DNRA bacteria. nirS and nirK, responsible for the reduction of NO2− to NO, were used as biomarkers for denitrifying bacteria58. The qPCR analysis was performed according to Zhao et al.1.

Amplification PCR and high-throughput sequencing and microbial association network construction

The nrfA gene and bacterial 16S-rRNA gene were amplified by PCR, using the barcode-primers set nrfAF2aw (5’-CARTGYCAYGTBGARTA-3’)/nrfAR1 (5’-TWNGGCATRTGRCARTC-3’) and 515F (5’-TWNGGCATRTGRCARTC-3’)/907R (5’-CCGTCAATTCMTTTRAGTTT-3’)28. PCR amplification products were detected by agarose gel electrophoresis and purified using a gel purification kit (AXYGEN, USA) for the fragment excision and recovery. A microplate reader (FLx800, BioTek, USA) and Quant-iT PicoGreen dsDNA Assay Kit were used to fluorescently quantify the amplified recovery products. The PCR amplicons were subjected to 2 × 300 bp paired-end sequencing using the Illumina NovaSeq platform. PICRUSt2 was then employed to predict the potential of each sample using 16S rRNA amplicon sequencing data. KEGG Orthology (KO) was utilized to classify all genes that share homology with specific genes known to function within the same category59. Microbial association network was constructed to assess the relationship of dominated genera. R language was employed to compute the number of nodes and edges, average degree (avgK), negative-positive linkage ratio (NP), and modularity. Subsequently, Gephi software (version 0.10.1) was utilized to produce visual graphics, presenting the resulting network diagram. The code used in R language was provided as below:

setwd(“***“)

library(igraph)

library(psych)

library(Hmisc)

library(vegan)

library(dplyr)

library(vegan)

library(reshape2)

otu=read.csv(“***.csv”,header = T,row.names = 1)

otu<-as.matrix(t(otu))

occor=corr.test(otu)

occor.r = data.frame(occor$r)

occor.p = data.frame(occor$p)

occor.r[occor.p>0.2|abs(occor.r)<0.5] = 0

diag(occor.r) <- 0

occor.r[upper.tri(occor.r)] <- 0

df=melt(as.matrix(occor.r))

df$Var1=as.character(df$Var1) #

df$Var2=as.character(df$Var2)

df1=subset(df,!df$Var1==df$Var2)

colnames(df1)=c(“Source”, “Target”, “Weight”)

df1=subset(df1,!df1$Weight==0)

write.csv(df1,“edge.csv”,row.names = F)

df2=data.frame(id=unique(c(df1$Source,df1$Target)))

write.csv(df2,“node.csv”,row.names = F)

Analysis of nrfA-functioning bacterium Bacteroides fragilis and transcriptome sequencing

Bacteroides fragilis was purchased from the Chinese Industrial Microbial Strain Collection Management Centre (strain number ATCC25285). The medium number is CM0786, Tryptone Soy Agar Medium (5% defibrinated sheep blood + TSA). The medium includes 15 g tryptone, 5 g soy peptone, 5 g NaCl, 15 g agar, and distilled water 1 L. The medium was sterilized at 121 °C for 15 min and 5% defibrinated sheep blood was added. Then the medium was added to the serum bottles of 250 mL volume. The whole process was carried out under anaerobic and aseptic conditions, and N2 was used as protective gas for configuration and transfer. The incubation temperature was 37 °C. The serum bottles were placed in a gas bath shaker at 180 rpm and 37 °C for 12 h to obtain the bacterial broth, which was centrifuged at 7000 rpm, then washed with isotonic solution and centrifuged again to obtain the bacterial broth. 50 mL of isotonic solution was added and placed in a refrigerator at 4 °C for 12 h. The purpose of refrigeration was to reduce cell metabolism while consuming ammonia, and the concentration of ammonia nitrogen in the broth decreased significantly after refrigeration. Before the start of each experiment, the resting bacterial solution was resuspended in the serum bottles. To find more suitable conditions for Bacteroides fragilis, the culture was carried out using different NO2−-N concentration gradients, using glucose as the carbon source and controlling the COD/NO3−-N ratio at 7:1. The irradiation experiments were performed with different IR-EMF irradiation using a photocatalytic reactor with an effective volume of 1 L. The water bath could effectively prevent the thermal effect of the IR-EMF apparatus. The experimental period was 2 h. IR-EMF intensities of 0, 0.04, and 0.06 μT were applied to investigate the effect of IR-EMF on Bacteroides fragilis. Then the bacterial solution was collected and centrifuged. After pouring off the supernatant, the bottom slime was placed in −80 °C refrigerator for transcriptome analysis.

Total RNA of Bacteroides fragilis was extracted using TransZol Up Plus RNA kit (Transgen, Beijing, China). Upon assessing the quality of RNA, mRNA molecules were enriched and fragmented into shorter fragments. These fragmented mRNA molecules were then utilized as templates for the construction of cDNA libraries, which were subsequently sequenced using Illumina HiSeq™ 2000. After removing the raw reads with adapter, the filtered reads were mapped to the Bacteroides fragilis reference genome (NCBI Accession No. ASM1688992v1) using HISAT software60. Based on the results of the alignment, the expression of each gene was calculated. The samples underwent additional analysis to examine expression differences, perform enrichment analysis, and conduct clustering analysis61. Finally, the results were visualized.

The effect of IR-EMF on functional bacteria and reverse transcription qPCR (RT-qPCR)

To further validate the mechanism proposed by the transcriptome analysis. Thauera sp. RT1901 and Shewanella oneidensis MR-1 were selected for nitrogen transformation experiments and specific functional genes were detected. Single colony of Thauera sp. RT1901 and Shewanella oneidensis MR-1 was inoculated into 300 mL the Luria broth medium in triplicate. The cultures were incubated at 30 °C for 24 h with agitation to facilitate growth. Then the cultures were centrifuged at 8000 rpm, followed by triple washes with sterile phosphate-buffered saline and ultrapure water. Subsequently, the cultures were resuspended in the nitrogen transformation medium. The nitrogen transformation medium for Thauera sp. RT1901 and Shewanella oneidensis MR-1 contained the following components: 0.5 g/L sodium acetate, 0.15 g/L NaNO2, 1.0 g/L NaCl, 0.5 g/L MgCl2, 0.3 g/L KCl, 0.015 g/L CaCl2, and 1 mL/L trace element solution42. Then the cultures were dispensed in a photocatalytic reactor with an effective volume of 1 L, and the concentration of the bacterial solution was adjusted at OD600 = 0.3 ~ 0.5. The water bath within the photocatalytic reactor could effectively prevent the thermal effect of the IR-EMF generator. By adjusting the distance between the photocatalytic reactor and the IR-EMF generator, the IR-EMF intensity at the center of the photocatalytic reactor was set to 0 μT, 0.04 μT, and 0.06 μT. At the end of the batch experiments, the cultures were centrifuged at 8000 rpm and then RNA was extracted and reverse transcribed using the PerfectStart Uni RT&qPCR kit (TransGen Biotech Co., Ltd. China). RT-qPCR was performed to quantify the gene expression level of functional genes by the Roche Light Cycler 480 Real-Time PCR system (Switzerland). The primer sequences of functional genes used in the RT-qPCR were placed in Table 1.

Table 1 Specific primers of target genes in RT-qPCR
Full size table