Characteristics and Nitrogen Removal Performance Optimization of Aerobic Denitrifying Bacteria Bacillus cereus J1 under Ammonium and Nitrate-Nitrogen Conditions

Abstract

A novel aerobic denitrifying bacterium Bacillus cereus J1 was isolated from a sewage treatment plant. Its characteristics under two distinct nitrogen sources were systematically investigated. According to the results of whole-genome sequencing, we inferred that strain J1 removes nitrogen through processes such as aerobic denitrification, dissimilatory nitrate reduction to ammonium, and ammonium assimilation. The degradation process of COD and total inorganic nitrogen (TIN) correlated to the zero-order degradation kinetics equation, and the maximum removal rate of NO₃⁻−N reached 3.17 mg/L/h and that of NH₄⁺−N was 3.79 mg/L/h. Utilizing single-factor experiments and response surface methodology, the optimal conditions for nitrate removal were determined as a shaking speed of 115 rpm, COD/nitrogen mass (C/N ratio) of 12.25, and salinity of 3.44 g/L, with the C/N ratio exerting the most significant influence. Similarly, for the maximum ammonium removal, the ideal conditions involved a shaking speed of 133 rpm, C/N ratio of 29, and salinity of 13.30 g/L, with the shaking speed exerting the most significant influence. These findings demonstrate that large amounts of ammonium and nitrate can be quickly removed with the help of Bacillus cereus J1, indicating that strain J1 may be applied to alleviate nitrogen pollution in aquatic environments.

1. Introduction

The rapid increases in industry and improvements in living standards have led to a significant release of nitrogen into the environment. High concentrations of nitrogen pollutants in water can cause eutrophication, lead to the proliferation of harmful algae, and harm the human body [1]. Therefore, it is crucial to remove nitrogen (i.e., ammonium−nitrogen NH₄⁺−N, nitrate-nitrogen NO₃⁻−N) from wastewater to protect the environment [2]. Traditionally, the construction of separate anoxic denitrification and aerobic nitrification systems has been required to achieve effective nitrogen removal [3]. However, this approach consumes more land, increases operating costs, and complicates the operational process. Aerobic denitrification microorganisms (ADMs) can utilize both oxygen and nitrate as electron acceptors simultaneously [4,5], effectively facilitating denitrification even in the presence of dissolved oxygen [6,7]. Applications of ADMs have been proven to be effective, promoting the formation of simultaneous nitrification/denitrification (SND) [8], short-cut nitrification/denitrification [9], and simultaneous nitrogen and phosphorus removal [10]. To maximize the performance of ADMs in practical applications, it is necessary to investigate the growth patterns [11], carbon metabolites [12], nitrogen metabolism [13], and optimum conditions [14] required to achieve the best performance for the two most common nitrogen pollutants in wastewater.

An unfavorable environment, strain competition, protozoan grazing, and the withdrawal phase of the reactor all lead to the failure of the bioaugmentation strategy of ADMs [15,16]. Therefore, the operating conditions should be optimized to achieve the practical application of ADMs. Several factors, such as dissolved oxygen concentration (DO), the ratio of COD to nitrogen mass (C/N ratio), salinity, and carbon source, play a significant role in influencing the aerobic denitrification process and nitrogen removal performance of ADMs [17,18]. Previous studies have revealed that different species of ADMs exhibit distinct responses to these factors [13,14,15,16]. For instance, Stenotrophomonas maltophilia achieved a maximum TN removal efficiency of 94.43% at 120 rpm and C/N = 7.5 [9]. Acinetobacter junii YB achieved the most efficient denitrification ability when using succinic acid as a carbon source at C/N 15, pH 7.5, 37 °C, and 200 rpm [19]. Moraxella sp. LT-01 achieved the highest ammonia removal capacity at 10 °C and C/N = 10, with sodium citrate as the carbon source and DO > 8 mg/L [20]. In addition, it is necessary to consider the combined effects of several factors in one system. Response surface methodology (RSM) can be used to design experiments, establish and analyze models, evaluate the interactive effects of multiple factors, and determine the optimal conditions; the methodology has been utilized in studies of AMDs [14,21,22,23].

In the present study, a novel aerobic denitrifying bacteria strain J1 was isolated from activated sludge and identified by its 16S rRNA. Whole-genome sequencing refers to the sequencing of the entire genome sequence of an organism, which can be used obtain complete genomic information. To further understand the carbon and nitrogen metabolism of strain J1, we annotated and organized the relevant genes. To compare the effect of two distinct nitrogen sources, strain J1 was inoculated into nitrification medium (NM) with ammonium-nitrogen as the sole nitrogen source and denitrification medium (DM) with nitrate-nitrogen as the sole nitrogen source, separately. Batch tests were then performed to evaluate the kinetics and optimal parameters for ammonium and nitrate removal. To optimize the nitrogen removal condition, Box–Behnken experiments were designed, and the RSM was executed. We thus aimed to investigate the effects of these two nitrogen sources on the growth, nitrogen metabolism, and carbon metabolism of strain J1, and to optimize the nitrogen removal conditions of strain J1 to improve its ability to treat nitrogen-containing wastewater.

2. Materials and Methods

2.1. Culture Medium and the Isolation and Identification of Strain J1

Luria–Bertani medium (LBM) was used for enriching bacteria; simultaneous nitrification and denitrification medium (SNDM) was used for culturing denitrifying bacteria; bromothymol blue medium (BTBM) was used for identifying denitrifying bacteria; denitrification medium (DM) and nitrification medium (NM) were used for investigation of the nitrogen removal performance of strain J1. The composition of the culture media is described in the Supplementary Materials.

The source of this isolate was activated sludge collected from the Lie De wastewater treatment plant in Guangzhou, China. Firstly, 10 mL of fresh sludge sample was incubated into a 250 mL shaking flask with 100 mL of SNDM on a rotary shaker (150 rpm) at 30 °C. Every 2 days, we inoculated 3 mL of suspension into the new SNDM. The culturing procedure was maintained for 2 weeks. The resulting bacterial suspensions were inoculated into fresh SNDM (3% inoculation), and the nitrogen removal efficiency of the upper supernatant was determined. Adopting the gradient dilution coating method, 0.2 mL of bacterial suspension was dispersed on an LB agar plate and incubated at 30 °C. We lined the cultivated colonies with different appearances on a fresh BTBM plate using an inoculation ring. We harvested several purified colonies through repeated scribing procedures and tested their denitrification performance using SNDM. Finally, a fast-growing isolate with excellent denitrification ability was obtained and stored in a bacterial suspension containing glycerol (25% v:v) at a temperature of −20 °C.

For identification, once the logarithmic growth phase was reached, the LBM enriched with strain J1 was sent to Sangon Biotech Co., Ltd., (Shanghai, China), who conducted the whole-gene sequencing. Furthermore, the acquired 16S rRNA sequence was compared with that of other microorganisms in the GenBank database via the online Nucleotide BLAST program (https://blast.ncbi.nlm.nih.gov/Blast.cgi) and was deposited in the NCBI database with accession number for nucleotide sequences: SUB13813896 Seq1 OR608012 (accessed on 28 July 2023). Additionally, a phylogenetic tree was constructed with MEGA version 11 software using the neighbor-joining method using the maximum composite likelihood model, with 1000 bootstrap replicates [24], and the gene annotation was executed with eggnog-mapper [25].

2.2. Kinetics Experiments

One loop of isolate dipped from a glycerol tube was inoculated into a 250 mL shake flask with 100 mL of LBM and activated for 24 h as the first-generation inoculum. Afterward, we extracted 1 mL to 100 mL of LBM from it and incubated it in a constant-temperature shaker at 150 rpm and 30 °C for 24 h as the second-generation inoculum. Injecting DM and NM with 1% of the inoculation amount, the initial OD₆₀₀ was about 0.020, and the bacterial population was around 8 × 10⁶. The following experiments were also the same. Samples were extracted at 0, 4, 8, 12, 16, 20, 24, 30, 42, 48, 66, and 72 h. After centrifugation (8000 rpm, 5 min) and filtration (by 0.45 μm PES membrane filters, Tian Jin JinTeng Co., Ltd., Zhejiang, China), the supernatant of the samples was used for detection, and the sum of NH₄⁺−N, NO₃⁻−N, and NO₂⁻−N was defined as the total inorganic nitrogen (TIN).

2.3. Optimization of Nitrogen Removal Performance

Single-factor and Box–Behnken experiments were performed at 30 °C and pH = 7.0 ± 0.2 with 100 mL of DM and NM. All samples in the single-factor and Box–Behnken experiments were extracted at 24 h. Four single factors, namely, shaking speed, C/N ratio, salinity, as well as carbon source, were considered in the single-factor experiments. The suitable range of the single factors was determined through single-factor experiments. Then, Box–Behnken experiments were employed to determine the optimal shaking speed (from 90 rpm to 150 rpm), C/N ratio (from 10 to 30), and salinity (0 g/L to 10 g/L). Consequently, we assessed the degree of the effect and the interaction of the three independent variables as well as the optimal conditions for strain J1 in DM and NM when sodium succinate was utilized as the carbon source.

2.4. Analytical Methods

The OD₆₀₀ and pH of the suspension as well as the concentrations of COD, TN, NH₄⁺−N, NO₃⁻−N, and NO₂⁻−N in the supernatant after filtration were measured. The biomass of strain J1 was determined with a UV–VIS spectrophotometer (EPOCH2 microplate reader, BioTek Instruments, Inc., Winooski, VT, USA) at 600 nm. The pH was determined using a pH electrode (PHS-3E, INESA Scientific Instrument Co., Ltd., Shanghai, China). The DO was measured with a portable dissolved oxygen analyzer (JPSJ-605F, Shanghai Yidian Scientific Instrument Co., Ltd., Shanghai, China). According to the State Environmental Protection Administration of China, the NH₄⁺−N concentration was measured using Nessler’s reagent colorimetric method at a wavelength of 420 nm; COD measurement was via rapid digestion spectrophotometry at 440 nm. NO₃⁻−N was measured with the phenol disulfonic acid photometry method; TN was determined according to the standard UV spectrophotometry method. NO₃⁻−N and TN were measured at 220 nm and 275 nm, respectively; the data in this experiment were analyzed with a one-way analysis of variance (ANOVA) test (p < 0.05) using SPSS Statistic version 26 software, and figures were made with Origin version Pro 2022 software. Each experiment had three repetitions, and the results are presented as mean ± standard deviation of the mean (SD).

3. Results

3.1. Isolation and Identification of Strain J1

After two weeks of culturing, strain J1 showed the best nitrogen removal performance and was therefore selected. The results of 16S rRNA sequencing demonstrated that strain J1 is affiliated with the genus Bacillus and is closely related to Bacillus cereus str. Schrouff and Bacillus cereus ATCC 14579 according to comparisons with data in the NCBI database. The genome features of strain J1 can be obtained from Table S1. Several aerobic denitrifying Bacillus bacteria have shown enormous potential for nitrogen removal [26,27]. To better demonstrate the phylogenetic relationship between strain J1 and other related strains, a neighbor-joining phylogenetic tree was constructed (Figure S1).

The results of the annotating carbon metabolism and nitrogen metabolism genes using eggNOG-mapper (http://eggnog-mapper.embl.de/ accessed on 28 July 2023) are shown in Table S2 [25]. In our study, various nitrate reductase genes such as narG, napA, and nasA were simultaneously identified in strain J1. The enzyme NapA edited by gene napA can assist with denitrification in the presence of oxygen [28,29]. Therefore, from a genetic point of view, we speculated that strain J1 is an aerobic denitrifier [30]. Furthermore, nitrite reductase gene nirS, nitric oxide reductase gene norB, and nitrous oxide reductase nosZ were annotated, demonstrating that strain J1 may have the capability to perform complete denitrification. Additionally, dissimilatory-type NADH-nitrite reductase genes nirB and nirD as well as ammonium transporters genes amtB and nrgA, which can reduce nitrite to ammonium, were annotated. Glutamine dehydrogenase coding gene glnD as well as glutamate synthetase coding genes gltB and gltD can convert to glutamine and further be used for the synthesis of proteins as well as nitrogen-containing metabolites like purines, pyrimidines, and amino sugars [31,32,33,34]. In conclusion, the annotation results showed that strain J1 has potential three nitrogen removal pathways (as depicted in Figure 1): the dissimilatory nitrate reduction to ammonia (DNRA) pathway (Nar/Nap/Nas ⟶ NirBD ⟶ NH₄⁺), the denitrification pathway (Nar/Nap ⟶ NirS ⟶ NorB ⟶ NosZ ⟶ N₂), and the GDH/GS-GOGAT assimilation pathway (GlnD/GltBD ⟶ glutamate/glutamine ⟶ proteins synthesis) [35].

In addition, we speculated that strain J1 can also metabolize various nitrogen sources such as urea (urease encoded by gene ure, urea ABC transporter encoded by gene urt), cyanate (enzyme Cyns encoded by cyns can catalyze the reaction of cyanate with bicarbonate to produce ammonia and carbon dioxide), allantoin (enzyme encoded by allA), formamide (specific formamide hydrolase encoded by gene amiF), and nitroalkanes (nitroreductases related to genes ydh, ydg, ydj, and yod).

3.2. Kinetics Experiments with Two Nitrogen Sources of Strain J1

3.2.1. Growth Characteristics of Strain J1

The growth rate of bacteria affects the start-up time of sewage treatment systems in practical applications, so it is important to investigate the growth characteristics of bacteria. The growth curve of strain J1, as represented by the OD₆₀₀, was fitted with a logistic model, as shown in Figure 2. In both DM and NM cultures, the fitting curves exhibited high correlation coefficients of 0.989, indicating that the logistic equation adequately captured the growth characteristics of strain J1, particularly within 0 to 48 h. Additionally, to better demonstrate the growth pattern of strain J1, we performed first-order differentiation on the fitted logistic curve. Based on the instantaneous growth velocity (ν), the growth characteristics of strain J1 were finally divided into four distinct phases [11]. In DM culture, strain J1 exhibited minimal growth during the initial 6 h (defined as the lag phase). After that, ν increased from 6 h to 14 h, which was defined as an exponential phase. From 14 h to 25 h, the growth rate of OD₆₀₀ slowed down, and ν decreased, indicating entry into the stationary phase. After 25 h, the death rate rapidly increased and entered the decline phase. Until 42 h, ν was close to zero, at which point OD₆₀₀ reached its highest value (1.032 ± 0.179). In NM culture, after a lag period of 2.2 h, strain J1 underwent swift proliferation. After an exponential phase of 10.8 h, OD₆₀₀ reached about 0.737 ± 0.084 and then continuously increased to the maximum of 1.627 ± 0.024 at 48 h.

Even with only 1% inoculation, strain J1 still exhibited a short lag period in DM and NM, suggesting its potential to reduce the start-up time of reactors. Compared to that in the DM culture, strain J1 demonstrated significantly increased biomass, shorter lag time, and longer exponential and stationary phases in the NM culture. Strain preference for NH₄⁺−N rather than NO₃⁻−N has also been reported [36,37,38]. This is attributed to the requirements for enzymes (like NapA, NarGHJ, and NirBD) and the additional and complex processes involved in the conversion of NO₃⁻−N to bio−N.

3.2.2. Characteristics of pH and Substrate Degradation for Two Nitrogen Sources

The results of the COD and nitrogen removal performance of strain J1 in 24 h for two nitrogen sources are summarized in Table S3. In terms of pH, the results are depicted in Figure 3A. Traditionally, denitrification, decarboxylation, and dehydrogenation processes result in an increase in pH along with the increased alkalinity, while the nitrification process is accompanied by the reduction in alkalinity, leading to a decrease in pH [39]. However, in our study, the pH in both media continuously increased, which verified that the nitrification process did not occur or occurred weakly. Therefore, we speculated that NH₄⁺−N was predominantly removed through the assimilation pathway.

As shown in Figure 3C,D, in DM, the initial TN was 97.86 ± 0.47 mg/L, which closely approximated the initial TIN of 97.30 ± 2.55 mg/L. Significantly, NH₄⁺−N was detected at both 4 h and 8 h, with concentrations of 4.20 ± 1.49 mg/L and 3.28 ± 1.27 mg/L, respectively, while the accumulation of NO₂⁻−N began at 12 h, reaching 1.13 ± 0.48 mg/L. Subsequently, 3.86 ± 0.38 mg/L NO₂⁻−N was detected at 16 h, 3.28 ± 0.48 mg/L at 20 h, and only a trace amount of NO₂⁻−N (around 0.1 mg/L) was detected. At 42 h, the concentration of NH₄⁺−N in the supernatant increased, reaching a final concentration of 4.89 ± 3.71 mg/L. The period from 0 to 8 h saw strain J1 primarily converting NO₃⁻−N into bio−N through the DNRA and assimilation pathways, as indicated by the presence of a small amount of NH₄⁺−N. Between 12 h and 20 h, the rapid decrease in nitrate (from 70.06 ± 7.09 mg/L to 17.25 ± 1.13 mg/L) could be attributed to two main factors: the substantial consumption of nitrogen sources due to biomass increase and the shift in the primary nitrogen removal pathway towards denitrification, resulting in faster nitrogen source removal. Additionally, the accumulation of nitrite during this stage suggested a higher activity of nitrate reductase (Nap) than of nitrite reductase (Nir) [40]. In NM, the initial concentration of NH₄⁺−N was measured as 96.08 ± 0.48 mg/L, which was comparable to the tested TIN concentration of 95.37 ± 2.60 mg/L. From 0 to 4 h, with NH₄⁺−N ranging from 95.30 ± 2.60 mg/L to 79.81 ± 2.52 mg/L, strain J1 mainly underwent assimilation and rapid amplification. From 8 h to 48 h, there was a significant production of NO₃⁻−N, with an average concentration of 3.29 mg/L. At 72 h, the final NO₃⁻−N concentration was 1.20 ± 0.06 mg/L, indicating the presence of a weak nitrification process. However, almost no NO₂⁻−N (less than 0.10 mg/L) was detected throughout the entire process, confirming that the majority of NH₄⁺−N was removed through the assimilation process.

3.2.3. The Fit Curve between TIN and COD

As presented in Figure 3B, in DM, the initial COD was 2411.73 ± 34.34 mg/L. The COD concentration decreased to 54.22 ± 8.60 mg/L at 24 h. In NM, the initial COD was 2287.17 ± 65.43 mg/L, and COD decreased to 184.35 ± 43.21 mg/L within 24 h. As shown in Figure 3E, from the initial inoculation to 24 h, a strong linear relationship between TIN and COD concentrations was observed (R² = 0.972 in DM, R² = 0.980 in NM). In DM, it took 22.22 mg/L COD to remove 1 mg/L TIN, whereas in NM, nearly 26.48 mg/L COD was required for the same TIN removal. This difference could be attributed to the fact that in DM, NO₃⁻−N was eliminated from the flask as N₂, while in NM, a significant portion of NH₄⁺−N was converted into bio−N, resulting in a lower TIN removal efficiency due to the mutual conversion between NH₄⁺−N and bio−N, and the higher biomass necessitated a larger carbon source for proliferation. In DM, after 24 h of cultivation, the COD removal efficiency reached its maximum value of 97.62% (degradation rate of 100.70 mg/L/h), and TIN achieved a maximum removal efficiency of 98.33% (degradation rate of 3.19 mg/L/h) at 30 h. In the initial stage, COD and TIN were rapidly adsorbed to fulfill the exponential growth requirements, resulting in a fast removal rate. Bacterial lysis led to the release of endogenous carbon source and bio−N when the substrate concentration limited the growth rate, leading to a slight increase in COD concentration. In NM, within 20 h, COD exhibited a rapid decrease, with a removal efficiency of 88.74% (degradation rate of 82.25 mg/L/h), and the TIN removal efficiency was 79.64% (degradation rate of 3.79 mg/L/h).

As shown in Figure 3F,G, in DM, the degradation curves of COD and TIN well fit the zero-order kinetic equations (COD: R² = 0.990, TIN: R² = 0.965) within 24 h. In NM, the zero-order kinetic equation (COD: R² = 0.967, TIN: R² = 0.967) was also well fit within 24 h. Compared to the degradation curves in NM, strain J1 degraded the carbon source faster and more of it with proliferation in DM. The amount of TIN degraded in DM exceeded that in NM at 9.97 h. When NO₃⁻−N served as the nitrogen source, strain J1 tended to use the denitrification pathway for energy metabolism, requiring more organic substrate to provide electron donors for denitrification.

3.3. Single-Factor Experiments for Two Nitrogen Sources with Strain J1

3.3.1. Shaking Speed

Maintaining an appropriate concentration of dissolved oxygen (DO) is crucial for growth and achieving optimal denitrification efficiency [41]. In this study, the effect of the DO value on the nitrogen removal of strain J1 was investigated by using different shaking speeds of 60, 90, 120, 150, and 200 rpm, which are equivalent to DO values of 1.51, 2.46, 3.23, 4.57, and 6.08 mg/L. As depicted in Figure 4, in DM and NM, there was a trend of first increasing and then decreasing with increasing shaking speed, reaching the highest growth rate at 120–150 rpm. The energy yield of aerobic respiration was higher than that of denitrification; therefore, with the increase in oxygen content, the energy utilization efficiency increased [42]. The oxygen support required by coenzymes in the TCA cycle and the NADH produced in the TCA cycle significantly affected the denitrification and bacterial growth processes [39,43,44]. Enzymes related to nitrogen assimilation, like Nas, are inhibited by a high DO, so the OD₆₀₀ at 200 rpm decreased [42]. In addition, the COD removal efficiency increased with the increase in shaking speed, because the increase in DO promoted strain J1 to use organic carbon sources as electron donors, consuming more of the carbon sources. In terms of nitrogen conversion, in DM, the bacteria exhibited better denitrification efficiency at 90 rpm to 150 rpm. Significant nitrite accumulation was observed at low shaking speeds as carbon sources were almost not consumed, leading to insufficient energy and electron supply. When the shaking reached 200 rpm, we found a more pronounced accumulation of ammonium in DM. Ammonium can only be produced through the DNRA or bacterial lysis process [45,46]. When bacteria were in the early stages of growth, the DNRA process for reproduction was the reason for this result, and by the time of bacterial decline, a large number of bacteria had already died and lysed. According to Figure 2 (entering the decline phase after 25 h), we think that excessive shaking speed accelerates bacterial death, leading to ammonium accumulation.

3.3.2. C/N Ratio

Heterotrophic bacteria rely on carbon sources as energy supply (ATP) and electron donors (NADH and FADH₂) for their growth and denitrification processes [47]. As depicted in Figure 5, in DM, OD₆₀₀ increased with the increase in C/N, while there was not much difference in the OD₆₀₀ when C/N > 10 in NM. The reason for this difference was that eight donors are needed in the reduction from NO₃⁻−N to NH₄⁺−N and then to bio−N, and there are more electron donors to reduce nitrate nitrogen with the increase in the C/N [48]. However, in NM, the deficiency of NH₄⁺−N became the main factor limiting bacterial growth, resulting in the OD₆₀₀ not showing a significant increase with the increase in the C/N. At C/N = 10, the COD removal efficiency reached its maximum value. However, we observed that at C/N = 3, the COD removal efficiency in NM was much higher than that in DM. This may have been because the lack of electron donors significantly inhibited the activity of nitrite reductase, resulting in poor denitrification and DNRA performance [49]. However, the lack of a carbon source had less of an impact on the removal of NH₄⁺−N than NO₃⁻−N. In DM, the optimal C/N should be between 10 and 20, while in NM, a larger carbon source is required as there is more biomass in NM.

3.3.3. Salinity

Salt stress induces ion imbalances and osmotic stress in bacteria, posing a threat to various metabolic processes and cell growth [50]. The “salt-in strategy” and “organic-osmolyte mechanism” are widely used by microorganisms to mitigate salt stress [51]. Previous research indicated that high-salt environments can hinder microbial biomass increase and electronic functioning, resulting in poor nitrate removal efficiency, increased nitrite accumulation, and N₂O release [52]. As shown in Figure 6, strain J1 displayed tolerance to salinity levels of 20–30 g/L in both DM and NM but exhibited higher biomass in NM. In DM, the OD₆₀₀ at a salinity of 30 g/L was only half of that at 20 g/L, indicating that strain J1 in DM was severely affected by the higher salinity. However, in NM, the OD₆₀₀ at 30 g/L was 93.38% of the value at 20 g/L, suggesting only slight inhibition. Similar trends were observed in the COD removal efficiency and nitrogen conversion ratio in both DM and NM. In DM, the COD removal efficiency at 30 g/L salinity was only 71.80% of that at a salinity of 20 g/L. At 20 g/L salinity, although the nitrate conversion ratio exceeded 94%, there was significant nitrite accumulation, indicating inhibition at the subsequent stages after nitrate reduction to nitrite due to high salinity. This observation aligns with those of previous studies showing a salinity sensitivity order of nosZ > cnorB ≈ nirS > napA [53]. In NM, the COD removal efficiency at 30 g/L salinity was 86.92% of that at 20 g/L, and there was almost no significant difference in nitrogen conversion, indicating that the assimilation process was not easily affected by salinity. In our study, genes amtB and otsA were annotated in strain J1. The presence of the Amt ammonium transporter family can alleviate the ammonia toxicity caused by salt stress, which may explain why strain J1 can tolerate higher salinity in NM [53]. Additionally, gene otsA is involved in osmo-protection via the biosynthesis of trehalose, which is beneficial for bacterial survival in higher-salinity environments [54,55].

3.3.4. Carbon Source and Carbon Metabolism Genes

ADMs possess multiple metabolic pathways for different carbon sources, but they vary in conversion efficiency and energy yield [56]. The degradability, chemical structure, and molecular weight of the carbon source can impact microbial enzyme activity, metabolic system, growth, and nitrogen degradation [41]. As shown in Figure 7, strain J1 exhibited distinct differences in the presence of different carbon sources in DM and NM. Firstly, in DM, strain J1 had the highest OD₆₀₀ in glucose, indicating that DNRA and the assimilation pathway can be stimulated by glucose, promoting the conversion of nitrate nitrogen to biological nitrogen. On the contrary, in NM, strain J1 had a higher OD₆₀₀ in organic acid carbon sources. Secondly, whether in DM or NM, when glucose was used as the carbon source, the pH was lower, which may have been due to the production of small organic acids during glucose metabolism, resulting in a slight decrease in pH [57]. In addition, in DM, there was no significant difference in the removal efficiency of strain J1 of these four types of carbon sources. However, in NM, strain J1 only removed 51.65% of glucose, while the removal efficiency of organic acid carbon sources reached over 90%. An interesting phenomenon was that strain J1 exhibited similar behavior in both DM and NM in sodium acetate and sodium succinate, but, in glucose and trisodium citrate, strain J1 exhibited significant differences in the nitrogen conversion and denitrification pathways. In DM, with glucose and trisodium citrate, nitrite accumulation and NH₄⁺−N generation (9.68% and 19.34%, respectively) were observed, demonstrating that denitrification was inhibited, and the DNRA process was activated.

As shown in Table S2 and Figure S2, strain J1 was confirmed to be a heterotrophic bacterium due to the lack of carbon fixation genes and the presence of organic carbon degradation genes. The glucose degradation pathway and TCA cycle are the central metabolic pathways involved in carbohydrate oxidation and energy production in most heterotrophic organisms, which apply to strain J1 [43]. Acetic acid, citric acid, and succinic acid can all directly enter the TCA cycle, but the metabolic process of glucose involves more enzymes such as glucose dehydrogenase, encoded by gcd, and phosphoglucomutase, encoded by pgm. Therefore, the efficiency of obtaining energy from glucose was lower than in the other three small-molecule carbon sources [18]. In the TCA cycle (refer to Table S2), we found that the pathway from citrate to oxaloacetate was infeasible, which may have affected the efficiency in the reversible conversion to oxaloacetate, resulting in insufficient NADH/H⁺ to continue reducing nitrite in DM. Therefore, when citrate was used as the carbon source, the accumulation of nitrite was observed. In DM and NM, strain J1 exhibited good performance with both sodium succinate and sodium acetate. Chen et al. pointed out that due to their acidic molecular structure, microorganisms can directly utilize the two carbon sources, which contributes to obtaining energy from these compounds and achieving higher denitrification efficiency [58]. In addition, strain J1 contains succinyl-CoA and acetoacetyl-CoA, which may also have been the reason for its good response to these two carbon sources.

3.4. Box–Behnken Design for Nitrogen Removal Condition Optimization

Previous studies have demonstrated the effectiveness of the RSM as an approach for modeling and optimizing multifactor experiments [14,36,59]. According to the response results of the Box–Behnken design experiments, two quadratic polynomial regression models for TIN removal efficiency (TIN RE%) in DM and NM were fitted as follows:

DM: TIN RE% = 50.64 + 0.80A − 0.30B + 1.23C + 0.0033AB − 0.0049AC + 0.00071BC − 0.0036A² − 0.0082B² − 0.088C² R² = 0.9864

(1)

NM: TIN RE% = −366.54 + 6.57A + 1.50B + 1.18C − 0.0078AB − 0.0054AC − 0.0020BC − 0.028A² − 0.0079B² − 0.017C² R² = 0.9963

(2)

The ANOVA of the quadratic parameters is provided in Tables S4 and S5. To further visualize these relationships, the response surfaces were plotted to show the effects of the tested factors on the TIN removal by strain J1. As illustrated in Figure 8, the mathematical model predicted that the maximum TIN removal efficiency was 95.01% in DM and 93.36% in NM. In DM, the order of importance of the three variables concerning denitrification efficiency was C/N ratio (p < 0.0001) > salinity (p = 0.0005) > shaking speed (p = 0.0140); the interaction of shaking speed and C/N ratio was significant (p = 0.006) and higher than the interaction of shaking speed and salinity (p = 0.0236). And, the interaction of carbon source and salinity (p = 0.8936) was not significant. In NM, the order of importance of the three variables concerning nitrogen removal was shaking speed (p < 0.0001) > C/N ratio (p = 0.0064) > salinity (p = 0.3621); the interaction of shaking speed and C/N ratio was significant (p = 0.0242). The interaction of shaking speed and salinity (p = 0.1799) and the interaction of carbon source and salinity (p = 0.8581) were not significant. By setting the goal of the model to maximize TIN RE, we obtained 100 solutions. In DM, the maximum TIN RE (96.02%) occurred with conditions: shaking speed of 115 rpm, C/N ratio of 12.25, and salinity of 3.44 g/L. Based on this condition, three more experiments were conducted: the final result for the TIN RE was an average of 95.14 ± 0.63%. In NM, the maximum TIN RE (96.46%) occurred with the following conditions: a shaking speed of 133 rpm, a C/N ratio of 29, and a salinity of 13.30 g/L. Based on these conditions, three more experiments were conducted: the final result for the TIN RE was an average of 93.61 ± 0.85%. These final results were closer to the predicted results.

4. Conclusions

A novel aerobic denitrifier strain, Bacillus cereus J1, was isolated and identified. We further investigated its nitrogen removal performance as well as optimum conditions in DM and NM. Strain J1 could remove nitrogen efficiently from the aquatic environments through the denitrification, DNRA, as well as GDH-GS/GOGAT pathways. The kinetics experiments showed that strain J1 can remove NO₃⁻−N and NH₄⁺−N at rates of 3.19 mg/L/h within 24 h and 3.79 mg/L/h within 20 h, respectively, with little nitrite accumulation. Furthermore, strain J1 exhibited better nitrogen removal performance in DM but a higher carbon source demand, better adaption to salinity shock, and a faster growth rate in NM. The RSM model predictions further verified that the maximum TIN removal efficiency could reach 95.14% in DM and 93.61% in NM within 24 h. These results suggest that strain J1 is a suitable candidate for nitrogenous wastewater treatment.