Culturable nitrogen-transforming bacteria from sequential sedimentation biofiltration systems and their potential for nutrient removal in urban polluted rivers

Novel heterotrophic bacterial strains—Bzr02 and Str21, effective in nitrogen transformation, were isolated from sequential sedimentation-biofiltration systems (SSBSs). Bzr02, identified as Citrobacter freundii, removed up to 99.0% of N–NH4 and 70.2% of N–NO3, while Str21, identified as Pseudomonas mandelii, removed up to 98.9% of N–NH4 and 87.7% of N–NO3. The key functional genes napA/narG and hao were detected for Bzr02, confirming its ability to reduce nitrate to nitrite and remove hydroxylamine. Str21 was detected with the genes narG, nirS, norB and nosZ, confirming its potential for complete denitrification process. Nitrogen total balance experiments determined that Bzr02 and Str21 incorporated nitrogen into cell biomass (up to 94.7% and 74.7%, respectively), suggesting that nitrogen assimilation was also an important process occurring simultaneously with denitrification. Based on these results, both strains are suitable candidates for improving nutrient removal efficiencies in nature-based solutions such as SSBSs.

www.nature.com/scientificreports/ transformations. Therefore, in the present study we focused on the characteristics of isolated bacterial strains capable of nitrogen removal. The nitrification involves two consecutive reactions (NH 4 + → NO 2 − → NO 3 − ), and it has been studied in different autotrophic strains: (i) the first reaction was described in ammonia oxidizing bacteria (AOB), in the genera Nitrosomonas, Nitrosospira (β-Proteobacteria) and Nitrosococcus (ϒ-Proteobacteria) 13,14 ; while (ii) the second reaction in nitrite oxidizing bacteria (NOB), in the genera Nitrobacter (α-proteobacteria), Nitrococcus (ϒ-Proteobacteria) and Nitrospina 15 . Nitrification also occurs in direct oxidation of NH 4 + → NO 3 − (complete ammonium oxidation, COMAMMOX) by autotrophic strains of Nitrospira spp. (Class Nitrospirae) 16,17 . Moreover, nitrification via the hydroxylamine (NH 2 OH) pathway, which is an intermediary product between the first nitrification reaction (ammonia oxidation to hydroxylamine), has also been described for Nitrosomonas 18 and heterotrophic strains of Acinetobacter 19 , Janthinobacterium 20 , Alcaligenes 21 , Enterobacter 22 , and Pseudomonas [23][24][25] . Denitrification-a dissimilatory nitrate reduction (DNR) pathway-involves four cascade reactions for the transformation of NO 3 − → NO 2 − → NO → N 2 O → N 2 , which was initially described for heterotrophic facultative anaerobic bacterial strains 26,27 . More recently, research has been focused in the identification of aerobic denitrifying strains that can perform parallel nitrification due to their potential utilization in waste water treatment plants (WWTPs) for the complete removal of nitrogen compounds. Several strains have been isolated and reported to perform simultaneous nitrification-denitrification (SNdN), with the most common genera represented by Acinetobacter, Agrobacterium, Alcaligenes, Bacillus, Klebsiella, Enterobacter and Pseudomonas 28 .
The majority of the above described nitrogen transforming bacteria have been isolated from sewage in WWTPs, constructed wetlands (CWs) or biofilm formations in experimental bioreactors 28 . To our knowledge, the bacteria carrying out nitrogen transformation processes have not yet been isolated and characterized within the SSBSs. Additionally, there is a limited number of studies discussing the nitrogen balance, most of which were in controlled experiments for selected bacterial strains, in order to confirm their preferred metabolic pathways [29][30][31][32][33] .
Therefore, the present study aimed to isolate and characterize heterotrophic bacterial strains that naturally occur in SSBSs, which are responsible for nitrogen transformation in nitrification and denitrification processes. To reach the objective, culturable bacteria were isolated from sediments, nitrogen transformation pathways were determined, and nitrogen balance was described. Additionally, the preference of the strains to perform nitrogen assimilatory over dissimilatory transformation processes was also investigated. Our results were compared with the nitrogen removal efficiency of other published isolated bacterial strains and discussed in the context of biotechnological potential of selected strains to improve the nutrient removal efficiency in NBS technologies.

Results and discussion
Selection and identification of potential nitrogen transforming bacteria. Initial screening of bacteria capable of nitrogen utilization. Ten bacterial strains were selected for their ability to transform nitrogen compounds and were summarized in Table 1. All mentioned strains were able to utilize NO 3 − in Giltay denitrifying medium (GiDM). Seven strains (Str21, Bzr07, Sok01, Sok03, Sok06, Sok20 and Sok41) presented no accumulation of NO 2 − , suggesting that it was further reduced by bacteria (Table 1). In contrast, three strains (Bzr02, Str01 and Sok05), only transformed NO 3 − to NO 2 − , which was then accumulated in the medium with no further utilization (Table 1). In turn, seven among 10 selected strains (Str21, Bzr02, Bzr07, Str01, Sok03, Sok05 and Sok41) were able to utilize NH 4 + on various nitrifying media with different carbon sources ( Table 1). The most efficient removal of NH 4 + was found in nitrifying medium containing glucose-GNM (up to 48 h for the strains Str21, Bzr02, Bzr07, Sok05 and Sok41; Table 1).
Proposed metabolic pathways for nitrogen transformation. Possible bacterial metabolic pathways for nitrogen transformation were described based on the amplification of key functional genes involved in the nitrogen cycling process (Table 1 and supplementary Fig S2). The strains Str01 and Sok05 were considered to be nitrate reducers, since NO 2 − was accumulated in GiDM (Table 1). The above suggestion was supported with the detection of the narG gene (respiratory nitrate reductase), which is involved in the reduction of NO 3 − → NO 2 − in anaerobic conditions (Table 1 and supplementary Fig S2). The strains Sok01, Sok06, and Sok20 were considered to be facultative anaerobic denitrifiers, since they were able to continue the reduction of NO 3 − to gas in GiDM, but could not utilize NH 4 + in any of the nitrifying media in aerobic conditions (Table 1 and supplementary Fig  S2). In contrast, the strains Sok41, Sok03 and Bzr07 were considered to be facultative anaerobic denitrifiers that could also utilize NH 4 + in aerobic conditions (Table 1). All six facultative anaerobic denitrifiers (Sok01, Sok41, Sok20, Sok06, Sok03 and Bzr07) presented the nosZ gene (Table 1), which is involved in the last step of denitrification, and therefore, suggested that they performed complete reduction of NO 3 − → N 2 .
Bzr02 (Citrobacter freundii) and Str21 (Pseudomonas mandelii), isolated from the Bzr-SSBS and Str-SSBS, respectively, presented the best results during the screening experiments on transformation of nitrogen compounds. Both strains were able to grow and remove NO 3 − and NH 4 + in a lower time of incubation in different culture media, and were observed with the highest number of studied key functional genes involved in assimilation, nitrification or denitrification processes (Table 1). Moreover, the Bzr02 was the only strain capable to utilize NH 4 + with the presence of hydroxylamine in GNM, suggesting that hydroxylamine could be an intermediary product in the nitrification process. Therefore, Bz02 and Str21 were selected for further quantitative experiments in nitrogen transformation assays.
Nitrogen transforming processes-strains Bzr02 and Str21. Ammonium transformation in nitrifying medium. Bzr02 and Str21 were cultivated in nitrifying medium (NM) under aerobic conditions, and their growth and utilization of N-NH 4 were followed for 24 h (Fig. 2a,b). The average and maximum removal rates of N-NH 4 for both strains were described in Table 2. Both strains were able to utilize N-NH 4 as a sole nitrogen source. Bzr02 presented a 4 h lag phase with minimal growth at the beginning of the assay (Fig. 2a). The log phase was observed after 4 h of incubation (Fig. 2a), which correlated with the maximum removal rate of N-NH 4 (16.17 ± 0.97 mg L −1 h −1 , Table 2). A stationary phase occurred between 12 and 18 h, however, the strain was able to remove 82.6% of N-NH 4 until 14 h of incubation (Fig. 2a). The maximum removal of N-NH 4 was observed at 22 h of incubation (99.0 ± 0.2%; Table 2). The average removal rate of N-NH 4 was 5.41 ± 0.13 mg L −1 h −1 (Table 2), which was significantly higher from other published strains: Alcaligenes denitrificans WY200811 (0.69 mg L −1 h −1 ) 34 , Klebsiella pneumonae EGD-HP19-C (2.29 mg L −1 h −1 ) 35 , K. pneumonae CF-S9 (4.3 mg L −1 h −1 ) 36 and Enterobacter cloacae CF-S27 (2.22 mg L −1 h −1 ) 22 .
Nitrate transformation in denitrifying medium. Bzr02 and Str21 were cultivated in denitrifying medium (DM) under aerobic conditions, and their growth and utilization of N-NO 3 were followed for 32 h (Fig. 2c,d). The average and maximum removal rates of N-NO 3 for both strains were described in Table 2. Bzr02 was not able to grow and transform N-NO 3 when it was added to the medium as the sole nitrogen source. Similar results were reported for Acinetobacter calcoaceticus HNR 19 , and it was proposed that the strain was sensitive to an initial high concentration of N-NO 3 (40 mg L −1 ) in denitrifying medium. The above observation suggests that Bzr02 was also sensitive to the high initial concentration of N-NO 3 (100 mg L −1 ) in DM.
On the contrary, Str21 was able to utilize N-NO 3 as a sole nitrogen source in DM (Fig. 2d). After a 6 h lag phase, the strain began to grow until the log phase was observed from 12 h of incubation (Fig. 2d). The maximum removal rate of N-NO 3 was 6.66 ± 0.27 mg L −1 h −1 ( Table 2). Str21 removed N-NO 3 to a maximum of 87.7 ± 0.16% during 28 h of incubation. The average removal rate of N-NO 3 was 3.89 ± 0.27 mg L −1 h −1 , which was significantly higher than other strains in the family Pseudomonadaceae: Pseudomonas sp. JQ-H3 (1.78 mg L −1 h −1 ) 33 , P. tolaasii Y-11 (2.04 mg L −1 h −1 ) 39 and P. stutzeri AD1 (1.98 mg L −1 h −1 ) 38 , and other bacteria: Klebsiella   36 and Bacillus cereus GS-5 (2.7 mg L −1 h −1 ) 31 . The formation of N-NO 2 was detected in DM, which was a result from the oxidation of N-NO 3 . A maximum concentration of N-NO 2 was observed at 16 h (18.66 ± 1.68 mg L −1 h −1 ) and decreased until it was completely utilized in 20 h of incubation (Fig. 2d). However, N-NO 3 was not completely removed at the end of the assay (13.54 ± 0.60 mg L −1 in 32 h; Fig. 2d), suggesting that the denitrification process by Str21 was partially inhibited by the aerobic condition.  www.nature.com/scientificreports/ Ammonium and nitrate transformation in simultaneous nitrifying-denitrifying medium. Bzr02 and Str21 were cultivated in simultaneous nitrification-denitrification medium (SNDM) under aerobic conditions, and their growth and utilization of N-NH 4 and N-NO 3 were followed for 36 h (Fig. 2e,f). The average and maximum removal rates of N-NH 4 and N-NO 3 for both strains were described in Table 2. Bzr02 was able to remove 94.1 ± 1.3% of N-NH 4 and 70.2 ± 3.6% of N-NO 3 after 36 h of incubation (Table 2). A total of 16.80 ± 1.24 mg L −1 of N-NO 3 was accumulated in SNDM after 24 h of incubation, with no further utilization by Bzr02 (Fig. 2e).
The formation of N-NO 2 was detected in SNDM, which was a result from the N-NO 3 oxidation. The concentration of N-NO 2 increased to a maximum of 20.02 ± 1.15 mg L −1 after 6 h, however, 9.48 ± 0.99 mg L −1 of N-NO 2 remained accumulated in SNDM from 24 h of incubation (Fig. 2e). The average removal rate of N-NH 4 (5.07 ± 0.09 mg L −1 ) was significantly higher than N-NO 3 (1.44 ± 0.16 mg L −1 ), which suggests that Bzr02 preferred to utilize N-NH 4 in SNDM (Table 2). Similarly, Str21 was able to remove a higher amount of N-NH 4 (95.6 ± 1.5%) than of N-NO 3 (75.4 ± 2.6%) ( Table 2), however, the utilization of N-NO 3 was not significant until after 12 h of incubation (Fig. 2f). The formation of N-NO 2 was detected in SNDM, which was a result from the reduction of N-NO 3 , however, some differences were observed when Str21 was compared to Bzr02: (i) the maximum concentration of N-NO 2 was lower (12.19 ± 0.77 mg L −1 ) and it was observed after 12 h of incubation, and (ii) N-NO 2 was almost completely utilized after 24 h of incubation (Fig. 2f). Moreover, a lower concentration of N-NO 3 (12.07 ± 0.91 mg L −1 ) was accumulated after 24 h of incubation (Fig. 2f), when compared to Bzr02. The average removal rate of N-NH 4 (3.35 ± 0.04 mg L −1 ) was higher than of N-NO 3 (2.29 ± 0.22 mg L −1 ), which also suggested that Str21 preferred to utilize N-NH 4 in SNDM (Table 2). Similar results for other strains, where the removal rate of N-NH 4 was faster than of N-NO 3 , have been described for Klebsiella pneumoniae CF-S9 (3.3 and 2.6 mg L −1 , respectively) 36 and Pseudomonas tolaasii Y-11 (2.13 and 0.52 mg L −1 , respectively) 39 . However, other strains have been found to remove N-NO 3 faster than of N-NH 4 , i.e.: Bacillus cereus GS-5 (2.94 and 2.69 mg L −1 , respectively) 31 and Janthinobacterium svalbardensis F19 (1.19 and 0.62 mg L −1 , respectively) 20 .
Hydroxylamine influence in the ammonium transformation by the strain Bzr02 in nitrifying medium. Bzr02 was cultivated in NM supplemented with NH 2 OH in different concentrations, and the growth and utilization of N-NH 4 and NH 2 OH were followed for 30 h (Fig. 3). The experiment was performed to corroborate the nitrification process by Bzr02 since the oxidized products (N-NO 2 and N-NO 3 ) were not observed during incubation with N-NH 4 as the sole nitrogen source. Bzr02 presented a log phase after 4 h of incubation in the control medium without hydroxylamine, which also corresponded with the maximum removal rate of N-NH 4 (23.80 ± 0.84 mg L −1 , Fig. 3a). When NH 2 OH was added to 10 mg L −1 in NM after 4 h of incubation, the log phase of Bzr02 was observed until after 6 h of incubation (Fig. 3b). The maximum removal of N-NH 4 was 8.03 ± 0.60 mg L −1 h −1 during the addition of 10 mg L −1 NH 2 OH, which was significantly lower when compared to the control (Fig. 3a,b). When 20 and 50 mg L −1 of NH 2 OH were added to NM after 4 h of incubation, the log phase was observed after 8 and 12 h of incubation, respectively (Fig. 3c,d). Moreover, the maximum removal rates of N-NH 4 were 2.05 ± 0.90 and 0.86 ± 0.67 mg L −1 , respectively, which were significantly lower when compared to the control (Fig. 3a,c,d). These results suggested that NH 2 OH, in high concentrations, significantly inhibited the growth of Bzr02, and in consequence, the removal of N-NH 4 . However, the transformation of N-NH 4 was resumed when significant amount of NH 2 OH was removed by Bzr02. Furthermore, N-NO 2 was not detected as product from the oxidation of NH 2 OH (Fig. 3b,c,d). Similar results in other strains have been reported for: Enterobacter cloacae CF-S27 22 , Alcaligenes faecalis 21 , and Thiosphaera pantotropha (formerly Paracoccus denitrificans) 42 .
Confirmation of bacterial nitrogen transforming pathways. The nitrogen balance during the transformation processes for Bzr02 and Str21 was calculated and presented in Table 3. The detection of key functional genes involved in nitrogen cycling was also summarized in Fig. 4, and the results were used to corroborate their nitrogen transforming pathways. For the ammonium transformation assay using NM, Bzr02 and Str21 utilized almost complete nitrogen and incorporated it into their cell biomass (94.7 ± 1.4 and 94.3 ± 2.0 mg L −1 , respectively) www.nature.com/scientificreports/ (Table 3). Only a small fraction of nitrogen was lost for Bzr02 and Str21 (0.75 and 1.25 mg L −1 , respectively; Table 3), suggesting that it was assimilated when N-NH 4 was given as the sole nitrogen source. The nitrification process seemed not to have occurred, especially because the products from the oxidation of N-NH 4 (N-NO 2 and N-NO 3 ) were not significantly detected through the entire assays (Fig. 2a,b). The nitrification process seemed to have occurred for Bzr02 when NH 2 OH was added to NM, which is another intermediary product during the first reaction of nitrification (NH 4 + → [NH 2 OH] → NO 2 − ). Bzr02 removed NH 2 OH from the NM while there was no significant bacterial growth or removal of N-NH 4 (Fig. 3), suggesting that NH 2 OH was oxidized (nitrification) rather than assimilated. Additionally, the detection of the gene hao (hydroxylamine oxidoreductase, HAO) supports the nitrification process by Bzr02 (Fig. 4a); however, the concentration of N-NO 2 -the product from NH 2 OH oxidation-was not significantly detected in all experiments (Fig. 3). These results are different from other strains that produced NO 2 − from the oxidation of NH 2 OH, i.e., Nitrosomonas europaea 18 and Pseudomonas PB16 23 . Other studies suggest that the enzyme HAO also catalyzes a different reaction where NH 2 OH is transformed to nitric oxide (NO) in Alcaligenes faecalis No.4 43 or reduced to N 2 in A. facecalis 44,45 and Acinetobacter calcoaceticus HNR 19 . The above results suggests that Bzr02 could have reduced NH 2 OH to a nitrogen gas (Fig. 4b), rather than being oxidized to NO 2 − in the process of nitrification. In the nitrate transformation assay in DM, only Str21 was able to grow and utilize N-NO 3 as the only nitrogen source (Fig. 2d). The initial nitrogen content in DM (105.0 ± 1.2 mg L −1 ) was utilized by Str21 until 11.40 ± 0.62 mg L −1 remained in the medium at the end of the experiment ( Table 3). The majority of nitrogen was detected in the cell biomass of Str21 (68.2 ± 1.2 mg L −1 ) and 25.4 mg L −1 was estimated to be lost ( Table 3). The above results suggested that Str21 transformed 89.1% of total nitrogen, from which 65.0% was assimilated and the remaining 24.1% was probably lost as a nitrogen gaseous form in the process of denitrification. Str21 was found to contain the gene nasA (assimilatory nitrate reductase, NAS; Fig. 4b) that confirmed the process of assimilatory NO 3 − reduction to NO 2 − , and subsequently to NH 4 + . The gene nasA is involved in the synthesis of cell biomass 46 (Fig. 4d). Moreover, Str21 was found to contain all studied genes involved in the process of denitrification (narG, nirS, norB and nosZ; Fig. 4b), suggesting that it is a facultative anaerobic denitrifier (Fig. 4d). The denitrification activity in anaerobic conditions for a similar strain-Pseudomonas mandelii strain PD30-has already been described with the gene expression of nirS and norB 47,48 . In the above research, it was argued that the gene expression was significantly inhibited in aerobic conditions, and therefore, it was concluded that P. mandelii PD30 performed denitrification in exclusive anaerobic conditions. In contrast, for other Pseudomonas strains, i.e., P. stutzeri YG-24 29 , P. sp. JQ-H3 33 and P. mendocina GL6 49 , the removal of nitrogen content as gas was up to 46.0-74.4% in aerobic conditions, suggesting that there was a similar preference for nitrogen denitrification and assimilation, and sometimes, denitrification could be significantly higher. The detection of the gene napA, rather than the gene narG, was probably the most important factor influencing aerobic denitrification in the above three mentioned strains. In the case of Str21, only the gene narG was detected (Fig. 4b), however, the process of denitrification was not completely inhibited when it was incubated in DM, during aerobic conditions (Fig. 2d). We believe that the aerobic conditions in the media could have partially influenced the reduction of N-NO 3 to subsequent forms of nitrogen for Str21, resulting in an evident preference to assimilate nitrogen rather than performing denitrification.
For the N-NH 4 and N-NO 3 transformation assays in SNDM, Bzr02 and Str21 were able to utilize N-NH 4 and N-NO 3 in aerobic conditions. For Bzr02, a total of 68.3 ± 1.2 mg L −1 of nitrogen was found in the cell biomass and 29.2 ± 2.1 mg L −1 remained in the medium ( Table 3). The remaining nitrogen was mostly from N-NO 3 and the accumulation of its reduction to N-NO 2 , that were not completely depleted by Bzr02 (Fig. 2e). The above results could be associated from the difficulty of Bzr02 to reduce NO 3 − to NO 2 − in aerobic conditions, as it was explained when it was incubated with higher N-NO 3 concentrations in DM (Fig. 2c). Despite the above, only 1.4 mg L −1 of nitrogen was lost ( Table 3), suggesting that the dominant metabolic pathway presented by Bzr02 was nitrogen assimilation (Fig. 4c). The gene nasA was not detected for Bzr02, indicating that N-NO 3 was rather reduced by a dissimilatory nitrate reductase (NAR or NAP), and then, part of N-NO 2 was incorporated into the cell biomass through the process of assimilatory nitrite reduction 46,50 (Fig. 4c).
Str21 presented 74.3 ± 1.6 mg L −1 of nitrogen in the cell biomass and 14.6 ± 0.2 mg L −1 remained in the medium with no further utilization (Table 3). A significant concentration of nitrogen (12.6 mg L −1 ) was lost at the end of incubation for Str21 (Table 3) in comparisson to Bzr02, suggesting that the process of denitrification took place. Moreover, the N-NO 2 -produced from the reduction of N-NO 3 -was not accumulated in Strs21 as it was observed for Bzr02 (Fig. 2e,f), also supporting that N-NO 2 was further reduced into nitrogen gaseous forms Table 3. Nitrogen balance of strains Bzr02 and Str21 during the nitrogen transformation. NM: nitrifying medium; DM: denitrifying medium; SNDM: simultaneous nitrifying-denitrifying medium. Values represent the mean and the standard error (n = 3).

Media
Strain Initial TN (mg L −1 ) www.nature.com/scientificreports/ in the process of denitrification. Similarly as it was described during the experiment in DM, the detection of nasA suggested that N-NO 3 was incorporated into the cell biomass through the process of assimilatory nitrate reduction, and the detection of all four nitrogen reductase genes (narG, nirS, norB and nosZ) supported that the lost nitrogen escaped as nitrogen gas during dissimilatory nitrate reduction (denitrification; Fig. 4d). The low denitrification activity by Str21 in SNDM was also the influece of the aerobic conditions, which could be appreciated for the long lag phase were N-NO 3 was not significantly utilized at the first 12 h of incubation (Fig. 2d).

Conclusion
Bzr02 and Str21 (isolated from SSBSs sediments), identified as Citrobacter freundii and Pseudomonas mandelii, respectively, were found to have potential applications in nature-based solutions to enhance nitrogen compounds removal, such as SSBSs. Nitrate reduction to nitrite in the denitrification process was found for both strains. Str21 www.nature.com/scientificreports/ seemed to be a facultative anaerobic denitrifier, and therefore, could participate in nitrogen cycling in SSBSs sediments, where oxygen limiting conditions occur. In turn, Bzr02 and Str21 were observed to significantly assimilate N-NH 4 and N-NO 3 into their cell biomass in aerobic conditions, which could subsequently help to improve the efficiency of SSBSs in the nitrogen removal with its sequestration in the sediments. Therefore the application of both strains could be recommended for sedimentation zones, where the release of nitrogen would be controlled by: i) other decomposing microbial communities dwelling in the sediments, and ii) the periodical removal of sediments to maintain the proper operation of SSBSs.

Materials and methods
Samples collection and isolation of bacteria. Sediment samples were collected from the sedimentation zone (August 2018) in three SSBSs constructed for different urban rivers: (i) the River Sokołówka (Sok-SSBS) and (ii) the River Bzura (Bzr-SSBS) in the city of Łódź, and (iii) the River Struga Gnieźnieńska (Str-SSBS) in the city of Gniezno, Poland. 9 Complete description of structure and function for Bzr-SSBS is detailed in Szulc et al. 51 and Jurczak et al. 8 , and for Sok-SSBS and Str-SSBS in Font-Nájera et al. 9 Sediment samples were suspended in sterile 0.75% NaCl w/v (10 g of sediment in 90 mL) and shacked for 30 min at 25 °C. Samples were allowed to settle for 15 min and supernatant was used to prepare serial dilutions (1 × 10 -1 -1 × 10 -6 ) according to Mankiewicz-Boczek et al. 52 . 100 µL of each dilution was plated on to Soil Extract Agar (SEA), a solid medium according to Hamaki et al. 53 , and incubated for seven days at 25 °C. For each SSBS, 50 heterotrophic bacterial isolates (150 in total) were randomly streaked out and re-plated on to nutrient agar solid medium (NA, Karl Roth).

Screening of nitrogen transforming bacteria.
A total of 150 well-separated bacterial colonies were picked from NA and checked for nitrogen transformation abilities in different culturable media (See also media description in supplementary material): (i) in Giltay denitrifying medium (GiDM) with high content of NO 3 − (N: 277 mg L −1 ) according to Alexander 54 , at 25 °C. Bacterial ability to reduce NO 3 − , under oxygen limited condition (Becton Dickinson Gas Pak System), was qualitatively monitored every 12 h with the semi-quantitative test strips QUANTOFIX nitrate/ nitrite (Macherey-Nagel) for 7 d. A total of 10 different bacterial strains were able to completely or partially reduce NO 3 − (denitrification process), and therefore, were selected for further experiments; (ii) the 10 selected bacterial isolates were incubated in 15 mL glucose nitrifying medium (GNM) described in Pahdi et al. 22 , with a small modification-KH 2 PO4 was used instead of NaH 2 PO4 (0.10 g MgSO 4 · 7H 2 O, 3.84 g K 2 HPO 4 , 1.5 g KH 2 PO 4, 0.802 g NH 4 Cl [N: 212 mg L −1 ], 5.3 g glucose C 6 H 12 O 6 [C: 2120 mg L −1 ]), and 2 mL of trace elements were added per 1000 mL of GNM, final pH was 7.2, shacked at 150 rpm and incubated at 25 °C. The trace element solution was prepared according to Pahdi et al. 22 . The effect of different carbon sources was also screened with changes to the nitrifying medium where glucose was replace by: (i) sodium succinate (11.9 g)-succinate nitrifying medium (SNM), (ii) sodium acetate (10.0 g) -acetate nitrifying medium (ANM), and (iii) sodium citrate (8.65 g)-citrate nitrifying medium (CNM). The carbon and nitrogen ratio was kept constant (C:N = 10) in all used media.
Bacteria were also tested for the transformation of NH 4 + under the presence of hydroxylamine in GNM. Cultures were grown in GNM for 6 h and spiked with high concentration of hydroxylamine (100 mg L −1 final concentration) according to Padhi et al. 22 . For the screening purpose, their ability to transform NH 4 + was qualitatively monitored with the semi-quantitative test strips QUANTOFIX ammonium (Macherey-Nagel), every 12 h during 7 d.
Additional methods to corroborate the taxonomical identification of two strains (Bzr02 and Str21) were described in supplementary material. The strain Bzr02 was incubated on GEN III Biolog MicroPlates with different carbon substrates, according to the manufacturer specifications 61 , and the taxonomic characteristics of bacterium were determined using the GEN III Biolog database. For Str21, the gene rpoB (coding for the β subunit of the RNA bacterial polymerase) was used as a molecular marker, since it has been recommended for the www.nature.com/scientificreports/ optimal differentiation between Pseudomonas species 62 . The DNA sequence of the rpoB gene was published in GenBank database for Str21 (MW286260).
Ammonium transformation. Bzr02 and Str21 were cultured overnight in LB at 25 °C and 120 rpm. Cells were harvested by centrifugation (8000 rpm, 10 min, 4 °C), and washed three times with sterile water. Then, each strain was inoculated into the nitrifying medium NM (0.1 final OD 600 ) with adjusted concentrations of NH 4 + (N: 100 mg L −1 ) and glucose (C: 1000 mg L −1 ), incubation was performed at 25 °C and 150 rpm. Bacterial growth (optical density OD 600 nm) was checked at 2 h intervals using an Eppendorf Biophotometer in a 24 h experiment 22 . Supernatant was also collected during each interval (13,000 rpm, 10 min, 4 °C) for the measurement of N-NH 4 , N-NO 3 , N-NO 2 and extracellular TN. The pellet was washed three times with sterile water and used to estimate intracellular TN 32 .
Nitrate transformation. Bzr02 and Str21 were inoculated into denitrifying medium (DM). The denitrifying media was similar to NM with the use of KNO 3 (N: 100 mg L −1 final concentration) as the source of nitrogen. The check of bacterial growth and the collection of samples were performed similarly as explained for the ammonium transformation assays in a 32 h experiment 22 . The supernatant was used to measure N-NH 4 , N-NO 3 , N-NO 2 and extracellular TN, and the bacterial pellet for intracellular TN.
Simultaneous ammonium and nitrate transformation. Bzr02 and Str21 were inoculated into the simultaneous nitrifying-denitrifying medium (SNDM). The media was similar to NM with the use of KNO 3 and NH 4 CL (N: 50 mg L −1 each; TN: 100 mg L −1 final concentration) as sources of nitrogen. The check of bacterial growth and the collection of samples was performed similarly as explained for the ammonium transformation assay, in a 50 h experiment 22 . The supernatant was used to measure N-NH 4 , N-NO 3 , N-NO 2 and extracellular TN, and the bacterial pellet for intracellular TN.
The impact of hydroxylamine for ammonium transformation. Bzr02 was the only strain capable of growth in the presence of hydroxylamine during the screening experiments (described in Chapter 2.2.). Therefore, in a parallel experiment, the transformation of ammonium by Bzr02 was also investigated with different concentrations of hydroxylamine (0, 10, 20 and 50 mg L −1 as final concentrations) added after 4 h of growth in NM. The bacterial growth and the collection of samples were performed similarly as explained for the ammonium transformation assay, at 0 and 2 h (before the addition of hydroxylamine), and 4, 8, 12, 24 and 30 h of incubation (after the addition of hydroxylamine) 22 . The supernatant was used to measure N-NH 4 , N-NO 3 , N-NO 2 and NH 2 OH concentrations.

Analytical methods. Concentration of nitrogen sources were measured with the Multiskan Sky Microplate
Spectrophotometer (Thermo Fisher Scientific) according to standard methods 63 : (i) N-NH 4 by the Nessler's colorimetric assay, (ii) N-NO 3 by the ultraviolet spectrophotometric method, and (iii) N-NO 2 by the Griess colorimetric assay. The Hydroxylamine was measured by indirect spectrophotometry 64 . The TN was calculated with the total Kjeldal reagent set 65 as follows: (i) using the supernatant for the extracellular TN, and (ii) reconstitution of the cell pellet with sterile water for intracellular TN 32 . All measurements were performed in triplicate.

Analysis of data. Nitrogen balance was monitored with the formula:
where N L is the loss of nitrogen at the end of the experiment, the TN Fe and TN Fi are the final extracellular and intracellular TN, respectively, and the TN Ie is the initial extracellular TN (adapted from Fidélis Silva et al. 32 ).
Bacterial removal rates for N-NH 4 − , N-NO 3 − and NH 2 OH (mg L −1 h −1 ) were estimated as follows: where C i and C f are the initial and final concentration of the nitrogen source, respectively, and the t is the final time of the experiment 29 . www.nature.com/scientificreports/