Co-culturing Hyphomicrobium nitrativorans strain NL23 and Methylophaga nitratireducenticrescens strain JAM1 allows sustainable denitrifying activities under marine conditions

Hyphomicrobium nitrativorans strain NL23 and Methylophaga nitratireducenticrescens strain JAM1 were the principal bacteria involved in the denitrifying activities of a methanol-fed, fluidized marine denitrification reactor. We believe that a tight relationship has developed between these two strains to achieve denitrification in the reactor under marine conditions. To characterize the potential synergy between strain JAM1 and strain NL23, we compared some of their physiological traits, and performed co-cultures. Pure cultures of strain JAM1 had a readiness to reduce nitrate (NO3−) with no lag phase for growth contrary to pure cultures of strain NL23, which has a 2-3 days lag phase before NO3− starts to be consumed and growth to occur. Compared to strain NL23, strain JAM1 has a higher μmax for growth and higher specific NO3− reduction rates. Antagonist assays showed no sign of exclusion by both strains. Planktonic co-cultures could only be performed on low NaCl concentrations for strain NL23 to survive. Denitrification rates were twice higher in the planktonic co-cultures than those measured in strain NL23 pure cultures. Biofilm co-cultures were performed for several months in a 500-mL bioreactor filled with Bioflow supports, and operated under fed-batch mode with increasing concentrations of NaCl for strain NL23 to acclimate to marine conditions. Under these conditions, the biofilm co-cultures showed sustained denitrifying activities and surface colonization by both strains. Increase in ectoine concentrations produced by strain JAM1 was observed in the biofilm with increasing NaCl concentrations. These results illustrate the capacity of both strains to act together in performing denitrification under marine environments. Although strain JAM1 did not contribute in better specific denitrifying activities in the biofilm co-cultures, its presence was essential for strain NL23 to survive in a medium with NaCl concentrations > 1.0%. We believe that ectoine is an important factor for the survival of strain NL23 in these environments.


INTRODUCTION
Denitrification takes place in bacterial cells where N oxides serve as terminal electron acceptor instead of oxygen (O 2 ) for energy production when O 2 depletion occurs, leading to the production of gaseous nitrogen (N 2 ). Four sequential reactions are required for the reduction of nitrate (NO 3 -) to gaseous nitrogen, via nitrite (NO 2 -), nitric oxide and nitrous oxide, and each of these reactions is catalyzed by different enzymes, namely NO 3reductases (Nar or Nap), NO 2reductases (Nir), nitric oxide reductases (Nor) and nitrous oxide reductases (Nos) [1][2][3].
We have been studying for many years the biofilm derived from a continuous, fluidized-bed methanol-fed denitrification reactor that treated a 3 million-L seawater aquarium that contains fish, birds and invertebrates. The fluidized carriers in the denitrification reactor were colonized by naturally occurring multispecies bacteria to generate a marine methylotrophic denitrifying biofilm [4,5], among which the methylotrophic bacteria Methylophaga spp. and Hyphomicrobium spp. accounted for 60 to 80% of the bacterial community [6]. Two bacterial strains representative of Methylophaga spp. and Hyphomicrobium spp. were isolated from the denitrifying biofilm. The first one, Methylophaga nitratireducenticrescens strain JAM1, is capable of growing in pure cultures under anoxic conditions by reducing NO 3 -to NO 2 -, which accumulates in the medium [7,8]. It was later shown to be able to reduce NO and N 2 O to N 2 [9]. These activities concur with the presence of gene clusters encoding two Nar reductases, two Nor reductases and one Nos reductase [10,11]. A dissimilatory NO-forming NO 2reductase gene (nirS or nirK) is absent, which correlates with accumulation of NO 2in the culture medium during NO 3reduction. The second strain, Hyphomicrobium nitrativorans strain NL23, is capable of complete denitrification from NO 3to N 2 , and possesses operons that encode for the four different nitrogen oxide reductases, among which a periplasmic Nap-type NO 3reductase [12][13][14]. These two species are the main bacteria responsible of the dynamics of denitrification in the biofilm.
Hyphomicrobium spp. are methylotrophic bacteria that are ubiquitous in the environment [15]. They have also been found in significant levels in several methanol-fed denitrification systems treating municipal or industrial wastewaters or a seawater aquarium, and they occurred often with other denitrifying bacteria such as Paracoccus spp., Methylophilales or Methyloversatilis spp. Their presence correlates with optimal performance of bioprocess denitrifying activities [16][17][18][19][20][21][22]. Methylophaga spp. are methylotrophic bacteria isolated from saline environments [23,24]. They have been found in association with diatoms, phytoplankton blooms and marine algae, which are known to generate C1 carbons [25][26][27]. Co-occurrence of Methylophaga spp. and Hyphomicrobium spp. has been shown in two other methanol-fed denitrification systems treating saline effluents [28,29]. Therefore, understanding how these two taxa collaborate could benefit in optimizing denitrification systems treating saline/brackish waters.
We hypothesize that a tight relationship has developed between H. nitrativorans strain NL23 and M. nitratireducenticrescens strain JAM1 to achieve denitrification in the original biofilm. For instance, because growth of strain NL23 is impaired in marine environments [12], strain JAM1 would protect strain NL23 from osmotic shock in the biofilm by producing, for instance, the osmoprotectant ectoine, as most of Methylophaga species can generate [23]. In the present report, we compared some physiological traits of both strains that can influence their cohabitation. We then used co-culture approach to characterize the potential synergy between these two strains to achieve denitrification. Our results showed that when both strains are present, sustainable denitrifying activities were achieved and that strain JAM1 is essential for strain NL23 to survive under marine conditions.

MATERIAL AND METHODS Determination of NO3 -, NO2and biomass.
NO 3 -and NO 2were measured either by ionic chromatography as described by Mauffrey et al. [9] or by colorimetry based on the method described by Schnetger and Lehners [30]. In this method, NO 3was reduced to NO 2with vanadium (III) chloride (VCl 3 ). Eight hundred mg VCl 3 were dissolved in 100 mL of 0.1 M HCl (Solution A). The solution was filtered (0.45 µm). Solution B was made of 200 mg N-1naphthylethylenediamine dihydrochloride in 100 mL H 2 O. Solution C was made of 2 g sulfanilamide in 100 mL HCl 10%. These solutions were kept in amber bottle at 4°C for one month. Solutions D and E were made fresh. For solution D, solutions A, B and C were mixed at the proportion of 5:1:1. For solution E, solutions B and C were mixed at 1:1 proportion. The assays were carried out in 96-well plates. To determine the NO x concentrations (NO 3 -+ NO 2 -), samples (120 µL) were added to one plate, then 100-µL solution D was quickly added and mixed with a multichannel micropipet. The plate was covered and immediately incubated for 60 min at 45°C. To measure NO 2already present in the medium, the same samples (200 µL) were added to another plate, then 20-µL solution E was quickly added and mixed with a multichannel micropipet. The plate was covered and immediately incubated for 30 min at 45°C. Both plates were read at 540 nm with a plate reader. The concentrations were determined with standard solutions. Linear response ranged from 0.1 to 2 mg-N/L (either NO 3 -or NO 2 -). Results from the NO 3reduction by VCl 3 generated the NO x concentrations (NO 3 -+ NO 2 -). Therefore, NO 3 -concentrations were calculated as NO x -NO 2 -.
The growth rates, the NO 3reduction rates and the denitrification rates were calculated by the linear portion of the culture growth (OD 600nm ), of the NO 3concentrations and of the NO x concentrations, respectively, over time for each replicate (OD h -1 ; NO 3 -mM h -1 , NO x mM h -1 ). The specific denitrification rates (NO x mM h -1 mg-protein -1 ) were reported as the denitrification rates divided by the quantity of biomass (mg protein) in a vial or in the reactor. The amount of protein in the biomass was determined by the Quick StartTM Bradford Protein Assay (BioRad, Mississauga, ON, Canada).

Culture media
The Instant Ocean (IO) seawater medium was bought from Aquarium systems (Mentor, OH, USA) and dissolved at 30 g/L. NaNO 3 was added at prescribed concentrations. The medium was adjusted at pH 7.5 and autoclaved.

Planktonic pure cultures and co-cultures
Generation of pre-cultures were derived from cultures of strain JAM1 and strain NL23 cultured under oxic conditions without NO 3 -(unless specified) in their respective medium (Methylophaga 1403 and 337a) and incubated at 30°C at 150 rpm. The cells were then centrifuged and dispersed in the prescribed medium before use. For anoxic cultures, the sterile media were distributed (60 mL) in sterile serologic vials. The vials were purged of oxygen for 10 min with nitrogen gas (Praxair, Mississauga, ON, Canada) and sealed with sterile septum caps. The planktonic co-cultures (anoxic conditions) were inoculated with pre-cultures of strain JAM1 and of strain NL23 with a JAM1/NL23 ratio of 1:10 and a final cell dispersion of 0.1 OD 600nm . Cultures were incubated for 1-7 days at 30°C without shaking.

Biofilm co-cultures in recirculating reactor
The reactor is illustrated in Figure 1. It consisted of 500ml working volume. A flow cell consisted of a 50-mL tube was added in the recirculating circuit where two microscope slides were added to monitor cell attachment on the surface. The system was carefully washed with water and ethanol. Two hundred and sixty acid-washed Bioflow 9 mm supports were added in the reactor. The media (Table 1) were sterilized before used, but was added not sterilely under oxic environment. Throughout all the different operating conditions, the medium was recirculated with a peristaltic pump at 30 mL/min, and the reactor was run at room temperature (ca. 22°C). Gas production was recorded by water displacement in a graduated cylinder (Fig. 1). Samples of the suspended biomass (1 mL) were taken at sampling ports and centrifuged; the supernatant was used to determine the NO 3and NO 2concentrations. The pellets of some of these samples were kept for DNA or protein (or both) extractions. From the DNA extracts, the concentrations of strain JAM1 and strain NL23 in the suspended biomass were determined by qPCR. The same reactor was used to carry out three runs sequentially. Reactors 1 and 2 were inoculated with strain JAM1 and strain NL23 (biofilm co-cultures). Reactor 3 was inoculated with only strain NL23 (biofilm mono-culture). Results from reactor 1 allowed to adjust the conditions for reactor 2. In reactors 2 and 3, protein and RNA extractions, and microscopic observations were performed; ectoine concentration was measured in the biofilm of reactor 2.
To initiate reactors 1 and 2 (co-cultures), inocula of strain NL23 (four 60 mL-vials) and strain JAM1 (three 60 mLvials) were centrifuged (ca. 3 g wet weight, each), dispersed in 20 mL 0.5% NaCl Methylophaga 1403 medium, and added to the reactor filled with 500 mL of the same medium without NO 3 -. To acclimate both strains to the reactor environment, the oxic conditions were kept by injecting air with a peristaltic pump for 3 days in the reactor, after which air was no longer supplied and NO 3was added to create anoxic conditions. For reactor 3, the inoculum of strain NL23 was added in the reactor with 0% NaCl Methylophaga 1403 medium, and the reactor was operated directly under anoxic conditions (no oxic phase). The reactors were then run for a week, with additions of NO 3when needed, after which the medium was replaced by a fresh one. Only bacteria attached to supports and vessel were kept in the reactor. The planktonic bacteria were discarded. Upon NO 3and NO 2reduction, the medium was changed with fresh one. Anoxic conditions and denitrifying activities were occurring as indicated by gas production and the reduction of NO 3and NO 2 -. When these production and reduction appeared stable, the reactor was sampled at regular intervals for at least 24 h to derive the dynamics of NO 3and NO 2reduction. Afterwards, one to two supports were collected, frozen at -20°C until use for the extractions of proteins, DNA and ectoine. In some of the operating conditions (Table 1), supports were collected for RNA extractions (reactors 2 and 3. To do so, fresh medium was added in the reactor. When NO 3was near completely consumed, 10 supports were collected in a glove box flushed with nitrogen gas and frozen immediately in liquid nitrogen, and kept at -70°C until use to extract total RNA. In all cases, the taken supports were replaced with virgin ones in the reactor. After these steps, we proceeded with new medium (increasing concentration of methanol/NO 3or NaCl) as described in Table 1. When the different operating runs for reactor 1 were finished, all the supports were frozen at -20°C for further analysis. The vessel was then washed thoroughly and sterilized. New virgin supports were added to start reactor 2. The same was done with reactor 3.

Ectoine
The protocol was based on Chen et al. [31] and Zhang et al. [32]. The biofilm on one support was directly extracted with 2 ml of methanol/chloroform/water (10:5:4) with intermittent strong agitation for 3 h. The extract was then centrifuged at 16 000 g for 10 min. One mL of the supernatant was extracted with 1 mL chloroform/water (1:1) with agitation for 30 min. The extract was then centrifuged at 16 000 g for 10 min. The water phase was collected and dried at 35°C with N 2 gas for 30 min. The extract was then dissolved in 1 mL water and filtered on 0.2 µm filter. The stable isotopically labelled internal standard 5,6,7,8tetradeutero-4-hydroxy-2-heptylquinoline (HHQ-d 4 ) was added in the samples (2.66 µg/mL final concentration). HHQ-d 4 was provided by the laboratory of Eric Déziel (INRS, Laval, QC, Canada). Ectoine ((S)-2-methyl-1,4,5,6tetrahydro-pyrimidine-4-carboxylic acid, 97% purity) was bought at Sigma Aldrich (Oakville, ON, Canada), and used to derive a standard curve.
Fifteen-µL of the samples were analysed by a HPLCcoupled to a mass spectrometer (Waters, Milford, MA, USA). The Waters 2795 HPLC system was equipped with a Kinetex (100 x 4.6 mm) 2.6-µm C8 reverse-phase LC column (Phenomenex, Torrance, CA, USA). The mobile phase was a gradient of 1% acetic acid in water (solvent A) and 1% acetic acid in acetonitrile (solvent B) programmed as follows: initial 2% solvent B (0-4 min), 2-70% solvent B (4-5 min), 100% solvent B (5-8 min), then hold 3 min and followed by 3 min of re-equilibration. HPLC flow rate was 400 µl/min split to 40 µl/min by a Valco tee splitter. The Quattro Premier XE mass spectrometer was equipped with a Z-spray interface using electro-spray ionization in positive mode (ESI-MS/MS). Multiple Reaction Monitoring (MRM) mode was used to quantify ectoine. MassLynx and QuanLynx software (Ver. 4.1) were used. The capillary voltage was set at 3.0 kV and the cone voltage at 30 V. The source temperature was kept at 120°C. Nitrogen was used as nebulising and drying gas at flow rates of 15 and 100 ml/min, respectively. In MRM mode, the following transitions were monitored: for ectoine 143→98 and for the internal standard HHQ-d 4 248→163. The pressure of the collision gas (argon) was set at 2 × 10 −3 mTorr and the collision energy at 30 V for all transitions. The area of each chromatographic peak was integrated and the ectoine/HHQ-D 4 response ratio was used to create the calibration curve (R 2 = 09945) and quantify the concentration of ectoine in the samples.

Antagonist assays
Cultures of strain JAM1 or strain NL23 (each OD 600nm 0.5) were first spread (200 µL) over Methylophaga 1403 (0.5% NaCl, no NO 3 -) agar plates with 0.3% methanol. From these, three protocols were carried out. In the first protocol, one-cm 2 blotting papers were soaked with either strain NL23 culture or strain JAM1 culture. The papers were then put on the inoculated agar plates. In the second protocol, one drop of strain JAM1 and strain NL23 cultures was put onto the inoculated agar plates. In the third protocol, a well was made in the inoculated agar plate and one drop of either strain JAM1 or strain NL23 cultures was put in it. These plates were incubated between 30 and 48 h at 30°C under oxic conditions. Clearing zone around the inoculated points or the absence of growth at these points were recorded.  The biofilm that colonized the microscope slides in the flow cell chamber was fixed with 4% cold paraformaldehyde in PBS for 60 min, and washed twice with water. The biofilm was dehydrated for 3 min successively in 50%, 80% and 95% ethanol/water, and then air dried. Hybridization buffer for both probes consisted of 20% formamide, 20 mM Tris-HCl pH 8.0, 0.01% SDS and 0.9 M NaCl. The probe concentrations used were between 10 to 20 ng/µL. Hybridization was carried out in an Omnislide in situ thermal cycler (Thermo Electron Corporation, Waltham, Mass., USA) for 3 h at 46°C, followed with by 2 min at 48°C. Slides were washed in 20 min at 48°C, in 20 mM Tris-HCl pH 8.0, 5 mM EDTA, 0.01% SDS and 210 mM NaCl. FISH samples were mounted with Prolong Gold agent containing DAPI (Molecular Probes, Fisherscientific, Ottawa, ON Canada). The probes for Methylophaga spp. (MPH-730-Atto488; 5' CAGTAATGGCCCAGTGAGTCGCC 3') [33] and for Hyphomicrobium spp.

DNA and RNA extraction, qPCR and RT-qPCR, and RNA sequencing
DNA and RNA extractions of the planktonic biomass and of the biofilm from 1 to 4 supports were performed as described [5,9,35]. Concentrations of strain JAM1 and strain NL23 were performed by qPCR with primers targeting narG1 and napA, respectively as described by Geoffroy et al. [10] (Table 2). Standard curves were performed with PCRamplified DNA fragments of narG1 and napA from strain JAM1 genome and strain NL23 genome, respectively ( Table  2). RT-qPCR were performed as described by Mauffrey et al. [9] with primers targeting narG1 and narG2 for strain JAM1, napA and nirK for strain NL23, and the reference genes rpoB and dnaG specific for each strain ( Table 2). The standard curves were performed with dilution of the respective genomes.
The RNA samples were sequenced at the Centre d'expertise et de services Génome Québec Montreal QC, (Canada) (Sequencing type: Illumina NovaSeq 6000 S4 PE100 -25M reads; Library Type: rRNA-depleted for bacteria). The sequencing data were uploaded to the Galaxy web platform, and we used the public server at usegalaxy.org to analyze the data [36]. Raw reads were filtered to remove low quality reads using FASTX toolkit by discarding any reads with more than 10% nucleotides with a PHRED score <20. The resulting reads were aligned to the genome of M. nitratireducenticrescens strain JAM1 (GenBank accession number CP003390.3) and to the genome of H. nitrativorans strain NL23 (CP006912.1) using Bowtie2 with default parameters. BEDtools were used to assign the number of reads to the respective genes in the genome, which were then normalized as transcripts per million (TPM). Because the samples were from one reactor run, statistical analysis could not have been performed. However, an arbitrary variation of 10% in the read values revealed that the ratio of TPM of a gene from one condition to another of >2 or <0.2 is significantly different. A. Growth under planktonic anoxic conditions (triplicate cultures). B and C. Growth rates and NO3reduction rates, respectively, at different NO3concentrations. Each point is the average of triplicate anoxic cultures. D. Specific NO3reduction rates. These rates were calculated with the NO3reduction rates by the generated culture biomass (OD600nm) at the end of the exponential phase. Data for strain JAM1 in panels C and D were taken from Geoffroy et al. [10].  (Fig. 2), with standard error in parentheses from triplicates cultures. a Data from Mauffrey et al. [35] and from new measurements.

Physiological differences between the two strains
We measured the dynamics of NO 3uptake by M. nitratireducenticrescens strain JAM1 and H. nitrativorans strain NL23. Planktonic anoxic pure cultures of strain JAM1 and strain NL23 were performed with different concentrations of NO 3to derive their respective maximum growth rates (µmax) and half-saturation constants of NO 3concentration for growth (Table 3). Strain NL23 cultures showed a 48 to 72-h lag before growth occurred, whereas strain JAM1 cultures presented no such lag (Fig. 2A). The µmax of strain JAM1 cultures were 44% higher than those of strain NL23 cultures (Fig. 2B, Table 3). To assess the affinity of these strains toward NO 3for growth, the µmax/Ks ratio was calculated [37] (Table 3). This ratio in strain JAM1 cultures was twice higher than that of strain NL23 cultures. The NO 3reduction rates in the NL23 cultures increased linearly with the increase of NO 3concentrations (from 2 to 40 mM) in the medium (Fig. 2C). These rates reached a plateau at 24 mM NO 3in strain JAM1 cultures, and showed no significant changes at higher concentrations. The specific NO 3reduction rates (rates normalized by the biomass) were however, relatively constant for both strains whatever the NO 3concentrations (Fig. 2D). These specific rates were six times higher in strain JAM1 cultures (average 4.5 ± 0.8 NO 3h -1 OD -1 ) than in strain NL23 cultures (average 0.77 ± 0.32 NO 3h -1 OD -1 ). All these results suggest that strain JAM1 has a higher dynamism towards NO 3than strain NL23.
The second physiological character that we tested was the capacity of both strains to produce antagonist molecules to exclude one to each other. Under the tested conditions (see M&M section), no sign of exclusion was observed from both strains.

Anoxic planktonic co-cultures
Co-cultures were first attempted under planktonic conditions to assess the effect of culturing both strains together on the denitrifying activities. Initially, we wanted to perform planktonic co-cultures under marine conditions as both strains originated from a marine denitrification system. However, strain NL23 planktonic growth is impaired when the culture medium > 1% NaCl [12], impeding planktonic cocultures under marine conditions. We found that the Methylophaga culture medium (Methylophaga 1403) with 0.5% NaCl instead of 2.4% can sustain both strains with minimal growth defect. This medium was used to perform our anoxic planktonic co-culture assays. Higher levels of strain NL23 in the inoculum (JAM1/NL23 ratio of 1:10) had to be used because of the higher dynamics of strain JAM1 towards NO 3as shown above. In the first co-culture attempts, both inocula were derived from oxic pre-cultures. Anoxic planktonic mono-cultures of both strains were performed in parallel also in this medium as control. Our results showed that these co-cultures failed to perform complete denitrification. The NO 3reduction rates were similar between the co-cultures (0.675 ± 0.008 mM-NO 3h -1 ) and strain JAM1 mono-cultures (0.815 ± 0.024 mM-NO 3h -1 ), and NO 2accumulated in the medium (Fig. 3A). The proportion of strain NL23 at the end of the co-cultures was 1.6% (Table 4).
Other planktonic co-cultures were performed by inoculating strain NL23 that was derived from pre-cultures cultured under anoxic conditions to stimulate its denitrification pathway. This time, these co-cultures were capable of full denitrification (Fig. 3B). The NO 3reduction rates were similar between the co-cultures (1.165 ± 0.054 mM-NO 3h -1 ) and strain JAM1 mono-cultures (0.995 ± 0.032 mM-NO 3h -1 ). NO 2accumulated transitorily after NO 3reduction in the co-cultures and then was completely consumed after 36h (Fig. 3B). The denitrification rates of the co-cultures were twice higher (t-test, p=0.0038) (0.912 ± 0.033 mM-NO x h -1 ) than those in strain NL23 mono-cultures (0.428 ± 0.019 mM-NO x h -1 ). The proportion of strain NL23 was about 12 times higher in these co-cultures than that in the first co-culture assays (19.7% vs 1.6%; Table 4). The total  concentrations of both strains in these co-cultures were three times higher than those in the first co-culture assays (Table  4).

Biofilm co-cultures
As mentioned before, planktonic co-cultures could not be performed under marine conditions without impairing strain NL23 growth. As strain JAM1 and strain NL23 were isolated from a biofilm, developing a biofilm co-culture provided a mean to assess the adaptability of strain NL23 to marine conditions. Two reactors with the Bioflow 9 mm supports and containing the Methylophaga 1403 medium at 0.5% NaCl were inoculated with both strains, and operated under fed-batch mode. The reactors were run for several weeks for the biofilm to build up and the denitrifying activities to stabilize. The biofilm was then acclimated with increasing concentrations of NaCl (Table 1). Finally, the Methylophaga 1403 medium was replaced with the Instant Ocean (IO) medium to mimic the original bioprocess. A third reactor was operated under the same conditions, this time with only strain NL23. The three reactors were run sequentially. Results from reactor 1 allowed us to adjust the conditions for reactor 2.
In reactor 1, the NO 3reduction rates and the denitrification rates (represented by the NO x reduction rates) ranged both between 1.4 and 5.1 mM h -1 (average 2.5 mM h -1 ) when operating with the Methylophaga 1403 medium at 0.5%, 1%, 2% and 2.75% NaCl (Table 5). Transitory NO 2accumulation was only observed in the reactor operated with 2.75% NaCl, which peaked after 8 h, suggesting a decrease in the denitrification performance. When the reactor was run with the IO medium, an 18-fold decrease occurred in the denitrification rates with accumulation of NO 2 -. The gas production was absent reflecting low bacterial activities. We suspected that the absence of the supplements (Wolf solution, solution T, and B 12 vitamin) that were present in the Methylophaga 1403 medium could have caused this defect. After 12 days following re-addition of these supplements, the activities were partially restored (Table 5).
In reactor 2, the NO 3reduction rates were higher than these measured in reactor 1 when operating with the Methylophaga medium at 0.5%, 1%, 2% and 2.75% NaCl (Table 5), and ranged between 2.6 and 7.1 mM h -1 (average 4.4 mM h -1 ). These higher NO 3reduction rates could have had an impact on the reactor dynamics by generated transitory NO 2accumulation that peaked after 2 to 4 hours under these four conditions. Consequently of these transitory accumulations, the denitrification rates in reactor 2 was lower than those in reactor 1 with values ranging between 0.9 and 2.2 mM h -1 (average 1.4 mM h -1 ). During these four operating conditions, reactor 2 showed increases from 4.1 to 8.9 mM h -1 mg protein -1 in the specific denitrification rates. To avoid the detrimental effects observed in reactor 1, reactor 2 was run with the IO medium containing the supplements. The gas production, the NO 3reduction and denitrification rates, and the transitory NO 2accumulation were similar than those observed in the other operating conditions (Table 5). However, the specific denitrification rates decreased by 4.5 times. Once again, to avoid detrimental effect on the denitrification activities, the supplements were removed one by one (except for trace metal elements) by operating the reactor for a few days between each removal. In all cases, complete denitrification occurred in less than 24 h, and gas production was constant throughout the operating conditions.
In reactor 3 (ran with only strain NL23), the NO 3reduction rates ranged between 2.7 and 3.2 mM h -1 (average 2.8 mM h -1 ) when operating with the Methylophaga 1403 medium at 0.5%, 1%, 2% and 2.75% NaCl (Table 5). However, the denitrification rates and the specific denitrification rates dropped by 5.1 and 6.2 times, respectively, when NaCl concentration reached 2.0%. When operated with 2.75% NaCl and in the IO medium, denitrification rates were further down with NO 2accumulation, as well as low gas production.
The proportions of strain NL23 in the suspended biomass of reactor 1 operated with the Methylophaga 1403 medium at 0.5%, 1%, 2% and 2.75% NaCl ranged from 2.3% to 21.4% ( Table 6). The proportion of strain NL23 was below 1% in reactor 1 operated with the IO medium. The proportions of strain NL23 in the suspended biomass of reactor 2 operated with the Methylophaga 1403 medium at 0.5%, 1%, 2% and 2.75% NaCl ranged from 1.3 to 5.9% (Table 6), and were at similar levels in the reactor operated with the IO medium (1.2 and 3.0%). The levels of both strains were also determined in the biofilm of reactor 2. The proportions of strain NL23 were 3-4 times higher in the biofilm (4.3 to 10.9%) than in the suspended biomass, suggesting better environment for strain NL23 to grow. In reactor 3, with the increase of NaCl concentrations in the medium, the concentrations of strain NL23 decreased in both the suspended biomass and the biofilm ( Table 6).
An indication of the level of specific denitrifying activities in the reactors was to divide the denitrification rates (mM-NO x h -1 ; in Table 5) by the estimated total number of cells (strain NL23 and strain JAM1) in the reactor (suspended biomass and biofilm) as determined by qPCR (Table 6). These specific denitrification rates, translated per billion of cells (or gene copies), were at similar levels in reactor 2 during the different operating phases (Table 6). In reactor 3, the denitrification rates per billion cells were between 12 to 25 times higher than in reactor 2. It was at its lowest with medium at 2.0% NaCl, suggesting decreases in denitrification efficiency by strain NL23 cells.

Ectoine in the biofilm co-culture
We measured the level of ectoine in the biofilm of reactor 2 ( Table 5). Increasing amount of ectoine was noticed with the increase of NaCl concentration in the reactor.  2 Values are those when the gas production stabilized. 3 Transitory NO2accumulation: Peak concentration of NO2 -(time in hours of the maximum peak measured). 4 NO2accumulation with low consumption (LC). 5 Estimated total protein in the reactor (suspended biomass and biofilm).  1 Determined by the multiplication of the gene copies/mL of both strains by 550 mL (volume of the reactor) plus the multiplication of the gene copies per support of both strains by 260 supports. 2 Specific rates were calculated by the division of the NOx reduction rates (Table 5) by the gene copies/reactor. Nm: not measured. NA: not applicable.

Biofilm structure
The colonization of surface by both strains was followed in the reactor with a flow chamber that was connected to the reactor and containing microscope slides (Fig. 1). Figure 4A and B showed uniform dispersed clusters of cells throughout the surface after one to five days of colonization. FISH assays showed that strain NL23 was lower in concentration than strain JAM1 in the biofilm (Fig. 4), which concurs with the qPCR results. No specific pattern of distribution of both strains was noticed in the biofilm.

Gene expression of denitrification genes in biofilm coculture and mono-culture
The impact of the operating conditions in the reactors on the expression of denitrification genes of key reductases were assessed by measuring in the biofilm the levels of transcripts of both nar genes (narG1 and narG2) for strain JAM1, and napA and nirK for strain NL23. The transcript levels of narG1 did not change in the biofilm co-culture of reactor 2 operated at 0.5% and 2.75% NaCl, and with the IO medium (Table 7). For narG2, the transcript levels in the biofilm coculture raised by about 2-3 times when the operating conditions in reactor 2 increased in NaCl concentration from 0.5% to 2.75% or when the medium was changed for the IO medium ( Table 7).
The napA transcript levels in the biofilm co-culture of reactor 2 had an 11-fold decrease when the operating conditions changed from 0.5% NaCl to 2.75% NaCl, or a 17fold decrease when the medium was changed for the IO medium ( Table 7). The napA transcript levels were similar in the biofilm mono-culture of reactor 3 operated with 0.5% and 2.0% NaCl. Finally, the nirK transcript levels showed no significant difference between the operating conditions in reactors 2 and 3, except in the biofilm co-culture of reactor 2 operated with the IO medium with a 6-fold decrease compared to the biofilm co-culture of reactor 2 operated with 0.5% NaCl.

The transcriptomic analysis of selected biofilm
The impact of the operating conditions in gene expression of the pathways of denitrification and nitrogen assimilation were assessed by deriving the transcriptome of biofilm samples taken from reactors B2 and B3 operated in different conditions (Fig. 5). No substantial changes were found in the relative transcript levels of the denitrification genes between the three operating conditions in reactor B2 (0.5% and 2.75% NaCl and IO) for both strains, and in the two operating conditions in reactor B3 (0.5% and 2.0% NaCl) for strain NL23.
The type of medium (Methylophaga 1403 medium and IO medium) had a great impact on the N-assimilation pathways in both strains. Higher relative transcript levels were observed in both strains in the reactor B2 and B3 that were running with the IO medium for genes encoding ammonium transporters, and genes involved in the NO 3transformation in ammonium. The Methylophaga 1403 medium contains ammonium, which is absent in the IO medium. Therefore, the biofilm cultured with the IO medium relied on N-assimilation pathways as nitrogen source.  Finally, the relative transcript levels of the four genes involved in the ectoine synthesis in strain JAM1 were higher in reactor 2 operated under marine conditions (2.75% NaCl and IO medium) compared to the reactor operated in low salt concentration (0.5% NaCl).

Figure 5. Changes in the relative transcript levels of genes involved in the denitrification and N-assimilation pathways, and in ectoine synthesis.
Total RNA were extracted from the biofilm samples taken from the bioreactors 2 and 3 operated at the different conditions. RNA samples were sequenced, and the reads associated to the denitrification genes, the N-assimilation genes and ecotine genes were transformed in transcripts per million (TPM). The TPM values of each gene operon were compared to those of the bioreactor 2, 0.5% NaCl, which was set to 1.0.

DISCUSSION
The marine and methylotrophic environment that prevailed in the original denitrification system favored the establishment in the biofilm of two methylotrophic species, H. nitrativorans and M. nitratireducenticrescens directly involved in the denitrifying activities. An equilibrium between these two species had to be reached in this biofilm where specific physiological characters of both species may have certainly played important roles in this equilibrium. One of these characters is the dynamics of NO 3uptake. In pure cultures, M. nitratireducenticrescens strain JAM1 has shown to have a readiness to reduce NO 3with no lag phase for growth contrary to H. nitrativorans strain NL23, which has a 2-3 days lag phase before NO 3starts to be consumed and growth to occur. These results suggest that the denitrification pathway is tightly regulated in strain NL23 under anoxic conditions, which is not in strain JAM1. Compared to strain NL23, strain JAM1 has a higher µmax for growth and higher specific NO 3reduction rates, suggesting that strain JAM1 cells have a higher dynamism of NO 3processing (e.g. NO 3intake and reduction) than strain NL23 cells. These dynamisms might be attributed to the different types of NO 3reductases used by these strains. Strain JAM1 possesses two Nar systems, which contribute both to the growth of strain JAM1 under anoxic conditions [35], whereas strain NL23 has one periplasmic Nap-type NO 3reductases [12]. The Nar system is found associated with the membrane facing the cytoplasm. NO 3reduction by this system can generate proton motive force (PMF) and thus energy for the cells.
Because of its location, the Nap system does not generate PMF upon NO 3reduction [3]. Furthermore, strain JAM1 possesses three NO 3transporters associated with the denitrification pathway. As the antagonist assays did not show exclusion from each others in the tested conditions, differences in the regulation and the kinetics of the denitrification pathway could therefore confer an advantage of strain JAM1 for growth over strain NL23. These features are reflected in the higher abundance of strain JAM1 than strain NL23 that we found in the co-cultures as revealed by qPCR and FISH assays.
The planktonic co-cultures assays revealed the importance of strain NL23 to have its denitrification pathway stimulated to compete with strain JAM1 for growth. This stimulation resulted in better denitrification rates in planktonic co-cultures compared to those measured in strain NL23 pure cultures. As the original denitrification system operated under continuous mode, with constant concentration of NO 3in the affluent, strain NL23 would have had its denitrification pathway stimulated, thus capable to sustain growth.
The planktonic co-cultures did not reflect the true cohabitation of both strains under marine conditions that prevailed in the original denitrification system, since the planktonic co-cultures were carried out at low NaCl concentration to allow strain NL23 to grow. Our hypothesis that strain JAM1 has to be present for strain NL23 to survive under marine conditions revealed to be true with the biofilm co-culture assays in the reactors. Acclimation of strain NL23 with increasing concentration of NaCl and then to the marine IO medium (medium used in the aquarium tanks that were treated by the original system) showed (i) sustained denitrifying activities, (ii) support colonization of strain by NL23 (FISH assays) and (iii) higher proportions of strain NL23 in the biofilm co-culture compared to the suspended biomass.
With the increase of NaCl concentration in reactor 3 (ran only with strain NL23), important decreases in denitrifying activities was noticed, specially when the NaCl concentration reached 2.0%. Martineau et al. [12] showed decreases in growth and denitrifying activities in pure cultures of strain NL23 cultured at 2% NaCl. However, compared to reactor 2 (co-cultures), reactor 3 had higher specific denitrification rates (at least in low NaCl concentrations) and higher denitrification rates per billions of cells. These differences in favor of strain NL23 can be explained by the lower level of biomass generated in this reactor (about 4 times less in protein; Table 5). Therefore, the presence of strain JAM1 did not contribute in better specific denitrifying activities in the reactor. However, its presence was essential for strain NL23 to operate at NaCl concentrations higher than 1.0%.
No substantial changes in the transcript levels of the measured denitrification genes in the biofilm co-culture of reactor 2 were noticed except for napA when the reactor was operated at 0.5% NaCl. In these conditions, the napA transcript levels were >10 times higher than those measured in the other operating conditions (2.75% NaCl and IO medium). However, such levels of napA transcripts were not observed in the biofilm mono-culture of reactor 3 operated with 0.5% NaCl, suggesting that strain JAM1 could have had an impact on the napA expression in the biofilm co-culture at this NaCl concentration. Wang et al. [38] showed decrease of about 10 times in denitrification activities of a denitrifying granular sludge when the NaCl concentration increased gradually from 0% to 10% in synthetic wastewater. The same study observed increases in lethality and DNA leakage in these cultures when the NaCl concentrations were above 2%. Not many studies were reported on the effect of salinity on the expression of denitrification genes. Gui et al. [39] observed decreases in the expression of the 4 denitrification genes (including napA) with the increase of NaCl concentration in the medium of Achromobacter sp. GAD-3 pure cultures.
We observed a drop of the denitrifying activities with the passage from the Methylophaga 1403 medium to the IO medium. Besides IO to be commercial medium which could have unknown additives, differences between the two media is the presence of phosphate, ammonium and BisTris (buffer, and can bind some ions) in the Methylophaga 1403 medium that are absent in the IO medium. Ammonium can be a source of N that is readily assimilated in the glutamine pathway for biomolecule synthesis. In IO medium, nitrate must be transformed in ammonium before this assimilation, which require energy from the cells. Indeed, we found higher levels of relative transcripts of genes associated with the Nassimilation pathways of both strains in the biofilms that were derived from reactors operated with IO.
As expected, increase in ectoine production in the biofilm co-culture correlated with the increase in NaCl concentration in reactor 2. This higher level concurs with the higher relative transcript levels of the genes involved in ectoine production in the biofilm derived from the reactor 2 operated with 2.75% NaCl and IO compared to this reactor operated with 0.5% NaCl. Acclimation by strain NL23 of increases of NaCl concentration in the biofilm co-cultures could be related with an increase in ectoine production by strain JAM1. It is possible that the level of strain NL23 that can be reached in the biofilm is restricted by the level of ectoine generated by strain JAM1. Many bacteria have developed uptake transporter systems to accumulate osmoprotectant such as ectoine in their cells. For instance, Corynebacterium glutamicum possesses four secondary carriers for compatible solutes among which two, ProP and EctP, can uptake ectoine. These transporters belong respectively to the major facilitator superfamily and to the sodium/solute symporter superfamily [40]. Genes associated with these types of transporters were found in the genome of strain NL23 [13]. Huang ZZ et al. [41] found high levels of intracellular ectoine and hydroxyectoine in a phenol-degrading microbial enrichment acclimated up to 3 M NaCl (17.5%). Demonstration that osmoproctectant uptake by bacteria in biofilm occurs was shown by Kapfhammer et al. [42] where a Vibrio cholera strain impaired in the production of ectoine can increase cell attachment and biofilm growth in presence of the osmoprotectant glycine betaine in high-osmolarity medium, but not when the betaine transporter was impaired too.

CONCLUSIONS
M. nitratireducenticrescens strain JAM1 have a higher dynamism of NO 3processing (e.g. NO 3intake and reduction) than H. nitrativorans strain NL23. Differences in the regulation and the kinetics of the denitrification pathway confer an advantage to strain JAM1 for growth over strain NL23. Better denitrifications rates in planktonic co-cultures were obtained in low salt medium. In reactors carrying the biofilm co-culture, acclimation of strain NL23 with increasing concentration of NaCl and then in the marine IO medium showed (i) sustained denitrifying activities, (ii) support colonization by strain NL23 and (iii) higher proportions of strain NL23 in the biofilm compared to the suspended biomass. Reactor ran with only strain NL23 failed to achieve such activities under marine conditions. Higher concentrations of ectoine were measured in the biofilm coculture with the reactor operated under marine conditions. Although the presence of strain JAM1 did not contribute in better specific denitrifying activities in the reactor, its presence was essential for strain NL23 to operate at NaCl concentrations > 1.0%. Our results highlighted some mechanisms of co-habitation from these two strains in achieving denitrifying activities, and can contribution in optimizing of denitrification systems treating saline/brackish waters.

FUNDING
This research was supported by a grant to Richard Villemur from the Natural Sciences and Engineering Research Council of Canada # RGPIN-2016-06061.