Oxygen tolerance and detoxification mechanisms of highly enriched planktonic anaerobic ammonium-oxidizing (anammox) bacteria

Oxygen is a key regulatory factor of anaerobic ammonium oxidation (anammox). Although the inhibitory effect of oxygen is evident, a wide range of oxygen sensitivities of anammox bacteria have been reported so far, which makes it difficult to model the marine nitrogen loss and design anammox-based technologies. Here, oxygen tolerance and detoxification mechanisms of four genera of anammox bacteria; one marine species (“Ca. Scalindua sp.”) and four freshwater anammox species (“Ca. Brocadia sinica”, “Ca. Brocadia sapporoensis”, “Ca. Jettenia caeni”, and “Ca. Kuenenia stuttgartiensis”) were determined and then related to the activities of anti-oxidative enzymes. Highly enriched planktonic anammox cells were exposed to various levels of oxygen, and oxygen inhibition kinetics (50% inhibitory concentration (IC50) and upper O2 limits (DOmax) of anammox activity) were quantitatively determined. A marine anammox species, “Ca. Scalindua sp.”, exhibited much higher oxygen tolerance capability (IC50 = 18.0 µM and DOmax = 51.6 µM) than freshwater species (IC50 = 2.7–4.2 µM and DOmax = 10.9–26.6 µM). The upper DO limit of “Ca. Scalindua sp.” was much higher than the values reported so far (~20 µM). Furthermore, the oxygen inhibition was reversible even after exposed to ambient air for 12–24 h. The comparative genome analysis confirmed that all anammox species commonly possess the genes considered to function for reduction of O2, superoxide anion (O2•-), and H2O2. However, the superoxide reductase (Sor)-peroxidase dependent detoxification system alone may not be sufficient for cell survival under microaerobic conditions. Despite the fact that anaerobes normally possess no or little superoxide dismutase (Sod) or catalase (Cat), only Scalindua exhibited high Sod activity of 22.6 ± 1.9 U/mg-protein with moderate Cat activity of 1.6 ± 0.7 U/mg-protein, which was consistent with the genome sequence analysis. This Sod-Cat dependent detoxification system could be responsible for the higher O2 tolerance of Scalindua than other freshwater anammox species lacking the Sod activity.


INTRODUCTION
Oxygen plays a key role in regulating the marine nitrogen cycle even in oxygen minimum zones (OMZs). The OMZs constitute only 0.1% of total ocean volume (if the OMZ is defined as O 2 ≦5 µM) but account for 20-40% of the oceanic nitrogen loss through denitrification and anaerobic ammonium oxidation (anammox) [1]. Such nitrogen loss (i.e., N 2 production) intensively occurs at the oxic and anoxic interface of the OMZs [2,3]. This indicates the importance of close microbial interactions among aerobic nitrification, denitrification, and anammox (anaerobically converts NH 4 + to N 2 with NO 2 − ). The effects of oxygen levels on these processes are of central importance for understanding their contributions to the marine nitrogen cycle. Especially, the estimated contribution of anammox may vary greatly depending on the oxygen tolerance capability of marine anammox bacteria which basically determines the OMZ water volume with active anammox activity [4,5]. Thus, their oxygen sensitivity and dynamic response to oxygen exposure need to be experimentally quantified.
A reported wide range of oxygen sensitivity could be attributed to many factors such as anammox species, degree of enrichment, form of biomass (planktonic or aggregated biomass), and measurement methods of anammox activity (e.g., NH 4 + and/or NO 2 − consumption or 14+15 N 2 gas production). In order to assess the reported aerotolerant natures more prudently, the anammox biomass used in those studies need to be clearly specified in detail since oxygen could be consumed by coexisting aerobes [4,6,13,15] or aerobes shield anammox consortia from oxygen exposure in microbial aggregates, biofilms, or marine snow, both of which could result in overestimation of oxygen tolerance [16]. Therefore, aggregated or flocculated biomass often exhibited higher oxygen tolerance than planktonic biomass [17,18]. In addition, there could be truly inter-species differences in oxygen sensitivity as suggested earlier by Yan et al. [19], which, however, has not been experimentally confirmed yet. Inherent oxygen inhibition kinetics (e.g., 50% inhibitory concentration (IC 50 ) and upper O 2 limits (DO max )) and recovery from O 2 inhibition of anammox bacteria have not been directly determined using welldefined laboratory enrichment cultures to date.
Furthermore, oxygen detoxification mechanisms are largely unknown for anammox bacteria. It is generally known that when molecular oxygen (O 2 ) diffuses into cells, reactive oxygen species (ROSs), such as superoxide anion (O 2 • -) and hydrogen peroxide (H 2 O 2 ) are generated as oxygen reduction by-products. Furthermore, O 2 •can react with H 2 O 2 to generate hydroxyl radicals (•OH) in cells, which are the most potent oxidants among ROSs and thus cause the damage on DNA and proteins. Thus, organisms need to detoxify these ROSs by anti-oxidative enzymes such as superoxide dismutase (Sod), catalase (Cat), and peroxidases to survive in the presence of oxygen [20]. However, it is widely known that strict anaerobes usually do not possess Sod and/or Cat because these enzymes generate O 2 , propagating the further production of ROS [21]. Since anammox bacteria are regarded as obligate anaerobic bacteria [22], it is vital to investigate their ROS detoxification mechanism if they possess. It was speculated that "Ca. Brocadia sp." most likely utilize a Sod-cytochrome c peroxidase system to detoxify ROS based on metagenomic and metatranscriptomic analyses [23]. However, abundance of "Ca. Brocadia sp." in the biomass used in their study was relatively low (< 50%), and the activities of anti-oxidative enzymes have never been experimentally determined.
In the present study, oxygen tolerance capability of one marine species ("Ca. Scalindua sp.") and four freshwater anammox species ("Ca. Brocadia sinica", "Ca. Brocadia sapporoensis", "Ca. Jettenia caeni", and "Ca. Kuenenia stuttgartiensis") were determined and then related to the activities of anti-oxidative enzymes (Sod, Cat, and peroxides). Free-living planktonic enrichment anammox cultures were further purified (> 99.8%) by applying Percoll density gradient centrifugation [24] and then subjected to these studies. The experimental results revealed that a marine anammox species, "Ca. Scalindua sp.", exhibited the highest oxygen tolerance, which is likely attributed to the higher Sod activity. To the best of our knowledge, this is the first time to quantitatively verify that marine anammox bacteria possess higher oxygen tolerance than freshwater species and its oxygen detoxification mechanisms.

Biomass preparation
One night before the start-up of batch incubation experiments, planktonic anammox biomass was collected from respective MBRs, harvested by centrifugation under anoxic conditions (20°C, ×10000 rpm, 6 min), and washed with anoxic inorganic medium without ammonium and nitrite (purged by N 2 gas for 1 h) for three times in an anaerobic chamber (Coy Laboratory Products, Michigan, USA). To further purify the MBR anammox cultures, a density gradient separation was performed using Percoll (Cytiva, Tokyo, Japan) as described previously [24]. The Percoll-purified biomass was then anoxically washed twice with inorganic nutrient medium (purged by N 2 gas for 1 h) without NH 4 + and NO 2 − in the anaerobic chamber, in which anammox bacteria accounted for more than 98% of the total biomass based on FISH analysis. The Percoll-purified biomass was anoxically kept in a 37°C incubator or room temperature (ca. 25°C) for overnight before oxygen inhibition or reduction experiments. Percoll purification has been confirmed to have no effect on anammox activity.

Oxygen inhibition experiments
To determine the inhibitory effect of DO concentrations (µmol L −1 ) on specific anammox activity (SAA), 15 N-labelling batch incubation experiments were performed at varying O 2 concentrations (%, v/v) in 20 mL headspace of 70 mL serum vials containing 50 mL of the inorganic medium. The experimental procedure of 15 N-labelling batch incubation experiments has been described in detail elsewhere [26]. Briefly, the inorganic medium containing 3 mM ( 14 NH 4 ) 2 SO 4 and 3 mM Na 15 NO 2 was firstly purged with N 2 gas for at least 30 min. Each vial was then sealed with a butyl-rubber stopper and aluminum cap and then autoclaved one night before the start-up of batch incubation. The headspace gas of vials was exchanged by applying 3 cycles of vacuuming (2 min) and purging (1 min at 1.5 atm) with highly pure helium gas (99.9999% He) by a gas exchange machine (Model IP-8, SANSHIN, Yokohama, Japan) to remove the residual air (oxygen) and to enhance the detection sensitivity of 14+15 N 2 produced by anammox bacteria. The initial gas pressure of the headspace was set at ca. 1.5 atm. A magnetic stir bar was placed inside the vial to gently stir the medium during batch incubation.
To determine the effect of O 2 concentrations on SAA, varying volumes of ambient air were injected into the headspace of sealed vials using a gas tight glass syringe (VICI, Baton Rouge, LA, USA) at least 4 h before the inoculation of biomass. The headspace O 2 concentrations varied from 0 to 3.5% O 2 (v/v), corresponding to theoretical saturated DO concentrations in the culture medium ranging from 0 to ca. 65 µM and 0 to ca. 53 µM at 25°C and 37°C (at 1.5 atm), respectively. To convert the headspace O 2 concentrations to DO concentrations in the medium, the standard curves of the measured DO concentrations vs. the headspace O 2 concentrations (%, v/v) were constructed in advance (Fig. S1). The DO concentrations in the medium (µM) were measured by a microsensor (Unisense oxygen needle sensor OX-N 13621) and the Winkler method and plotted against the headspace O 2 concentrations (%) (Fig. S1). The measured DO concentrations were well correlated with the theoretical O 2 concentrations at 1.5 atm for both temperatures (25 and 37°C). The DO concentrations in the following O 2 inhibition and reduction experiments were determined using the regression lines for microsensor's data.
For O 2 inhibition experiments, 1 mL of Percoll-purified biomass suspension was inoculated into the medium in sealed vials to obtain a final biomass concentration of 0.1-0.15 mg-protein mL −1 . Immediately after biomass inoculation, the air sample (at 0 h) was collected from the headspace, and 14+15 N 2 concentrations were measured by a gas chromatography and mass spectrometry (GC-MS, SHIMADZU). The vials with "Ca. B. sinica" and "Ca. J. caeni" were incubated at 37°C in an incubator, and the other anammox species were incubated at room temperature (ca. 25°C) in dark. The 14+15 N 2 concentrations were measured S. Okabe et al.
hourly, and the maximum specific anammox activity (MSAA, expressed as µmole 14+15 N 2 mg-protein −1 h −1 ) were calculated by dividing the slope of linear regression of initial 14+15 N 2 -production as a function of time by the amount of biomass (expressed as protein) in the vial at the beginning of the assay. The lag-phase was considered in the determination of 14+15 N 2production rates. The 15 N-labelling batch incubation experiments were performed at varying O 2 concentrations at least triplicate for each species.
The percentages of the MASS (relative MASS) that were maintained after exposure to varying O 2 concentrations were calculated with respect to the average MASS of nonexposed samples (anoxic cultures) as follows: where MSAA and MSAA anoxic is the average of MSAA of O 2 -exposed samples and nonexposed samples (anoxic cultures) (µmole 14+15 N 2 mgprotein −1 h −1 ), respectively.

Oxygen inhibition model
An oxygen inhibition model with two parameters (IC 50 : the DO concentration that causes 50% inhibition of MSAA (µM) and DO max : the DO concentration above which anammox activity is completely inhibited (µM)) were used to quantitatively evaluate the effect of DO concentration on anammox activity [33].
where DO is dissolved oxygen concentration in culture medium (µmole L −1 ). This model was used because the upper limits of DO have been previously reported for marine environments [10,11] and engineered systems [6]. Values for IC 50 and DO max were obtained by curve fitting the data.

Oxygen reduction experiments
The selected anammox species ("Ca. B. sinica", "Ca. K. stuttgartiensis" and "Ca. Scalindua sp.") were transferred into 30 mL of the inorganic medium in 100 mL serum vials containing 70 mL headspace. The residual DO was removed and the headspace was exchanged with highly pure helium gas (99.9999% He) at 1.5 atm as described above. O 2 gas (99%, 200 µL) was injected into the headspace with a gas tight syringe 3 h before the addition of 15 NH 4 Cl and Na 14 NO 2 (final concentration of 3 mM each). Thereafter, the vials were incubated at room temperature (ca. 25°C). The air samples were taken from the headspace over the course of the incubation time, and 14+15 N 2 and O 2 concentrations were measured by a gas chromatography and mass spectrometry (GC-MS, SHIMADZU). The O 2 concentrations (%) were converted to DO concentration using the standard curves of the measured DO concentrations vs. the headspace O 2 concentrations (%, v/v) (Fig. S1) as described above. The O 2 reduction rates were determined by dividing the slope of linear regression of DO concentration as a function of time by the protein concentration and are expressed as µmole DO per g-protein per hour.

Recovery after oxygen exposures
Anammox biomass was exposed to varying headspace O 2 concentrations (0 (anoxic), 0.7, 1.4, 2.8, and 21% (v/v)) for 12 h and 24 h in the absence of NH 4 + and NO 2 − as described above. The 21% headspace O 2 was achieved by unsealing the vials (exposure to ambient air). Unexposed vials (anoxic) were also prepared as a positive control. The oxygen recovery assays were performed in 12 ml vials with a liquid volume of 6 mL at least triplicate biological samples.
After exposure to oxygen for 12 h and 24 h, the headspace gas was exchanged by applying 3 cycles of vacuuming (2 min) and purging (1 min at 1.5 atm) with highly pure helium gas (99.9999% He) by a gas exchange machine (Model IP-8, SANSHIN, Yokohama, Japan). Thereafter, anoxically prepared (N 2 gas purged) stock solution of ( 14 NH 4 ) 2 SO 4 and Na 15 NO 2 were supplemented to obtain a final concentration of 3 mM each. 14+15 N 2 gas production was monitored up to 8-10 h, which was defined as "immediate recovery". The vials were also continuously incubated for 24 h under anoxic conditions, and then the same amounts of substrates were supplied again (a final concentration of ( 14 NH 4 ) 2 SO 4 and Na 15 NO 2 , 3 mM each). The 14+15 N 2 gas production was monitored from 24 h to 32 h as described above, which was defined as "secondary recovery".

Comparative genome analysis
Comparative genome analysis was carried out to study the types and distributions of genes encoding anti-oxidative enzymes among the anammox species studied. Presence/absence of the genes were examined by performing a blastp search (threshold e-value of blastp search 10 −15 ) with 31 anammox bacterial genomes affiliated into 18 bacterial species in the bacterial order Brocadiales (Table S1). The 18 bacterial species cover all the bacterial species in the order Brocadiales defied in the GTDB database (Release RS95) [34]. The genome sequences include those obtained from the enrichment culture of "Ca. B. sinica", "Ca. B. sapporoensis", "Ca. J. caeni", "Ca. K. stuttgartiensis" and "Ca. Scalindua sp." used in the present study [35]. Multiple sequence alignment used the ClustalW 1.83 (Gap opening and extension penalties in a pairwise alignment; 10 and 0.05, respectively) [36], and visualized using ESPrint 3.0 [37]. A phylogenetic tree was calculated in MEGA 11.0.8 using maximum likelihood (Jones-Taylor-Thornton model) and neighbor joining methods (Poisson model) [38].

Anti-oxidative enzyme activity assays
Planktonic anammox cells were collected from respective stock MBRs and incubated in 70 mL vials after Percoll density gradient centrifugation. During incubation, anammox cells were subjected to two patterns of oxygen exposure: (i) exposure to different O 2 concentrations (0, 0.7, 1.4, and 2.8% (v/v)) for 12 h; or (ii) exposure to air-saturating DO concentrations (ambient air) for different periods of time (0, 0.5, 1 and 2 h). After each oxygen exposure, biomass was harvested by centrifugation (4°C, 10,000 rpm, 6 min) and concentrated 25-fold by resuspending in an icecold potassium phosphate buffer (50 mM, pH 7.4). To prevent protein degradation, 0.1 mM phenylmethylsulfonyl fluorid (PMSF) was added as a protease inhibitor. The cell suspension was disrupted by passing thrice through a French pressure cell press unit (AVESTIN, ON, Canada) at homogenizing pressure of 1200 MPa. Cell debris was removed by centrifugation (4°C, 4500 rpm, 60 min). The resulting supernatant fraction was transferred to 15 mL sterile tubes and stored in an ice-cold bath. All cell-free extracts were tested within a day. At least, three independent determinations were performed for each enzyme activity in all conditions. All anti-oxidative enzyme activities were expressed as units of enzyme activity per milligram of protein of the cell-free extract.
Superoxide dismutase (Sod) activity was assayed spectrophotometrically at 560 nm by measuring its ability to inhibit the photochemical reduction of nitroblue tetrazolium (NBT) (superoxide anion O 2 •reacts with NBT and reduces the yellow tetrazolium to a blue precipitate) according to the protocol previously described [39]. Since the assays were conducted in microplates, the reaction mixture contained in 200 µL: 50 mM potassium phosphate buffer (pH 7.8), 10 mM L -methionine, 60 µM NBT, 8 µM EDTA, 1.6 µM riboflavin, and 22 µL of cell-free extract (0.05-0.06 mg-protein). Riboflavin was added at the end, and the tubes were mixed by shaking [40]. The microplate was illuminated by a 4000 Lux cool LED lightboard (AXEL, Tokyo, Japan) for 20 min, and the absorbance at 560 nm was recorded at 2, 4, 6, 8, 10, 15, and 20 min in a multilabel plate counter (PerkinElmer, Waltham, USA). The samples containing the same components were placed in dark, which were used as a blank reference. One activity unit (1 U) of Sod was defined as the amount of enzyme required to inhibit the rate of riboflavin and illumination dependent NBT reduction by 50% at 25°C, pH 7.8, and a light intensity of 4000 Lux [40].
Catalase (Cat) activity was assayed spectrophotometrically at 240 nm by measuring its ability to decompose H 2 O 2 as described elsewhere [41]. The reaction mixture contained in 1 mL: 50 mM potassium phosphate buffer (pH 7.0), 10 mM H 2 O 2 , and 0.1 mL of cell-free extracts (0.2-0.3 mg-protein). The reaction was initiated by adding H 2 O 2 into a quartz cuvette containing phosphate buffer and enzyme extracts. One activity unit (1 U) of Cat was defied as the amount of enzyme required to catalase the decomposition of 1 µmol H 2 O 2 in 1 min at 25°C and pH 7.0.
Cytochrome c peroxidase (Ccp) activity was assayed spectrophotometrically at 550 nm by monitoring the loss of reduced cytochrome c according to the protocol provided by Munkres et al. [42]. The reaction mixture contained in 1 mL: 50 mM acetate buffer (pH 6.0), 1 mM EDTA, 1 mM sodium azide, 20 µL of a reduced cytochrome c solution (1.2 mg mL −1 ), 180 µM H 2 O 2 , and 0.1 mL of cell-free extracts (0.2-0.3 mg-protein). The net rate of absorbance change was obtained by subtracting the rate of a control without cell-free extract. One activity unit (1 U) of Ccp was defended as the amount of enzyme required to catalase the oxidation of 1 µmol cytochrome c in 1 min at 25°C and pH 6.0.
Glutathione peroxidase (Gpx) activity was assayed spectrophotometrically at 340 nm by monitoring the loss of NADPH according to the protocol provided by Drotar et al., [43]. The reaction mixture contained in 1 mL: 50 mM potassium phosphate buffer (pH 7.0), 2 mM EDTA, 2 mM reduced glutathione, 0.1 mM NADPH, 2.5 units of glutathione reductase, 0.09 mM H 2 O 2 , and 0.1 mL of cell-free extracts (0.2-0.3 mg-protein). The net rate of absorbance change was obtained by subtracting the rate of a control without cell-free extract. One activity unit (1 U) of Gpx was defined as the amount of enzyme required to catalase the reduction of 1 µmol NADPH in 1 min at 25°C and pH 7.0.

Chemical analyses
For determination of 14+15 N 2 gas and O 2 concentrations, 50 μL of headspace gas was collected from the tested vial with a gas tight syringe (VICI, Baton Rouge, LA, USA) and injected into a gas chromatography mass spectroscopy (GCMS-QP2010SE, Shimadzu, Japan) equipped with a CP-Pora Bond Q fused silica capillary column (Agilent Technologies, Santa Clara, CA, USA) as described previously [25]. The specific anammox activity (SAA) was calculated by dividing the 14+15 N 2 gas production rate by biomass (protein) concentration. In each experiment, SAA tests were performed in triplicate at least.
Concentrations of DO were measured by a Unisense oxygen needle sensor (OX-N 13621, Aarhus, Denmark). The oxygen sensor was calibrated using a two-point calibration (zero and a saturation point). The deionized water was purged with pure N 2 gas for at least 1 h or dissolving 1 g of sodium sulfite (Na 2 SO 3 ) to obtain a zero-DO solution and bubbled with air for at least 1 h to achieve the saturated oxygen concentration at 25°C and 37°C, respectively.
Biomass concentration was determined as protein concentration with the DC Protein Assay Kit (Bio-Rad Laboratories, Munich, Germany) using the bovine serum albumin (BSA) as the protein standard. In brief, 1 mL of cell suspension was collected, centrifuged at 12,000 rpm for 15 min, resuspended in the same amount of 10% (w/v) sodium dodecyl sulfate (SDS) solution and incubated for 30 min at 99°C (Note this procedure is unnecessary for the protein measurement of cell-free extract in enzyme activity tests). Then, 5 µL of the cell suspension was incubated with Bio-Rad protein assay reagent in microplate for 15 min at 25°C. The absorbance was measured at 750 nm.

RESULTS AND DISCUSSION
Effect of O 2 concentration on specific anammox activity (SAA) It was confirmed that the Percoll-purified anammox biomass were highly enriched and well dispersed (Fig. 1). The high purity (> 98%) was also confirmed for all anammox species by FISH and the measurements of 16 S rRNA gene copy numbers by qPCR.
The Percoll-purified planktonic anammox cells were cultured with 14 NH 4 + and 15 NO 2 − in sealed vials with headspace containing different O 2 concentrations (up to 3.32%, corresponding to ca. 60 µM and 50 µM dissolved O 2 (DO) at 25°C and 37°C at 1.5 atm, respectively), and then 14+15 N 2 production rates were measured for five anammox species (Fig. 2). DO concentrations in the vials gradually decreased during the batch incubations for all batch experiments (Fig. S2). The onset of 14+15 N 2 gas production was gradually delayed (lag-time), and the 14+15 N 2 gas production rates decreased along with the increase of O 2 concentration for 4 freshwater anammox species (Fig. 2). However, a marine species "Ca. Scalindua sp." could steady produce 14+15 N 2 up to 1.65% O 2 (DO = ca. 30 µM) without any lag time.
The effect of O 2 concentration on the MSAA, as indicated by relative MSAA (%), was evaluated (Fig. 3)  The DO concentrations that cause 50% inhibition of MSAA (IC 50 ) and the maximum DO concentrations above which anammox is completely inhibited (DO max ) were estimated for each anammox species based on the inhibition model Eq. (2). All freshwater  anammox species showed similar responses to increased O 2 (high sensitivity). The IC 50 values were in the range 2.7 -4.2 µM, and DO max values were in the range 10.9 -26.6 µM. In contrast, a marine anammox species, "Ca. Scalindua sp." exhibited much higher values of IC 50 = 18.0 µM and DO max = 51.6 µM without significant lag times. The MSAA of "Ca. Scalindua sp." was inhibited by only ca. 30% at 10 µM DO (the DO upper limit of suboxic condition [44]).
A wide range of oxygen tolerance capabilities have been reported for freshwater anammox species so far [6]. For example, DO max was reported to be~200 µM for "Ca. K. stuttgartiensis", < 63 µM for "Ca. B. sinica", < 1 µM for "Ca. B. anammoxidans", 120 µM for "Ca. B. caroliniensis", and 70 µM for "Ca. B. fulgida", respectively. On the other hand, complete inhibition was observed at much lower levels of DO (1.25-3.75 µM) in lab-scale bioreactors [6,[12][13][14]. The reason for the variation in reported values could be partly attributed to formation of microbial aggregates and the presence of coexisting aerobes (i.e., the purity of anammox biomass). Therefore, intrinsic oxygen tolerance capabilities of these anammox bacteria cannot be simply evaluated and compared with these reported data.
In the present study, it was confirmed that since highly enriched (>98%) and well-dispersed cells were subjected to the oxygen inhibition studies, the influence of oxygen consumption by coexisting aerobes and/or oxygen shielding effects of aerobes in microbial aggregates can be excluded. Therefore, the observed oxygen tolerances could conceivably be their intrinsic properties. Anammox activities of freshwater species ("Ca. B. sinica", "Ca. B. sapporoensis", "Ca. J. caeni", and "Ca. K. stuttgartiensis") were completely inhibited at~25 µM DO with IC 50 of 2.7-4.2 µM DO, which are in the middle of the reported range.
Another reason could be that since planktonic free-living "Ca. Scalindua sp." has been continuously cultured for more than 10 years in MBRs in our laboratory, in which it is difficult to maintain strict anoxic conditions, oxygen insensitive Scalindua cells could have been selectively enriched during such a long cultivation period. Since anammox bacterial cells in our MBRs are highly enriched and planktonic biomass, not aggregated one, as shown in FISH images (Fig. 1), oxygen shielding effect and/or oxygen consumption by coexisting planktonic aerobes seems to be minimum.
The experimental results clearly suggest that a marine anammox species "Ca. Scalindua sp." intrinsically possess higher oxygen tolerance than freshwater species. More importantly, the upper DO limit for "Ca. Scalindua sp." was much higher than the values reported so far (~20 µM). This might suggest that the ocean volume, where N 2 production (N loss) by anammox is potentially expected, could be larger than we have been led to believe so far.
Oxygen reduction rates "Ca. B. sinica", "Ca. K. stuttgartiensis", and "Ca. Scalindua sp." were cultured in the presence of oxygen (ca. 0.18-0.20% headspace O 2 ). For all three anammox species, O 2 gradually decreased with the incubation time while producing 14+15 N 2 in batch cultures (Fig. 4). The O 2 reduction rates were determined by dividing the slope of linear regression of DO concentration by the protein concentration. The oxygen reduction rates of "Ca. B. sinica" and "Ca. Scalindua sp." were similar (0.26 nmole O 2 /gprotein/h), whereas "Ca. K. stuttgartiensis" reduced O 2 at a more rapid rate (0.53 nmole O 2 /g-protein/h). The specific 14+15 N 2 production rates of "Ca. B. sinica", "Ca. K. stuttgartiensis", and "Ca. Scalindua sp." were 3.31 µmole N 2 /g-protein/h, 3.91 µmole N 2 /g-protein/h, and 3.44 µmole N 2 /g-protein/h, respectively. The O 2 reduction rates were 4 orders of magnitude lower than the specific 14+15 N 2 production rate, suggesting that O 2 could be reduced for detoxification not for respiration. As discussed below (Section of comparative genome analysis), O 2 can be enzymatically reduced to H 2 O by nonrespiratory proteins for detoxification.

Recovery of anammox activity after O 2 exposure
After 12-h exposure to varying headspace O 2 concentrations (0, 0.7, 1.4, 2.1, and 21% (ambient air) O 2 at 1.5 atm) in the absence of NH 4 + and NO 2 − , the headspace gas was exchanged with highly pure Helium gas (> 99.9999%) by vacuuming and purging 3 times to restore anoxic conditions. ( 14 NH 4 ) 2 SO 4 (3 mM) and Na 15 NO 2 (3 mM) were supplemented at 0 h, and the recovery of SAA ( 14+15 N 2 production rate) was examined (Fig. S4). The 14+15 N 2 production was immediately detected without any lag time for "Ca. Scalindua sp." even after exposure to ambient air (21% O 2 ) for 12 h. Subsequent anoxic incubation resulted in almost full recovery: 94 ± 3%, 95 ± 13%, 97 ± 3%, and 65 ± 36% in MSAA after exposure to 0.7%, 1.4%, 2.1 %, and 21% (ambient air) O 2 for 12 h,  (Fig. S1, see Materials and Methods). The protein concentration was 0.28, 0.15, and 0.47 mg-protein/mL for "Ca. B. sinica", "Ca. K. stuttgartiensis", and "Ca. Scalindua sp.", respectively. The batch experiments were conducted in duplicates. One of representative data sets is presented here. respectively (Fig. 5A). In contrast, "Ca. B. sinica" and "Ca. K. stuttgartiensis" were severely inhibited by all O 2 concentrations and could not recover within 7 h anoxic incubation. However, "Ca. B. sinica" exposed to 21% O 2 for 12 h regained 80 ± 15% of the initial activity 13 h after the restoration of anoxic incubation, whereas "Ca. K. stuttgartiensis" could not recover at all, indicating irreversible inhibition (Fig. S5). "Ca. J. caeni" gradually recovered 42 ± 2% of initial activity after exposed to 21% O 2 with lag times (ca. 0.5-1.5 h). It should be noted that the Jettenia biomass concentrations were about 10 times higher than other species, which could result in the quicker and better recovery. "Ca. B. sapporoensis" exhibited an immediate recovery of 14+15 N 2 production, however, this could be due to the formation of small flocs during the recovery test, in which O 2 transfer limitation could shield anammox bacteria from O 2 exposure and thus anammox bacteria could remain active inside.
After 24-h anoxic incubation, the SAAs further recovered > 40% of anoxic control activity, except for "Ca. K. stuttgartiensis" (Fig. 5B). This high oxygen tolerance is partly because the oxygen sensitivity and reversibility of the oxygen inhibition were assessed in the absence of NH 4 + and NO 2 − in the present study, indicating that cells were metabolically inactive. The metabolically inactive cells are likely less susceptible to oxygen inhibition than active cells because of the minimum production of NAD(P)H, which is probably acting as electron donor for O 2 reduction (i.e., production of reactive oxygen species, ROS) [49].
The effect of longer O 2 exposure time (24 h) on the MSAA recovery was also examined in the absence of NH 4 + and NO 2 − under ambient air (Fig. S6). The 14+15 N 2 profiles after 24 hexposure to ambient air exhibited a recovery trend similar to 12 hexposure (Fig. S5). "Ca. Scalindua sp." immediately reinitiated almost the same activities as anoxic controls even after 24 h exposure to ambient air. Interestingly, "Ca. B. sinica" gradually increased to 80 ± 15% of the initial activity after 18 h lag time, whereas "Ca. K. stuttgartiensis" could not recover during > 30 h of anoxic incubation. These results coincided with those of the 12 hexposure test (Fig. S5).
These experimental results suggest that anammox bacteria were aerotolerant anaerobes, although anammox bacteria are previously classified as strict anaerobes. A marine species, "Ca. Scalindua sp.", exhibited the highest aerotolerance and reversibility among anammox species studied. Furthermore, the oxygen inhibition was reversible except for "Ca. K. stuttgartiensis". The reversibility seems to depend on the exposure conditions; reversible at low DO levels (0.25-2% O 2 ) but probably irreversibly inhibited at high DO levels (> 46.9 µM or 20% O 2 ) for freshwater anammox species [12][13][14]. The O 2 toxicity arises directly from itself or from reactive oxygen species (ROS). O 2 and/or ROS reacts unspecifically with catalytic centres or accessory metals of redox enzymes and proteins. For example, key enzymes involved in energy metabolism in anaerobes such as pyruvate ferrodoxin oxidoreductase (PFOR), pyruvate formate lyase (PFL), CO-dehydrogenease, 4-hydroxy-butyryl-CoA dehaydratase, and proteins containing low potential iron-sulfur (Fe-S) clusters are known to be O 2 sensitive [50,51]. However, unfortunately transcriptomic and/ or proteomic studies have not been conducted to confirm that these enzymes or proteins were indeed inactivated by molecular O 2 and repaired after the restoration of anoxic incubation in the present study.
In the present study, since the anammox biomass was highly enriched and well dispersed, the obtained responses to O 2 exposure and recovery from O 2 inhibition are most likely intrinsic oxygen tolerance capability of anammox species. For granular biomass, co-existing micro-aerobic heterotrophic bacteria and O 2 transfer limitation could shield anammox bacteria from O 2 exposure, resulting in higher resilience as compared to planktonic free-living anammox bacteria. For example, floc-style "Brocadia" biomass (dominated only 75%) reinitiated activity to 55-80% of pre-exposed activity after exposure to air-saturated DO (ca. 250 µM) for 24 h [14].
Comparative genome analysis ROS detoxification mechanisms. Upon exposure to O 2 , anaerobes need to detoxify O 2 and generated ROS (e.g., superoxide anion (O 2 •-) and hydrogen peroxide (H 2 O 2 )) for survival [52]. O 2 can be enzymatically reduced to H 2 O by nonrespiratory flavodiiron proteins (Fdp), named rubredoxin:oxygen oxidoreductase (Roo) [50,53], reverse rubrerythrins (revRbr) [50,54], and/or terminal oxidase [55] (Fig. S7A). Fdps (Roo) and revRbr are widely distributed among anaerobic or microaerophilic bacteria and archaea but not in aerobes [50].  (Fig. S7B). The oxidized rubredoxin (Rd ox ) is reduced back to the . The initial MSAA measured under anoxic condition before O 2 exposure was defined as 100% (Anoxic). The data represent the average ± standard deviations of triplicate samples. It should be noted that "Ca. B. sapporoensis" formed small flocs in this recovery experiments, and that the biomass concentrations of "Ca. J. caeni" were about 10 times higher than other species. These could be reasons for higher recovery from O 2 exposure.
The electron donor of Ccps and Rbrs is cytochrome c and Rd red , respectively. The Sor-Ccp and/or Rbr detoxification system confers a selective advantage on anammox bacteria because O 2 is not generated.
All anammox species possess the genes encoding the class A Fdp [57] with the N-terminal methallo-β-lactamase domain and the C-terminal flavodoxin domain and rubrerythrin, which were considered to function in O 2 and/or NO detoxification [58,59] ( Table 1, Table S1). Intriguingly, only the Scalindua genomes (except for the Scalindua sp001828595) contained an operon encoding Fdp and rubredoxin (Rd) which is a potential electron carrier of Fdp. In the operon, the rd gene is located downstream of the fdp gene, suggesting the expression of Scalindua fdp and rd is regulated under the same regulation factor. As for the terminal oxidase, the Scalindua and Kuenenia genomes have the genes encoding cbb 3 -type cytochrome c which can be involved in O 2 reduction [60]. Cytochrome ba3 oxidase and cytochrome oxidase were not identified in all anammox species. In addition, the gene encoding a flavorubredoxin (flavoRd) with a NO-binding nonheme diiron center was identified in all anammox genomes, which possibly function in NO and/or O 2 reduction and detoxification [61,62]. The gene encoding Nror with N-terminal FAD/NADbinding domain and C-terminal Rd-binding domain [63,64] was identified in the Scalindua, Jettenia, and Brocadia genomes.
The gene encoding putative Sod (the kustd1303 protein) was widely conserved in the genomes of B. sinica, B. sapporoensis, J. caeni, and K. stuttgartiensis, and also in the other Brocadiaceae genomes except for those affiliated into the Brocadiaceae-family (Table S1). However, only the Scalindua sp. (affiliated into the species SCALAELEC01 sp004282745) and other Scalindua genomes (Scalindua sp001828595 and Scalindua japonica) have a gene encoding typical Fe/Mn-type Sod (sodA or sodB) ( Table 1). Multiple sequence alignment of the anammox bacterial Sod showed that all the metal binding sites were conserved among the Scalindua Sod (Fig. S8). On the other hand, all the Brocadiaceae Sod lacked the histidine residues requiring for the metal binding (i.e., His 24 and His 75 ), and N-terminal alphahairpin domain was also not conserved in the Brocadiaceae Sod. Furthermore, phylogenetic analysis of anammox bacterial Sod revealed that Scalindua Sod and Brocadiaceae Sod are affiliated into different phylogenetic clades (Fig. S9). These evidences indicate the Brocadiaceae Sod is likely not to function.
As for O 2 •reduction to H 2 O 2 by Sor, the gene encoding neelaredoxin (Nlr) with functionally-important amino acid residues [65] (Fig. S10) was widely conserved among the anammox bacterial genomes except for Jettenia (Table S1). As for reduction of the generated H 2 O 2 , all genera possess the genes encoding cytochrome c peroxidase (Ccp) that obtains reducing equivalents from cytochrome c, whereas B. sinica, K. stuttgartiensis, and Scalindua sp. possess the genes encoding catalase (Cat). The B. sinica and Scalindua sp. Cat were affiliated into a phylogenetic clade apart from that of K. stuttgartiensis Cat (Fig. S11). The genes encoding another peroxidase, rubrerythrins (Rbr) that obtains reducing equivalents from reduced rubredoxin (Rd red ) and glutathione peroxidase (Gpx) that obtains reducing equivalents from glutathione, were conserved in only Scalindua sp.
Based on the comparative genome analysis, it was clearly confirmed that all anammox species commonly possess the genes considered to function for O 2 reduction (i.e., fdpA and fdpF), O 2 •reduction (i.e., nlr), and H 2 O 2 reduction (i.e., rbr and ccp) for survival under microaerobic conditions. The Sorperoxidase dependent O 2 •and H 2 O 2 reduction system is much advantageous to anammox bacteria because O 2 is not generated in these reduction reactions. However, the Sor-peroxidase dependent detoxification system alone may not be sufficient for cell survival under high O 2 conditions. Intriguingly, only "Ca. Scalindua sp." possess the genes for a classical Sod-Cat dependent O 2 •and H 2 O 2 detoxification system (i.e., sodA or sodB) and a functional Fdp-rubredoxin operon, which could be responsible for the higher O 2 tolerance than other freshwater anammox species. The presence of Sod-Cat system on the Scalindua genome triggered the experimental investigation of those enzymatic activities.

Activities of anti-oxidative enzymes in different anammox bacteria
To explain the inter-species difference in O 2 tolerance, the activities of major anti-oxidative enzymes (Sod, Cat, Ccp, and Gpx) of "Ca. Scalindua sp.", "Ca. J. caeni", "Ca. B. sinica", "Ca. B. sapporoensis" and "Ca. K. stuttgartiensis" were assayed for their cell-free extracts prepared from biomass collected from respective anaerobic MBR cultures. Only Scalindua exhibited high Sod (converts O 2 •to H 2 O 2 and O 2 ) activity of 22.6 ± 1.9 U/mg-protein with relatively low Cat (converts H 2 O 2 to O 2 ) activity of 1.6 ± 0.7 U/mg-protein (Fig. 6A). This Sod activity level is similar to those of aerobic and facultative bacteria (10.9-49.7 U/mg-protein) [66], but slightly higher than aerotolerant anaerobic bacteria (0.44-19.6 U/mg-protein) and higher than intermediate and extremely oxygen sensitive (obligately) anaerobic bacteria (almost not detected) [67]. In contrast, other four freshwater anammox species possessed very low or virtually no Sod activity with moderately high levels of Cat activity (5.9-16.5 U/mg-protein), which might reflect their lower oxygen tolerance capability. The gene encoding catalase was not identified in the J. caeni and B. sapporoensis genomes, which may have been a result of the incomplete nature of the genomes. Since the activities of Sod and Cat were highly dependent on medium compositions and culture conditions [68], the effect of salinity on these activities needs to be investigated in the future. Activities of cytochrome c peroxidase (Ccp converts H 2 O 2 to H 2 O) were detected in all species except for "Ca. B. sapporoensis" but 3-4 orders of magnitude lower than the Cat activities (Fig. 6B). Activities of glutathione peroxidase (Gpx converts H 2 O 2 to H 2 O) were barely detected only in "Ca. Scalindua sp." and "Ca. J. caeni". Effect of exposed O 2 concentrations and exposure time on the activities of anti-oxidative enzymes Cell-free extracts were prepared from anaerobic MBR cultures of "Ca. Scalindua sp.", "Ca. B. sinica", "Ca. J. caeni", and "Ca. K. stuttgartiensis" were subjected to two patterns of O 2 exposure: (1) exposed to different O 2 concentrations (0, 0.7, 1.4 and 2.1% O 2 ) for 12 h and (2) exposed to ambient air (21% O 2 ) for different periods of time (0, 0.5, 1 and 2 h). High activity levels of Sod were detected in all samples of "Ca. Scalindua sp.", but no significant effects of exposed O 2 concentrations and duration of O 2 exposure on the Sod activity levels were observed even after exposure to ambient air for 2 h (Fig. S12 and Fig. S13). The Cat activity in "Ca. Scalindua sp." was also remained unchanged for different O 2 concentrations and durations of O 2 exposure. Similarly, O 2 concentrations and O 2 exposure times did not significantly affect the activity levels of all anti-oxidative enzymes for other three freshwater species. These results revealed that the activities of all anti-oxidative enzymes could be constitutively expressed and active in all anammox bacteria. "+"; not found in the genome sequence but found in other genome sequence affiliated into the same bacterial species. *The protein with N-terminal FAD/NAD-binding domain and rubredoxin-binding C-terminal domain.
The oxygen tolerance is directly related to the ability of the bacteria to reduce (scavenge) O 2 and to detoxify O 2 •-(namely Sod activity). Thus, Sod activity is a primary important determinant of oxygen tolerance because O 2 •is more toxic than H 2 O 2 [66], whereas Cat is the secondary importance since catalase activity showed no clear correlation to oxygen tolerance [66]. "Ca. Scalindua sp." possessed significantly higher Sod activity and therefore exhibited higher oxygen tolerance than other freshwater species. It is likely that O 2 •was primarily detoxified by Sod in "Ca. Scalindua sp.", whereas other four freshwater species cannot efficiently detoxify O 2 •due to lack of Sod activity. The generated H 2 O 2 could be converted to O 2 and H 2 O by Cat and/or to H 2 O by peroxidases (Ccp, Gpx, and/or rubrerythrin (Rbr)) ( Fig. S7). However, since Ccp and Rbr were active only in the absence of O 2 with low levels of H 2 O 2 and also unable to degrade H 2 O 2 quickly [69,70], it is not clear if Ccp and Rbr play a vital role in H 2 O 2 decomposition or not in "Ca. Scalindua sp.". Catalase is the most prominent of the stationary-phase induced scavengers and able to degrade H 2 O 2 more quickly at higher concentrations, whereas rubrerythrins are used to scavenge low levels of H 2 O 2 [70]. This suggests that "Ca. Scalindua sp." most likely employs different peroxidases including Cat depending on H 2 O 2 concentration [51]. Further study is definitely required to identify which enzyme(s) acts effectively in the H 2 O 2 detoxification in "Ca. Scalindua sp.".
In the presence of oxygen, anammox bacteria reduce O 2 to form toxic ROS, or reduce to H 2 O by Sor. If bacteria reduce no oxygen and therefore no toxic ROS is generated, resulting in higher oxygen tolerance [66]. "Ca. Scalindua sp." with high Sod activity reduced O 2 at a relatively slow rate (0.26 µmole O 2 /gprotein/h), whereas "Ca. K. stuttgartiensis" that has substantially no Sod activity reduced O 2 at a more rapid rate (0.53 µmole O 2 /gprotein/h). Although "Ca. B. sinica" has no Sod activity, they reduced O 2 at a relatively slow rate (0.26 µmole O 2 /g-protein/h). These results suggest that "Ca. K. stuttgartiensis" generate SOR rapidly but could not detoxify it quickly, resulting in lower O 2 tolerance and irreversible recovery. It should be also noted that the oxygen reduction rate is an important factor in determining oxygen tolerance.
In conclusion, a marine species, "Ca. Scalindua sp.", exhibited the higher aerotolerance and reversibility than other freshwater anammox species. This is primarily because "Ca. Scalindua sp." possesses the classical Sod-Cat ROS detoxification system in addition to the Sor-peroxidase dependent O 2 and ROS reduction system. The upper DO limit for "Ca. Scalindua sp." (~51.6 µM) was much higher than the values reported so far (~20 µM). This might suggest that the contribution of anammox to oceanic nitrogen loss could be larger than we have ever thought.