Transcriptomic Response of Nitrosomonas europaea Transitioned from Ammonia- to Oxygen-Limited Steady-State Growth

Nitrification is a ubiquitous microbially mediated process in the environment and an essential process in engineered systems such as wastewater and drinking water treatment plants. However, nitrification also contributes to fertilizer loss from agricultural environments, increasing the eutrophication of downstream aquatic ecosystems, and produces the greenhouse gas nitrous oxide. As ammonia-oxidizing bacteria are the most dominant ammonia-oxidizing microbes in fertilized agricultural soils, understanding their responses to a variety of environmental conditions is essential for curbing the negative environmental effects of nitrification. Notably, oxygen limitation has been reported to significantly increase nitric oxide and nitrous oxide production during nitrification. Here, we investigate the physiology of the best-characterized ammonia-oxidizing bacterium, Nitrosomonas europaea, growing under oxygen-limited conditions.

Ϫ to N 2 O via NO (22)(23)(24)(25). The first pathway is the dominant process at atmospheric O 2 levels, while the latter is more important under O 2 -limited (hypoxic) conditions (26,27), where NO 2 Ϫ and NO serve as alternative sinks for electrons generated by NH 3 oxidation.
Nitrosomonas europaea strain ATCC 19718 was the first AOB to have its genome sequenced (28) and is widely used as a model organism in physiological studies of NH 3 oxidation and NO/N 2 O production in AOB (27,(29)(30)(31)(32)(33)(34)(35)(36). The enzymatic background of NO and N 2 O production in N. europaea is complex and involves multiple interconnected processes (Fig. 1). Most AOB harbor a copper-containing nitrite reductase, NirK, which is necessary for efficient NH 3 oxidation by N. europaea at atmospheric O 2 levels. NirK is also involved in but not essential for NO production during nitrifier denitrification in N. europaea (26,27,29,35) and is upregulated in response to high NO 2 Ϫ concentrations (37). Moreover, two forms of membrane-bound cytochrome (cyt) c oxidases (cNOR and sNOR) and three cytochromes, referred to as cyt P460 (CytL), cyt c= beta (CytS), and cyt c 554 (CycA), have been implicated in N 2 O production in N. europaea and other AOB (12,24,32,(38)(39)(40). However, the involvement of cyt c 554 in N 2 O production has recently been disputed (41). Finally, recent research has confirmed that the oxidation of NH 3 to NO 2 Ϫ in AOB includes the formation of NO as an obligate intermediate, produced by NH 2 OH oxidation via the hydroxylamine dehydrogenase (HAO) (20). The enzyme responsible for the oxidation of NO to NO 2 Ϫ (the proposed nitric oxide oxidase) has not yet been identified (40).
The production of NO and N 2 O by N. europaea, grown under oxic as well as hypoxic (oxygen-limited) conditions, was previously demonstrated and quantified in multiple batch and chemostat culture studies (11,12,34,35,42,43). Furthermore, recent studies have investigated the instantaneous rate of NO and N 2 O production by N. europaea during the transition from oxic to oxygen-limited or anoxic conditions (12,35,36). Despite this large body of literature describing the effect of oxygen (O 2 ) limitation on NH 3 oxidation and NO/N 2 O production in N. europaea, little attention has been paid to the regulation of other processes under these conditions. Previous studies have utilized reverse transcription-quantitative PCR (RT-qPCR) assays to examine transcriptional patterns of specific mainly N cycle-related genes in AOB grown under O 2 -limited conditions (34,36,44). To date, no study has evaluated the global transcriptomic response of N. europaea to O 2 -limited growth. However, research on the effect of stressors other than reduced O 2 tension have demonstrated the suitability of transcriptomics for the analysis of physiological responses in AOB (43,(45)(46)(47)(48).
N. europaea utilizes the Calvin-Benson-Bassham (CBB) cycle to fix inorganic carbon (28,49). Whereas all genome-sequenced AOB appear to use the CBB cycle, differences exist in the number of copies of ribulose-1,5-bisphosphate carboxylase/oxygenase (RuBisCO) genes encoded as well as the presence or absence of carbon dioxide (CO 2 )-concentrating mechanisms (50)(51)(52). N. europaea harbors a single form IA greenlike (high-affinity) RuBisCO enzyme and two carbonic anhydrases but no carboxysomerelated genes (28). RuBisCO is considered to function optimally in hypoxic environments, as it also uses O 2 as a substrate and produces the off-path intermediate 2-phosphoglycolate (53,54). However, the effects of O 2 limitation on the transcription of RuBisCO-encoding genes and resulting growth yield in AOB are still poorly understood.
In this study, we expand upon previous work investigating the effects of O 2 limitation on N. europaea by profiling the transcriptomic response to substrate (NH 3 ) versus O 2 limitation. N. europaea was grown under steady-state NH 3 -or O 2 -limited conditions, which allowed for the investigation of differences in transcriptional patterns between growth conditions. We observed a downregulation of genes associated with CO 2 fixation as well as increased expression of two distinct heme-copper-containing cytochrome c oxidases (HCOs) during O 2 -limited growth. Our results provide new insights into how N. europaea physiologically adapts to thrive in O 2 -limited environments and identified putative key enzymes for future biochemical characterization.

RESULTS AND DISCUSSION
Growth characteristics. N. europaea was grown as a continuous steady-state culture under both NH 3 -and O 2 -limited growth conditions. During NH 3 -limited steadystate growth, the culture was kept oxic with a constant supply of filtered atmospheric air, was continuously stirred (400 rpm), and contained a standing NO 2 Ϫ concentration of ϳ60 mmol liter Ϫ1 . N. europaea grown under NH 3 -limited conditions consumed ϳ98% of substrate provided; therefore, cultures were considered to have nonlimiting amounts of O 2 (Table 1). In contrast, during O 2 -limited steady-state growth, no additional air inflow was provided, but the stirring was increased (800 rpm) to facilitate O 2 transfer between the headspace and growth medium. As a consequence of O 2 limitation, the medium contained standing concentrations (ϳ30 mmol liter Ϫ1 ) of both NH 4 ϩ and NO 2 Ϫ (Fig. 2; Table 1). During NH 3 -limited steady-state growth (days 7 to 16) (Fig. 2), N. europaea stoichiometrically oxidized all supplied NH 4 ϩ to NO 2 Ϫ (N balance ϭ 61.0 Ϯ 1.7 mmol liter Ϫ1 ) and maintained an optical density at 600 nm (OD 600 ) of 0.15 Ϯ 0.01 (Table 1). During O 2 -limited steady-state growth (days 23 to 32) (Fig. 2) (12,35,42,55).
The dilution rate (0.01 h Ϫ1 ) of the chemostat was kept constant during both NH 3and O 2 -limited growth, and resulted in 14.4 mmol day Ϫ1 NH 4 ϩ delivered into the chemostat. On days 9, 10, and 11, which were sampled for NH 3 -limited growth transcriptomes, N. europaea consumed NH 3 at a rate (q NH3 ) of 24.73 Ϯ 0.53 mmol g (dry cell weight) Ϫ1 h Ϫ1 with an apparent growth yield (Y) of 0.40 Ϯ 0.01 g (dry cell weight) mol Ϫ1 NH 3 . During days sampled for O 2 -limited growth transcriptomes (days 28, 29, and 30), the q NH3 was significantly higher (28.51 Ϯ 1.13 mmol g [dry cell weight] Ϫ1 h Ϫ1 ; P Յ 0.05), while Y was significantly lower (0.35 Ϯ 0.01 g [dry cell weight] mol Ϫ1 NH 3 ; P Յ 0.05). When the whole 10-day NH 3 -and O 2 -limited steady-state growth periods were considered, the q NH3 and Y trends remained statistically significant (P Յ 0.05) ( Table 1). Overall, NH 3 oxidation was less efficiently coupled to biomass production under O 2 -limited growth conditions. Global transcriptomic response of N. europaea to growth under NH 3 -versus O 2 -limited conditions. Under both NH 3 -and O 2 -limited growth conditions, transcripts mapping to 2,535 of 2,572 protein-coding genes (98.5%) and 3 RNA-coding genes (ffs, rnpB, and transfer-messenger RNA [tmRNA]) were detected. Many of the 37 genes not detected encode phage elements or transposases, some of which may have been excised from the genome in the Ͼ15 years of culturing since genome sequencing (see Data Set S1 in the supplemental material). In addition, no tRNA transcripts were detected. The high proportion of transcribed genes is in line with recent N. europaea transcriptomic studies, where similarly high fractions of transcribed genes were detected (43,48). A significant difference in transcript levels between growth conditions was detected for 615 (ϳ24%) of transcribed genes (see Fig. S1). Of these 615 genes, 435 (ϳ71%) were present at higher levels, while 180 (ϳ29%) were present at lower levels during O 2 -limited growth. Genes encoding hypothetical proteins with no further functional annotation accounted for ϳ21% (130) of the differentially transcribed genes (Data Set S1). Steady-state growth under O 2 -limited conditions mainly impacted the transcription of genes in clusters of orthologous groups (COGs) related to transcription and translation, ribosome structure and biogenesis, carbohydrate transport and metabolism, and energy production and conversion (Fig. 3).
Universal and reactive oxygen stress. The transcript levels of various chaperone proteins and sigma factors considered to be involved in the general stress response in N. europaea (45) differed between NH 3 -and O 2 -limited growth, with no discernible trend of regulation (see Table S2; Data Set S1). Overall, prolonged exposure to O 2 limitation did not seem to induce a significantly increased general stress response in  Table S1 in the supplemental material). concentrations.
d Letters A and B represent highly significant differences (P Յ 0.01), and letters C and D represent significant differences (P Յ 0.05) within parameters. Capital letters represent comparisons between 10-day periods, whereas lowercase letters represent comparisons between 3-day periods.
Oxygen-Limited Growth of Nitrosomonas europaea N. europaea. Key genes involved in oxidative stress defense (superoxide dismutase, catalase, peroxidases, and thioredoxins) were transcribed at lower levels during O 2limited growth, as expected (Table S2; Data Set S1). Surprisingly, rubredoxin (NE1426) and a glutaredoxin family protein-encoding gene (NE2328) did not follow this trend and were transcribed at significantly higher levels (2.8-and 1.8-fold, respectively) during O 2 -limited growth (Table S2). Although their role in N. europaea is currently unresolved, both have been proposed to be involved in cellular oxidative stress response (56,57), iron homeostasis (58, 59), or both. Carbon fixation and carbohydrate and storage compound metabolism. There was a particularly strong effect of O 2 -limited growth on the transcription of several genes related to CO 2 fixation (Fig. 3B). The four genes of the RuBisCO-encoding cbb operon (cbbOQSL) were among the genes displaying the largest decrease in detected transcript numbers ( Fig. 4; Table S2). Correspondingly, the transcriptional repressor of the cbb operon (cbbR) was transcribed at 4.5-fold higher levels ( Fig. 4; Table S2). This agrees with the previously reported decrease in transcription of the N. europaea cbbOQSL operon in O 2 -limited batch culture experiments (60). The reduced transcription of RuBisCO-encoding genes potentially reflects a decreased RuBisCO enzyme concentration needed to maintain an equivalent CO 2 fixation rate during O 2 -limited growth. Since O 2 acts as a competing substrate for the RuBisCO active site, the CO 2 -fixing carboxylase reaction proceeds more efficiently at lower O 2 concentrations (53,61,62). When N. europaea is grown under CO 2 limitation, the transcription of RuBisCO-encoding genes increases significantly (43,60,63). Due to the absence of carboxysomes, N. europaea appears to regulate CO 2 fixation at the level of RuBisCO enzyme concentration.
Genes encoding the remaining enzymes of the CBB pathway and carbonic anhydrases were not significantly differentially regulated, with the exception of the transketolase-encoding cbbT gene (Table S2). Likewise, almost no differences in transcription were observed for the majority of genes in other central metabolic pathways (glycolysis/gluconeogenesis, tricarboxylic acid [TCA] cycle) (Data Set S1). As the specific growth rate of N. europaea was kept constant during both NH 3 -and O 2 -limited growth, it is not surprising that genes associated with these core catabolic pathways were transcribed at comparable levels.
Differential transcription of polyphosphate (PP) metabolism-related genes suggests an increased accumulation of PP storage during O 2 -limited growth. Transcripts of the polyphosphate kinase (ppk) involved in PP synthesis were detected in significantly higher numbers (2.1-fold), while transcription of the gene encoding the PP-degrading exopolyphosphatase (ppx) did not change (Table S2). Indeed, N. europaea was previously shown to accumulate PP when ATP generation (NH 3 oxidation) and ATP consumption become uncoupled and surplus ATP is available (64). As the specific growth rate was kept constant throughout the experiment, PP accumulation could be a result of increased efficiency in ATP-consuming pathways, such as CO 2 fixation or oxidative stress-induced repair. A decrease in the reaction flux through the energetically wasteful oxygenase reaction catalyzed by RuBisCO could result in surplus ATP being diverted to PP production. Energy conservation. Genes encoding the known core enzymes of the NH 3 oxidation pathway in N. europaea were all highly transcribed during both NH 3 -and O 2limited growth (Table S2). These included ammonia monooxygenase (AMO; amoCAB operons and the singleton amoC gene) and the genes encoding HAO (haoBA) and the accessory cyt c 554 (cycA) and cyt c m552 (cycX). Due to a high level of sequence conservation among the multiple AMO and HAO operons (65), it is not possible to decipher the transcriptional responses of paralogous genes in these clusters. Therefore, we report the regulation of AMO and HAO operons as single units (Table S2). The transcript numbers of genes in the AMO operons decreased up to 3.3-fold during O 2 -limited growth, while transcripts of the singleton amoC were present at 1.9-fold higher levels. However, these transcriptional differences were not statistically significant. The HAO cluster genes were also not significantly differentially transcribed (Table S2).
Previous research has shown that transcription of AMO, and to a lesser extent of HAO, is induced by NH 3 in a concentration-dependent manner (66). In contrast, other studies have reported an increase in amoA transcription by N. europaea following substrate limitation (44,67). Furthermore, N. europaea has been reported to increase amoA and haoA transcription during growth under low-O 2 conditions (34). However, exposure to repeated transient anoxia did not significantly change amoA or haoA mRNA levels (36). As both NH 3 and O 2 limitation were previously shown to induce transcription of AMO-and HAO-encoding genes, the high transcription levels observed here under both NH 3 -and O 2 -limited steady-state growth conditions are not surprising.
The periplasmic red copper protein nitrosocyanin (NcyA) was among the most highly transcribed genes under both NH 3 -and O 2 -limited growth conditions (Table S2). Nitrosocyanin has been shown to be expressed at levels similar to those of other nitrification and electron transport proteins (68) and is among the most abundant proteins commonly found in AOB proteomes (47,69). To date, the nitrosocyaninencoding gene ncyA has been identified only in AOB genomes (24) and has been proposed as a candidate for the nitric oxide oxidase (40). However, as comammox Nitrospira do not encode ncyA (2,3,13), nor do all genome-sequenced AOB (70), nitrosocyanin cannot be the NO oxidase in all ammonia oxidizers. In this study, a slight (1.7-fold) but not statistically significantly higher number of ncyA transcripts was detected during O 2 -limited growth (Table S2). This agrees with a previous study comparing ncyA mRNA levels in N. europaea continuous cultures grown under highand low-O 2 conditions (44). However, N. europaea performing pyruvate-dependent NO 2 Ϫ reduction also significantly upregulated ncyA, while transcription of amoA and haoA decreased (44). Overall, there is evidence for an important role of nitrosocyanin in NH 3 oxidation or electron transport in AOB, but further experiments are needed to elucidate its exact function.
Three additional cytochromes are considered to be involved in the ammoniaoxidizing pathway of N. europaea: (i) cyt c 552 (cycB), essential for electron transfer; (ii) cyt P460 (cytL), responsible for N 2 O production from NO and hydroxylamine (39); and (iii) cyt c=-beta (cytS), hypothesized to be involved in N oxide detoxification and metabolism (24,71). All three were among the most highly transcribed genes (top 20%) under both growth conditions (Table S2). In this study, cytS was transcribed at significantly lower levels (2.3-fold) during O 2 -limited growth. However, transcription levels of cycB and cytL were not significantly different (Table S2). While the in vivo function of cytS remains elusive, it is important to note that in contrast to ncyA, the cytS gene is present in all sequenced AOB and comammox Nitrospira genomes (12,13,52). The ubiquitous detection of cytS in genomes of all AOB, comammox Nitrospira, and in methaneoxidizing bacteria capable of NH 3 oxidation (72) indicates that cyt c=-beta might play an important yet unresolved role in bacterial aerobic NH 3 oxidation.
Nitrifier denitrification. During O 2 -limited growth, N. europaea either performs nitrifier denitrification or experiences a greater loss of N intermediates such as NH 2 OH (73) or NO (20), which leads to the observed N imbalance between total NH 4 ϩ consumed and NO 2 Ϫ produced ( Fig. 2; Table 1). The Cu-containing NO 2 Ϫ reductase NirK and the iron-containing membrane-bound cyt c-dependent NO reductase (cNOR; NorBC) are considered to be the main nitrifier denitrification enzymes (24,35). N. europaea NirK plays an important role in both nitrifier denitrification and NH 3 oxidation (27) and is known to be expressed during both O 2 -replete and -limited growth (29,30,35). However, under O 2 -limited conditions, nirK was among the genes with the largest decrease in transcript numbers (4.2-fold) observed in this study ( Fig. 5; Table S2). In N. europaea, nirK transcription is regulated via the nitrite-sensitive transcriptional repressor nsrA (30). Thus, in contrast to the nirK of many denitrifiers (74), nirK transcription in N. europaea is regulated in response to NO 2 Ϫ concentration and not NO or O 2 availability (31,34,48). The reduced O 2 supply during O 2 -limited growth resulted in an ϳ50% decrease in total NH 3 oxidized and an ϳ60% reduction in steady-state NO 2 Ϫ concentration ( Fig. 2; Table 1). The decrease in NO 2 Ϫ concentration during O 2 -limited growth likely induced the transcription of nsrA, which was significantly (2.1-fold) upregulated ( Fig. 5; Table S2). Therefore, the large decrease in nirK transcription observed here was likely due to the lower NO 2 Ϫ concentrations and not a direct reflection of overall nitrifier denitrification activity. In more natural nitrifying systems (e.g., agricultural soils or wastewater treatment plants [WWTPs]) changes in NO 2 Ϫ concentration could have a greater effect on AOB nirK expression than O 2 availability. However, it should be noted that environmental NO 2 Ϫ concentrations are unlikely to reach those observed in this study (30 to 60 mmol liter Ϫ1 NO 2 Ϫ ). Regulation of nirK transcription in response to primarily NO 2 Ϫ and not O 2 concentration is consistent with the observation that NirK is not essential for NO 2 Ϫ reduction to NO in N. europaea. This supports the hypothesis that a not-yet-identified nitrite reductase is present in this organism. Previously, it was shown that N. europaea nirK knockout mutants are still able to enzymatically produce NO and N 2 O (29, 35), even if hydrazine is oxidized by HAO instead of hydroxylamine as an electron donor (35). In addition, NO and N 2 O formation have also been observed in the AOB Nitrosomonas communis that does not encode nirK (12). The other three genes in the NirK cluster (ncgCBA) were differentially transcribed, with ncgC and ncgB being transcribed at lower levels (2-and 1.3-fold, respectively), while ncgA was transcribed at a significantly higher level (2.6-fold) during O 2 -limited growth. The role of ncgCBA in N. europaea has not been fully elucidated, but all three genes were previously implicated in the metabolism or tolerance of N oxides and NO 2 Ϫ (31). In contrast, transcripts of the norCBQD gene cluster, encoding the iron-containing cyt c-dependent cNOR-type NO reductase, were present at slightly higher (1.2-to 1.5-fold) but not significantly different levels during O 2 -limited growth ( Fig. 5; Table S2). Previous research has demonstrated that in N. europaea, cNOR functions as the main NO reductase under anoxic and hypoxic conditions (35). Interestingly, all components of the proposed alternative heme-copper-containing NO reductase (sNOR), including  Table S2.
Oxygen-Limited Growth of Nitrosomonas europaea the NO/low-oxygen sensor senC (24), were transcribed at significantly higher levels (2.7to 10.8-fold) during O 2 -limited growth ( Fig. 6; Table S2). Therefore, it is possible that the phenotype describing cNOR as the main NO reductase in N. europaea (35) was a product of short incubation times and that during longer term O 2 -limited conditions, sNOR contributes to NO reduction during nitrifier denitrification. Another possibility is that the increased transcription of sNOR observed here during O 2 -limited growth is primarily related to respiration and not NO reductase activity.
Respiratory chain and terminal oxidases. N. europaea harbors a low-affinity cyt c aa 3 (A1 type) HCO but not a high-affinity cbb 3 -type (C type) cyt c HCO harbored by other AOB such as N. eutropha or Nitrosomonas sp. GH22 (28,50,52). Significantly higher numbers of transcripts (1.7-to 3.0-fold) of all three subunits of the cyt c aa 3 HCO and the cyt c oxidase assembly gene ctaG were detected during O 2 -limited growth ( Fig. 6; Table S2). Increased transcription of the terminal oxidase was expected, as it is a common bacterial response to O 2 limitation (75). In addition, transcripts of all three subunits of the proton translocating cyt bc-I complex (complex III) were present in higher numbers (Table S2). The genes encoding NADPH dehydrogenase (complex I) and ATP synthase (complex V) were transcribed at similar levels during both growth conditions (Table S2).
As mentioned above, transcripts of both subunits of sNOR (norSY, previously called coxB 2 A 2 ), and the NO/low-oxygen sensor senC were present at significantly higher numbers (2.7-to 10.8-fold) during O 2 -limited growth ( Fig. 6; Table S2). The NO reductase function of the sNOR enzyme complex was proposed based on domain similarities between NorY and NorB (24,32). Yet, norY phylogenetically affiliates with and structurally resembles B-type HCOs (76). In addition, NorY does not contain the five well-conserved and functionally important NorB glutamate residues (77), which are present in the canonical NorB of N. europaea. All HCOs studied thus far can reduce O 2 to H 2 O and couple this reaction to proton translocation, albeit B-and C-type HCOs translocate fewer protons per mole O 2 reduced than A-type HCOs (78). Notably, NO reduction to N 2 O is a known side reaction of the A2-, B-, and C-type but not A1-type HCOs (79)(80)(81). The transcriptional induction of sNOR during O 2 -limited growth reported here, as well as the high O 2 affinity of previously studied B-type HCOs (82), indicates that sNOR might function as a high-affinity terminal oxidase in N. europaea and possibly other sNOR-harboring AOB. Furthermore, functionally characterized B-type HCOs display a lower NO turnover rate than the more widespread high-affinity C-type HCOs (79,80). Taken together, these observations indicate that B-type HCOs, such as sNOR, are ideal for scavenging O 2 during O 2 -limited growth conditions that coincide with elevated NO concentrations, which would impart a fitness advantage for AOB growing under these conditions. Lastly, the NOR of Roseobacter denitrificans structurally resembles cNOR but contains an HCO-like heme-copper center in place of the heme-iron  Table S2. center of canonical cNORs. Interestingly, this cNOR readily reduces O 2 to H 2 O but displays very low NO reductase activity (83,84). Therefore, in line with previous hypotheses (79,83), the presence of a heme-copper center in NOR/HCO superfamily enzymes, such as the sNOR of N. europaea, may indicate O 2 reduction as the primary enzymatic function. Notably, a recent study provided the first indirect evidence of NO reductase activity of sNOR in the marine NOB, Nitrococcus mobilis (85). However, further research is needed to resolve the primary function of sNOR in nitrifying microorganisms.
Conclusions. In this study, we examined the transcriptional response of N. europaea to continuous growth under steady-state NH 3 -and O 2 -limited conditions. Overall, O 2 -limited growth resulted in a decreased growth yield but did not invoke a significant stress response in N. europaea. On the contrary, a reduced need for oxidative stress defense was evident. Interestingly, no clear differential regulation was observed for genes classically considered to be involved in aerobic NH 3 oxidation. In contrast, a strong decrease in transcription of RuBisCO-encoding genes during O 2 -limited growth was observed, suggesting that control of CO 2 fixation in N. europaea is exerted at the level of RuBisCO enzyme concentration. Furthermore, the remarkably strong increase in transcription of the genes encoding sNOR (B-type HCO) indicates this enzyme complex might function as a high-affinity terminal oxidase in N. europaea and other AOB. Overall, despite lower growth yield, N. europaea successfully adapts to growth under hypoxic conditions by regulating core components of its carbon fixation and respiration machinery.
For steady-state growth, a flowthrough bioreactor (Applikon Biotechnology) with a 1-liter working volume was inoculated with 2% (vol/vol) of an exponential-phase N. europaea batch culture. The bioreactor was set to "batch" mode until the NH 4 ϩ concentration reached Ͻ5 mmol liter Ϫ1 (6 days) (see Table S1 in the supplemental material). Subsequently, the bioreactor was switched to continuous flow "chemostat" mode, at a dilution rate/specific growth rate () of 0.01 h Ϫ1 (doubling time ϭ ϳ70 h), which was controlled by a peristaltic pump (Thermo Scientific). The culture was continuously stirred at 400 rpm, and the pH was automatically maintained at 7.0 Ϯ 0.1 by addition of sterile 0.94 mol liter Ϫ1 (10% [wt/vol]) Na 2 CO 3 solution. Sterile-filtered (0.2 m) air, at a rate of 40 ml min Ϫ1 , was supplied during batch and NH 3 -limited steady-state growth. Once NH 3 -limited steady-state was reached (day 7), the chemostat was continuously operated under NH 3 -limited conditions for 10 days. To transition to O 2 -limited steady-state growth, after day 16, the air input was stopped, and the stirring speed was increased to 800 rpm to facilitate gas exchange between the medium and the headspace. The headspace was continuously Ϫ in the growth medium. The culture was continuously grown under these conditions for 10 days.
Sterile samples (ϳ5 ml) were taken on a daily basis. Culture purity was assessed by periodically inoculating ϳ100 l of culture onto lysogeny broth (Sigma-Aldrich) agar plates, which were incubated at 30°C for at least 4 days. Any observed growth on agar plates was considered contamination, and those cultures were discarded. NH 4 ϩ and NO 2 Ϫ concentrations were determined colorimetrically (86), and cell density was determined spectrophotometrically (Beckman) by making optical density measurements at 600 nm (OD 600 ) (Table S1). Total biomass in grams (dry cell weight) per liter, substrate consumption rate (q NH3 ), and apparent growth yield (Y) were calculated as described in Mellbye et al. (43). To test for statistically significant differences in NH 4 ϩ to NO 2 Ϫ conversion stoichiometry, q NH3 , and Y between NH 3and O 2 -limited steady-state growth, a Welch's t test was performed.
Depleted RNA quality was assessed using the Bioanalyzer 6000 Nano Lab-Chip kit (Agilent Technologies). Sequencing libraries were constructed from at least 200 ng rRNA-depleted RNA with the TruSeq targeted RNA expression kit (Illumina), and 100-bp paired-end libraries were sequenced on a HiSeq 2000 (Illumina) at the Center for Genome Research and Biocomputing Core Laboratories (CGRB) at Oregon State University.
Transcriptome analysis. Paired-end transcriptome sequence reads were processed and mapped to open reading frames (ORFs) deposited at NCBI for the N. europaea ATCC 19718 (NC_004757.1) reference genome using the CLC Genomics Workbench (CLC bio) under default parameters as previously described (43). Residual reads mapping to the rRNA operon were excluded prior to further analysis. An additive consensus read count was manually generated for all paralogous genes. Thereafter, mapped read counts for each gene were normalized to the gene length in kilobases, and the resulting read per kilobase (RPK) values were converted to transcripts per million (TPM) (87). To test for statistically significant differences between transcriptomes obtained from NH 3 -and O 2 -limited steady-state growth, TPMs of biological triplicate samples were used to calculate P values based on a Welch's t test. The more stringent Welch's rather than the Student's t test was selected due to the limited number of biological replicates (88). Additionally, linear fold changes between average TPMs under both growth conditions for each expressed ORF were calculated. Transcripts with a P value of Յ0.05 and a transcription fold change of Ն1.5ϫ between conditions were considered present at significantly different levels.
Data availability. All retrieved transcriptome sequence data have been deposited in the European Nucleotide Archive (ENA) under the project accession number PRJEB31097.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. FIG S1, PDF file, 1.7 MB.