Preparation for Denitrification and Phenotypic Diversification at the Cusp of Anoxia: a Purpose for N₂O Reductase Vis-à-Vis Multiple Roles of O₂

ABSTRACT

Adaptation to anoxia by synthesizing a denitrification proteome costs metabolic energy, and the anaerobic respiration conserves less energy per electron than aerobic respiration. This implies a selective advantage of the stringent O₂ repression of denitrification gene transcription, which is found in most denitrifying bacteria. In some bacteria, the metabolic burden of adaptation can be minimized further by phenotypic diversification, colloquially termed “bet-hedging,” where all cells synthesize the N₂O reductase (NosZ) but only a minority synthesize nitrite reductase (NirS), as demonstrated for the model strain Paracoccus denitrificans. We hypothesized that the cells lacking NirS would be entrapped in anoxia but with the possibility of escape if supplied with O₂ or N₂O. To test this, cells were exposed to gradual O₂ depletion or sudden anoxia and subsequent spikes of O₂ and N₂O. The synthesis of NirS in single cells was monitored by using an mCherry-nirS fusion replacing the native nirS, and their growth was detected as dilution of green, fluorescent fluorescein isothiocyanate (FITC) stain. We demonstrate anoxic entrapment due to e⁻-acceptor deprivation and show that O₂ spiking leads to bet-hedging, while N₂O spiking promotes NirS synthesis and growth in all cells carrying NosZ. The cells rescued by the N₂O spike had much lower respiration rates than those rescued by the O₂ spike, however, which could indicate that the well-known autocatalytic synthesis of NirS via NO production requires O₂. Our results bring into relief a fitness advantage of pairing restrictive nirS expression with universal NosZ synthesis in energy-limited systems.

IMPORTANCE Denitrifying bacteria have evolved elaborate regulatory networks securing their respiratory metabolism in environments with fluctuating oxygen concentrations. Here, we provide new insight regarding their bet-hedging in response to hypoxia, which minimizes their N₂O emissions because all cells express NosZ, reducing N₂O to N₂, while a minority express NirS + Nor, reducing NO₂⁻ to N₂O. We hypothesized that the cells without Nir were entrapped in anoxia, without energy to synthesize Nir, and that they could be rescued by short spikes of O₂ or N₂O. We confirm such entrapment and the rescue of all cells by an N₂O spike but only a fraction by an O₂ spike. The results shed light on the role of O₂ repression in bet-hedging and generated a novel hypothesis regarding the autocatalytic nirS expression via NO production. Insight into the regulation of denitrification, including bet-hedging, holds a clue to understanding, and ultimately curbing, the escalating emissions of N₂O, which contribute to anthropogenic climate forcing.

INTRODUCTION

In the energy hierarchy of dissimilatory processes, denitrification is second only to aerobic respiration. The four-step process where nitrate (NO₃⁻) is reduced to dinitrogen (N₂) via nitrite (NO₂⁻), nitric oxide (NO), and nitrous oxide (N₂O) is driven by the metalloenzymes nitrate (Nar/Nap), nitrite (NirK/NirS), nitric oxide (c/q/qCu_ANor), and nitrous oxide (NosZ I or II) reductase (1–3). Thermodynamics dictate that the reduction of nitrate to N₂ releases 95% of the energy per e⁻ compared to the reduction of O₂ (4), but canonical denitrification yields only ~60% of the charge separation per electron transported (5, 6). Thus, it makes sense that denitrifiers in general, being facultative anaerobes, strongly favor O₂ as electron acceptor, only switching to reduction of alternative e⁻ acceptors when conditions become anoxic. This transition is orchestrated by a massive, variably tuned apparatus transcriptionally regulated by O₂- and N oxide-sensing systems, e.g., consisting of Fnr/Crp-type proteins (5). Many denitrifying microorganisms have a truncated denitrification pathway because they lack one or more of the genes coding for the four reductases of the full-fledged pathway (7). However, a truncated pathway may even occur in organisms equipped with all of the genes due to repressed gene expression or posttranscriptional interference with the synthesis of functional enzymes (8). These premises give rise to a spectrum of physiological variants (7), with implications for the emission of gaseous intermediates, such as the ozone-depleting greenhouse gas N₂O (9, 10).

Since the metabolic cost of synthesizing proteins is huge compared to the costs of replicating the genes, organisms need transcriptional regulation that secures protein synthesis only when the enzyme is needed (11). The cost of synthesizing the entire denitrification apparatus is high, as it requires the expression of more than 50 functional and ancillary genes (1). Thus, when anoxia is imminent, denitrifying bacteria face the conundrum of when and whether to make the investment of synthesizing the denitrification proteome. Most denitrifying bacteria are nonfermenting, relying on respiration for generation of ATP, and these organisms must synthesize a minimum of denitrification enzymes before oxygen is depleted to avoid entrapment in anoxia without energy to synthesize such enzymes. To our knowledge, Højberg et al. (12) were the first to demonstrate such entrapment, achieved by inflicting sudden anoxia. This shows that excessively stringent transcriptional oxygen repression of the denitrification genes implies a risk for entrapment in anoxia. Conversely, lack of negative regulation, resulting in production of the entire denitrification proteome under high O₂ tension, is wasteful and may reduce fitness. The dilemma is exacerbated by the fact that the organisms cannot sense the future oxygen trajectory of their habitat: the imminent anoxia may initiate a long-lasting anoxic spell or just a transient depression of oxygen concentration. This dilemma and a compromise manifest in the alphaproteobacterium Paracoccus denitrificans. The model organism carries the full set of denitrification genes, encoding Nar and Nap, NirS, cNor, and NosZ I, but when aerobic cultures face anoxia, all cells express nosZ, while only a minor subpopulation synthesizes NirS (13). This was attributed to FnrP-mediated transcription of nosZ under semioxic conditions (14) but a low probability of nir (+ nor) transcription in the presence of O₂, accompanied by the requirement for autocatalytic induction by NO (via Nnr) for full expression (15, 16). Hence, P. denitrificans displays early induction of nosZ but strong repression of nirS resulting in phenotypic diversification, colloquially termed bet-hedging. This is likely to increase fitness on a population level. In mixed communities, expressing nosZ enables the cells to scavenge N₂O emitted by others, thus maintaining metabolic activity and avoiding entrapment in anoxia. Should oxygen return, the cells have limited their investment, but in the event of persistent anoxia, respiration of N₂O should enable them to eventually synthesize a full denitrification proteome.

It is reasonable to assume that anoxic entrapment underlies the cell differentiation observed in clonal populations of P. denitrificans. Previous experiments where O₂ or N oxides were introduced to nondifferentiated cultures of P. denitrificans after prolonged periods of electron acceptor deprivation (hours to days) showed rapid recovery of respiration rates (unpublished data). This indicates that spikes of O₂ would represent a second chance for entrapped cells to synthesize a complete denitrification proteome. The effect of oxygen is two-sided, however, as it modulates transcription (suppressing denitrification) and acts as an e-acceptor, but it is not known to play any additional roles during de novo synthesis of N-oxide reductases or ancillary factors. Thus, if the required machinery (NosZ) is present, O₂ should be fully replaceable by N₂O as the e-acceptor for establishing a full-fledged denitrification proteome. Unlike O₂, N₂O has no known regulatory role and should thus facilitate nirS expression in all cells carrying NosZ. However, this has yet to be explored.

A prerequisite for studying entrapment in anoxia is the exposure of aerobic cultures to sudden and complete anoxia, which poses a challenge in liquid cultures, as micromolar concentrations of O₂ remain after sparging or He washing (Molstad et al. [17] and subsequent sections). We implemented a procedure for removal of residual O₂ after He washing, prior to inoculation, using glucose oxidase + catalase (GOX) (18). Remaining O₂ after GOX treatment was ≤10 ppmv in headspace (corresponding to ≤0.013 μM in the liquid), resulting in complete anoxic entrapment of aerobically raised cells. We tracked NirS (mCherry-nirS fusion replacing native nirS) and anaerobic growth (dilution of fluorescein isothiocyanate [FITC] stain) in single cells in cultures of P. denitrificans during anoxic entrapment and spiking with O₂ or N₂O. Besides using fluorescence microscopy, we also established a readout method for the detection of physiological expression levels of NirS in flow cytometry using a high sensitivity charge-coupled-device (CCD) camera-based flow cytometer. Cells remained viable during prolonged periods of entrapment and were readily recruited to denitrification when supplied with pulses of O₂, albeit displaying the typical bet-hedging. Provision of N₂O led to universal NirS synthesis in entrapped cells carrying NosZ but with lower initial cell-specific NirS activity compared to O₂-spiked cells. Thus, O₂ is replaceable as an e-acceptor during the oxic-anoxic transition, and its conflicting roles make it a less-effective facilitator of NirS synthesis than N₂O. However, the comparably weaker induction of nirS in N₂O-spiked cells indicate an additional role of oxygen in enhancing nirS expression, possibly linked to NO and Nnr.

RESULTS

The glucose oxidase-catalase system.

The glucose oxidase-catalase system (GOX) was tested with respect to the kinetics of O₂ removal, inactivation of glucose oxidase, specificity (NO or N₂O scavenging), acidification of medium, and toxicity to anaerobically respiring cells. The O₂ removal capacity of GOX was limited by turnover-dependent inactivation of glucose oxidase, thus near-anoxia (≤10 ppmv O₂ in headspace and ≤0.013 μM in the liquid) could only be achieved after prior depletion of O₂ by He washing as described. The GOX reactants did not react with NO or N₂O, did not affect the pH of the medium significantly (GOX treatment of He washed, sterile medium reduced the pH by 0.08 ± 0.02 pH units), and had no adverse physiological effect on anaerobically respiring cells. For details, see Supplemental Item 1 (SI 1) at https://data.mendeley.com/datasets/zpwkwwg5xz.

Flow cytometry for detection of mCherry and FITC fluorescence.

Flow cytometry was used in the O₂-spiking experiment to distinguish between active and inactive subpopulations. Cells separated well from background by forward scatter (FSC) and side scatter (SSC) (Fig. 1A). To further exclude coincident events and aggregates, we used FSC intensity versus FSC aspect ratio, which is a morphometric parameter available on CellStream. Using the imagery captured by the system’s CCD camera and object detection/masking, the software allows for evaluation of the ratio of the minor axis divided by the major axis of an ellipse fitted around the detected objects (Fig. 1B). The Amnis CellStream allowed for the detection of weak fluorescence in single bacterial cells (tentative approximation of mCherry-NirS proteins per NirS+ cell in order of magnitude, 10³; cell size, 1 to 2 μm by 0.5 to 1 μm), thus distinguishing between cells with and without mCherry-NirS in a culture after depletion of O₂. Likewise, growing/nongrowing subpopulations in a bet-hedging culture could be distinguished by FITC intensity, which is retained in nongrowing cells but diluted by growth (13). Specificity of the signals and detection sensitivity were assessed using appropriate control samples (Fig. 1C and D).

FIG 1.

Gating strategy and detection of mCherry and FITC. (A) Discrimination of bacterial population from background noise in FSC-456/51 versus SSC-773/56 scatterplot (20,000 observations). (B) Identification of singlets within the bacterial population, FSC-456/51 versus aspect ratio_FSC-456/51. (C) Distribution of FITC fluorescence intensity within the singlet population of an unstained culture (FITC −, gray), stained aerobic inoculum (FITC +, blue), and bet-hedging population treated similarly to vial 1.2 in Table 1 (red). (D) Distribution of mCherry fluorescence intensity within singlet population of an aerobic (mCherry negative) culture (mC −, blue), a bet-hedging culture (green), and a near 100% NirS-positive control that was grown anaerobically over several batches (mC +, red).

Two main experiments: entrapment in anoxia and spiking with O₂ or N₂O.

The subsequent paragraphs summarize the results of the two experiments gauging NirS expression and anaerobic growth in single cells of P. denitrificans (i) during exposure of aerobic cultures to sudden anoxia and partial or complete entrapment followed by O₂ spiking and (2) after N₂O spiking following sudden or gradual (N oxide deficient) transition to anoxia.

(i) Experiment 1: entrapment in anoxia and response to subsequent O₂ spiking. This experiment was designed to secure a fast transition to complete anoxia, i.e., rapid depletion of residual O₂, by using a large inoculum. This would entrap a majority of cells in anoxia based on the theory of a probabilistic initiation of nirS transcription in response to hypoxia and that cells which fail to synthesize NirS before O₂ is depleted will be unable to do so later because they lack the energy (13). According to this theory, (i) a minority of the cells would avoid entrapment in He-washed vials (containing only 450 ppmv O₂ in the headspace), (ii) an even lower number of cells (if any) would make it in the vials pretreated with GOX (containing ≤10 ppmv O₂), and (iii) a subsequent spike of O₂ would enable NirS synthesis in at least a fraction of the cells. The gas kinetics lend strong support to this theory. (i) In the He-washed vials, there was detectable and exponentially increasing N₂ production after depletion of the residual O₂, but the calculated electron flow rates to denitrification indicated that less than 10% of the cells had switched to anaerobic respiration and growth. (ii) In the vials pretreated with GOX, N₂ production remained below the system’s detection limit throughout the 120 h of incubation. (iii) A spike of oxygen after 69 h (6 μM in the liquid, depleted within 20 to 30 h) induced significant N₂ production in the GOX-treated vials after depletion of the oxygen and resulted in enhanced N₂ production (after O₂ depletion) in the He-washed vials. These phenomena were observed both for unstained and FITC-stained cells. Detailed gas data and electron flow rates for FITC-stained and -unstained cells are shown in SI 2, and the apparent anaerobic growth rates as calculated by nonlinear regression of the N₂ production rate against time are shown in SI 4 (both found at https://data.mendeley.com/datasets/zpwkwwg5xz).

(a) FITC-stained cultures, gas kinetics, and flow cytometry. The accumulation of gaseous N oxides and N₂ and optical density at 660 nm (OD₆₆₀) measurements were used to estimate average e flow rates per cell to denitrification (v_e-dT, fmol e⁻ cell⁻¹ h⁻¹) (Fig. 2A). In the He-washed vials untreated by GOX, v_e-dT fluctuated below 0.1 fmol e⁻ cell⁻¹ h⁻¹ for 40 h before gradually increasing, exceeding 1.5 fmol e⁻ cell⁻¹ h⁻¹ toward the end of the experiment. Injection of O₂ to such vials after 69 h led to transient suppression of v_e-dT followed by a sharp increase to higher levels than in the vials that were not spiked with O₂. The cultures in GOX-treated vials remained largely inactive (except for minimal accumulation of NO and N₂O; see SI 2 at https://data.mendeley.com/datasets/zpwkwwg5xz) unless spiked with O₂ after 69 h, which induced a subsequent exponential increase in v_e-dT, reaching 0.6 fmol cell⁻¹ h⁻¹ at the end of the experiment (Fig. 2A).

FIG 2.

Entrapment in anoxia and subsequent recruitment of cells to denitrification by O₂ spiking. Observed activity (gas kinetics) and discrimination of active and inactive subpopulation by flow cytometry in FITC-stained P. denitrificans at selected time points. Complete flow cytometry data are found within Supplementary Item 2 (https://data.mendeley.com/datasets/zpwkwwg5xz). (A) Electron flow rate per cell in the total population (v_e-dT = V_e-D/N_T, where V_e-D is the electron flow rate to denitrification [mol e⁻ vial⁻¹ h⁻¹] and N_T is the total number of cells in the vial) plotted against time for four different treatments (Table 1); 1.2, He-washed vials; 1.3, He-washed vials spiked with O₂ after 69 h; 1.8, completely anoxic vials (GOX pretreated); 1.9, completely anoxic vials spiked with O₂ after 69 h. The oxygen spike is shown as a shaded area (complete data in SI 2). (B) Cumulative cells mL⁻¹ and the fraction of actively denitrifying cells as observed by flow cytometry (population “A,” upper, left quadrants in panel C). (C) Upper two rows show He-washed (not GOX-treated) vials (1.2 and 1.3 in panel A), and lower two rows show GOX-treated cultures with and without O₂ spiking (1.8 and 1.9 in panel A). Complete flow cytometry data are shown in Movies 1 to 4 and in SI 2 found at https://data.mendeley.com/datasets/zpwkwwg5xz.

The low initial v_e-dT, and its increase with time was expected, assuming that a majority of the cells became entrapped in anoxia without NirS, while a minority synthesized NirS in time, sustaining subsequent anaerobic respiration and growth. This was corroborated by inspecting FITC and mCherry-NirS fluorescence in the cells by flow cytometry as shown in Fig. 2C. In these plots of mCherry intensity versus FITC intensity, all cells are expected to be in the lower right quadrant initially (no mCherry and high FITC), and as they synthesize mCherry-NirS, they move to the upper right quadrant and, if growing, they dilute FITC and move toward the upper left quadrant.

In the He-washed vials (not GOX treated) (vial 1.2 in Fig. 2A and C), ~8% of the cells had expressed NirS (mCherry positive) and had grown (diluted FITC) after 65 h, and the fraction increased gradually throughout the rest of the incubation, while a very low fraction resided in the upper right quadrant, i.e., cells that had expressed nirS but had not grown. In the vial spiked by O₂ after 69 h (vial 1.3), two NirS-positive populations were detected after 91 h, one with a very diluted FITC signal and one with a much higher signal. The latter plausibly represents cells that had been entrapped in anoxia but enabled to synthesize NirS due to the O₂ spike. This extra recruitment to denitrification was also seen as an increase in the average cell-specific electron flow to denitrification (v_e-dT), subsequent to the depletion of the O₂ spike (vial 1.3 versus vial 1.2) (Fig. 2A).

In the GOX-treated vials without O₂ spiking (vial 1.8) (Fig. 2A and C), essentially all cells remained in the lower right quadrant, i.e., without NirS and with no growth, indicating that the entire population was entrapped in anoxia without NirS. If spiked with O₂, however (vial 1.9), a new population emerged with full expression of NirS and with a subsequent gradually declining FITC signal, indicating growth by anaerobic respiration.

In vials without GOX, cumulative cell densities (based on direct counts by flow cytometry corrected for dilution by liquid sampling) correlated well with growing subpopulations of active cells (F_A) (Fig. 2B, top). In GOX-treated vials, F_A never increased to levels sufficient for any detectable increase in cell density (Fig. 2B, bottom).

(b) Estimation of active population in unstained cultures. While flow cytometry provided direct observation of the fraction of actively denitrifying cells (F_A = the fraction of mCherry-positive, FITC diluting cells), F_A can also be estimated from measured gas kinetics and cell density, provided that we know the cell-specific rate of electron flow to denitrification in the active cells = v_e-dA (mol e⁻ cell⁻¹ h⁻¹):

$$F_A=\frac{V_{e-D}}{v_{e-\text{d}A}\text{*}{N}_T}$$ (1)

where V_e-D is the measured rate of electron flow to denitrification in the vial (mol e⁻ vial⁻¹ h⁻¹) as calculated from the measured gas kinetics, and N_T is the total number of cells (cells vial⁻¹) as measured by OD. The experiments provide no direct measurements of v_e-dA, however, and to estimate this, we fitted (by least square) F_A= (V_e-D/v_e-dA)/N_T (equation 1) to F_A measured by flow cytometry (adjusting v_e-dA) with the generalized reduced gradient solver in Excel using data from vials 1.2, 1.3, 1.8, and 1.9 (Table 1). The result (Fig. 3A) shows a reasonable fit throughout, with v_e-dA = 2.5 fmol e⁻ cell⁻¹ h⁻¹. This means that the measured electron flow kinetics could be used to estimate F_A throughout the batch cultivation of unstained cultures as shown in Fig. 3B. This shows that unstained cultures responded similarly to the FITC-stained cells. In the He-washed vials (not GOX treated), recruitment to denitrification was evident immediately after inoculation, followed by a gradually increasing fraction of active cells. The cells in the GOX-treated vials remained inactive until spiked with O₂, which induced recruitment to denitrification, and the active fraction subsequently increased to reach ~0.8 (80%) toward the end of the experiment. The apparent growth rates of the active populations in this experiment were estimated by nonlinear regression of the electron flow rates to denitrification against time (see SI 4 at https://data.mendeley.com/datasets/zpwkwwg5xz; Table 1), and were 0.038 h⁻¹ (± 0.001 h⁻¹; n = 4) for the He-washed vials without GOX (Fig. 3B).

TABLE 1.

Treatment of individual vials in main experiments^a

Expt no. and condition	Vial no.	FITC	Gas spikes	Liquid sampling
Expt 1, O2 spiking
Helium washed	1.1	+
	1.2	+		+
	1.3	+	O2	+
	1.4
	1.5			+
	1.6		O2	+
GOX pretreated	1.7	+
	1.8	+		+
	1.9	+	O2	+
	1.10
	1.11			+
	1.12		O2	+
Expt 2, N2O spiking
Initially 0.25 vol % O2	2.1-2	+
	2.3-4			+
	2.5-8		N2O
	2.9-11		N2O	+
GOX pretreated	2.12-13
	2.14-15			+
	2.16-17		N2O	+/−
	2.18-19		N2O + O2	+/−
	2.20-21		O2	+/−
	2.22-23	+	O2 + N2O	+/−

^aFor experiment 1, O₂-spiking, 18 vials with 50 mL Sistrom’s medium were either only helium washed (i.e., with 200 to 400 ppmv O₂) or made completely anoxic by the GOX pretreatment and inoculated either with unstained cells (n = 9) or cells stained with FITC (n = 9). KNO₂ (100 μmol vial⁻¹) was after 5 h. In an additional experiment, KNO₂ was added immediately after inoculation, without any consequences for the results (see SI 2, Fig. IV and V versus VI and VII at https://data.mendeley.com/datasets/zpwkwwg5xz). A spike of O₂ (15.5 μmol O₂) was added to selected vials after 69 h. For experiment 2, N₂O-spiking, 23 vials with Sistrom’s medium that had been stripped of nitrite and nitrate, either completely anoxic (pretreated with GOX, n = 12) or with 0.25 vol % O₂ in the headspace (n = 11), were all inoculated with FITC-stained cells. A total of 100 μmol KNO₂ was injected after 14 h. Selected vials were spiked with either O₂ or N₂O. For both experiments, liquid sampling for OD measurement, microscopy (Exp. 2) and flow cytometry (Exp. 1), was made throughout in some vials, while others were left untouched to obtain undisturbed measurement of the gas kinetics. GOX-pretreated results for experiment 2 are reported within supplementary items only.

FIG 3.

Fraction of active cells (F_A) in unstained cultures. (A) Fraction of active cells (F_A) throughout the incubation of FITC-stained cells estimated by measured electron flow rates (equation 1), fitted by least-squares to F_A measured by flow cytometry. The fitting resulted in an estimated cell-specific electron flow rate, v_e-dA = 2.5 fmol e⁻ cell⁻¹ h⁻¹, in active cells. (B) F_A in unstained cultures throughout the incubations, calculated by equation 1, assuming v_e-dA = 2.5 fmol e-cell⁻¹ h⁻¹. Open black circles, He-washed vials (not GOX treated; n = 2 replicate vials) with and without O₂ spiking; open red circles, GOX-treated vials (n = 2 replicate vials) without O₂ spiking; closed red circles, GOX-treated cultures (n = 2) spiked with O₂ at 69 h.

(ii) Experiment 2: entrapment in anoxia and response to subsequent N₂O spiking. In this experiment, aerobically raised, FITC-stained cells either faced sudden anoxia (inoculated to GOX-treated vials) or transient hypoxia (initial O₂ in headspace of ~0.25 vol %) in a medium that had been stripped for nitrate and nitrite. Nitrite was added 14 h after inoculation, i.e., after depletion of O₂. Selected vials were spiked with N₂O or O₂ (Table 1).

We hypothesized that cells exposed to sudden anoxia (GOX-pretreated vials) would have little or no NosZ; hence, they would be unable to utilize a spike of N₂O to generate energy for NirS synthesis. This was verified both by the observed gas kinetics (N₂O was not reduced, and the injection induced no detectable N₂ production from NO₂⁻) and by microscopy (no cells expressed NirS, and all retained a high FITC signal). Details are shown in SI 3 (found at https://data.mendeley.com/datasets/zpwkwwg5xz).

In contrast, cells that went through transient hypoxia (initially 0.25 vol % O₂) had evidently synthesized NosZ and were thus poised for NirS synthesis when spiked with N₂O. The N₂O spike was quickly reduced (within ~2 h), inducing a subsequent high rate of N₂ production from NO₂⁻, and the microscopy revealed that all of the cells expressed NirS (became mCherry positive) and grew (diluted the FITC signal) (Fig. 4; see also Movies 5 and 6, with more detailed gas kinetics and microscopy data shown in SI 3 at https://data.mendeley.com/datasets/zpwkwwg5xz). Prior to the N₂O spiking, however, only a marginal fraction of the cells had synthesized NirS as evidenced by very low electron flow to denitrification, v_e-dT < 0.1 fmol e⁻ cell⁻¹ h⁻¹ (Fig. 4A), which assuming v_e-dA = 2.5 fmol e cell⁻¹ h⁻¹ is equivalent to F_A < 4%.

FIG 4.

Effect of N₂O spiking on NirS synthesis and activity in cells carrying NosZ. Vials (n = 11) with nitrite/nitrate-free medium and 0.25% O₂ in headspace inoculated with FITC-stained aerobic cells. Seven vials were spiked with 20 μmol N₂O after ~26 h. (A) Average electron flow to N oxides (v_e-dT, fmol e⁻ cell⁻¹ h⁻¹) during ~50 h of anoxia in cultures with (red circles) and without (black open circles) N₂O spiking (gray area). v_e-dT is the electron flow rate per cell = V_e-D/N_T, where V_e-D is the electron flow rate in the whole vial (mol e⁻ h⁻¹) for each time increment between two gas samplings, and N_T is the cell number per vial for the same time interval. (B) FITC and mCherry fluorescence in single cells after 27 and 73 h in vials spiked with N₂O (top) and vials not spiked (bottom). Time resolved development of subpopulations is summarized in Movies 5 and 6. (C) Fraction of total population carrying mCherry-NirS (as observed by fluorescence microscopy; diamonds) compared to fraction of actively denitrifying cells (F_A; lines) as estimated based on measured electron flow rates and equation 1, assuming that the electron flow rate per active cell v_e-dA = 2.5 fmol e⁻ cell⁻¹ h⁻¹, as determined previously for oxygen-spiked cells (Fig. 3).

Considering that all cells expressed NirS in response to the N₂O spike (Fig. 4C, right), one would expect that the average electron flow rate per cell (v_e-dT) should equal v_e-dA = 2.5 fmol cell⁻¹ h⁻¹ as determined for the active cells in experiment 1 (see Fig. 3). During depletion of the N₂O spike, v_e-dT was indeed very close to 2.5 fmol cell⁻¹ h⁻¹ (the single high value in Fig. 4A), but once the N₂O was depleted, i.e., when the cells were forced to use NO₂⁻, average v_e-dT fell to 0.6, increasing gradually to 2.0 fmol cell⁻¹ h⁻¹ during the subsequent 30 h. This suggests that (i) all cells had expressed nosZ prior to N₂O spiking and (ii) all cells synthesized a minimum of NirS in response to the spiking and that the amount of NirS per cell increased gradually thereafter.

These results demonstrate that O₂ and N₂O are entirely interchangeable as e-acceptors to provide entrapped cells with energy to synthesize NirS. But O₂ was less efficient at the population level than N₂O, plausibly due to its role as a repressor of the transcription of nirS. We speculated that O₂ could have a secondary effect on NirS synthesis by dampening the positive feedback via NO, i.e., that O₂ induces a sudden and complete shutdown of NO production, and tested this in an additional experiment where actively denitrifying cultures were spiked with N₂O and O₂, monitoring NO and the rate of NO₂⁻ reduction in response to this spiking. The results lend no support to this hypothesis, however. While V_eNIR dropped in response to O₂, the shutdown was complete in response to N₂O but only partial by O₂, and the NO declined to similar levels in response to both (see SI 5 at https://data.mendeley.com/datasets/zpwkwwg5xz).

An alternative explanation to the strong effect of N₂O injection on the synthesis of NirS (Fig. 4) could be a regulatory effect of N₂O as such via an unknown N₂O sensor protein. To test this, we conducted an experiment with an nosZ-deficient mutant and the wild type as control, both incubated with and without 0.4 vol % N₂O in the headspace (~120 μM N₂O in the liquid) and with 0.5 vol % O₂ in the headspace (see SI 3, Fig VI at https://data.mendeley.com/datasets/zpwkwwg5xz). To ensure that most cells became entrapped in anoxia, the cultures were provided with NO₂⁻ at a time when oxygen was nearly depleted. The measured kinetics of anaerobic respiration was used to estimate the fraction of cells expressing NirS (F_den), by fitting a simplified version of the model by Hassan et al. (15). For the wild type, F_den increased from 0.01 (± 0.0025) in the cultures without N₂O to 0.13 (±0.015) in those provided with N₂O. In contrast, F_den of the nosZ mutant remained low in both treatments as follows: 0.0043 (± 0.0012) without N₂O and 0.0075 (± 0.24) with N₂O (see SI 3, Fig VII at https://data.mendeley.com/datasets/zpwkwwg5xz).

DISCUSSION

We have previously demonstrated that P. denitrificans displays phenotypic diversification (bet-hedging) when preparing for anoxia, i.e., only a fraction of the population expresses nirS but all cells synthesize NosZ (13). Although we had a tentative explanatory model, a refined approach was necessary to understand (i) whether cells that fail to synthesize NirS before O₂ depletion become entrapped in anoxia due to energy deprivation and (ii) the efficacy of N₂O and O₂ spiking in promoting NirS synthesis. We found compelling evidence for anoxic entrapment and a fitness advantage of early NosZ expression, but we also made observations propounding reflections on the roles of O₂ and NO in transcriptional and metabolic regulation of denitrification.

The transcriptional regulation of denitrification has been extensively studied (19–21), but the respective roles of the three Fnr-type regulators FnrP, Nnr, and NarR and their effectors (O₂ and N oxides) are still not entirely clear. Part of the explanation for this is the indistinguishable binding sites of the respective factors (22). Oxygen limiting conditions activate FnrP, which positively or negatively regulates the expression of many genes. Among those positively regulated in P. denitrificans are the cbb₃ high-affinity oxidase (23) and nar and nos (24). The nir genes have been reported to be subject to negative regulation by FnrP (20), most likely through cross talk with Nnr. FnrP is found in higher numbers compared to Nnr in P. denitrificans (20), and unlike Nnr and NarR, it only requires release from O₂ suppression for its activation. Thus, FnrP constitutes the majority of active Fnr-type proteins in cells facing anoxia and may contribute strongly to (leaky) repression of nir expression. Besides repression by FnrP, other O₂ responsive systems may be involved in nir regulation, such as the recently described denR-NirR system (25). The current wisdom is that in hypoxic cells where nir escapes repression, production of a fully functional NirS pool is driven by the positive feedback loop with NO via Nnr and fueled by respiration of the remaining O₂ (16). Of note, the NO signal is subject to quenching by NO scavenging proteins such as the flavohemoglobin Hmp (26) and eventually the respiratory NO reductase, Nor. This could further restrict the initial synthesis of NirS.

Denitrification is also regulated at the metabolic level. It is well known that actively denitrifying cells shut down denitrification almost instantaneously if provided with O₂, which was clearly the case in our experiment: spiking with O₂ lowered the denitrification rate to a minimum (Fig. 2A) until the concentration of O₂ in the liquid reached <0.5 μM. This response can be ascribed to a competition for electrons between terminal oxidases and the N-oxide reductases, as studied by Kucera and Sedlacek (27), who found that in P. denitrificans, the high-affinity cbb3-type oxidases played a key role in drawing electrons away from the N-oxide reductases. Our experiments demonstrated (see SI 5 at https://data.mendeley.com/datasets/zpwkwwg5xz) that the electron flow is drawn effectively away from NirS by terminal oxidases (in response to O₂ spiking) and even more effectively by N₂O reductase (in response to N₂O spiking), while the NO concentrations reached similarly low levels by O₂ and N₂O spiking. This phenomenon has implications for our understanding of NO’s role (if any) in the regulatory biology underpinning bet-hedging in P. denitrificans.

The bet-hedging of P. denitrificans in response to O₂ depletion has previously been ascribed to a low probability for a cell to initiate nirS transcription (as discussed above), resulting in two populations after oxygen depletion, one actively denitrifying population with NirS and one entrapped in anoxia without energy for synthesizing NirS (13, 15, 16). Direct evidence for energy-dependent entrapment in anoxia was lacking however, and the role of O₂ as repressor and NO as an inducer of the initial transcription of nirS remained unclear. Our results lend strong support to the energy-dependent entrapment by demonstrating that the entrapped cells could indeed synthesize NirS if provided with a spike of either N₂O or O₂. Further, the fact that all the entrapped cells synthesized NirS in response to the N₂O spike, while a minority did so in response to the O₂ spike can be taken to illustrate the role O₂-responsive factors as repressors of nirS under hypoxia, resulting in a very low probability for a cell to initiate transcription of nirS. The role of NO is still elusive, however. In theory, NO could play two roles as follows: one is to secure autocatalytic synthesis of NirS once the first molecules of NirS in a cell become active, producing NO, and a second role could be that NO produced by the cells with NirS could induce the initiation of nirS transcription in cells without any NirS. Such signaling does not seem to occur, however. A tentative explanation could be that during the transition to anoxia, such NO signaling is effectively quenched by the NO-scavenging protein Hmp, which is most active under aerobic conditions (26). Under anoxic conditions, Nor is likely the main NO sink. Whatever causes NO scavenging, the concentration of NO did decline to similarly low levels in response to O₂ and the N₂O spike (see SI 5 at https://data.mendeley.com/datasets/zpwkwwg5xz), while bet hedging only occurred in response to the O₂ spike.

It is tempting to speculate that NO as such is not the inducer of transcription of nirS but that the culprit is one of its possible products within the cell (28). If so, this unknown ^xNO^x derived from NO would be able to secure a positive feedback of nirS transcription within a cell but plausibly not affect the transcription in other cells if retained within the cell or its half-life is too short for it to reach out to other cells. Supposing that the formation of our hypothetical ^xNO^x requires O₂, the positive feedback loop of nirS transcription would not be effective in anoxia as appeared to be the case in the cultures spiked with N₂O: the cell-specific NirS activity increased gradually throughout a period of 20 h. Peroxynitrite (ONOO⁻/ONOOH, pK_a = 6.8) is a compound that could fill these criteria: it is readily formed from NO by a very fast chemical reaction with superoxide (O₂⁻) (28), and O₂⁻ is a by-product of reactions between O₂ and complexes within the e-transport chain. As a result, NO production within a cell will generate peroxynitrite under hypoxia, but less (if any) if oxygen is absent. At cytoplasmic pH, which is typically 7 to 7.5, peroxynitrite (pK_A = 6.8) will be predominantly anionic and, thus, retained within the cell. While highly speculative, this hypothesis is tantalizing because it can explain the observed phenomena, which appears to conflict with the conventional concept of NO as the signal molecule sensed by Nnr. Of note, the identification of Nnr as an NO sensor was deduced from in vivo experiments, and there is no direct evidence that Nnr reacts directly with NO (22, 29). Ironically, our refined in vivo experiments can be taken to suggest that NO is not the substance sensed by Nnr.

In conclusion, we have strong evidence in favor of our longstanding hypothesis that NosZ can act as an energy efficient safety valve in substrate-poor systems where oxygen fluctuates. In the case of a short-lived anoxic spell, the cells limit their investment by restrictive NirS production but secure the option of continued respiration and growth through scavenging N₂O emitted by the surrounding microbial community. As such, NosZ provides a second chance to cells that initially “opted out” of full-fledged denitrification, e.g., the energy to eventually produce the complete set of denitrification proteins. In addition to this insight into the fitness advantage of bet-hedging, a novel understanding of the role of oxygen vis-à-vis the autocatalytic regulation of NirS is emerging. This multiplicity, where oxygen is concurrently suppressor, e-acceptor, and enhancer of gene expression speaks to the intricacy of the denitrification regulatory network, and it will fuel further enquiries into the interplay between NO and reactive oxygen species.

MATERIALS AND METHODS

Organism.

All experiments were performed on P. denitrificans Pd1222 where an mCherry-nirS fusion gene replaces the native nirS gene, which allows the tracking of NirS in individual cells by fluorescence microscopy (13). For testing the potential regulatory effect of N₂O, Pd1222 wild type and an nosZ mutant derived from this strain were used (20).

Medium.

Sistrom’s medium as described by Lueking et al. (30) with 34 mM succinate, pH = 7, was used throughout, and KNO₂ (1 or 2 mM) was added as an e-acceptor for anoxic respiration. For experiments where complete absence of nitrate and nitrite was essential, we stripped the medium for these nitrogen oxyanions prior to use as described in detail by Bergaust et al. (24). In short, the stripping was achieved by anaerobic incubation of medium with ~10⁷ Paracoccus denitrificans cells mL⁻¹, followed by filtration and autoclaving.

Helium washing.

In preparation for incubation experiments, 120-mL serum vials filled with 50 mL Sistrom’s medium and a magnetic Teflon stirring bar were sealed with butyl rubber septa and aluminum crimp caps and washed with helium repeatedly as follows: 7 cycles of evacuation followed by filling with helium using an automated system described by Molstad et al. (17). This fails to remove all O₂, however, and the O₂ concentration increases gradually during the first 10 to 20 h after He washing due to the release of O₂ (and N₂) from the Teflon magnet and the septum (17), stabilizing at 200 to 400 ppmv (0.6 μM in the liquid) after 20 h. The standard procedure is thus to He wash the vials >24 h before inoculating the vials. The He washing leaves an overpressure in the vials, and this is released to reach 1 atmosphere after temperature equilibration in the water bath by piercing the septum with a syringe filled with 70% ethanol (no piston).

Glucose oxidation treatment for removal of residual O₂.

To obtain completely anoxic conditions, the residual O₂ in He-washed vials was removed by glucose oxidase + catalase as described by Thorndycroft et al. (18). This was done by injecting 1 mL of the mixture of the two enzymes (200 u/mL glucose oxidase, 1,000 u/mL catalase) and subsequently 1 mL glucose solution (800 mM glucose) into each vial with 50 mL medium. The efficiency of the GOX treatment was tested thoroughly, including vials with different initial O₂ concentrations of 1 to 21 vol % in the headspace, to determine the oxygen scavenging kinetics and the decay rate of the enzymes. We also tested if GOX had any effect on NO and N₂O by treating vials with these gases in the headspace. Since H₂O₂ is an intermediate in the glucose oxidase + catalase reaction, a transient accumulation of H₂O₂ during O₂ depletion could theoretically have some toxic effects. To avoid this, the vials were GOX treated 1 day before being inoculated. We tested if this pretreatment with GOX left any residual physiological effect by inoculating vials (untreated and GOX treated) with anaerobically raised cells of P. denitrificans.

FITC staining.

Aerobically raised cells were harvested by centrifugation at 4°C and stained with fluorescein isothiocyanate (FITC) as previously described by Lycus et al. (13). Briefly, cells were incubated with 0.1 mg/mL FITC for 10 min at 4°C and dispersed by pumping through a 0.5-mm needle to ensure the even uptake of the stain. Excess stain was removed by washing the cells three times with 30 mL Sistrom’s medium that was nitrite and nitrate free if required.

Incubation and monitoring of gas kinetics.

Incubations were carried out in a water bath at 17°C if not specified otherwise. Helium-washed vials were placed in the incubation system described by Molstad et al. (17). The concentrations of gases in the headspace (O₂, CO₂, NO, N₂O, and N₂) were monitored by repeated sampling through the rubber septum. The gas was sampled with a peristaltic pump that returned an equal amount of He to the headspace, ensuring a constant pressure. The dilution of headspace gases by He was accounted for in data analysis. The autosampler is coupled to a chemiluminescence NO/NOx analyzer (Teledyne 200E) and a GC (Agilent GC-7890A) with a PLOT column for separation of CH₄, CO₂, and N₂O and a Molsieve for separating O₂ and N₂. The GC has a flame ionization (FID), a thermal conductivity (TCD), and an electron capture (ECD) detector. N₂O is detected by both ECD and TCD to ensure accurate measurements at both near-ambient (ECD) and higher concentrations. Continuous stirring at 600 rpm ensured near equilibrium between gas concentrations in liquid and headspace.

Analyses of cell-specific electron flow rates to denitrification.

Given the fact that the cultures were provided with NO₂⁻ (not NO₃⁻) and that the incubation system provided frequent measurements of NO, N₂O, and N₂, thus monitoring all reduction steps, the electron flow rate to denitrification could be calculated for each time increment between two gas samplings (V_e-D, mol e⁻ vial⁻¹ h⁻¹). For each time increment, we could also estimate the average electron flow rate per cell v_e-dT = V_e-D/N_T, where N_T is the total number of cells in the vial, estimated from measured optical density (OD₆₆₀). OD was measured with lower frequency than the gas measurements, but by interpolating with the SRS1 cubic spline function (see SI 4 at https://data.mendeley.com/datasets/zpwkwwg5xz), we could estimate N_T for each time interval between two gas samplings. Assuming that the cell-specific electron flow to denitrification in active cells, v_e-dA, is known, we have the fraction of active cells within the whole population of cells, F_A = v_e-dT/v_e-dA. No direct measurements of v_e-dA could be made, but estimates were obtained by fitting F_A = v_e-dT/v_e-dA to F_A as measured by flow cytometry.

Liquid sampling.

Liquid samples (3 mL) were taken intermittently during incubations for measurement of OD₆₆₀ and analyses by microscopy or flow cytometry. The bottles were briefly inverted, and the samples were taken through the septum with a syringe. To secure constant liquid volume throughout, an equal volume of He-washed Sistrom’s medium was injected immediately prior to sampling. The OD₆₆₀ of the sample was measured, and a 1.8-mL sample was fixed by adding 38% formalin to a final concentration of 4% pending flow cytometry or microscopy.

Flow cytometry.

Formalin-fixed samples were diluted to 10⁶ cells/mL with filtered (0.1 μm) Milli-Q water before loading into the flow cytometer (Amnis CellStream, Luminex). The system was run at slow speed. Laser powers were adjusted to 20% for forward scatter (FSC), 50% for side scatter (SSC), 100% for 488 nm, and 100% for 561 nm. FSC, SSC, and FITC intensity was measured using the 528/46 filter, and mCherry intensity was detected using the 611/31 channel. Gating was applied to specify the single bacteria and distinguish between active (mCherry positive, diluting FITC) and inactive (mCherry negative, retaining FITC) subpopulations.

Fluorescence microscopy.

Formalin-fixed samples were washed three times with phosphate-buffered saline (PBS) (137 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, 1.8 mM KH₂PO₄, pH = 7) and inspected with fluorescence microscopy, using a Zeiss Axio Observer with ZEN Blue software. The UV exposure times were 1,000 ms for mCherry and 750 ms for FITC, and an HXP 120 Illuminator (Zeiss) was used as the fluorescence UV light source. Images were acquired with an ORCA-flash 4.0 V2 digital complementary metal oxide semiconductor (CMOS) camera (Hamamatsu Photonics), and images were analyzed using the ImageJ plugin MicrobeJ (31). Cell outlines were detected in the phase contrast images and the FITC and mCherry signal intensities of each cell were determined in the corresponding fluorescent images (13).

Two main experiments: anoxic entrapment and the effect of spiking with O₂ or N₂O.

(i) O₂ spiking. Vials with 50 mL Sistrom’s medium without NO₂⁻ or NO₃⁻ were inoculated with ~4 × 10⁹ cells that had been raised under strict aerobic conditions to secure absence of any denitrification enzymes. Prior to inoculation, the vials were either He washed (2 to 400 ppmv residual O₂ in the headspace) or He washed and GOX treated (completely anoxic), and NO₂⁻ was injected after 5 h. The purpose of these treatments was to inspect the effect of exposing cells to sudden anoxia (without NO₂⁻) on their ability to express denitrification enzymes. Some vials were inoculated with FITC-stained cells, allowing the distinction between growing and nongrowing cells. Others were inoculated with equal amounts of unstained cells, to check if FITC staining would influence the respiration kinetics. Spikes of O₂ (~15 μmol = 350 μL vial⁻¹ or ~7.35 μM in liquid) were injected to some vials after 69 h to assess the capacity of cells that were apparently entrapped in anoxia (without mCherry, retained FITC stain). Table 1 summarizes the various treatments.

Flow cytometry was used to discriminate between active and inactive cells based on red and green fluorescence as described above.

(ii) N₂O spiking. Since normal Sistrom’s medium contains traces of NO₃⁻, which was suspected to influence the percent of cells expressing denitrification enzymes in the first experiment, we used Sistrom’s which had been stripped for NO₂⁻ and NO₃⁻ in the second experiment. Vials were either completely anoxic (He washed + GOX treated) or provided with 0.25 vol % O₂ in headspace prior to inoculation with ~3 × 10⁹ cells vial⁻¹ (aerobically raised, FITC stained). NO₂⁻ was injected (100 μmol vial⁻¹) after 14 h when the vials with 0.25 vol % O₂ had become completely anoxic. During the anoxic phase of the incubation, vials were spiked with either ~15 μmol O₂ or ~20 μmol N₂O (250 μL vial⁻¹; ~77 μM in liquid). The amount of O₂ injected during the spiking experiments was chosen based on previous observations of maximal nirS (+ norB) transcription at an O₂ of <15 μM in liquid (14), thus ensuring provision of electron acceptor while avoiding severe suppression of nir and nor expression. The amount of N₂O was selected to ensure adequate provision of electron acceptor, albeit for a shorter time window compared to the oxygen spike. Table 1 summarizes the treatment of the individual vials.

Presence of NirS (mCherry, red fluorescence) and growth (dilution of FITC stain, green fluorescence) in single cells was monitored in selected liquid samples by fluorescence microscopy and image analysis as described above and in Lycus et al. (13).

Data availability.

Raw data from gas measurements, microscopy and flow cytometry can be made available upon request.

Supporting information referred to in the paper thoroughly describes the data and the analyses and can be found at https://data.mendeley.com/datasets/zpwkwwg5xz.

In Movies 1 to 6, red (mCherry-NirS expression) and green (FITC, growth) fluorescence in single cells is captured by flow cytometry (Movies 1 to 4, O₂ spiking experiment) or fluorescence microscopy and image analyses (Movies 5 and 6; N₂O spiking experiment). Movie 1, vial 1.2; Movie 2, vial 1.3; Movie 3, vial 1.8; Movie 4, vial 1.9; Movie 5, vials 2.3 and 2.4; Movie 6, vials 2.7 and 2.11.

For Supplemental Items (SI) 1 to 5, Supplemental Item 1 is a description and qualification of the glucose oxidase-catalase (GOX) approach used for the removal of residual oxygen in experimental vials before inoculation. In Supplemental Item 2, we display the gas and flow cytometry data in the O₂ spiking experiment, including the data shown in Fig. 2 and 3 in the paper. In Supplemental Item 3, we show the gas kinetics and microscopy analyses from the N₂O spiking experiment, which in the paper is summarized in Fig. 4. We also show the lack of positive effect from N₂O addition on Nir expression in an NosZ-deficient mutant. In supplemental item 4, we describe the steps taken to estimate apparent specific growth rates in single vials and cell yield per mol electron to N oxides. In Supplemental Item 5, we describe a simple experiment where nitrite-reducing cultures of Paracoccus denitrificans was spiked with N₂O and O₂, and the subsequent rate of nitrite reduction was assessed.