Denitrification by Bradyrhizobia under Feast and Famine and the Role of the bc1 Complex in Securing Electrons for N₂O Reduction

ABSTRACT

Rhizobia living as microsymbionts inside nodules have stable access to carbon substrates, but also must survive as free-living bacteria in soil where they are starved for carbon and energy most of the time. Many rhizobia can denitrify, thus switch to anaerobic respiration under low O₂ tension using N-oxides as electron acceptors. The cellular machinery regulating this transition is relatively well known from studies under optimal laboratory conditions, while little is known about this regulation in starved organisms. It is, for example, not known if the strong preference for N₂O− over NO₃⁻ reduction in bradyrhizobia is retained under carbon limitation. Here, we show that starved cultures of a Bradyrhizobium strain with respiration rates 1 to 18% of well-fed cultures reduced all available N₂O before touching provided NO₃⁻. These organisms, which carry out complete denitrification, have the periplasmic nitrate reductase NapA but lack the membrane-bound nitrate reductase NarG. Proteomics showed similar levels of NapA and NosZ (N₂O reductase), excluding that the lack of NO₃⁻ reduction was due to low NapA abundance. Instead, this points to a metabolic-level phenomenon where the bc1 complex, which channels electrons to NosZ via cytochromes, is a much stronger competitor for electrons from the quinol pool than the NapC enzyme, which provides electrons to NapA via NapB. The results contrast the general notion that NosZ activity diminishes under carbon limitation and suggest that bradyrhizobia carrying NosZ can act as strong sinks for N₂O under natural conditions, implying that this criterion should be considered in the development of biofertilizers.

IMPORTANCE Legume cropped farmlands account for substantial N₂O emissions globally. Legumes are commonly inoculated with N₂-fixing bacteria, rhizobia, to improve crop yields. Rhizobia belonging to Bradyrhizobium, the microsymbionts of several economically important legumes, are generally capable of denitrification but many lack genes encoding N₂O reductase and will be N₂O sources. Bradyrhizobia with complete denitrification will instead act as sinks since N₂O-reduction efficiently competes for electrons over nitrate reduction in these organisms. This phenomenon has only been demonstrated under optimal conditions and it is not known how carbon substrate limitation, which is the common situation in most soils, affects the denitrification phenotype. Here, we demonstrate that bradyrhizobia retain their strong preference for N₂O under carbon starvation. The findings add basic knowledge about mechanisms controlling denitrification and support the potential for developing novel methods for greenhouse gas mitigation based on legume inoculants with the dual capacity to optimize N₂ fixation and minimize N₂O emission.

INTRODUCTION

Bacteria in most natural and engineered environments are faced with fluctuating availability of nutrients and need to adapt to a “feast and famine” lifestyle. While many soil types are rich in total organic carbon, the concentration of bioavailable carbon substrate is low, particularly in nonrhizosphere soil where lack of substrate is a major factor limiting the growth of heterotrophic bacteria (1). It is likely that bacteria in soil are starved most of the time (2) and only experience infrequent episodes of ample provision of carbon substrate, for example, as exudates from a root or organic material released during decay of dead (micro)organisms. Bacteria have developed several strategies to survive extended periods of starvation, such as the development of high-affinity uptake systems to scavenge alternative carbon sources from the surroundings, as well as changes in cell morphology, condensation of the nucleoid, and decreased protein synthesis to adapt to a low metabolic activity (3). Many bacteria produce carbon-rich storage materials such as poly-3-hydroxyalkanoates (PHA) and glycogen during periods of substrate availability, which can be utilized to sustain a minimum of metabolic activity when deprived of carbon and energy (4, 5).

Denitrification is the reduction of NO₃⁻ to N₂ through anaerobic respiration, where the N-oxides are used as terminal electron acceptors when O₂ becomes scarce. This process can be performed by a diverse range of heterotrophic bacteria, archaea, and fungi, which use various forms of organic compounds as electron donors to obtain energy (6) or, in some cases, H₂ (7). The last step of denitrification is the reduction of N₂O, a strong climate gas, to harmless N₂, catalyzed by the N₂O reductase (Nos) (8, 9). It is found in a diverse range of prokaryotic organisms but has not been reported in eukaryotes. Some denitrifying prokaryotes can perform all steps of denitrification, others only some, and lack of the last step is common due to absence of the nosZ gene coding for NosZ, or lack of other essential gene(s) in the nos operon (10, 11), but the amounts of N₂O released from denitrification in relation to N₂ (the N₂O/N₂ product ratio) is also influenced by transcriptional and posttranscriptional control mechanisms and by various environmental factors (6, 12–19). Denitrification in agricultural soils is a major source of N₂O, accounting for more than 60% of the global anthropogenic emissions (20, 21). A steady increase in atmospheric N₂O has been recorded since the start of industrialization, largely driven by increasing and excessive use of synthetic fertilizers (22, 23) and these emissions are predicted to continue to increase unless novel mitigation options are developed (24, 25).

Although the addition of reactive nitrogen compounds via synthetic fertilizers accounts for the main part of the N₂O emissions from agricultural soil, the N₂O emitted from legume cropped fields is far from negligible. A compilation of data from ca 70 legume cropped fields estimated ca 1.29 kg N₂O−N ha⁻¹ during one growing season, while the corresponding data for N-fertilized crops and pastures showed emissions of 3.22 kg N₂O−N ha⁻¹ (26). Legumes do not have to rely on uptake of reactive N such as NH₄⁺ or NO₃⁻ but can acquire N through their symbiotic relationship with some groups of bacteria, collectively called rhizobia, which elicit the production of root nodules on the plant inside which the N₂ fixation takes place. In this process, the rhizobia reduce N₂ to NH₃, which the plant cells reduce to glutamine and use to produce amino acids and eventually proteins (27). When this N-rich plant material is degraded, organic N is released and mineralized to NH₃/NH₄⁺ which will be oxidized to NO₃⁻ by nitrifying organisms. The O₂ consumption by the nitrifiers, together with the production of NO₃⁻ and the availability of organic compounds from the degraded plants, creates conditions that are conducive to denitrification, likely leading to increased N₂O production.

A novel approach to minimize N₂O emissions from agricultural soil is to enhance the populations of N₂O-reducing bacteria (25). In the case of legume crops, which are often inoculated with rhizobia to optimize the N₂ fixation, there are a few promising studies reporting decreased N₂O emissions from soybean fields inoculated with rhizobia with the dual capability of efficient N₂ fixation and N₂O reduction (28, 29). Consequently, selection of rhizobial strains for development of commercial inoculants should, ideally, take both these aspects into account. One problem is, however, that far from all rhizobia carry the nosZ gene that encodes Nos. There are relatively few surveys of denitrification genes in different groups of rhizobia. Complete denitrification, which includes all four reduction steps of NO₃⁻ to N₂, has so far mainly been reported for the genus Bradyrhizobium, which is the microsymbiont of a range of economically important legume crops such as soybean, cow pea, and peanut (30, 31). A full set of denitrification reductases in bradyrhizobia include, with few exceptions, the periplasmic NO₃⁻ reductase NapA; the Cu-containing NO₂⁻ reductase NirK, the bc-type NO reductase cNor, and a NosZ belonging to clade I (19, 32). We recently investigated the denitrification capacity of bradyrhizobia from two strain collections, one obtained from nodules of legume trees and herbs growing in Ethiopia, and another mainly consisting of strains isolated from nodules of peanut growing in China (18, 19). In these collections, 50 and 37% of the isolates, respectively, were complete denitrifiers, while the others generally lacked NosZ and thus were potential N₂O sources. Common to all strains with complete denitrification was a strong preference for N₂O− over NO₃⁻ reduction. Transcription analysis and proteomics showed comparable expression levels of NapA and NosZ, suggesting a control mechanism at the metabolic level based on competition for electrons between the electron pathways to NosZ and NapC, in which the pathway to NosZ apparently wins. This supports the hypothesis proposed by Mania et al. (18): electrons to both pathways are delivered from the tricarboxylic acid (TCA) cycle via NADH dehydrogenase to the quinone/quinol pool and channeled either to NapC, which delivers electrons to NapA (via the enzyme NapB), or to the bc1 complex. Cytochromes will receive electrons from the bc1 complex and deliver them to NosZ, and also to nitrite reductase (NirK or NirS depending on the organism) and nitric oxide reductase Nor (cNor or qNor). The competition for electrons between the pathways to NapA and NosZ most likely takes place at the first branching point where the membrane-bound bc1 complex competes very efficiently with the membrane-bound NapC for the electrons.

The results presented by Mania et al. and Gao et al. (18, 19) were based on experiments with organisms provided with ample amounts of C substrate (electron donor), which is likely to reflect the situation in legume nodules where the microsymbiont receives C from the plant. It can, however, be expected that rhizobia, which may survive for many years in soil (33), spend a large part of their life cycle as free-living organisms in soil where they will experience lack of available C substrate most of the time (2). Rhizobial inoculants that carry NosZ are thus potentially important sinks for N₂O produced both by themselves and by other soil microbes. It is, however, not known if the competition for electrons favoring N₂O reduction in well-fed cultures is retained during substrate limitation (“starvation”). Here, we exposed cultures of Bradyrhizobium strain HAMBI 2125, also studied by Gao et al. (19), to shorter and extended periods of starvation and analyzed the denitrification kinetics, including the electron flow rates to the individual denitrification reductases. We also quantified the cellular abundancies of Nap, Nir, and Nos. The results have practical implications, supporting that these organisms can act as sinks for N₂O under natural conditions. Moreover, the results are ecologically interesting since they show that cultures exposed to extended starvation divided into two opposite denitrification phenotypes, one with very slow metabolism, the other with retained metabolism, possibly reflecting a strategy to increase the chances for survival during periods of starvation.

RESULTS

Denitrification kinetics in cultures prepared following bioassay 1.

Two bioassays were developed for starvation experiments, bioassays 1 and 2 (Fig. 1A and B). In a first experiment, following bioassay 1, the denitrification gas kinetics, concentrations of NO₃⁻, NO₂⁻, and electron flow rates to the different reductases were compared for well-fed versus starved cultures (Fig. 2A to D). In this bioassay the cultures were allowed to synthesize the denitrification reductases in the presence of ample amounts of substrate. Cultures were raised under fully oxic conditions in yeast mannitol broth (YMB) medium. When the optical density at 600 nm (OD₆₀₀) reached 0.1, the headspace was replaced with He. Then, 1% O₂ was injected, corresponding to an initial concentration of 10 μM in the liquid, to allow transition to anaerobic respiration in response to a gradual depletion of O₂. After centrifugation/washing, pellets were pooled and inoculated into flasks containing YMB (9.9 × 10⁸ cells flask⁻¹) or buffer (5.0 × 10⁹ cells flask⁻¹), supplemented with 1 mM KNO₃ and 0.25 mM KNO₂, and with He plus 1 mL N₂O (around 80 μmol N) in headspace. The initial O₂ concentration in these flasks was 0.5 μM for cultures with YMB and 0.2 μM in cultures with buffer.

FIG 1.

Bioassays for assessing the effect of starvation on electron flow to denitrification enzymes. (A) Bioassay 1: starvation of cells with a previously expressed denitrification proteome. Cells were raised from stock cultures in fully oxic flasks containing YMB medium. When OD₆₀₀ reached ~0.1, the flasks were made anoxic by He washing, then supplemented with 1% O₂ in the headspace and 1 mM NO₃^– in the liquid. The cultures were allowed to deplete the O₂ and to initiate denitrification when growing in YMB medium, after which they were centrifuged (10,000 × g at 4°C for 10 min) and washed twice in sterile ddH₂O. The pellets from triplicate flasks were pooled and used to inoculate flasks containing either C-free buffer or YMB medium (well-fed control), in both cases supplemented with 1 mM KNO₃ and 0.25 or 0.5 mM KNO₂, and with He and 1 mL N₂O (around 80 μmol N flask⁻¹) in headspace. The starving cells, incubated in buffer, had a low respiratory electron flow rate (mol electrons cell⁻¹ h⁻¹), initially being 10 to 18% that of the well-fed cultures, and then decreasing to reach around 4% after 20 h. Results from experiments using bioassay 1 are presented in Fig. 2, 3, and 4. (B) Bioassay 2: denitrification induced during starvation. Cells were raised from stock cultures in fully oxic flasks containing YMB medium. When OD₆₀₀ reached ~0.1, the cultures were centrifuged (10,000 × g at 4°C for 10 min) and washed twice in sterile ddH₂O. The pellets from triplicate flasks were pooled, then evenly divided and used to inoculate fully oxic flasks containing C-free buffer. These cultures were incubated for 20 h, then centrifuged, after which the pellets were pooled and divided evenly before being inoculated into flasks containing C-free buffer provided with 1 mM KNO₃, and with He and 1 mL N₂O (around 80 μmol N flask⁻¹) in headspace. The respiration rate (mol electrons cell⁻¹ h⁻¹) of the starving cultures was 1 to 4% compared to that of well-fed cultures. All cultures in bioassays 1 and 2 were incubated at 28°C, and with vigorous stirring (650 rpm). Results from experiments using bioassay 2 are presented in Fig. 5 and Fig. S2 in the supplemental material.

FIG 2.

Denitrification kinetics as affected by starvation in cultures with a complete denitrification proteome (bioassay 1). Cells were allowed to develop a full denitrification proteome under well-fed conditions and were then washed twice in buffer prior to inoculation to flasks with YMB (well-fed control) or buffer (starved), supplemented with 1 mM KNO₃ and 0.25 mM KNO₂, and with He plus 1 mL N₂O (around 80 μmol N) in headspace. A larger inoculum was given to the flasks with buffer (9.9 × 10⁸ cells flask⁻¹) than to the flasks with YMB (5.0 × 10⁹ cells flask⁻¹), to secure measurable activity in the starved cells and adequate time resolution of the denitrification kinetics in the well-fed cells. The denitrification kinetics in well-fed (A) and starved (B) cultures are shown. The flasks were practically anoxic from the start with an initial O₂ concentration in the liquid of <0.52 μM which decreased to approximately 0 μM (insets in A and B). (C and D) Cell-specific electron flow rate. V_{e total den} designates the total electron flow to the denitrification reductases and V_{e total incl O2} the total electron flow, including that to denitrification and to O₂. The electron flow rate to the individual reductases is designated V_eNap, V_eNir, V_eNor, and V_eNos. The rates of NO₃^– reduction (V_eNap), which showed that no NO₃^– reduction took place in the starved cultures throughout the incubation period, were calculated by N-mass balance for each time increment, as done previously (19, 35). V_eNir and V_eNor were practically identical and cannot be distinguished from one another in the figure. Inserted panels show V_eNap throughout, including negative values which are due to slight errors in determination of N₂ and N₂O (V_eNap was calculated by N-mass balance). Bars in all graphs show standard deviation (n = 3).

The O₂ was depleted within the first 5 h in both treatments (insets in Fig. 2A and B). The provided NO₂⁻ and N₂O were reduced simultaneously from the beginning of the incubation in both treatments. The NO₃⁻ was left untouched in the well-fed cultures until the exogenous N₂O was reduced (Fig. 2A), also seen from lack of electron flow to Nap except for a small peak in electron flow early in the anoxic incubation (Fig. 2C). No NO₃⁻ reduction took place in the starved cultures throughout the incubation period, as seen from the lack of electron flow to Nap (Fig. 2D). The negative V_eNap estimated for the starved cultures (inset in Fig. 2D) and the initial phase of the well-fed cultures are probably due to minor errors in calibration of N gas measurements as well as in parameters used to calculate rates of N transformation (sampling loss and N₂ leakage) (see Molstad et al. [34]), which amounts to substantial errors in the estimates of NO₃⁻ reduction rates because they are based on N-mass balance (explained in detail by Lim et al. [35]). This implies a relatively high detection limit for NO₃⁻ reduction, and the negative values cannot be taken as evidence for the complete absence of any electron flow to NO₃⁻. However, the measured NO concentrations lend some support to the claim that V_eNap was ~0 until depletion of the externally provided N₂O. In the well-fed culture, the concentration of NO declined to zero in response to depletion of NO₂⁻, and increased soon after, as V_eNap increased (Fig. 2A and C). Likewise, the NO concentration declined to very low values in response to depletion of NO₂⁻ in the starved culture (Fig. 2B).

Fig. 2C and D show the calculated electron flow rates per cell to the individual reductases, and their sum, illustrating their competition for electrons. This shows that the well-fed cells (Fig. 2C) sustained a nearly constant total electron flow rate around 9 fmol e⁻ cell⁻¹ h⁻¹ throughout, but allocated to different reductases depending on the availability of electron acceptors: as long as NO₂⁻ and N₂O were both present, Nos captured around 50% of the electrons (V_eNos ~ V_eNir + V_eNor), increasing to 100% when NO₂⁻ was depleted after 6 to 7 h, while the electron flow rate to Nap remained insignificant until the externally provided N₂O became depleted. The total electron flow rate per cell of the starving cultures was an order of magnitude lower than that of well-fed cultures cells (Fig. 2D) and declined gradually throughout the incubation. In the starving cultures ≫50% of the electrons were captured by Nos during the first 10 h (V_eNos ≫ V_eNir + V_eNor), but this dropped to ~50% towards the end of the incubation.

A second experiment was set up according to bioassay 1 to determine if the individual denitrification reductases were functional during starvation if the appropriate N-oxide was present. The cultures were incubated anoxically ([O₂] <0.21 ± 0.04 μM) in C-free buffer, like in the first experiment, with the difference that only one of the N-oxides (NO₃⁻, NO₂⁻, or N₂O) was provided as the initial electron acceptor (Fig. 3A to C). This experiment established that starved cultures readily reduced NO₃⁻ from the start when no other N-oxide was provided (Fig. 3A). It also demonstrated that the cell specific electron flow to N-oxides was practically unaffected by the type of electron acceptor provided, except for the slightly lower rates initially for flasks with NO₂⁻. For all treatments, the cell-specific electron flow rate decreased gradually throughout, and the levels are very similar to those observed in the first experiment (Fig. 3D).

FIG 3.

Bioassay 1 with single nitrogen oxides. Experimental conditions are as in Fig. 2, but the starving cells (in buffer) were provided with either NO₃^–, NO₂^–, or N₂O in individual flasks (n = 3 for each treatment). (A to C) Gas kinetics in flasks provided with 1 mM NO₃^– (A), 0.5 mM NO₂^– (B), or 70 μmol N₂O-N added to the headspace (C). (D) Cell-specific electron flow rate (V_e) measured in buffer supplemented with NO₃^–, NO₂^–, or N₂O; the O₂ concentration (left y axis); and V_e as percentage of the rates in well-fed cultures (right y axis). The number of cells inoculated into the incubation flasks at the start (0 h of incubation in anoxic buffer) was 3.6 × 10⁹ for all treatments. Bars show standard deviation (n = 3).

Cell size and PHA accumulation of starved versus well-fed cultures.

The morphologies of single bacterial cells from different treatments (starved/well-fed and anoxic/oxic) were analyzed by phase-contrast microscopy (see Fig. S1A in the supplemental material). By quantifying cell dimensions, we found marginal but statistically significant effects (Mann-Whitney test, P < 0.01) of starvation; while there was no significant difference in cell area of individual cells (Fig. S1B), starved cells were on average ~12% longer and ~5% thinner than well-fed cells (Fig. S1B). By further calculating cell volumes, a slight reduction in average cell volume was observed upon starvation of aerobic cells (P = 0.057), but not for cells grown under anoxic conditions (Fig. S1C). Furthermore, a qualitative analysis of the presence of PHA granules was done by staining the cells with Nile Red, a lipophilic dye with high affinity for PHA (36). Large foci corresponding to PHA granules were seen under all conditions. It should be noted that, as a lipophilic dye, Nile Red will also bind nonspecifically to lipids and membrane, and based on the imaging performed here, it could not be concluded whether there were any significant differences in PHA content between the conditions.

Denitrification kinetics after providing starving cultures with an artificial electron donor.

A third incubation experiment was performed based on Bioassay 1, but TMPD (N, N, N′, N′-tetramethyl-p-phenylenediamine) and ascorbate were added to starved cultures (100 μM and 10 mM, respectively, in the culture medium) to provide cytochrome c with an excess of electrons (37, 38). Since cultures treated according to bioassay 1 were able to provide electrons for denitrification (10 to 18% of the electron flow in well-fed cells during the first 5 h, then decreasing to ca 4%; Fig. 2D), we expected that the TMPD treatment would reduce or eliminate the oxidation of quinol by the bc1 complex and that this would allow electrons to flow to Nap via NapC, resulting in NO₃⁻ reduction. We also expected that loading cytochrome c with electrons would relieve or weaken the competition for electrons between Nir, Nor, and NosZ. The results (Fig. 4) lend little support to the former since the electron flow to Nap remained insignificant until N₂O had been depleted. But the results confirm an effect of TMPD on the competition between Nos and Nir: V_eNir, V_eNor, and V_eNos were similar and high (4 to 5 fmol e⁻ cell⁻¹ h⁻¹) until NO₂⁻ was depleted. Thus, Nir and Nos competed equally well when the cytochrome c pool was fully reduced by TMPD. After depletion of NO₂⁻, V_eNos leaped to its maximum level, 13 to 14 fmol N cell⁻¹ h⁻¹ and kept this rate until all the exogenous N₂O was depleted.

FIG 4.

Denitrification in starved cells after addition of TMPD as an external electron donor. Preparation of the cultures followed bioassay 1 (Fig. 1A), except that 100 μM TMPD and 10 mM ascorbate were injected into the flasks with starving cells 15 min before the first gas sampling. Each flask was inoculated with 5.28 E + 08 cells (n = 4 replicate flasks). The flasks were initially provided with 1 mM initial NO₃^–, 0.25 mM NO₂^–, and 1 mL N₂O (around 80 μmol N₂O–N). The initial O₂ concentration was 0.35 μM and decreased to approximately 0 within 5 h (inset in A). Bars show standard deviation (n = 4). (A) Kinetics of NO₂^–, NO, N₂O, and N₂ (O₂ in inserted panel). (B) Calculated cell-specific electron flow rate (V_e, fmol e^- cell⁻¹ h⁻¹) to each of the reductases Nap, Nir, Nor, and NosZ, and the total electron flow rate (V_etotal). The electron flow to Nir was practically identical to the electron flow to Nor (miniscule amounts of NO), and the two are shown as a single graph (V_eNir/Nor). Inset plots show V_eNap throughout, including negative values which are due to slight errors in determination of N₂ and N₂O (V_eNap was calculated by N-mass balance).

Denitrification kinetics and reductase abundancies of cultures exposed to extended starvation following bioassay 2.

While bioassay 1 successfully induced starvation in terms of a downshift in respiration, the rates did not reach a stable level during the assay but declined gradually throughout, apparently approaching more stable low levels after 20 h. On this background, we introduced a more severe starvation assay (bioassay 2; Fig. 1B) by including a 20-h aerobic incubation in buffer prior to the starvation-denitrification assay to reach a lower and more stable rate of respiration than in the first experiment. In addition, bioassay 2 was designed to force the cells to synthesize the denitrification proteome while starving. In several preliminary experiments with bioassay 2 (results not shown), we found a conspicuous variability between flasks regarding the cell-specific respiration rate where one or two out of three replicate flasks had four to six times higher cell respiration rates than the other(s). At first, we suspected that it could be due to impurities of flasks or magnets, but meticulous acid washing failed to remove the stochastic variation. Being convinced that the stochasticity reflects a real regulatory switch of the cultures, we performed a final experiment in which 15 replicate flasks were monitored for denitrification kinetics (Fig. 5).

FIG 5.

Bioassay 2: stochasticity of starvation response, response to input of organic C, and quantification of denitrification enzymes. Altogether, 15 flasks were prepared following bioassay 2 (see Fig. 1B). All flasks were anoxic (<0.5 μM at the start of the incubation) and contained 50 mL buffer supplemented with 1 mM KNO₃ and with He plus 1 mL N₂O in the headspace. The NO₂^– concentrations, measured during the first 5 h after incubation in the buffer, were approximately 0.59 ± 0.20 μM (not shown). The cultures separated into two distinct phenotypes: 9 flasks had slow respiration (A) and 6 flasks had fast respiration (B). The electron flow to the individual reductases (fmol e⁻ cell⁻¹ h⁻¹) for the flasks with slow and fast respiration are shown in C and D, respectively. Cultures with slow respiration rate had a total electron flow rate of maximum 0.27 e⁻ cell⁻¹ h⁻¹ (C) and cultures with fast respiration rate had a total electron flow rate of 1.0 to 1.8 fmol e⁻ cell⁻¹ h⁻¹ (D). The inset plot in C shows the electron flow to the individual denitrification reductases after the carbon addition. The cultures (entire flasks, 50 mL) were sampled for proteomics analyses at different time points (E and F). At time points throughout, marked by dashed vertical lines in A and B, three flasks were harvested for proteomics analysis (including 0 h). The six flasks with fast respiration were harvested for proteomics analysis at 5 and 20 h of incubation in these anoxic buffers (triplicates at each sampling point). Of the nine flasks with slow respiration, three were harvested at 10 and three at 20 h. The remaining three slow respiration flasks were supplemented with YMB at 24.8 h, at a concentration which made the buffer a half-strength YMB medium, containing 5 g mannitol l⁻¹ and 0.25 yeast extract l⁻¹, and then harvested for proteomics at 27 h. The gas measurements and electron flows are averages from the three flasks (n = 3) for each phenotype, that were left untouched until the end of the incubation when they were sampled for proteomics. Proteomics analyses were done for triplicate flasks (n = 3) at each sampling point. Standard deviations are indicated as bars in all graphs (in several cases not visible due to low variation).

The cultures received initially 1 mM NO₃⁻ in the buffer (but no NO₂⁻), and He plus 1 mL N₂O (80 μmol N) in headspace. The O₂ concentration at the time of inoculation was <0.5 μM in the liquid. The flasks separated into two distinct denitrification phenotypes (Fig. 5A and B). Nine flasks showed “low” cell-specific respiration rates (total electron flow maximum 0.27 fmol e⁻ cell⁻¹ h⁻¹), corresponding to approximately 2.7% compared to well-fed cultures (Fig. 5A), while the other six showed a “fast” respiration rate (total electron flow 1.0 to 1.8 fmol e⁻ cell⁻¹ h⁻¹). Both phenotypes reduced N₂O from the beginning of the incubation, showing a strong preference for N₂O over NO₃⁻. In the flasks with fast respiration, the cells started to reduce NO₃⁻ in response to depletion of the externally provided N₂O. In the flasks with slow respiration, N₂O was not depleted within the time frame of the experiment, and NO₃⁻ reduction remained negligible.

To investigate if cell lysis, and thus release of available C, could explain why some cultures showed the “fast” growth phenotype, we compared the OD₆₀₀ and the number of viable cells in “slow” versus “fast” cultures. The samples were taken after incubation for 3.1 h in anoxic buffer, when the two phenotypes were clearly distinct (Fig. S2A) but when some possible growth of cells in the “fast” cultures would not yet hide if lysis had occurred. The OD₆₀₀ spanned from 0.12 to 0.13 with no statistical difference (P > 0.3) between cultures with fast and slow respiration. Similarly, no difference (P = 0.4) was found for the viable counts which showed 1.21 E + 10 to 1.33 E + 10 CFU flask⁻¹ for the “fast” cultures and 1.23 E + 10 to 1.34 E + 10 CFU flask⁻¹ for the “slow” cultures (Fig. S2B).

We also tested the metabolic integrity of the cells in the flasks with slow respiration by injecting C substrates (YMB) to the flasks after 24.8 h, which proved their capacity to quickly regain activity approaching that of well-fed cells (Fig. 5A). A close inspection of the cell-specific electron flow rate showed an immediate increase in V_eNos. To investigate if the observed divergencies in phenotypes were due to differences in denitrification reductase abundancies, we quantified the relative abundances of Nap, NirK, and NosZ in samples taken at different time points throughout the incubations (Fig. 5E and F), together with the corresponding denitrification kinetics (Fig. 5A and B) and cell-specific electron flow to reductases (Fig. 5C and D). The membrane-bound NO reductase (cNor) could not be extracted quantitatively (the results showed 1,000 times lower abundancies than for the other denitrification reductases) and is therefore not shown. The inoculum had been cultured aerobically for 3 to 4 generations, never permitting the OD₆₀₀ to exceed 0.1, to ensure that any denitrification enzymes would be diluted to extinction by aerobic growth, assuming that the transcription of all genes is effectively repressed by oxygen. This strategy was apparently successful since Nap and Nir were undetectable at the start of the anoxic incubation. NosZ, on the other hand, was detected also in the aerobic cultures, suggesting that the nosZ gene is constitutively transcribed at low levels in these organisms.

After transferring the cells to anoxic buffer, the abundance of all three reductases increased both in cultures with slow and fast respiration. Cultures with slow respiration rate synthesized less denitrification reductases than those with fast respiration rate during the first 20 h. The relative abundances of the different reductases also differed between the two phenotypes. In cultures with slow respiration, NosZ was significantly more abundant than Nap and Nir (P < 0.01), which were comparable (Fig. 5E). Cultures with fast respiration instead contained higher abundancies of NosZ and Nir compared to Nap at 5 h. After this, the abundance of Nir increased more than that of the others, and at 20 h, the label-free quantification (LFQ) of Nir was 0.20 ± 0.07, while the abundancies of NosZ and Nap were approximately half of that (Fig. 5F). After addition of carbon substrate to the cultures with slow respiration, a rapid synthesis of all three reductases took place. This synthesis was most prominent for Nir which, after a couple of hours, had increased 3-fold, reaching a relative abundance that was twice as high as Nap and NosZ (Fig. 5E), resembling the abundance profile of the cultures with fast respiration (Fig. 5F).

DISCUSSION

Detailed eco-physiological studies during the past decades have revealed several aspects of how the denitrification process is regulated in different organisms (6, 13, 39). Most of this knowledge is, however, based on laboratory studies where cultures have been grown under optimal conditions, while less attention has been paid to understanding how denitrification, and particularly the release of N₂O, is affected if cells are starved, which is the normal state of cells in most natural environments. We focused on how starvation, i.e., lack of carbon substrate, affects denitrification and N₂O release. The general notion has been that N₂O reductase is less competitive for electrons than the other denitrification reductases, leading to emissions of N₂O when the availability of electron donors is low. This is largely based on a single study of Alcaligenes faecalis (40) and has been supported by some studies of complex communities but contested by others (41–43). A. faecalis can perform partial denitrification reducing NO₂⁻ to N₂ using NirK, cNor, and NosZ clade I, but this organism lacks dissimilatory NO₃⁻ reductases (Nar or Nap). The study by Schalk-Otto (40), performed in continuous cultures, showed increased N₂O release under low substrate concentrations, and it was suggested that N₂O reductase did not compete successfully with the other reductases for electrons from cytochrome c, possibly due to lower affinity for the electron donor. This conclusion needs, however, further verification by more detailed studies of the mechanism involved. Moreover, such studies need to be extended to other groups of denitrifying microorganisms and should include organisms carrying a complete denitrification pathway (thus with Nar and/or Nap). Research over the past decades has revealed diverse denitrification phenotypes among even closely related bacteria, with implications for their accumulation of denitrification intermediate products (15, 44).

The denitrification phenotype described for a range of taxonomically diverse Bradyrhizobium strains with a complete denitrification pathway is characterized by a strong preference for N₂O over NO₃⁻ when grown in full-strength YMB medium under denitrifying conditions (18, 19). The present study provides compelling evidence that this preference is retained when the organisms are starved for carbon and energy, with practically no electron flow to Nap when N₂O was present (Fig. 2A to D). It seems unlikely that this would be due to too low levels of Nap, both because the cultures showed well-functioning Nap activity if NO₃⁻ was provided as the only initial electron acceptor (Fig. 3A) and because the cultures were able to produce comparable amounts of Nap, Nir, and Nos even when grown under severe starvation conditions using bioassay 2 (proteomics results; Fig. 5E and F). Therefore, the lack of NO₃⁻ reduction during the period when there was N₂O in the system could not be explained by a lack of Nap molecules. Instead, the results suggest the same metabolic-level phenomenon as found for well-fed cultures in this study (Fig. 2A) and earlier (18, 19), where NosZ outcompetes Nap for electrons, leaving Nap virtually without electrons so long as exogenous N₂O is available. In theory, the observed lack of NO₃⁻ reduction in the presence of N₂O could be due to a direct inhibition of NapA by N₂O, but this is refuted by the results of Mania et al. (18), who observed high NapA activity in cells exposed to 10 vol% N₂O, when NosZ was inhibited by acetylene (18). These results are analyzed and discussed in more detail Fig. S3.

The attempt to tweak the electron flow toward Nap by the addition of TMPD and ascorbate did not result in measurable NO₃⁻ reduction in the starved cultures (Fig. 4A). A recent study by Mania et al. (18) demonstrated that well-fed Bradyrhizobium cells did reduce some NO₃⁻ in the presence of N₂O, if provided with TMPD and ascorbate. This suggested that a cytochrome c pool that was strongly reduced by TMPD allowed Nap to receive a minimum of the electrons from quinol, delivered from the TCA cycle. The specificity of the electron delivery from TMPD to cytochrome c cannot be taken for granted, however, and the result by Mania et al. (18) could instead reflect a minimum of electron flow from TMPD to quinone or to NapC, directly or indirectly. In the present experiment with starved cells, TMPD plus ascorbate failed to induce measurable electron flow to Nap in the presence of N₂O (Fig. 4B). The contrasting results for well-fed versus starved cells probably reflects the marginal electron flow from the TCA cycle to the quinone/quinol pool in the starved cells. A separate experiment supported this, showing that when carbon substrate (YMB) was added to starved cultures, this provided enough electrons to support some Nap activity, although NosZ activity still dominated (Fig. S4).

At the same time, the result refutes the concerns regarding unspecific electron delivery of electrons from TMPD to quinone. Functional Nap was apparently present, and NO₃⁻ reduction started when N₂O was almost depleted, with V_eNap being 2 fmol e⁻ cell⁻¹ h⁻¹, which was twice as high as V_eNir/Nor (Fig. 4B). The latter is as expected, since the reduction of 1 mol of NO₃⁻ to NO₂⁻ requires 2 mol electrons, while reduction of NO₂⁻ and NO requires 1 mol electrons. Furthermore, the electron flow to NosZ was the same as to Nir and Nor, suggesting that Nir and NosZ competed equally well for electrons from cytochrome c. The maximum total electron flow of 13 to 14 fmol e⁻ cell⁻¹ h⁻¹ when the cytochrome c pool was saturated with electrons from TMPD is likely to be close to the maximum capacity of the denitrification pathway. This electron flow rate is higher than the total electron flow rate in the well-fed cells, which was 8 to 10 fmol e⁻ cell⁻¹ h⁻¹ (Fig. 2C). The factor limiting V_eNos to 13 to 14 fmol cell⁻¹ h⁻¹ is plausibly the rate of electron delivery from cytochrome c and/or k_cat for Nos. Of note, Nir and NosZ received equal shares of the electron flow when delivered by TMPD via cytochrome c (Fig. 4B), while NosZ clearly outcompeted Nir under normal respiration, when electrons were delivered via quinol (Fig. 2C and D). This contrast indicates that the competitive edge of NosZ versus Nir under normal respiration is due to an alternative electron flow to NosZ via NosR, as suggested previously (18).

In another experiment, we added YMB medium to “slow respiration” phenotypes of the cultures that had been exposed to extended starvation (Fig. 5A, C, and E). These cultures had some NosZ activity (V_eNos 0.1 to 0.3 fmol e⁻ cell⁻¹ h⁻¹), while Nap and Nir activities could not be detected. Addition of the electron donor (YMB) led to an immediate upshoot in the activities of all reductases. This could not have been the case if the reductases were not present already, which was proven by the proteomics results. Taken together, the results support that the absolute preference for N₂O over NO₃⁻ was due to competition for electrons, also under severe starvation conditions.

It is well known that bacteria have developed a range of physiological responses to tackle starvation. Some survive by forming spores, but most bacteria survive by strongly reduced metabolic rates and minimizing the synthesis of some proteins while upregulating others, such as genes for high-affinity transporters, essential repair mechanisms, and alternative energy sources (3, 45, 46). Some such changes have been observed in rhizobia belonging to Rhizobium leguminosarum, which stayed viable for long periods (55 days) of C starvation (47). Changes in cell size are common during long-term starvation, sometimes leading to the formation of small or even ultramicrocells, which may have increased tolerance to antibiotics and other stresses (2, 46). Another way for many bacteria, including rhizobia, to survive is to use carbon stored as polyhydroxyalkanoates (PHA) or glycogen, formed during periods of ample nutrient abundance (48). In the present study, the metabolic activity, measured as anaerobic respiration rate, decreased to between 1 and 18% that of well-fed cultures, depending on which starvation bioassay was used. We did not, however, detect any obvious decrease in cell size when comparing cells starved for 24 h, although cell morphologies were slightly altered (Fig. S1A and B). PHA was observed in the starved cells as well as well-fed ones, as shown with Nile Red staining (Fig. S1A) and no obvious reduction in PHA could be observed during starvation from our assays. Since this was a comparably short period of starvation, it is conceivable that the cells would make use of the stored carbon if the starvation was prolonged. It could be speculated that one reason for not using the PHA reserves (and glycogen) immediately, is that these are saved to be used during bacteroid formation (48). However, to clarify these issues, more detailed studies are needed.

A striking separation into two phenotypes during starvation, as shown in Fig. 5, was observed in repeated experiments (bioassay 2), each time with about two thirds of the flasks showing a “fast” and the others “slow” respiration: 1.0 to 1.8 fmol e⁻ cell⁻¹ h⁻¹ versus 0.1 to 0.3 fmol e⁻ cell⁻¹ h⁻¹, respectively. Phenotypic heterogeneity has been observed in single-strain cultures of various bacteria when exposed to carbon substrate deficiency and may or may not be due to mutations during starvation for 1 week or more (3, 49). Mutations in the entire population in several replicate flasks are unlikely and cannot, however, have caused the rapid diversification into slow or fast respiration in the present study. It may instead reflect a stochastic phenomenon, or that the culture contained different subpopulations. We speculated that the “fast respiration” phenotype may be due to a fraction of the cells dying, allowing the other cells to survive on nutrients released from lysed cells, as seen for other bacteria (3). However, this would require lysis of a substantial fraction of the cells, which is refuted by the observation that the flasks with slow and fast phenotypes had practically identical numbers of viable cells (Fig. S2). Thus, further studies are needed to understand this phenomenon of different respiration rates.

Although the starvation bioassays developed for this study cannot be regarded as a close mimicking of the conditions in natural environments, it is conceivable from the experiments that these organisms are potentially strong sinks for N₂O when living in soil under fluctuating availability of carbon substrate. Bioassay 1 is probably closer to a real-world situation than bioassay 2, since it is likely that denitrifying bacteria experience regular fluctuations in oxygen and thus are not devoid of denitrification reductases if they enter starvation. On the other hand, bioassay 2, where the cells had to produce the denitrification proteome in the absence of external electron donors (C substrate), showed that, even under these conditions, N₂O reduction strongly dominated over NO₃⁻ reduction.

MATERIALS AND METHODS

Bacterial strain and culture preparations.

Bradyrhizobium strain HAMBI 2125, originally isolated from nodules of Arachis hypogaea growing in Sichuan, China (50), was used in all experiments. This strain, which is closely related to Bradyrhizobium ottawaense, contains the genes needed for complete denitrification (19). A culture was raised from one single colony after streaking on agar plates. After checking the purity by sequencing the 16S rRNA gene (19), portions were preserved in 15% glycerol at –80°C. Cultures for all the experiments were started from the –80°C stocks and raised under fully oxic conditions in 120 mL serum flasks containing 50 mL yeast mannitol broth (YMB): 10 g L⁻¹ d-Mannitol, 0.5 g L⁻¹ K₂HPO₄, 0.2 g L⁻¹ MgSO₄·7H₂O, 0.1 g L⁻¹ NaCl and 0.5 g L⁻¹ yeast extract (51). YMB medium was used in all incubations of well-fed cultures. A buffer consisting of YMB without mannitol and yeast extract was used for all incubations of starved cultures. All incubations were done at 28°C. Medical flasks (120 mL) were used in all experiments. A magnet in each flask secured vigorous stirring (600 to 700 rpm) to avoid cell aggregation and to optimize the gas exchange between liquid and gas phases (18). To prevent that the cells experienced anoxia during this oxic incubation, and thus to avoid de novo synthesis of denitrification reductases, portions were regularly transferred to new flasks containing fresh medium, so that the OD₆₀₀ was never allowed to exceed 0.1 (19). These aerobically grown cultures were used as inoculants in the starvation bioassays described below.

Flasks for denitrification experiments were prepared as described by Mania et al. (18). Briefly, flasks (120 mL) containing 50 mL buffer (or medium) were capped with sterilized, gas-tight butyl rubber septa (Matrix AS, Norway). The air was removed by applying vacuum repeatedly (6 × 360 s) and He was then filled for 30 s, after which the overpressure was released. The flasks were left for 2 days to equilibrate the gases between the headspace and liquid (52). Before starting the experiment, 0.7 mL or 1 mL O₂ (equal to 1 or 1.5 vol %) and 1 mL N₂O (70 to ~80 μmol N flask⁻¹) was injected into the headspace and sterile filtered solutions of KNO₃ (and sometimes KNO₂) were added to the liquid reaching initial, desired concentrations (0.25 to 1 mM).

Starvation bioassays.

Starvation bioassays were established, which followed one of the procedures described in Fig. 1. In bioassay 1 “mild starvation,” cultures were allowed to make the transition to denitrification in full-strength YMB before being exposed to starvation. Oxically grown, well-fed cultures were incubated for 48 h, after which the headspace was replaced with He and 1% O₂, and 1 mM KNO₃ was added to the medium. When O₂ was depleted, the cultures were centrifuged (10,000 × g at 4°C for 10 min) and washed twice in sterile ddH₂O. The pellets (triplicates) were pooled to reduce bias in the form of variations due to the centrifugation/washing. Each pellet was divided into three and used to inoculate flasks containing 50 mL C-free buffer supplemented with 1 mM KNO₃ and 0.25 or 0.5 mM KNO₂. These flasks had been made anoxic (O₂ <0.5 μM in the liquid) and contained He and/or 1 mL N₂O (around 80 μmol N flask⁻¹) in the headspace. When incubated in the buffer, the respiration rate of the cultures was 10 to 18% that of well-fed cultures during the first 15 h, then it decreased to about 4%.

In Bioassay 2 (extended starvation), denitrification was instead induced after having exposed the cells to starvation for 20 h. The cultures were raised in fully oxic flasks containing YMB medium. When OD₆₀₀ reached ~0.1, the cultures were centrifuged (10,000 × g at 4°C for 10 min) and washed twice in sterile ddH₂O. The pellets (triplicates) were pooled, after which they were evenly divided and used to inoculate fully oxic flask containing C-free starvation buffer. These cultures were incubated for 20 h, then centrifuged, after which the pellets were pooled and divided evenly before being inoculated into flasks containing C-free starvation buffer with 1 mM KNO₃, and with He and 1 mL N₂O (around 80 μmol N flask⁻¹) in headspace. These flasks had been made anoxic (O₂ <0.5 μM in the liquid). The respiration rate of cultures exposed to extended starvation was ca 1 to 4% compared to that of well-fed cultures. Results from the different starvation bioassays were compared to those from well-fed cultures. To avoid biases, the treatments, which were in all cases set up at least in triplicates (n ≥ 3), were the same regarding centrifugations and washings until the last step, when pooled cells were inoculated to anoxic flasks containing either YMB or buffer. The carbon source in the YMB medium comprised >200 times surplus of electron donor compared to electron acceptors throughout all incubations, thus ensuring that the electron donor was not depleted. All cultures were incubated at 28°C and with vigorous stirring (600 to 700 rpm).

Addition of YMB medium or TMPD as electron donors to starved cultures.

Experiments were performed to investigate how starved cultures of Bradyrhizobium strain HAMBI 2125 responded to the addition of an electron donor, either provided as an artificial electron donor (Fig. 4) or as YMB medium (2 mL mannitol/yeast solution providing them with a substrate concentration corresponding to half-strength YMB, thus 5 g mannitol and 0.25 g yeast L⁻¹) (Fig. 5). As an artificial electron donor, we used sodium TMPD (N, N, N´, N´-tetramethyl-p-phenylenediamine) which, in the presence of ascorbate, donates electrons to cytochrome c, thus providing electrons to Nir and NosZ (18, 37). Ascorbate and TMPD (both from Sigma-Aldrich, Germany) were dissolved in ddH₂O or 96% ethanol, respectively, and filter sterilized. The solutions were added to the incubation flasks 10 to 15 min before gas sampling. The effect of different concentrations of TMPD (100, 250, and 500 μM in the culture buffer) combined with 10 mM ascorbate on N₂O reduction was checked prior to the main experiments. This showed that 500 μM TMPD had an obvious inhibition effect on N₂O reduction, while 100 μM and 250 μM showed no inhibitory effect (not shown). To minimize other possible effects we used 100 μM TMPD for the experiments (as done in Mania et al. [18]). In another experiment, shown in Fig. S4, 4 mL of a mannitol/yeast solution was added to flasks containing 50 mL starved cultures, providing them with a substrate concentration corresponding to full-strength YMB (10 g mannitol and 0.5 g yeast L⁻¹).

Monitoring of gas kinetics, NO₃^– and NO₂^– concentrations and electron flow rates.

The culture flasks were placed in a robotized incubation system and the headspace gas was sampled frequently for N₂, N₂O, NO, and O₂ measurements (34). Gas losses caused by sampling were considered when calculating the production and consumption of gases, as described by Molstad et al. (34) and Mania et al. (18). The concentration of O₂ in the anoxic incubation flasks was <0.6 μM at the start of the experiment, which was well below the level for initiating denitrification (4.6 μM) (19).

The NO₂^– concentrations were monitored as described in Mania et al. (18). Briefly, samples (0.1 to 0.5 mL) were taken every 1 or 2 h (n = 3) from the liquid phase through the septum of the flasks using a sterile syringe. To avoid that gas kinetics being affected by the sampling, a set of flasks was dedicated to NO₂^– measurements and a parallel set was left untouched for gas measurements. NO₂^– was determined using a chemoluminescence NOx analyzer (Sievers NOA 280i, GE Analytical Instruments) after first reducing the NO₂^– to NO by adding 10 μL liquid sample into a purging device containing a reducing agent (50% acetic acid with 1% [wt/vol] NaI) (53, 54). The NO₂^– concentrations were determined against a standard curve (range 0 to 2 mM NO₂^–; r² = 0.999).

The cell-specific electron flow rates (V_e, mol e⁻ cell⁻¹ h⁻¹) for each time increment between two gas samplings were calculated as V_eflask(t)/(N(0)+E_cum(t)*Y), where V_eflask(t) is the electron flow rate in the flask (mol e⁻ flask⁻¹ h⁻¹) calculated from measurements, N(0) is number of cells in the flask at time = 0, E_cum(t) is the cumulated electron flow (mol e⁻ flask⁻¹) at time = t, and Y is the yield per mol electron (cells mol⁻¹ e⁻) for HAMBI 2125, as measured previously (19). The conversion factor 5.8 E8 cells mL⁻¹ *OD₆₀₀⁻¹ was used to convert OD₆₀₀ to cell numbers (19).

Proteomics.

The abundances of Nap, Nir, and NosZ were quantified in starved cultures treated, as described for bioassay 2 (extended starvation). Altogether, 18 flasks were prepared according to Fig. 1B. After the 20-h aerobic incubation in C-free buffer, three flasks were harvested for proteomics analysis and, following centrifugation/washing, the three cell pellets from these flasks were frozen individually at –20°C. The cultures in the other flasks were pooled three by three and after centrifugation/washing, each cell pellet was divided in three and used to inoculate new flasks containing C-free buffer with NO₃^– in the liquid and with He and 1 mL N₂O in the headspace (Fig. 1B). These flasks (15 in total) were placed in the robotic incubation system for monitoring of gas kinetics and NO₂^– concentrations, as described above. The entire culture volume (50 mL) was harvested from each of the triplicate flasks at different time points during the anoxic incubation. The cultures in six of the flasks showed a fast respiration rate (total electron flow 1.0 to 1.8 e⁻ cell⁻¹ h⁻¹) and were harvested at 5 and 20 h of incubation in anoxic buffer (triplicates at each sampling point). The cultures in the other nine flasks showed a slow respiration rate (maximum total electron flow rate was 0.27 fmol e⁻ cell⁻¹ h⁻¹). Six of them were harvested in triplicates after 10 and 20 h in anoxic buffer. The remaining three flasks from the cultures with slow respiration rate received a portion of YMB at 24.8 hand were harvested at 27 h. Harvested cell cultures were centrifuged and stored as pellets at –20°C. The protein extraction was as described in Gao et al. (19). Briefly, the thawed cell pellets were resuspended in lysis buffer (20 mM Tris–HCl pH 8, 0.1% vol/vol Triton X-100, 200 mM NaCl, 1 mM DTT). They were then subjected to bead beating (3 × 45 s) with glass beads (particle size ≤ 106 μm; Sigma) using a MP Biomedicals FastPrep-24 (Thermo Fischer Scientific) at maximum power and with cooling on ice between the cycles. After centrifugation to remove cell debris (10,000 × g, 5 min), the supernatant, containing water soluble proteins, was used for proteomic analysis using an Orbitrap mass spectrometer (19, 55). Quantification was based on LFQ in MaxQuant (56), and the relative abundance of the individual reductases was calculated as percentages of the sum of all protein abundances for each time point. The nitric oxide reductases NorB/C were not measured since only a small fraction of these membrane-bound enzymes can be accurately obtained with this extraction protocol (approximately 1/1,000^th the quantity of the other reductases).

Viable counts and microscopy.

The number of viable cells in starved cultures showing slow and fast respiration rates, observed in flasks prepared according to bioassay 2, was determined by plating dilutions of the cultures on yeast-mannitol agar (YMA). The morphology of cells from well-fed cultures and starved cultures was compared using phase-contrast microscopy and a qualitative determination (presence/absence) of PHA was done by staining with Nile Red (Sigma-Aldrich) followed by fluorescence microscopy. Microscopy was performed on a Zeiss AxioObserver with an Orca-Flash 4.0 CMOS camera (Hamamatsu Photonics) controlled by the ZEN Blue software. Images were taken with a 100× phase contrast objective. An HPX-120 illuminator was used as light source for fluorescence microscopy. Images were prepared using ImageJ and analysis of cell sizes was done using the ImageJ-plugin MicrobeJ (57).

Data availability.

Raw data from gas measurements and microscopy can be made available upon request. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE (58) partner repository with the data set identifier PXD038844.