Microaerobic Lifestyle at Nanomolar O2 Concentrations Mediated by Low-Affinity Terminal Oxidases in Abundant Soil Bacteria

ABSTRACT High-affinity terminal oxidases (TOs) are believed to permit microbial respiration at low oxygen (O2) levels. Genes encoding such oxidases are widespread, and their existence in microbial genomes is taken as an indicator for microaerobic respiration. We combined respiratory kinetics determined via highly sensitive optical trace O2 sensors, genomics, and transcriptomics to test the hypothesis that high-affinity TOs are a prerequisite to respire micro- and nanooxic concentrations of O2 in environmentally relevant model soil organisms: acidobacteria. Members of the Acidobacteria harbor branched respiratory chains terminating in low-affinity (caa3-type cytochrome c oxidases) as well as high-affinity (cbb3-type cytochrome c oxidases and/or bd-type quinol oxidases) TOs, potentially enabling them to cope with varying O2 concentrations. The measured apparent Km (Km(app)) values for O2 of selected strains ranged from 37 to 288 nmol O2 liter−1, comparable to values previously assigned to low-affinity TOs. Surprisingly, we could not detect the expression of the conventional high-affinity TO (cbb3 type) at micro- and nanomolar O2 concentrations but detected the expression of low-affinity TOs. To the best of our knowledge, this is the first observation of microaerobic respiration imparted by low-affinity TOs at O2 concentrations as low as 1 nM. This challenges the standing hypothesis that a microaerobic lifestyle is exclusively imparted by the presence of high-affinity TOs. As low-affinity TOs are more efficient at generating ATP than high-affinity TOs, their utilization could provide a great benefit, even at low-nanomolar O2 levels. Our findings highlight energy conservation strategies that could promote the success of Acidobacteria in soil but might also be important for as-yet-unrevealed microorganisms. IMPORTANCE Low-oxygen habitats are widely distributed on Earth, ranging from the human intestine to soils. Microorganisms are assumed to have the capacity to respire low O2 concentrations via high-affinity terminal oxidases. By utilizing strains of a ubiquitous and abundant group of soil bacteria, the Acidobacteria, and combining respiration kinetics, genomics, and transcriptomics, we provide evidence that these microorganisms use the energetically more efficient low-affinity terminal oxidases to respire low-nanomolar O2 concentrations. This questions the standing hypothesis that the ability to respire traces of O2 stems solely from the activity of high-affinity terminal oxidases. We propose that this energetically efficient strategy extends into other, so-far-unrevealed microbial clades. Our findings also demonstrate that physiological predictions regarding the utilization of different O2 concentrations based solely on the presence or absence of terminal oxidases in bacterial genomes can be misleading.

based solely on the presence or absence of terminal oxidases in bacterial genomes can be misleading. KEYWORDS terminal oxidase, oxygen, acidobacteria, kinetics, transcriptomics O xygen (O 2 ) has a high redox potential (E 0 9 = 10.82 V), which, together with its ubiquity, makes it a favorable electron acceptor for energy generation. The concentration of O 2 across numerous microbial habitats can vary from saturation to anoxia (1). It is believed that aerobic microorganisms meet these fluctuating conditions by harboring low-and high-affinity terminal oxidases (TOs), presumably allowing them to use a wide range of O 2 concentrations.
Terminal oxidases, which mediate the final redox reaction in the electron transport chain (ETC) during aerobic respiration, are grouped into three superfamilies: (i) hemecopper oxidases (HCOs), (ii) cytochrome bd-type oxidases, and (iii) alternative oxidases. HCOs are multisubunit complexes and function as cytochrome c or as quinol oxidases, contributing to energy conservation, the generation of a proton motive force, O 2 scavenging, and maintaining redox homeostasis (2,3). Based on overall amino acid similarities of the catalytic subunits and differences of the proton channels, the HCO superfamily is classified into three families: A (subfamilies A1 and A2), B, and C (4). Family A oxidases have a low affinity for O 2 , with a reported Michaelis-Menten constant (K m ) for O 2 of 200 nmol O 2 liter 21 (5). HCO families B and C are considered high-affinity TOs with high catalytic activity at low O 2 concentrations but reduced proton-pumping efficiency (6), with K m values for the family C cbb 3 -type oxidases of 7 to 40 nmol O 2 liter 21 (7)(8)(9). The high-affinity cytochrome bd-type oxidase encoded by the cydAB genes (10)(11)(12) has reported K m values of 3 to 8 nmol O 2 liter 21 (13). Cytochrome bd-type oxidases do not pump protons across the membrane but contribute to proton motive force by using electrons from the extracytoplasmic side and protons from the cytoplasmic side (11).
High-affinity TOs are believed to sustain energy conservation at diminishing concentrations by enabling respiration at trace amounts of O 2 (i.e., micromolar O 2 concentrations) (14)(15)(16). Although there has been some suggestion that low-affinity TOs are present at micromolar O 2 concentrations in addition to high-affinity TOs (5), it remains unclear if the low-affinity TOs can actively and even solely contribute to respiration at these O 2 concentrations. At nanomolar O 2 concentrations, microorganisms transition from aerobic respiration to anaerobic-based metabolism (substrate-level phosphorylation or anaerobic respiration), referred to as the Pasteur point (17,18). To the best of our knowledge, gene expression-based investigations of terminal oxidases at nanomolar O 2 concentrations are scarce (e.g., Gong et al. reported expression at O 2 levels of #200 nmol [19]), and therefore, it is mostly speculated that the high-affinity terminal oxidases are primarily responsible for energy production at low-nanomolar O 2 concentrations.
In soil, O 2 availability can be spatially and temporally dynamic, depending on the edaphic properties and microbial activity (20,21). As such, microbial survival in soil is dependent on the ability to adapt to changes in local O 2 conditions. Environmental data and genome surveys suggest that both low-and high-affinity TOs are widely distributed in soils (16). Acidobacteria represent one of the most abundant and phylogenetically diverse phyla in soils worldwide (22)(23)(24) and are assigned a central role in carbon mineralization and plant polymeric carbon degradation (25,26). Genes encoding high-and low-affinity TOs have been identified in several genomes of the phylum Acidobacteria (27), suggesting the capacity to respire across a wide gradient of O 2 concentrations. As respiratory flexibility can be attained through branched respiratory chains that terminate in multiple oxidases with different affinities for O 2 (15), this facet might be key to their ecological success in soil.
Using Acidobacteria as model soil organisms, we explored respiratory kinetics and evaluated their gene expression using whole-transcriptome sequencing and reverse transcription-quantitative PCR (RT-qPCR) across decreasing low-micromolar to nanomolar O 2 concentrations. As such, we could test the hypothesis that at micro-to nanomolar O 2 concentrations, aerobic respiration is mediated by high-affinity TOs. Our data demonstrate that O 2 concentrations down to the nanomolar level can be respired by low-affinity TOs, an unexpected physiological response, suggesting that the ability to respire O 2 under micro-to nanooxic conditions is not exclusively based on the presence and activity of high-affinity TOs.

RESULTS
Distribution of low-and high-affinity terminal oxidases. Five acidobacterial strains were chosen to explore their respiratory kinetics, and of these strains, three were chosen to explore their TO expression patterns across nanomolar O 2 concentrations. All strains harbored branched respiratory chains terminating in multiple oxidases ( Fig. 1; see also Data Set S1 in the supplemental material). They differed in their distributions of low-and high-affinity TOs (complex IV) as well as of complexes III (cytochrome bc 1 complex and/or alternative complex III [ACIII]) ( Fig. 1; Data Set S1).
There was consistent gene synteny for the A1 caa 3 HCO, A2 caa 3 HCO, C cbb 3 HCO, and bd-type quinol oxidases and the adjacent complex III genes among the acidobacterial strains (Fig. 1). Genes for the A1 caa 3 HCO were always located in an operon upstream of the genes encoding the bc 1 complex (described here as a "superoperon") ( Fig. 1). The A2 caa 3 HCO also occurred in a superoperon with the genes encoding ACIII, instead of the bc 1 complex, and were located downstream of the ACIII genes ( Fig. 1). Additional, single homologs of either the A1 or A2 caa 3 -type oxidases were detected in the genomes of Acidobacteriaceae bacterium KBS 83 (Fig. 1a), Terriglobus sp. TAA 43 (Fig. 1b), and T. roseus KBS 63 (Fig. 1c). T. roseus KBS 63 (Fig. 1c) and A. capsulatum 161 (Fig. 1e) contained cbb 3 operons consisting of genes for cbb 3 subunits N and O as well as an additional cco gene of unknown function. Edaphobacter sp. TAA 166 (Fig. 1d) and A. capsulatum 161 (Fig. 1e) contained both cydA and cydB subunits for the bd-type quinol oxidase.
Assessment of O 2 respiratory kinetics. We determined the O 2 respiration rates and population apparent K m (K m(app) ) values for the five acidobacterial strains with differing distributions of high-and low-affinity TOs in exponential phase (non-energy limited) with only O 2 -limiting respiration rates (Fig. 2 (Fig. 2d), respectively. The maximum population respiration rate (V max ) of Acidobacteriaceae bacterium KBS 83 was on average 355 6 12 nmol O 2 liter 21 h 21 , and the maximum respiration rates per cell (R max ) progressively decreased over time from 9.8 to 6.8 6 0.4 fmol O 2 cell 21 h 21 ( Fig. 2a; Table S1). The V max of Terriglobus sp. TAA 43 was 998 6 6 nmol O 2 liter 21 h 21 , and the R max was constant at 2.  (Table S2), with a final K m(app) value 1 order of magnitude lower than the values of the other investigated strains (Fig. 2c). In addition, the V max and R max of A. capsulatum 161 progressively increased during the whole period of measurements from 2,150 6 156 to 3,609 6 430 nmol O 2 liter 21 h 21 (Table S2) and from 0.17 6 0.01 to 0.26 6 0.03 fmol O 2 cell 21 h 21 , respectively (Fig. 2c). The respiration rates rose to a maximum as O 2 concentrations increased and then descended to a nonzero asymptote. Additionally, the velocity curves saturated rapidly, compared to the other strains (Fig. 2e).
Differential gene expression due to changing O 2 concentrations. Of the five strains, we selected three that encompass different combinations of low-and high-affinity TOs to compare changes in gene expression levels when exposed to different, decreasing O 2 concentrations. Transcriptome analysis of Acidobacteriaceae bacterium KBS 83, T. roseus KBS 63, and A. capsulatum 161 showed that in the course of the time series, 5,121 (93% of all annotated genes), 4,239 (97%), and 3,321 (97%) genes, respectively, were transcribed at least at one time point across the O 2 concentrations (Table S5).
The decrease from 10 to 0.1 mmol O 2 liter 21 had the greatest impact on the transcriptomes of all three strains, with the highest number of significantly differentially expressed genes observed (Fig. 3a). Among 1,602 (31%) differentially expressed genes of Acidobacteriaceae bacterium KBS 83, 16% were upregulated and 15% were downregulated upon the transition from 10 to 0.1 mmol O 2 liter 21 after cells equilibrated for 60 min at each respective O 2 concentration (Fig. 3b). For T. roseus KBS 63 and A. capsulatum 161, 38% (20% upregulated and 18% downregulated) and 81% (41% upregulated and 40% downregulated), respectively, were differentially expressed upon this transition from 10 to 0.1 mmol O 2 liter 21 (Fig. 3b). Comparatively, there were few to no significant expression changes when transitioning from 0.1 to 0.001 mmol O 2 liter 21 regardless of the equilibration time at the lower O 2 concentration; similar patterns were Their O 2 uptake rates were higher than the O 2 inflow rate, thereby creating an apparent anoxic environment below our detection limit of 0.0005 mmol O 2 liter 21 ( Fig. 4c and e).
Transcriptional responses of branching electron transport chain key genes and terminal oxidases to decreasing O 2 concentrations. We further explored the transcriptional changes of TOs (complexes III and IV) of the ETC by focusing on key functional genes of these complexes ( Fig. 4; Data Set S2).
(i) Acidobacteriaceae bacterium KBS 83. Continuous expression of two out of the four low-affinity caa 3 -type cytochrome c oxidases, one of the bc 1 -A1 caa 3   and the ACIII-A2 caa 3 superoperon, was observed across all investigated O 2 concentrations, even after exposure to 0.001 mmol O 2 liter 21 for an extended period of time ( Fig. 4b; Data Set S2); similar patterns were observed by RT-qPCR (Fig. S1a). All genes of superoperon ACIII-A2 caa 3 exhibited significantly lower expression levels at 0.001 than at 10 mmol O 2 liter 21 (P , 0.05), yet the catalytic subunit ctaD of the A2 HCO was consistently highly expressed across O 2 concentrations and not significantly downregulated (Fig. 4b). In contrast, ctaD of the A1 HCO complex together with petC of the bc 1 complex were significantly upregulated at 0.001 mmol O 2 liter 21 (P , 0.05). The transcription level of the electron-receiving subunit II (ctaC) was higher than that of the rest of the bc 1 -A1 caa 3 superoperon and remained high upon transitions to lower O 2 concentrations (Fig. 4b); the same responses were observed within the first 10 min after shifts of oxygenation by RT-qPCR (Fig. S1a). We still observed gene expression 15 min after the O 2 supply was ceased (Fig. S1a). Even then, the O 2 concentration did not fall below 0.01 mmol O 2 liter 21 (Fig. 4a), and Acidobacteriaceae bacterium KBS 83 was still expressing its TOs after 3 h at 0.01 mmol O 2 liter 21 (Fig. 4b). Of the other complexes IV, only ctaE that encodes subunit III of the single complex IV exhibited high expression levels (Fig. 4b).
(ii) T. roseus KBS 63. The expression levels (transcripts per million [TPM]) of the catalytic subunit of the cbb 3 -type high-affinity TO (ccoN) across the investigated O 2 concentrations were low ( Fig. 4d; Data Set 2) and too low for reliable quantification by RT-qPCR (Fig. S1b). The catalytic subunits of both low-affinity A2 HCO TOs (ctaD) exhibited the highest expression levels and were transcribed at significantly higher levels (P , 0.0001) at 0.001 than at 10 mmol O 2 liter 21 (Fig. 4d). All other genes of the ACIII-A2 caa 3 superoperon were also upregulated (Fig. 4b). After a shift to anoxic conditions, the single ctaD gene was still expressed and upregulated (Fig. S1b).
(iii) A. capsulatum 161. The cbb 3 -type high-affinity TO was transcribed at low levels at 10 mmol O 2 liter 21 and was significantly downregulated (P , 0.0001) at all subsequent lower concentrations (down to 0.001 mmol O 2 liter 21 ) (Fig. 4f; Data Set S2). Expression of the cbb 3 -type high-affinity TO by RT-qPCR was seen from 10 to 0.1 mm O 2 but only for 10 min at this concentration as measured by RT-qPCR (Fig. S1c). In contrast, the bd-type TO (cydAB) was expressed at all investigated O 2 concentrations (10 to 0.001 mmol O 2 liter 21 ). The relative abundance of cydA transcripts was very high under all O 2 tensions (58-fold higher than that of the rpoB gene) (Data Set S2). RT-qPCR showed a clear and significant (P # 0.05) upregulation of the catalytic subunit cydA (Fig. S1c). Furthermore, cydA transcription levels were always high, even under anoxic conditions. The ctaD gene, encoding the catalytic subunit of the low-affinity A1 HCO, was continuously transcribed across all O 2 concentrations as detected by transcriptomics and RT-qPCR ( Fig. 4f; Fig. S1c). However, the proportion of ctaD transcripts decreased under anoxic conditions (Fig. 4f).

DISCUSSION
Members of an abundant soil phylum, the Acidobacteria, respire environmentally relevant micro-and nanomolar O 2 concentrations with the use of low-affinity TOs. Respiratory kinetics were determined using highly sensitive optical sensors, which allowed us to study the O 2 kinetics with a high degree of accuracy. Our findings extend the current knowledge on O 2 kinetics to species outside the Proteobacteria.
Acidobacteria harbor branched respiratory chains terminating in multiple complexes IV with either low or high affinities for O 2 . Branched ETCs terminating in differing terminal electron acceptors (such as O 2 , NO 3 , or NO 2 ) are typically found in bacteria, providing flexibility when exposed to various environmental conditions (14,15). Enzymatic redundancy in using a single electron acceptor (such as O 2 ) can provide additional flexibility due to varying substrate affinities, allowing the microorganism to respire most efficiently across different concentrations, as seen in organisms living at the oxic-anoxic interface (28)(29)(30)(31)(32)(33)(34). This flexibility extends to our investigated soil acidobacterial strains, as many of them have branched ETCs that terminate in multiple complexes IV with either low or high affinities for O 2 (Fig. 1). Furthermore, in select strains, genes for complex IV were detected in superoperons together with genes for complex III, either bc 1 or alternative complexes III (Fig. 1), as previously seen in other members of the Acidobacteria and further phyla (35,36), potentially functioning as respiratory supercomplexes (37)(38)(39). Although the physiological relevance of supercomplexes is still unclear (40), we suggest that this physical association might provide additional metabolic flexibility in the acidobacteria. The close association could allow a more favorable transfer between complexes, bypassing soluble electron carriers (39). Nevertheless, follow-up investigations will be needed to elucidate the advantage of the supercomplexes. The complex IV genes were also found independent from complex III genes in three strains (Fig. 1).
The conventional high-affinity cbb 3 -type TO does not actively contribute to the capacity to respire O 2 at nanomolar concentrations. High-affinity TOs are historically believed to enable respiration and provide the capacity for energy conservation at trace concentrations of O 2 , a physiology that was shown to be widespread among bacteria and archaea of diverse environments, as suggested by genome surveys (16). Yet in the investigated acidobacterial strains, the cbb 3 -type high-affinity TO did not impart the capacity to respire O 2 at nanomolar concentrations. In our experimental setup, strains harboring high-affinity TO genes had the potential to develop low apparent K m values by expressing these TO genes under O 2 -limited conditions, as in our incubations, the cells were exposed to multiple oxic-to-anoxic gradients over a 24-h period. Furthermore, our investigated strains harbor the minimal core, the CcoNO protein dyad (41), for the functionality of the enzyme (Fig. 1). Expression of the cbb 3 -type oxidase could not be detected below 10 mmol O 2 liter 21 in both strains T. roseus KBS 63 and A. capsulatum 161 (Fig. 4), although they indeed consumed O 2 down to (apparent) anoxia. Compared to reported K m(app) values for O 2 of Proteobacteria strains harboring cbb 3 -type oxidases measured by the same method (42), the K m(app) value of T. roseus KBS 63 was high (113 nmol O 2 liter 21 ) (Fig. 2b). This further provides evidence for the activity of the low-affinity oxidase(s) and suggests that it might be used for respiration in environments with low O 2 concentrations, such as the heterogeneous soil environment. O 2 fluctuations in soil are dynamic, and exposure to low-nanomolar O 2 concentrations might be temporally limited to short intervals (43). Therefore, we hypothesize that the investment in the expression of a less-energy-efficient TO (the high-affinity cbb 3 type) (44, 45) will not provide any competitive advantage for these investigated time intervals. At this time, it is unclear if the cbb 3 -type oxidase has lost its function to generate proton motive force in these strains. Alternatively, cbb 3 TO expression in T. roseus KBS 63 and A. capsulatum 161 could be triggered by other factors, such as nutrient limitation or carbon depletion, as recently reported for Shewanella oneidensis (46).
Utilization of acidobacterial bd-type oxidases at nanomolar O 2 concentrations. The bd-type oxidases are another type of high-affinity TO, which are less efficient at creating the charge gradient for ATP generation as they do not pump protons across the membrane but generate a proton motive force by transmembrane charge separation (12). Expression data showed a clear and significant upregulation of the catalytic subunit cydA gene as O 2 concentrations decreased in A. capsulatum 161 ( Fig. 4f; see also Fig. S1c in the supplemental material). This suggests that the bd-type oxidase contributed to the respiratory activity under trace O 2 conditions. In contrast, the cbb 3 type was transcribed only at low levels at 10 mmol O 2 liter 21 and was significantly downregulated (P , 0.0001) at all subsequent lower concentrations ( Fig. 4f; Fig. S1c). However, the use of the bd-type oxidase for respiration activity appears to be strain dependent. In another strain harboring a high-affinity bd-type oxidase (Edaphobacter sp. TAA 166), the expression of cydA could not be detected at any examined O 2 concentration; rather, the low-affinity TOs were expressed across these O 2 concentrations (RT-qPCR data not shown). Here, the bd-type oxidase could be contributing to physiological functions other than respiratory O 2 reductions, such as reactive oxygen species (ROS) stress, iron deficiency, or nitric oxide stress responses (11,12,47).
Although the bd-type oxidases are not as efficient at creating a charge gradient, these oxidases have functional and structural characteristics that favor a faster electron flux than cbb 3 -type oxidases (11,12), which could be advantageous under conditions with plentiful reducing potential stemming from carbon surplus. For instance, they receive electrons directly from the quinol pool and thereby take a shortcut through the branched ETC, bypassing any complexes III (Fig. 1). In support of this conjecture, bd-type oxidase genes were found to be more prevalent in environments where carbon is in excess, such as host-associated environments and carbon-rich forest soils compared to carbon-poor agricultural soils (16). As our investigated conditions were a combination of carbon surplus and O 2 limitation, we therefore hypothesize that this selected for the utilization of the bd-type oxidase compared to the cbb 3 type in A. capsulatum 161.
The strain expressing the bd-type oxidase under low O 2 concentrations (A. capsulatum 161) was the only one that was inhibited by high O 2 concentrations at its maximum respiration rate (R max ) (Fig. 2c) (.250 nmol O 2 liter 21 ). Furthermore, its K m(app) value decreased over multiple oxic-anoxic shifts (n = 17) within 24 h, indicating a need for less substrate and, therefore, an adaptation to these conditions. This temporal kinetic development was previously observed for marine Proteobacteria (42). The final estimated K m(app) value of A. capsulatum 161 (37 nmol O 2 liter 21 ) suggests a mixed activity of low-and high-affinity TOs (Fig. 2c), with its high-affinity TO contributing a large portion of the K m(app) value. This respiratory kinetic activity of A. capsulatum 161 suggests that this strain can use different O 2 concentrations due to its enzymes' O 2 affinities. Presumably, this strain has a different strategy to exploit microoxic niches compared to the other investigated strains, which also could be advantageous in the soil when exposed to spatiotemporal gradients and diffusion limitations.
Acidobacterial low-affinity TOs are used at nanomolar O 2 concentrations. Acidobacterial low-affinity caa 3 -type HCOs are functioning at previously unknown nanomolar O 2 concentrations, as shown in the investigated strains ( Fig. 4; Fig. S1). The use of low-affinity A HCOs at low concentrations of O 2 is energetically favorable, as they have more free energy available for driving proton translocation due to poor O 2 binding (44, 45) and a more efficient, and thus favorable, gating for proton leakage (44) than high-affinity TOs. High-affinity C HCOs typically exhibit higher catalytic activity at lower O 2 concentrations due to a different redox-driven proton-pumping mechanism that allows an increased electron transfer rate and a faster reduction of O 2 (48). Still, these high affinities come with a reduced proton-pumping efficiency (6,44).
Many of the genes for the A2 caa 3 HCO in T. roseus KBS 63 were not only expressed across varying O 2 concentrations but in some cases also even upregulated at lower O 2 concentrations ( Fig. 4d; Fig. S1b). A continuous expression of low-affinity caa 3 -type TOs at low O 2 concentrations was previously reported in aerobic marine bacterial species (19,49); however, in that study (19), the high-affinity cbb 3 -type TO was upregulated at ,0.2 mmol O 2 liter 21 . In our study, we did not observe any measurable contribution via transcriptomics or qPCR of the high-affinity cbb 3 -type TO in any of the strains at 10 to 0.001 mmol O 2 liter 21 , although we cannot completely rule out the possibility of a minor contribution (undetectable with our current methods) of the cbb 3type TO. Likewise, it is conceivable that high-affinity cbb 3 -type TOs function only at extremely low concentrations of O 2 (,1 nmol O 2 liter 21 ), which we currently cannot establish, maintain, and measure in the laboratory. Nevertheless, it appears that at the low O 2 concentrations (down to 1 nmol O 2 liter 21 ) investigated in this study, T. roseus KBS 63 definitely prioritizes the low-affinity TOs. The energetic advantage of the low-affinity TOs might explain the strategy of T. roseus KBS 63 to invest in the high expression and upregulation of A2 caa 3 HCOs, compared to its cbb 3 -type high-affinity TO (Fig. 4d;  Fig. S1b).
In contrast, Acidobacteriaceae bacterium KBS 83 harbored only low-affinity TOs (caa 3 type) and was able to respire at O 2 concentrations of 10 mmol O 2 liter 21 and lower. Below 0.01 mmol O 2 liter 21 , it consumed O 2 at a respiration rate lower than the rate at which O 2 was supplied, causing concentrations not to reach anoxic conditions (Fig. 4a). However, complete consumption to anoxia was reached during the kinetics measurement experiments, reflecting the capacity to respire O 2 at trace concentrations. This difference could be explained by a lower cell density in the incubations for transcriptome analysis, not allowing these incubations to reach anoxia during the time course of the incubations simply due to cell number. Alternatively, O 2 diffusion could explain this discrepancy; this is unlikely as it was not observed in other incubations of the investigated acidobacteria. Efficient energy conservation (generating more ATP/ electron) would be a vital survival strategy in times of substrate limitation in environments such as soil. It therefore might be an advantage to use low-affinity TOs even at nanomolar O 2 concentrations as they, despite their lower reaction rate, ultimately drive more charges across the membrane per mole of O 2 , making them more efficient in energy conservation.
It appears that the capacity of Acidobacteriaceae bacterium KBS 83 to respire O 2 under low concentrations was limited, as seen by the decreasing V max and R max over time (Table S1). Its K m(app) value (166 nmol O 2 liter 21 ) is lower than and in contrast to the previously reported K m value for the caa 3 -type oxidase of Pseudomonas aeruginosa (4,300 nmol O 2 liter 21 ) (8) but in the same range as the one for the low-affinity cytochrome bo 3 ubiquinol oxidase of Escherichia coli (200 nmol O 2 liter 21 ) (5). Although it is difficult to compare K m values across studies as the determined K m values can differ dramatically depending on the applied approach (8,50), we want to stress the fact that one has to be careful with historically set benchmarks that propagate in the literature. The determined K m(app) values of our study represent ecophysiologically relevant estimates as we used whole populations and intact cells as well as highly sensitive optical sensors with an extremely low detection limit.
Conclusion. Microorganisms frequently have to cope with changing O 2 tensions; therefore, having the flexibility to use a wide range of O 2 concentrations is beneficial (16). Here, we show that members of a dominant and ubiquitous soil phylum (22,24,26), the Acidobacteria, have branched ETCs that terminate in multiple oxidases (highand low-affinity TOs), providing them with respiratory flexibility and adaptability to environmental changes (14)(15)(16). More specifically, their low-affinity TOs are functioning at nanomolar O 2 concentrations, presumably providing a great benefit for soil acidobacteria as they are more efficient in generating ATP than high-affinity TOs (44). We hypothesize that this strategy could be employed by other bacterial clades in soil as well as other habitats. Follow-up work is needed to ascertain if respiration at nanomolar O 2 concentrations allows biomass production or population growth in the long run during exposure to such low O 2 levels. In addition, low O 2 concentrations and nutrientrich conditions selected for the expression of the high-affinity bd-type oxidase rather than the cbb 3 type, which presumably provides a more optimal balance of substrate oxidation and ATP production under these conditions. Follow-up studies are needed to elucidate the conditions under which acidobacterial cbb 3 -type TOs are employed for respiration. Our results extend the current knowledge on the respiratory flexibility of the prevalent Acidobacteria, which could help explain their success in the heterogeneous soil environment.
"Microaerobes" were previously defined as microorganisms that harbor high-affinity TOs in their genomes, either alone or in combination with low-affinity TOs, and use them to respire O 2 in microoxic environments (16). However, "microoxic" or subatmospheric concentrations of O 2 could be anything below 21% (vol/vol) O 2 , and within this range, the response of TOs can vary dramatically. In our study, we pushed microoxic to nanooxic conditions and explored the transcriptional response combined with enzyme kinetics to obtain a state-of-the-art assessment of their response to O 2 tension. We detected high-and low-affinity TOs in multiple acidobacterial genomes and respiration at nanomolar O 2 concentrations across the investigated strains. Yet our gene expression data did not indicate any detectable contribution of the cbb 3 -type high-affinity TOs at these O 2 concentrations; only one strain had contributions from the high-affinity bd-type TO. This suggests that the capability for microaerobic respiration in these acidobacteria is not solely due to the presence and associated activity of high-affinity TOs.
Instead, the acidobacterial microaerobic lifestyle seems to also be imparted by low-affinity caa 3 -type TOs that enable them to respire O 2 at nanomolar concentrations. This illustrates that the presence of a high-affinity TO in a genome is not a prerequisite for microaerobic respiration. To that end, we would like to amend the definition of microaerobe to encompass microorganisms that are capable of respiring O 2 under microoxic conditions via the utilization of high-or low-affinity TOs. Furthermore, these findings demonstrate that it can be challenging to make predictions on the ecophysiology and lifestyle of microorganisms based solely on their genomic information, even for a process as well studied as aerobic respiration.  (51,52) amended with 10 mM glucose as the sole carbon source at pH 6 or 5 (A. capsulatum 161). Additional information on the strains was reported previously (27,(52)(53)(54).

MATERIALS AND METHODS
Setup and incubation for respiratory kinetic parameters. The details of the setup and experimental procedure were previously described (34,42,55). Briefly, the incubations were conducted in custommade 500-or 1,100-ml glass bottles, which had been sequentially rinsed with a solution containing 0.1 M NaOH, 0.1 M HCl, and autoclaved water to prevent contamination. A continuous flow of N 2 was maintained while filling the bottles with N 2 -purged medium and subsequent sealing with ground-glass stoppers. Exponential-phase acidobacterial cells were injected into these bottles (2 to 3 replicates/strain), while glass-coated magnetic stirrers homogenized the suspension. The O 2 concentration was optically determined every 20 s by luminescence-based O 2 sensors (Lumos) with sensor spots (measurement range, 0.5 to 1,500 nmol O 2 liter 21 ) (56) glued onto the inside of the bottles. Bottles were incubated at room temperature and shielded from light for 24 h. Air-saturated water (4 to 5 ml) was repeatedly injected into the bottles after anoxia was reached by cell respiration, with peak concentrations ranging from 600 to 1,620 nmol O 2 liter 21 . One milliliter of the cell suspension was collected and fixed with 1% glutaraldehyde (Sigma-Aldrich, St. Louis, MO, USA) to determine cell numbers as described previously (42). After the incubations were completed, O 2 sensors were calibrated with oxygenated water and sodium dithionite.
Calculation of kinetic parameters. O 2 consumption rates were calculated from linear regression of O 2 concentrations over time in intervals of 6 min from the highest O 2 concentration down to anoxia. Kinetic parameters, the apparent half-saturation constant (K m(app) ) and the maximum respiration rate (V max ) of the Michaelis-Menten equation, were estimated by performing nonlinear parametric fits on the respiration-versus-O 2 -concentration curves for each replicate. V max and K m(app) were varied iteratively until the best fit was obtained by least-square fits using Solver in Microsoft Excel (57). Maximum respiration rates per cell (R max ) were calculated by dividing the population respiration rate (V max ) by cell numbers. Michaelis-Menten plots of respiration rates versus O 2 concentrations were obtained by fitting a Michaelis-Menten model to the data using the equation 21 , where V is the rate, V max is the maximum rate (nanomoles of O 2 per liter per hour), K m is the half-saturation constant (nanomoles of O 2 per liter), and [O 2 ] is the substrate concentration (nanomoles of O 2 per liter). Additional modifications of the Michaelis-Menten equation and further corrections can be found in Text S1 (Supplemental Materials and Methods 1) and Tables S1 and S2 in the supplemental material.
Transcriptional profiling incubations. Acidobacteriaceae bacterium KBS 83, T. roseus KBS 63, and A. capsulatum 161 were grown in biological quadruplicates in glass bottles (Schott) containing 1 liter of VSB minimal medium amended with 10 mM glucose under fully aerated conditions. Once cells reached exponential phase, they were transferred into HCl-sterilized and autoclaved-water-rinsed glass bottles equipped with internally preglued sensing spots. Incubations were run for 225 min and split into four discrete, declining O 2 concentrations (10 mmol O 2 liter 21 , 1mmol O 2 liter 21 , 0.1 mmol O 2 liter 21 , and 0.001 mmol O 2 liter 21 ) down to anoxia (0 mmol O 2 liter 21 is ,0.0005 mmol O 2 liter 21 ) obtained by purging with N 2 -air mixtures (Table S3). O 2 concentrations were monitored by two Lumos systems with different sensitivity ranges (0.5 to 1,500 and 10 to 20,000 nmol O 2 liter 21 ) (56). At every time point (Table S3), 30 to 50 ml of the culture was collected for RNA extractions by syringes prefilled with a phenol-stop solution (58). The sensor spots were calibrated after the incubations with oxygenated water and sodium dithionite. Additional details can be found in Text S1 (Supplemental Materials and Methods 2).
RNA extraction and purification. Total RNA was extracted from frozen cell pellets using an acidic phenol-chloroform-isoamyl alcohol protocol as described previously (59), with mechanical disruption (FastPrep-24 bead beater; MP Biomedicals, Heidelberg, Germany). The extraction supernatant was purified using standard chloroform-isoamyl alcohol purification, and RNA was precipitated using a polyethylene glycol (PEG) solution and RNA-grade glycogen by centrifugation (21,130 Â g for 1 h at 4°C). Coextracted DNA was digested using a Turbo DNA-free kit (Thermo Fisher), and complete DNA removal was verified by failure to obtain quantitative PCR (qPCR) amplification products with the purified RNA template, targeting the rpoB gene encoding the b subunit of the DNA-directed RNA polymerase, under the qPCR conditions described in Table S4. A more detailed protocol can be found in Text S1 (Supplemental Materials and Methods 3).
Primer design, cDNA synthesis, RT-qPCR, and data analysis. Specifications of the newly designed primers targeting the catalytic subunits (subunit I) of the TOs are listed in Table S4. See Text S1 (Supplemental Materials and Methods 4) for details on primer design, cDNA synthesis, reverse transcription-qPCR (RT-qPCR), and data analysis.
Transcriptome sequencing. Triplicate total RNA samples of Acidobacteriaceae bacterium KBS 83, T. roseus KBS 63, and A. capsulatum 161 from selected O 2 concentrations and time points were sent to the Vienna BioCenter Core Facilities. rRNA was depleted using the New England BioLabs (NEB) Ribo-Zero rRNA removal kit for bacteria. Sequencing was performed on an Illumina NextSeq 550 system, resulting in a total of 36 samples with 8.2 million to 18.2 million 75-nucleotide reads each.
Transcriptome data processing and statistical analyses. Raw reads were trimmed of sequencing adapters and low-quality 39 ends using BBduk (BBtools v37.61; https://jgi.doe.gov/data-and-tools/ bbtools/) with default parameters and error corrected using the Bayes-Hammer module of SPAdes assembler version 3.13.0 (60). Any reads mapping to either SILVA small-subunit (SSU) or large-subunit (LSU) release 132 (61) or the 5S rRNA database (62) with a sequence identity of .70% (performed with BBmap and BBtools; https://jgi.doe.gov/data-and-tools/bbtools/) were removed from the data set. The remaining reads were mapped to the publicly available genomes of the acidobacterial strains (53). The RNA reads per gene were summarized using the featureCounts tool from the Subread package v1.6.2 (63). Based on the generated read count tables, transcripts per million were calculated in R v3.6.0. Differential expression analyses, such as calculations of log 2 fold changes of relative transcript abundances and the significance of these changes, were performed in DESeq2 v1.26.0 using default parameters and a P value cutoff of 0.05 (64).
Data availability. The raw transcriptomic reads are available under BioProject accession number PRJNA635786. The code and pipelines used for data analysis are available upon request.