Identification of Key Factors for Anoxic Survival of B. cenocepacia H111

Burkholderia cenocepacia is an opportunistic pathogen that can lead to severe infections in patients suffering from cystic fibrosis (CF) and chronic granulomatous disease. Being an obligate aerobe, B. cenocepacia is unable to grow in the absence of oxygen. In this study, we show that the CF isolate B. cenocepacia H111 can survive in the absence of oxygen. Using a transposon sequencing (Tn-seq) approach, we identified 71 fitness determinants involved in anoxic survival, including a Crp-Fnr family transcriptional regulatory gene (anr2), genes coding for the sensor kinase RoxS and its response regulator RoxR, the sigma factor for flagella biosynthesis (FliA) and subunits of a cytochrome bd oxidase (CydA, CydB and the potentially novel subunit CydP). Individual knockouts of these fitness determinants significantly reduced anoxic survival, and inactivation of both anr copies is shown to be lethal under anoxic conditions. We also show that the two-component system RoxS/RoxR and FliA are important for virulence and swarming/swimming, respectively.


Introduction
Despite recent improvements in treatment regimens, pulmonary disease remains the leading cause of morbidity and mortality in patients with cystic fibrosis (CF) [1]. The pathophysiology of this autosomal recessive inherited disease is based on a single mutation in the cystic fibrosis transmembrane conductance regulator gene (CFTR), resulting in the thickening of mucus [2,3], which becomes prone to colonization by opportunistic pathogens. Chronic infection can lead to bronchiectasis (lung damage) and respiratory failure, often with deadly outcomes [4]. The local oxygen concentrations in the CF lung range from aerobic to micro-oxic, and even anoxic pockets in mucus plugs or in deep biofilm layers have been reported [5,6]. Thus, bacteria colonizing the CF mucus must be able to adapt their metabolism to different oxygen levels. Indeed, it was shown that the facultative anaerobe Pseudomonas aeruginosa, the predominant infectious agent invading the airways of CF patients [7], is able to grow in oxygen-limited environments by using a combination of micro-oxic respiration, denitrification and arginine and pyruvate fermentations [5]. Denitrification, which is the complete reduction of nitrate (NO 3 − ) to molecular nitrogen (N 2 ), is an important alternative energy-conserving pathway used in the absence of oxygen. The energy yields of denitrification consist of approximately nine ATPs per two molecules of nitrate [8] in contrast to the 36 ATP produced per one glucose consumed in aerobic respiration. In the absence of oxygen and nitrate, P. aeruginosa is able to ferment arginine to ornithine, thereby generating one molecule of ATP per molecule of arginine via substrate-level phosphorylation [9]. P. aeruginosa is also able to survive anoxic conditions by producing lactate via a mixed-acid fermentation of pyruvate [10]. Another important anoxic survival mechanism is the reduction of phenazines, small excreted molecules capable of extracellular electron transfer, which can be coupled to ATP generation. P. aeruginosa reduces phenazines, which then get oxidized by oxygen, other oxidants, or by P. aeruginosa [11,12]. CF patients suffering from P. aeruginosa infections are often co-infected with other opportunistic pathogens. Of particular concern are infections with strains belonging to the Burkholderia cepacia complex (Bcc), which can cause a lifethreatening systemic infection known as cepacia syndrome [13]. The Bcc comprises more than 25 closely related species [14], and their prevalence in CF patients is between 2 and 4% [15]. Members of the Bcc are highly versatile and adaptable and have been isolated from diverse habitats such as soil ecosystems (plant rhizosphere), aquatic environments and infected humans [14,16,17]. B. cenocepacia is an obligate aerobic opportunistic pathogen of the Bcc group that requires an oxygen concentration of at least 0.1% for growth [18]. Specific high oxygen affinity cytochrome c oxidases, known as cytochrome cbb 3 oxidases, which play important roles in the adaptation to hypoxic conditions, are almost exclusively found in Proteobacteria [19,20]. However, no genes coding for cytochrome cbb 3 oxidases were identified in the genus Burkholderia [18]. Moreover, B. cenocepacia H111 is unable to denitrify due to a lack of the gene clusters coding for enzymes required for the reduction of nitrate to dinitrogen, such as the nitrate reductase gene cluster (nar), the nitric oxide reductase (nor) and the nitrous oxide reductase (nos) cluster. However, B. cenocepacia H111 contains a cluster that potentially encodes the nitrite reductase (I35_RS24165-24180). The only Burkholderia sensu stricto species capable of denitrification belong to the Burkholderia pseudomallei complex (Bpc) [21]. A previous study showed that in the Bpc strain Burkholderia thailandensis E264, the first step in the denitrification process, namely nitrate reduction, generates sufficient energy for anoxic growth [21]. A global gene expression analysis of B. cenocepacia J2315 under nine different growth conditions, including growth in an environment with reduced oxygen (6% O 2 ), identified a specific locus carrying genes involved in central metabolism, stress response and transport that was highly up-regulated under micro-oxic conditions. This low-oxygen-activated locus (lxa) was shown to be involved in the anoxic persistence of B. cenocepacia strains J2315 and K56-2 [22]. However, the absence of the lxa locus in B. cenocepacia H111 suggests that this strain uses other strategies to survive in the absence of oxygen.
In this study, we show that B. cenocepacia H111 is able to survive without oxygen and applied a high-throughput parallel sequencing approach based on Tn-seq in order to identify fitness determinants for anoxic survival. We identified a total of 71 fitness determinants for anoxic survival, which include genes encoding regulatory proteins such as Anr and the two-component regulatory system RoxS/RoxR. Deletion of both anr copies in B. cenocepacia H111 completely abolished the ability to survive in the absence of oxygen, and mutation of the roxS, the cytochrome bd oxidase cydA and the fliA mutant, significantly reduced survival relative to the wild-type by 10%, 25% and 40%, respectively. Furthermore, we show that RoxS/RoxR is important for the virulence of B. cenocepacia H111 and that the FliA sigma factor is essential for swarming and swimming

Identification of Potential Fitness Determinants for the Survival of B. cenocepacia H111 in Anoxic Condition
The genome of B. cenocepacia strain H111 consists of three replicons with a total of roughly 8 million base pairs encoding for approximately 6700 proteins [23]. Previously, our lab constructed a B. cenocepacia H111 Tn23 transposon mutant library [24], which was utilized in this study to determine fitness determinants important for anoxic survival of B. cenocepacia H111 in ABC minimal medium with and without nitrate, the latter to take the possibility of yet unidentified nitrate reductases in this organism into account. Aerobically grown B. cenocepacia H111 precultures were diluted in fresh ABC medium with and without nitrate and incubated for five days at 37 • C in the absence of oxygen, which was followed by an aerobic enrichment step to reach the cell densities required for Tn-seq analysis. Aerobic control samples were grown in ABC in the presence of oxygen for five generations ( Figure 1A). The presence of nitrate in the growth medium did not enhance the capacity of B. cenocepacia H111 to survive without oxygen ( Figure 1B).

Figure 1.
Overview of the preparation of the B. cenocepacia H111 samples used for Tn-seq analysis. (A) The B. cenocepacia Tn23 mutant library was used to inoculate the control sample (grown aerobically until an OD 600 of 0.5 at 37 • C in ABC) and the test samples (anoxically for 5 days at 37 • C in ABC or ABC with 10 mM NaNO 3 ). Control and test samples were prepared for Tn-seq analysis, as described in Material and Methods. Different bacterial mutant strains are indicated with different colors. (B) Colony forming units (CFU) mL −1 at the start and end of the anoxic survival in ABC and ABC with 10 mM NaNO 3 for B. cenocepacia H111 wild-type. Error bar = standard deviation (SD), ns = not significant, unpaired t-test (two-tailed).
After sequencing and mapping the transposons sites against the B. cenocepacia H111 genome, a fitness analysis was performed, in which the transposon insertion abundance in the control sample (aerobic in ABC medium) was compared to the transposon insertion abundance of the anoxic test sample (anoxic in ABC with or without nitrate). If the insertion frequency significantly decreases in the anoxically grown test sample compared to the aerobic sample, the gene is important under the experimental conditions and is considered a potential fitness determinant. For the aerobic control sample and the anoxic test sample in ABC medium, we obtained approximately one million unique insertions counts (UIC), whereas, for the anoxic sample in ABC with nitrate, 840,000 UICs were mapped (Table 1). When applying the thresholds as detailed in the Material and Methods, we identified a total of 61 and 30 potential fitness determinants for the anoxic survival of B. cenocepacia H111 in ABC and ABC with nitrate, respectively ( Table 1). Comparison of the two anoxic data sets showed that 20 determinants were shared between both conditions, whereas 41 were unique for anoxic survival in ABC and 10 were only found to be important for the anoxic survival in ABC with nitrate, resulting in a total of 71 anoxic fitness determinants (Figure 2A, Tables S2 and S3).  The anoxic fitness determinants identified in ABC and ABC with nitrate were classified into a cluster of orthologues groups (COG) by using the software EggNOG 4.5.1 [25]. After performing the Fisher tests, an overrepresentation of genes belonging to the COG category N (cell motility) was observed among the genes important for anoxic survival ( Figure S1). Most of the 20 shared fitness determinants, between both anoxic conditions ( Figure 2A and Table 2), were either assigned to the categories of cell motility (COG: N), transcription (COG: K) or signal transduction (COG: T). Strikingly, 7 out of the 20 shared anoxic fitness determinants were annotated as regulatory genes (COG K or T). For example, I35_RS16810 is an ortholog of B. cenocepacia J2315 BCAM0049, which codes for a Crp-Fnr family transcriptional regulator. Additionally, a twocomponent regulatory system consisting of the two genes roxR (I35_RS15820), coding for the response regulator, and roxS (I35_RS15825), encoding the sensor kinase, were identified. Another regulatory gene found in the list of the 20 shared anoxic determinants was fliA (I35_RS00760), encoding a sigma factor orthologous to the B. cenocepacia J2315 BCAL0144, the flagellar biosynthesis sigma factor 28. Besides fliA, the gene flhC (I35_RS00665) coding for another transcriptional regulator was identified. FlhC forms a complex with FlhD (FlhDC) and acts as a positive transcriptional activator of the fliA gene expression [26]. However, flhD (I35_RS00660) was not identified as a shared fitness determinant for the anoxic survival of B. cenocepacia H111 but as a potential fitness determinant under anoxia in ABC with nitrate. A gene located upstream of the Crp-Fnr regulator encodes a universal stress protein I35_RS16815, which appears to be important in both anoxic conditions.   3 The log 2 fold change (FC) of the normalized UID (nUID) of the control sample (ABC aerobic) and the test sample (ABC or ABC with 10 mM NaNO 3 in anoxic condition). 4 Gene name from B. cenocepacia J2315 [27] or through homology search (anr and rox system).
We also identified genes unique for survival in ABC (41 genes, Figure 2A, Table S2) or ABC with nitrate (10 genes, Figure 2A). Interestingly, 3 out of the 10 genes only found in ABC with nitrate are part of the same operon (COG C, Figure 2B, Table S3). The cluster (I35_RS14430-I30_RS14440) contains two genes coding for subunits of a cytochrome bd oxidase (cydA: I35_RS14435 and cydB: I35_RS14440), which are preceded by the gene coding for the putative small membrane protein CydP (I35_RS14430) (Figure 3). Among the anoxic fitness determinants identified only in the absence of nitrate, we found an AraC family transcriptional regulator (I35_RS17480) and a formyltetrahydrofolate deformylase (purU, I35_RS14250), which catalyzes the hydrolysis of 10-formyltetrahydrofolate to formate and tetrahydrofolate in the purine metabolism [28].

Anr Is Essential for the Survival of B. cenocepacia H111 under Oxygen Depletion
We identified the regulatory gene I35_RS16810 (coding for a Crp-Fnr family transcriptional regulator) as a potential fitness determinant for the anoxic survival of B. cenocepacia H111 in the presence and absence of nitrate. A protein sequence homology search revealed that I35_RS16810 has a sequence identity of 43% to the P. aeruginosa PAO1 Anr (PA1544). A search into the B. cenocepacia H111 genome revealed that I35_RS16810 is paralogous to I35_RS23120, which is also located on chromosome 2. Alignment of the two regulatory proteins showed a 97% coverage and 53% identity [29]. From here on, we refer to I35_RS16810 as anr 2 and to I35_RS23120 as anr 1 . Despite their high sequence conservation, only anr 2 was identified by our Tn-seq analysis ( Table 2).
In order to further assess the importance of anr 2 for anoxic survival, we constructed a deletion mutant strain (∆anr 2 ) and tested it for anoxic survival. The ∆anr 2 strain was still able to survive but to a significantly lower extent compared to the B. cenocepacia H111 wild-type in ABC (55% survival relative to the wild-type) and in ABC with nitrate (64% survival relative to the wild-type) ( Figure 4A,B). When both anr regulators (∆anr 1 -anr 2 ) were deleted, the cells were unable to survive anoxic conditions ( Figure 4A,B). However, the ∆anr 1 -anr 2 double mutant was not affected in growth under aerobic and micro-oxic conditions in minimal and rich media relative to the wild-type strain (data not shown). . Y-axis shows the normalized survival in % relative to the wild-type survival according to the start and end number of CFU. Three biological replicates (n = 3) were performed for each survival experiment. The SD is shown by the error bars. A non-matching paired one-way ANOVA (Dunnett) was performed with confidence interval of 99%. *** = p-value < 0.001. ** = p-value < 0.01.
As mentioned above, both components of a two-component regulatory system (roxR and roxS) were identified as fitness determinants for survival under anoxic conditions ( Table 2). A homology search revealed that I35_RS15825 and I35_RS15820 of B. cenocepacia H111 shared 34% and 52% of the protein identity with RoxS (PA4494; 94% coverage) and RoxR (PA4493, 98% coverage) of P. aeruginosa PAO1, respectively. Accordingly, we named the I35_RS15820-25 B. cenocepacia two-component system RoxR/RoxS. To validate the Tnseq results, a ∆roxS mutant was constructed and tested for survival under anoxic conditions. Viability of the ∆roxS mutant was reduced by 91% and 95% relative to the wild-type in ABC and ABC with nitrate, respectively ( Figure 4A,B).
The cluster encoding a cytochrome bd oxidase (cydAB, Figure 3) was found to be important for anaerobic survival in the presence of nitrate (Figure 2A). The subunits CydA and CydB of B. cenocepacia H111 are highly homologous to the corresponding subunits of the E. coli K-12 cytochrome bd oxidase (CydA: 98% coverage, 68% identity and CydB: 100% coverage and 54% identity) [30]. However, the cytochrome bd oxidase of E. coli contains an additional and essential subunit called CydX [31], which was not annotated in the B. cenocepacia H111 reference genome. However, when the E. coli CydX sequence was blasted against the B. cenocepacia H111 genome, a cydX homolog was identified in the cytochrome bd oxidase operon ( Figure 3). Interestingly, the gene coding for the small membrane protein (I35_RS14430), identified as a fitness determinant by Tn-seq, that is upstream of cydA was not found in the cydABX operon of E. coli or the cioAB operon of P. aeruginosa. A BLAST search showed that this gene is highly conserved within the order of Burkholderiales but not found outside of it. We name it here cydP as it is a putative subunit of the B. cenocepacia cytochrome bd oxidase. Next, we tested if the cytochrome bd oxidase is essential for survival under anoxic conditions by creating a deletion mutant of the cytochrome bd oxidase subunit CydA (∆cydA). Relative to the wild-type, only 35% of the ∆cydA mutant cells survived in ABC and 21% in ABC with nitrate ( Figure 4).
For validation of the importance of the gene encoding the FliA sigma factor for anoxic survival, an insertional mutant was constructed. The fliA insertional mutant showed reduced anoxic survival (35% and 45% relative to the wild-type in the absence or presence of nitrate, respectively) ( Figure 4).

RoxS Is Required for Pathogenicity in a Galleria Mellonella Infection Model
Since anoxic survival is a strategy used by several opportunistic pathogens to establish chronic infections in their host, the fitness determinants identified in this study were tested for their virulence in a G. mellonella infection model [32,33]. G. mellonella larvae were injected with approximately 2 × 10 5 cells, and their ability to survive was monitored over 72 h ( Figure 5). The ∆roxS mutant was most strongly attenuated (60% survival), followed by the ∆cydA mutant (36% survival), ∆anr 1 -anr 2 double mutant (33% survival) and the anr 2 mutant (20% survival) ( Figure 5). By contrast, the fliA mutant killed G. mellonella as efficiently as the wild-type ( Figure 5). I35_RS00760, identified here as a potential fitness determinant for the anoxic survival of B. cenocepacia H111, encodes a protein that is similar to the P. aeruginosa PAO1 sigma factor FliA (PA1455) (47% identity and 59% coverage). In P. aeruginosa, FliA is involved in the regulation of the biosynthesis of flagella [34], and recent studies suggested that FliA is not only involved in motility but also in virulence [35]. B. cenocepacia H111 fliA is located within a cluster of genes required for flagella biosynthesis (flhABGF, Figure 3) [27]. FliA mutants of P. aeruginosa were reported to be non-flagellated and consequently are nonmotile [35]. Transmission electron microscopy was performed to investigate flagellation of the wild-type, the fliA mutant and the fliA complemented strain (fliA + ) ( Figure 6). Whereas B. cenocepacia H111 wild-type and the complemented fliA + strain possessed one flagellum on the pole (monopolar), the mutant was non-flagellated.
The fliA mutant showed reduced swimming ability in all tested conditions (aerobic, micro-oxic, with and without nitrate) compared to the wild-type strain. This defect was restored in the fliA + complemented strain (Figure 7).
The B. cenocepacia H111 fliA mutant was also not able to swarm in any of the tested conditions (aerobic, in micro-oxic condition, with and without nitrate). However, the swarming phenotype was rescued in the fliA + complemented strain (Figure 8). However, during the characterization of the fliA mutant strain, we realized that the swarming of B. cenocepacia H111 wild-type was (i) inhibited by the presence of nitrate in aerobic and micro-oxic conditions and (ii) increased in micro-oxic conditions (Figure 8).   We also tested the strains for biofilm formation in the ABC medium in the presence and absence of nitrate. We observed a 26% reduction in biofilm formation with the fliA mutant (unpaired t-test, p-value < 0.0001) compared to the wild-type in the presence of nitrate, whereas in the absence of nitrate, biofilm formation in the fliA mutant was reduced by 16% relative to the wild-type (unpaired t-test, p-value < 0.0194) ( Figure S2). The defect in the biofilm was restored to wild-type levels in the complemented strain fliA + ( Figure S2).

Discussion
In the absence of oxygen, energy can be generated by either fermentation or reduction of alternative electron acceptors such as nitrate and sulfate. In the genus Burkholderia, only specific strains from the Bpc are able to grow in the absence of oxygen by performing denitrification, i.e., reducing nitrate into dinitrogen [21], and most other Burkholderia species are considered obligate aerobes [22]. B. cenocepacia H111 is able to grow at a low oxygen concentration of 0.1%, and we show here that it is able to generate enough energy to survive in the absence of oxygen, an environment that can be found in deep biofilm layers in the lungs of CF patients [18]. How B. cenocepacia H111 is able to survive without oxygen is not understood; however, it was shown that the strain is unable to ferment arginine or reduce nitrate to nitrite or dinitrogen for energy production [18,22]. In this study, we applied Tnseq to identify potential fitness determinants for the anoxic survival of B. cenocepacia H111. Nitrate was used in order to accompany any non-annotated nitrate reductases. Twentyeight of the seventy-one identified genes were reported to be up-regulated under low oxygen conditions [22], and ten belong to the shared genetic determinants for the anoxic survival identified in this study, including anr 2 , fliA and genes encoding flagella components (i.e., flhC, motB, flgG, flgF). Among the genes required for anoxic survival in the absence of nitrate was the gene coding for the universal stress protein A (uspA, I35_RS16815), which is located upstream of anr 2 ( Figure 3) and was shown to be highly up-regulated in low oxygen conditions in B. cenocepacia J2315 [22]. Interestingly, the lxa locus of B. cenocepacia encodes six universal stress proteins that are almost exclusively up-regulated in response to low oxygen [22]. Furthermore, the analysis of sequential B. cenocepacia isolates from two adult male siblings with CF revealed that proteins encoded on the lxa locus, including the six USPs, are consistently up-regulated across both sets of patient isolates over time, suggesting that the uspA genes are important in vivo and likely to be important for niche adaptation to the hypoxic CF lung [36]. Some of the genes (anr 2 , roxS/roxR and fliA) identified in this study were also shown to be up-regulated in micro-oxic conditions (0.5% O 2 ) using a combined transcriptomics and proteomics approach [18]. An anr 2 homolog was also identified in the study of Lieberman et al., who sequenced the genomes of 114 Burkholderia dolosa strains isolated from 14 CF patients over 16 years. The large number of mutations in the homolog and other oxygen-sensing regulators indicate that these genes are under selective pressure and are potentially important for chronic infection of B. dolosa, another Bcc member [37]. An anr 2 ortholog also exists in B. thailandensis E254 and was shown to be required for growth under denitrifying conditions [21]. Interestingly, although the anr 2 paralog anr 1 was not detected in our Tn-seq analysis, an anr 1 mutant showed impaired anaerobic survival relative to the wild-type (data not shown). Given that the anr 2 anr 1 double mutant is less viable than the single mutants in the absence of oxygen, the two anr genes may be functionally redundant. The reason that only anr 2 provides a fitness benefit is currently unknown, and further investigations are required. In B. dolosa, the FixL/J two-component system is activated under low oxygen conditions. It was shown that the FixL senses the oxygen tension and activates the response regulator FixJ, which induces transcription of fixK [38], which is a homolog of anr 2 . Interestingly, B. cenocepacia H111 has genes coding for a homologous FixL/J system (I35_RS10705-I35_RS10710), which showed lower transposon insertion frequencies (but was below the threshold value for our Tn-seq fitness gene list) under anoxic conditions, suggesting a potential role of this two-component system in anaerobic survival. Many bacteria can adapt their metabolism to low oxygen tension by expressing terminal oxidases with high oxygen affinities [39,40]. Many bacteria express B-and C-class heme copper oxidases such as cytochrome cbb 3 or ba 3 oxidases as high (oxygen) affinity terminal oxidases [41][42][43][44]. However, these enzymes are absent in Burkholderia [45], raising the question of which oxidases are used under low oxygen tensions. Transcriptomic studies in B. cenocepacia found that the cytochrome bd oxidase is overexpressed under micro-oxic conditions [18,22], indicating that the cytochrome bd oxidase is used for survival at low oxygen tensions. Here, we identified the cytochrome bd oxidase genes cydP, cydA and cydB as fitness determinants for anoxic survival and a ∆cydA mutant showed significantly lowered survival under anoxic conditions. Together with the transcriptomic studies, our results highlight the importance of the bd oxidase for the adaption of B. cenocepacia to micro-oxic and anoxic conditions. Such conditions are often found in host-associated situations such as CF, and indeed, we could show that the ∆cydA mutant decreases the pathogenicity of B. cenocepacia H111 in G. mellonella. Since both the cytochrome bd oxidase cluster and the anr 2 gene were found to be involved in anoxic survival, we looked at potential binding boxes for Anr in the upstream region of the cytochrome bd oxidase cluster. Indeed, we were able to find a potential consensus sequence (TTGATCTCGATCAA) [46] in the promoter region, suggesting direct control of the cytochrome bd oxidase (I35_RS14430-14440) by Anr ( Figure S3). Given the fact that Anr is known to positively regulate the cbb 3 oxidase-2 of P. aeruginosa [41], which is overexpressed under low oxygen conditions, it is possible that Anr has a similar regulatory role in the regulation of cytochrome bd oxidase expression in B. cenocepacia. Further studies are needed to elucidate the role of cytochrome bd oxidase in the presence of nitrate. We performed a nitrate reduction test, which showed that nitrate is still present after five days of anoxic incubation, and no nitrite was produced (data not shown), further corroborating the fact that B. cenocepacia H111 is unable to reduce nitrate to nitrite. Another regulatory system identified in this study is the two-component system RoxS/RoxR, which is homologous to RoxS/RoxR of B. thailandensis E264 [21] and P. aeruginosa PAO1 [29]. In B. thailandensis, the roxS mutant was unable to grow in the presence of nitrite under denitrifying conditions and showed an accumulation of N 2 O, indicating an involvement of this two-component system in the regulation of nitrite reduction [21]. The RoxS/RoxR system in P. aeruginosa positively regulates the expression of the cyanide insensitive oxidase genes [20]. The cytochrome bd oxidase cluster identified in our studies (I35_RS14430-14440) possesses a potential RoxR consensus sequence (GCGGCAATTTAGAGC) [47] in the promoter region ( Figure S3). This suggests that in addition to Anr, RoxS/RoxR may directly control the expression of the cytochrome bd oxidase in B. cenocepacia H111. In B. pseudomallei, the RegA/RegB (homologous to RoxS/RoxR) is essential for anoxic growth in the presence of nitrate. Additionally, it was shown that this two-component system is also required for full B. pseudomallei virulence in the murine infection model [48]. In line with this observation, we observed reduced virulence of the ∆roxS mutant in B. cenocepacia H111 using a G. mellonella infection model, suggesting that this two-component system is controlling the virulence of B. cenocepacia H111. Bernier et al. showed that the RoxS/RoxR system is also involved in the resistance of B. cenocepacia K56-2 towards phenazines produced by P. aeruginosa [49]. The identification of flagella biosynthesis sigma factor fliA as a fitness determinant for anoxic survival may be due to the importance of flagella in swimming toward the surface where the possibility to encounter oxygen is higher. Despite the fact that the FliA regulator of P. aeruginosa plays a role in virulence by influencing colonization ability and hemolytic activity, the absence of the FliA regulator did not affect B. cenocepacia H111 virulence in G. mellonella [35]. During our experimental work, we observed an induction of swarming when B. cenocepacia H111 wild-type was exposed to micro-oxia, and this induction was independent from the regulators we investigated in this study (Anr, RoxR/S, FliA). We further observed that nitrate inhibits swarming under aerobic conditions and in micro-oxic conditions. However, nitrate has no negative or positive effect on the swimming ability of B. cenocepacia H111 in aerobic and micro-oxic conditions. In P. aeruginosa, nitrate has no effect on swarming [50]. We also identified the flagellar master operon flhDC to be important for anoxic survival. Studies in E. coli K-12 showed that loss of FlhD altered not only the expression of genes involved in flagella biosynthesis but also influenced anoxic respiration by regulating the expression of the oxygen sensor Aer, the nap gene cluster (periplasmatic nitrate reductase) and a cluster encoding for a nitrite reductase (nrfABCDE). Because FlhD/FlhC also negatively regulates the genes involved in aerobic respiration, such as cytochrome bo 3 ubiquinol oxidase (cyoBCDE) or succinate dehydrogenase (sdhCDAB), the FlhD/FlhC complex may play a role as a switch in the transition of aerobic to anoxic respiration [51]. An important survival mechanism of P. aeruginosa under anoxic conditions is the production of excreted phenazines allowing bacteria to generate ATP by reducing phenazines and re-oxidizing them far away by oxygen or other oxidants [12]. Some strains of the Bcc, such as Burkholderia lata ATCC 17760, contain the core phenazine biosynthesis genes encoded in a conserved seven gene operon (phzABCDEFG) [52], and phenazine production has been reported for B. cepacia strain 5.5B and B. cenocepacia strain K56-2 [53,54]. B. cepacia is also known to produce other redox-active compounds, such as pyrrolnitrin [55], which might have similar functions as phenazines. Therefore, it is possible that B. cenocepacia H111 might employ phenazine or phenazine-like compounds for survival under anoxic conditions, and this is an interesting question for future studies.

Bacterial Strains, Media and Growth Conditions
Bacterial strains, oligonucleotides and plasmids used throughout this study are listed in Table S1. Bacterial cultures for routine culture were grown either in lysogeny broth (LB) [56] or AB medium [57]

Tn-seq Technology
The Tn5-based transposon library of B. cenocepacia H111 used in this study contains approximately 1,000,000 unique insertion mutants and was previously constructed and used by our group [24]. A washed preculture of the B. cenocepacia H111 Tn23 library was inoculated in ABC (OD 600 of 0.02, corresponding to roughly 10 7 CFUs per ml) supplemented with 0.2% L-rhamnose and left to grow until the end of the exponential phase (OD 600 = 0.8). This preculture was then used to inoculate a 100 mL flask containing 25 mL ABC with 0.2% L-rhamnose to a starting OD 600 of 0.02, and the culture was grown until mid-exponential phase (OD 600 = 0.5) (control sample, approximately 5 generations). For the test samples (ABC or ABC with 10 mM NaNO 3, both amended with 0.2% L-rhamnose), 6 times 6 mL in 14 mL capped glass tubes were inoculated with the same preculture to the same starting OD 600 as the control sample. The tubes were anoxically incubated in a jar for 5 days. Thereafter, the six tubes for each condition were combined in a 100 mL sterile flask and grown until OD 600 0.5 under aerobic conditions (220 rpm, 37 • C). The cells were then pelleted and stored at −20 • C. The genomic DNA from the control and the test samples was extracted using the GeneElute TM Bacterial Genomic DNA kit (Sigma-Aldrich, St. Louis, MO, USA, NA2110) and prepared for Illumina sequencing using the circle method [59], as described by [21]. In brief, the genomic DNA was sheared by using Covaris (E220 focused ultrasonicator, Covaris TM , MA, USA) into 300 bp long fragments. The NEBNext Ultra II DNA Library Prep Kit (New England Biolabs (NEB), Ipswich, MA, USA, E7645S) was used for end repair. Adaptors were ligated, followed by a restriction digestion at the BamHI site within the transposon. Circularization followed by exonuclease treatment with ExoI (NEB, M0293L), T7 Gene 6 Exo (NEB, M0263S) and Lambda Exo (NEB, M0262S) occurred with size selected DNA fragments. DNA was sequenced (Miseq, Ilumina, San Diego, CA, USA) by using the paired end (2 times 150 bp) Miseq Reagent kit V2 using 300 cycles (Ilumina, San Diego, CA, USA, MS-102-2002). Analysis of the sequenced DNA libraries and a fitness analysis was performed as previously described [21]. In brief, the open-source software Tn-seq Explorer [60] was used to calculate the unique insertion density (UID), which is the unique insertion counts (UIC) of the transposon in a gene divided by its gene length in base pairs to normalize for gene sizes. Thereafter, the UID was normalized (nUID) by dividing the UID by the total UIC per sample (Table 1). After the fitness analysis, a list of potential fitness determinants was obtained by using the following parameters: log 2 fold change (FC) of at least −1 (FC = nUID of the test sample divided by the nUID of the control sample) and a difference in nUID between the control and the test samples of at least 0.009. Those parameters are strict but ensure a restricted list of fitness determinants with increased likelihood to show a strong fitness effect in anoxic conditions. Clusters of Orthologous Groups (COG) were added to the fitness list by using the online tool EggNog version 4.5.1 [25]. The heatmap was created in Rstudio (PBC, Boston, MA, USA) version 4.0.3 using the following packages: readxl and pheatmap with the command cluster_rows = TRUE.

Construction of B. cenocepacia H111 Mutant Strains
To validate the Tn-seq data, three deletion mutants of gene anr 2 (I35_RS16810), cydA (I35_14435) and roxS (I35_RS15825) were generated according to the deletion mutagenesis system established by Flannagan and colleagues [61]. The genomic DNA of B. cenocepacia H111 wild-type as PCR template was extracted by using the GenElute TM Bacterial Genomic DNA Kit. PCRs for the up-and downstream regions were carried out by using HF phusion polymerase (Sigma, St. Louis, MO, USA, F-530L). The following primer pairs were used for the upstream flanking site amplification of the above mentioned genes: anr 2 up_Fw_KpnI and anr 2 up_Rv_XbaI (size: 654 bp), cydA_Rv_up_NdeI and cydA_Fw_up_XbaI (size: 647 bp), roxSup_Fw_KpnI and roxSup_Rv_XbaI (size: 581 bp). For the downstream regions the primers anr 2 dw_Rv_KpnI and anr 2 dw_Fw_EcoRI (size: 574 bp), cydA_Fw_dw_NdeI and cydA_Rv_dw_EcorRI (size: 615 bp), roxSdw_Fw_EcoRI and roxSdw_Rv_KpnI (size: 505 bp) were used. The upstream and downstream fragments were ligated into the suicide plasmid, pGPI-SceI, which was previously extracted by using the following kit: QIAprep Spin Miniprep Kit (Qiagen, Hilden, Germany, Ref: 27106). The ligation mixture was transformed into competent E. coli SY327 λpir cells, and the transformants were plated on selective plates. The plasmids (pGPI-SceI_anr 2 updw, pGPI-SceI_cydAupdw, pGPI-SceI_roxSupdw) carrying up-and downstream regions were integrated into B. cenocepacia H111 wild-type via triparental mating (TPM). Double strand break at the I-SceI endonuclease site was initiated by introducing the plasmid pDAI-SceI E. coli pDAI-SceI S17λpir via biparental mating with the transconjugants. The double mutant of the anr genes was made in three steps. First, a deletion mutant of anr 1 (I35_RS23120) was constructed by using the Gateway Technology (Invitrogen, Waltham, MA, USA). In brief, the 5 and 3 regions of anr 1 were amplified by using primers pairs: anr 1 up_F and anr 1 up_R, and anr 1 dw_F and anr 1 dw_R. The kanamycin resistance cassette (KRC) was amplified from the plasmid pKD4 with the primers pKD4rev-2 and pKD4fwd-2. These three PCR products were annealed by using Ex Taq Polymerase (Takara Bio, San Jose, CA, USA), and then amplified with the primer set GWattB1 and GWattB2 by PCR. The resultant fragment consisting of 5 region, KRC and 3 region was cloned into pDONR221 using the BP clonase II kit (Invitrogen, Thermo Fisher Scientific, Waltham, MA, USA), which was then transformed into competent E. coli Top 10 cells. LR clonase kit II (Invitrogen, Thermo Fisher Scientific, MA, USA) was used to genetically transfer the insert from pDONR221 into the suicide vector pAUC40, which was transformed into E. coli cc118λpir. The mutagenesis targeting pAUC40_anr 1 updw plasmid was transferred into B. cenocepacia H111 by TPM to generate the anr1 deletion mutant. To mutate anr 2 (I35_RS16810), a 259 bp region of anr 2 was first amplified by using primers anr 2 _IM_F and anr 2 _IM_R, which was subsequently ligated directly into pGEM T easy vector (Promega, Madison, WI, USA). The fragment was released by digestion with XhoI and BglII and then subcloned into cut pSHAFT2 [21] to generate the anr 2 insertional mutagenesis construct. Plasmid pSHAFT2_anr 2 was conjugated into the wild-type and the deletion mutant anr 1 by TPM. The fliA (I35_RS00760) mutant was created by insertional mutagenesis using the suicide plasmid pSHAFT2. An internal part of fliA was amplified with the following primers fliA_Fw_EcoRI and fliA_Rv_XbaI (size: 238 bp), ligated into pSHAFT2 and later transformed into E. coli cc118λpir. The correct plasmid was called pSHAFT2_fliA and brought into the B. cenocepacia H111 wild-type genome by TPM. Complementation of the fliA was obtained by amplifying the entire gene using the primers fliA_comp_Rv_XbaI and fliA_comp_Fw_EcoRI (size: 749 bp) and inserting it into the plasmid pBBR1MSC-5. The plasmid pBBR1MSC-5_fliA with the entire gene was transferred into the fliA mutant via TPM resulting in the complemented strains fliA + .

Phenotypical Analyses
For the anoxic survival experiment, bacterial cultures were adjusted to a starting OD 600 of 0.02 in ABC or ABC supplemented with 10 mM NaNO 3 in 14 mL capped glass tubes to a final volume of 6 mL (approximately 2.5 × 10 7 cells per mL). The adjusted cultures were incubated at 37 • C statically in anoxic condition for 5 days as described above. Colony forming units were determined on LB plates by serial dilutions in 0.9% NaCl before and after 5 days of anoxic incubation. The normalized survival percentage relative to the wild-type was calculated by dividing the CFU pro mL after the anoxic experiment for each tested strain by the average of CFU pro ml at the beginning of the inoculum, multiplied by 100. Swarming and swimming were performed as previously described [18,62]. In brief, adjusted cultures (OD 600 = 0.5) were spotted onto 0.4% or 0.2% ABC or ABC with 10 mM NaNO 3 agar plates supplemented with 0.1% Bacto TM casamino acids (BD, Waltham, MA, USA) and incubated at 37 • C for three days or 24 h either in aerobic or in micro-oxic conditions. For the Transmission Electron Microscopy (TEM), the overnight culture of bacteria was ten-fold diluted in saline solution. The 20 µL of bacterial suspension was placed on a glow-discharged formvar-coated 300-mesh copper grid (Plano GmbH, Wetzlar, Germany). After 1 min, the excess liquid was removed using filter paper. The grid was negatively stained with 1% uranyl acetate for 1 min and dried with filter paper in a glass petri dish for 30 min. The sample was visualized using an FEI Tecnai G2 Spirit TEM (FEI, Hillsboro, OR, USA) at 120 kV acceleration voltage with the detector side-mounted digital camera Gatan Orius 1000 (4 k × 2.6 k pixels). Biofilm formation was performed as previously reported [58] in ABC medium or in ABC amended with 10 mM NaNO 3 . Wild-type and mutants were stained with crystal violet for 30 min. After washing the 96-well plates with deionized water, the attached crystal violet was solubilized by the addition of 120 µL of DMSO (Sigma, St. Louis, MO, USA) and left for 20 min at room temperature (RT). Biofilm formation was measured by a plate reader (TECAN, Infinite M200 pro, tecan biotechnology and life science, Taunton, MA, USA), which was measured at 570 nm.

Pathogenicity Experiments Using Galleria mellonella
G. mellonella larvae were purchased from BioSystems Technology (TruLarv, Exeter, UK) or the fishershop in Zurich (Reptile-Food.ch) and used immediately after arrival or kept at 16 • C for a maximum of 2 days. The infection assay was carried out as described previously with the following modifications [33]. The cultures were diluted to an OD 600 of 0.0625 (± 2-3 × 10 7 CFU per mL) in 10 mM MgSO 4 , and thereafter, 10 µL were injected into G. mellonella (approximately 200,000 cells) with a 27 gauge needle (BD Microlance, Waltham, MA, USA, Ref: 302,200). Three or more biological replicates were performed with 10 larvae for each sample. The statuses of the larvae were checked to be alive (responsive) or dead (non-responsive) every 24 h.

Statistical Analyses
GraphPad prism version 5.01 (San Diego, CA, USA) was used to analyze the data obtained during this study. One-way ANOVA was used (Dunnett) with a confidence interval of 99%. For Biofilm formation, unpaired t-test was used.