Notice: Wiley Online Library will be unavailable on Saturday 30th July 2016 from 08:00-11:00 BST / 03:00-06:00 EST / 15:00-18:00 SGT for essential maintenance. Apologies for the inconvenience.
The anaerobic metabolism of the opportunistic pathogen Pseudomonas aeruginosa is important for growth and biofilm formation during persistent infections. The two Fnr-type transcription factors Anr and Dnr regulate different parts of the underlying network in response to oxygen tension and NO. Little is known about all members of the Anr and Dnr regulons and the mediated immediate response to oxygen depletion. Comprehensive transcriptome and bioinformatics analyses in combination with a limited proteome analyses were used for the investigation of the P. aeruginosa response to an immediate oxygen depletion and for definition of the corresponding Anr and Dnr regulons. We observed at first the activation of fermentative pathways for immediate energy generation followed by induction of alternative respiratory chains. A solid position weight matrix model was deduced from the experimentally identified Anr boxes and used for identification of 170 putative Anr boxes in potential P. aeruginosa promoter regions. The combination with the experimental data unambiguously identified 130 new members for the Anr and Dnr regulons. The basis for the understanding of two regulons of P. aeruginosa central to biofilm formation and infection is now defined.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
In the absence of oxygen and the presence of nitrate or nitrite P. aeruginosa is able to grow by denitrification (Zumft, 1997). Moreover, it performs an arginine fermentation to produce energy under anaerobic conditions in the absence of suitable alternative electron acceptors (Vander Wauven et al., 1984). We showed that P. aeruginosa also performs pyruvate fermentation, which does not allow anaerobic growth but is essential for long-term anaerobic survival (Eschbach et al., 2004).
Recently, the regulatory network underlying the onset of nitrate respiration was elucidated (Sharma et al., 2006; Schreiber et al., 2007). In this process nitrate is reduced to dinitrogen via four consecutive steps with nitrite (NO2-), nitric oxide (NO) and nitrous oxide (N2O) as intermediates. The regulators Anr, Dnr, NarL in concert with IHF activate transcription of the narK1K2GHJI operon encoding nitrate reductase and two transporters in response to oxygen limitation, nitrate and N-oxides (Schreiber et al., 2007). Expression of the nitrate responsive two-component regulatory system genes narXL itself is enhanced under anaerobic conditions by Anr and Dnr (Schreiber et al., 2007). Anr in turn activates transcription of dnr (Arai et al., 1997).
Dnr in turn activates transcription of the other denitrification genes like nirS, norCB and nosZ (Arai et al., 1997; 2003). Dnr carries a moderate amino acid sequence similarity of 25% to Anr and is presumed to detect NO (Arai et al., 1999). Recently, the structure of the sensory domain and the complete Dnr protein were published (Giardina et al., 2008; 2009). Surprisingly, the sensory domain was found significantly rotated compared with the orientation of this domain in other Fnr/Crp family regulators. Dnr was shown to act via DNA binding sites indistinguishable to those described for Anr. A recent bioinformatics investigation of Fnr/Crp regulatory proteins and their binding sequence also failed to detect differences between the P. aeruginosa Anr and Dnr binding sites (Rodionov et al., 2005). In accordance with previous experimental investigations the nir, nor and nos genes were proposed as Dnr regulon members. An Anr regulon was not proposed. Additionally, genes regulated by both Anr and Dnr via one Anr box were described (Rompf et al., 1998; Schreiber et al., 2007).
A number of recent papers have been published presenting transcriptome and proteome data of P. aeruginosa cells under anaerobic conditions and during biofilm growth (Wagner et al., 2004; Filiatrault et al., 2005; Waite et al., 2005; Wu et al., 2005; Schreiber et al., 2006; Alvarez-Ortega and Harwood, 2007; Platt et al., 2008). However, no information is available on the Anr- and Dnr-dependent regulatory networks in P. aeruginosa. Therefore, we initiated a whole genome approach using Affymetrix DNA microarrays, a limited proteomic analysis via two-dimensional gel electrophoresis in combination with mass spectrometry and regulatory binding site prediction by the use of bioinformatics tools. The aim of this study was the elucidation of the Anr- and Dnr-dependent regulons.
Results and discussion
Experimental rationale for the discrimination of the Anr- and Dnr-dependent anaerobic regulons
The Anr and Dnr regulatory proteins share an indistinguishable DNA consensus binding site, but activate different promoters with high specificity. Although the nirS (5′-TTGAT-N4-GTCAA-3′) and arcD (5′-TTGAC-N4-ATCAG-3′) promoters contain almost identical Anr boxes, nirS is exclusively activated by Dnr while the arcD locus is solely activated by Anr (Gamper et al., 1991; Arai et al., 2003). Furthermore, Anr and Dnr form a regulatory cascade in which Anr activates transcription of the dnr gene (Arai et al., 1997) (Fig. 1A). As a consequence an anr mutant strain phenotypically behaves like an anr-dnr double mutant (Fig. 1C). To distinguish between Anr- and Dnr-dependent genes we compared gene expression in anr and dnr mutant strains with that in the wild type strain (Fig. 1A–E). A typical Dnr-dependent gene such as nirS showed decreased expression in both mutants. A typical Anr-dependent gene such as arcD exhibited decreased expression solely in the anr mutant while expression levels in the dnr mutant strain were comparable to the wild type levels (Fig. 1D and E).
As both anr and dnr mutant strains do not grow anaerobically in the presence of nitrate, anaerobic shift experiments were applied to determine the transcriptome profile of both mutants (see Experimental procedures). To verify proper aerobic and anaerobic conditions and a functional Anr-Dnr cascade, expression of the Dnr-dependent nirSMC and the Anr-dependent arcDABC operon were monitored by Northern blot analysis.
Transferring the P. aeruginosa cells to a sealed serum bottle generates a strictly anaerobic environment within 5 min via the consumption of the residual oxygen through respiration (Eschbach et al., 2004). Efficient expression of Anr-dependent arcDABC was used as marker for anaerobic conditions and for functional Anr in the dnr mutant strain. Accordingly, we detected four different transcripts derived from the arcDABC operon after 2 h anaerobic incubation as described previously (Gamper et al., 1992), with the 2500 nt arcAB transcript yielding the strongest signal (Fig. 2). The 2450 nt nirSMC mRNA was detectable 30 min after anaerobic induction and reached a maximum expression after 2 h. The presence of the nirSMC mRNA confirmed proper anaerobic conditions as well as a functional Anr-Dnr cascade. Cells of the aerobic control culture of the shift experiments were grown to an OD578 of only 0.3 where no nirSMC and arcDABC transcripts were detectable (Fig. 2B). Therefore, this 2 h anaerobic incubation period was chosen for analysis of the Anr- and Dnr-dependent regulons. This condition was also determined suitable for the comparison of the anr mutant, the dnr mutant and the wild type as all investigated strains exhibited no anaerobic growth at this time point.
Gene expression in response to anaerobic growth conditions and growth arrest
At first, we compared the whole genome expression profile of an aerobically grown P. aeruginosa PAO1 wild type reference culture with the expression profile of a 2 h anaerobically incubated PAO1 culture containing 50 mM nitrate. This analysis resulted in 117 induced and 197 repressed genes under anaerobic conditions (Tables S1 and S2). When we compared the expression profile of the anr and dnr mutant strains with the wild type after the shift to anaerobic denitrifying conditions, we found no difference in gene expression regarding the repressed genes (data not shown). This observation indicates that Anr and Dnr are not directly involved in anaerobic gene repression in P. aeruginosa. In our experimental setup, the shift to anaerobiosis also resulted in growth arrest. The OD578 only increased from 0.35 to 0.38 within the 2 h anaerobic incubation period (Fig. 2A). During this period, P. aeruginosa adapts its biochemical machinery to generate energy under anaerobic conditions by denitrification and fermentation processes. After 7 h, cell growth resumes with a generation time of 2.3 h, which reflects exponential growth under anaerobic denitrifying conditions. Therefore, changes in gene expression after the anaerobic shift are not only a result of limited oxygen tension, but are also caused by stresses due to growth arrest. In accordance with this assumption 29% (59 genes) of the anaerobically repressed genes could be assigned to the functional classes of ‘translation’ and ‘transcription’ (Fig. 3). Genes of those categories encode various proteins of the 30S and 50S ribosome subunits, seven aminoacyl-tRNA synthetases as well as the α-subunit of the RNA polymerase (Table S2). Furthermore, after the anaerobic denitrifying shift experiment, 17 genes were upregulated more than fourfold independently of Anr or Dnr. Among these is PA3049 encoding a homologue of the E. coli ribosome modulation factor (Rmf, Table S1) (Wada et al., 1990). In E. coli this protein has been shown to generate ribosome dimers in response to growth arrest. The eightfold induced gene PA3126 encodes a protein with high similarity to the E. coli heat shock protein IbpA. While heat shock proteins are usually not produced in response to anaerobiosis in E. coli (Neidhardt et al., 1984), there are other examples including Bacillus subtilis known to induce heat shock proteins in response to oxygen limitation (Hecker et al., 1996). In the close relative Pseudomonas putida, IbpA is induced in response to toluene and other stresses, probably because the stress distorts the membranes and it is sensed as an oxygen deficiency condition (Segura et al., 2005; Dominguez-Cuevas et al., 2006; Volkers et al., 2006; Matuszewska et al., 2008). Transcription of the open reading frame PA0459, which encodes a putative Clp-type protease, was induced threefold. Enhanced protein degradation by Clp proteases and the biosynthesis of chaperones imply drastic changes in overall protein and corresponding amino acid metabolism. Accordingly, we detected genes of enzymes involved in amino acid metabolism anaerobically upregulated in an Anr/Dnr-independent manner including PA5495 (thrB, homoserine kinase), PA0266 (gabT, 4-aminobutyrate aminotransferase) and PA5304 (dadA, d-alanine dehydrogenase). Obviously, the observed changes in gene expression included not only the responses to anaerobiosis but also a response to growth arrest. Protein biosynthesis is the biosynthetic process consuming most of the energy under logarithmic growth conditions. Our data indicate a reduced protein biosynthesis due to downregulation of genes involved in translation and biosynthesis of amino acids.
Anaerobic gene expression in P. aeruginosa was recently investigated by a solid comprehensive transcriptome and proteome analyses (Platt et al., 2008). However, the employed growth conditions varied significantly from those used in this investigation. Platt and colleagues (2008) used strict anaerobically growing cultures that are already fully adapted and compared them with aerobically growing counterparts. In contrast, we used shift experiments from aerobic to anaerobic growth conditions to study the anaerobic transcriptome as bases for the investigation of anaerobically growth deficient anr and dnr mutants. Additionally, in this study a defined minimal medium was used instead of rich medium. In accordance with these drastic differences between the two studies (and as Platt et al. only considered a fold change above 30 in their microarray analysis) we found eight genes regulated similarly under anaerobic conditions (marked as ‘T+’ in Table S1 and ‘T-’ in Table S2) (Platt et al., 2008). The intensive proteome analysis of Platt et al. additionally identified 35 proteins differentially synthesized during anaerobic growth (Platt et al., 2008). We found the corresponding genes for 12 of these proteins differentially expressed by our transcriptome analysis (marked as ‘P+’ in Table S1 and ‘P-’ in Table S2). Consistent with results from Platt et al. the central genes or proteins required for denitrification (nir, nor, nos) and arginine fermentation (arcDABC) and genes such as fhp, adhA and katA were found highly induced during anaerobiosis (Platt et al., 2008). However, we failed to detect the nitrate reductase encoding genes (narGHI). A comparison of our data with those previously reported is given in Tables S1 and S2.
Anr is the major positive oxygen regulator of P. aeruginosa
Out of the 117 genes that were upregulated in the wild type, 52 were at least twofold less induced in the anr mutant strain compared with only 10 genes that were at least twofold less induced in the dnr mutant strain (Table 1). Clearly, Anr represents the general positive regulator of the anaerobic response in P. aeruginosa while Dnr only induces a limited set of genes. Out of 52 Anr-dependently induced genes 29 transcriptional units contained Anr binding sites in the corresponding putative promoter regions as determined by the Virtual Footprint Software (Münch et al., 2003; 2005) (Table S4). Fifteen of the twenty-nine Anr binding sites have been described in the literature before (see Tables S3 and S5). We also detected Anr binding sites in five putative promoter regions of anaerobically induced genes, which were not significantly downregulated in the anr mutant strains in our experiment, as narL and nosZ (see below), two genes encoding hypothetical proteins PA3880, PA4610 and one weakly conserved Anr binding site in the putative katA promoter region. However, a detailed inspection of gene expression revealed that all five genes showed some downregulation either above our cutoff value of 0.5-fold or with a slightly increased probability of differential expression (pde) value (shown in Tables S1 and S4). In addition, our proteome analysis (see below) confirmed an Anr-dependent anaerobic expression of katA.
Table 1. Numbers of Anr- and Dnr-dependent genes in Pseudomonas aeruginosa.
Anr- or Dnr-dependent genes include all genes that are differentially regulated in the respective mutant strains.
These genes contain an Anr box in their putative promoter regions; they belong to 29 transcriptional units (see Table S4).
No Anr box was identified in the promoter regions of these genes.
As discussed above, none of the downregulated genes was connected to Anr-dependent regulation. Nevertheless, there are few examples of genes in the literature, which are considered to be repressed by Anr, as the cioAB genes, encoding the cyanide insensitive oxidase (Cooper et al., 2003), whose expression is activated by cyanide via the regulatory protein RoxR (Comolli and Donohue, 2002). Our cultures did not contain cyanide or an activated cyanide biosynthesis operon (hcnABC), which requires high cell densities and quorum sensing for induction (Pessi and Haas, 2000), which would allow for induction of the cioAB operon. Therefore, we did not observe any Anr-dependent cioAB downregulation in our DNA microarrays. Nevertheless, our observation of a solely positive control mediated by Anr in response to anaerobic conditions is in clear contrast to the described functions of the E. coli counterpart Fnr, where repression of transcription is also mediated by Fnr in response to anaerobic conditions. However, the Fnr binding sites for transcriptional activation and repression are located in different regions of the promoter (Spiro, 1994; Guest et al., 1996; Salmon et al., 2003; Constantinidou et al., 2006). Nevertheless, the location of some predicted Anr binding sites in P. aeruginosa would also indicate repression of Anr (see below).
The Anr-dependent stress response to anaerobiosis
We detected several genes involved in stress response. Five genes of the known Anr regulon, uspL (PA1789), uspK (PA3309), uspM (PA4328), uspN (PA4352) and uspO (PA5027) encode members of the universal stress protein family (Boes et al., 2006; 2008; Schreiber et al., 2006). Usp-type proteins in E. coli like UspA or UspE are upregulated in response to growth arrest caused by a variety of different stress conditions (Kvint et al., 2003). Recently, we have shown that two Usp-encoding genes PA3309 (uspK) and PA4352 (uspN) are important for survival in the anaerobic stationary phase and during anaerobic long-term survival (Boes et al., 2006; Schreiber et al., 2006) and that the anaerobic expression of both usp genes is under the control of Anr. Moreover, we also showed Anr dependence of PA1789, PA4328 and PA5027 (Boes et al., 2008). However, the biological function of Usp-type stress proteins is still enigmatic.
We further detected a group of Anr-dependent genes induced during the aerobic to anaerobic transition phase, which encode proteins with functions typically required for survival under microaerobic conditions or which confer protection to peroxides. Pseudomonas aeruginosa carries two genes involved in aerotaxis, aer and aer-2 (Hong et al., 2004), of which aer (PA1561) expression was increased after the shift from aerobic to anaerobic growth conditions. The aer gene product allows detection and movement towards oxygen, the preferred electron acceptor. A strong Anr-dependent induction was further detected for the genes PA1557, PA1556 and PA1555, which encode subunits of cbb3-2 type cytochrome c oxidase (Comolli and Donohue, 2004) with a high affinity for oxygen. Interestingly, PA1673 exhibits domains of and similarities to the non-haeme iron, oxygen-transport protein haemerythrin (Stenkamp et al., 1985). Genes encoding proteins that confer resistance towards peroxides were also found upregulated and dependent on Anr regulation. One of these genes is ccpR, the cytochrome c551 peroxidase precursor (PA4587). The second gene is the major cytoplasmic catalase KatA (PA4236), which disproportions peroxide to water and O2. While we detected a slight Anr dependence of katA in our transcriptome analysis, we had to exclude this data due to an increased pde value. However, we independently confirmed an Anr-dependent production of KatA in our proteomic analysis (see below).
The initial Anr-dependent fermentative response of P. aeruginosa to anaerobiosis
Our experimental setup resembles a situation where P. aeruginosa is suddenly exposed to an anaerobic environment. Clearly, the expression profiles we obtained are not comparable to expression profiles of P. aeruginosa under already continuously anaerobic growth conditions. In our experiment, we observed Anr to regulate an immediate response to oxygen deprivation. As the cells are confronted with a sudden loss of the central electron acceptor of aerobic respiration, major enzyme systems central to energy generation under anaerobic conditions have to be activated.
Interestingly, arcDABC, coding for enzymes of the arginine deiminase pathway was by far the highest anaerobically induced Anr-dependent operon. The enzymes of the arginine deiminase pathway are essential for arginine fermentation and catalyse the conversion of arginine to ornithine with the concomitant formation of ATP (Gamper et al., 1991). In addition to Anr, anaerobic induction of arcDABC is enhanced by a second regulator, ArgR, which detects arginine (Lu et al., 1999; 2004). However, the used medium did not contain arginine. One can argue that arginine released from the proteolysis of dispensable proteins might be the substrate for arginine fermentation and the signal for ArgR. In agreement with this hypothesis is the Anr-dependent induction of a gene encoding a putative ClpA/B protease ATP binding subunit (PA0459). The induction of the arginine fermentation pathway even in the presence of nitrate might be due to the period of time required for production of the complete denitrification pathway (see below). In this transition period, anaerobic energy generation relies partly on the arginine fermentation pathway. Accordingly, we detected upregulation of a set of genes that encode proteins involved in classical fermentation processes including two putative alcohol dehydrogenases PA5427 (adhA) and PA2119, a phosphate acetyltransferase, PA0835 (pta) and an acetate kinase, PA0836 (ackA). Previously, we demonstrated that pta and ackA are required for anaerobic pyruvate fermentation in P. aeruginosa (Eschbach et al., 2004). In agreement with our current results, transcription of the ackA-pta operon is anaerobically induced by Anr (Eschbach et al., 2004). Obviously, these fermentation pathways also contribute to energy generation under the outlined conditions.
Onset of denitrification is controlled by Anr and Dnr
As the growth medium contained nitrate as alternative electron acceptor we expected the induction of the membrane-bound nitrate reductase genes (narGHJI) by Anr, Dnr and NarXL. This was observed recently by reporter gene testing of the narK1K2GHJI promoter (Schreiber et al., 2007). Surprisingly, we did not detect a significant induction of the nitrate reductase operon narK1K2GHJI (PA3877–PA3872). However, we detected a moderate induction of the narL response regulator that is located in the opposite direction to the narK1K2GHJI operon. Nonetheless, we detected the NarH subunit in our proteome approach 6 h after a shift to anaerobic conditions (see below), indicating a delayed induction of the narK1K2GHJI operon. Furthermore, we determined the consumption of 2.2 mM nitrate after 2 h of anaerobic incubation and detected the presence of 110 µM of formed nitrite. Therefore, a nitrate reductase activity is present during the first 2 h after the shift experiment, resulting in the generation of NO, which in turn is necessary for the activation of the Dnr regulator. This can be accounted for either by an already low expression of the nar operon or the production of nitrite is catalysed by the soluble nitrate reductase encoded by the napEFDABC operon (PA1177–PA1172). In agreement with the latter, a low napEFDABC expression independent of oxygen tension was observed recently (Schreiber et al., 2007).
Several components of the denitrification machinery require haeme or sirohaeme as cofactor. A key regulatory point of haeme biosynthesis is the oxidative decarboxylation of coproporphyrinogen IX. Formation of an oxygen-dependent (HemF) and independent coproporphyrinogen IX (HemN) oxidase is controlled by Anr and Dnr. Our DNA arrays and proteome data are in good agreement with an earlier report (Rompf et al., 1998). The gene for the independent enzyme HemN is found induced during the shift to anaerobic conditions. Interestingly, we identified the HemF protein in our proteome analysis (Table S5). Thus, the bacterium is capable of utilizing all residual oxygen prior complete shift to anaerobic catalysis.
Denitrification is strictly controlled by Dnr
The microarray analysis in combination with our bioinformatics approach revealed that 7 out of 10 Dnr-dependent genes belong to three known Dnr-dependent operons. Part of these are nirSMC, three genes of the large nir operon, nirSMCFDLGHJEN (PA0519–PA0509) encoding the nitrite reductase, c-type cytochromes and the enzymes of haeme d1 biosynthesis (Kawasaki et al., 1997). Moreover, induction of nirQ (PA0520), the norCB operon (PA0523–PA0524) and the nosZ (PA3392) gene of the nosRZDFYL operon (PA3391–PA3396), coding for nitric oxide and the nitrous oxide reductases are Dnr-dependent (Tables S1 and S4). We found the nir and nor genes also to be downregulated in the anr mutant strain, as expected from the Anr-Dnr regulatory cascade (see first chapter). Only nosZ, which showed the weakest induction of twofold was missing a significant reduction in the anr mutant strain due to low gene expression (0.61-fold). The induction of the dnr-dependent nir, nor and nos operons but not the narK1K2GHJI encoding the first enzyme of the denitrification pathway indicates an ordered sequence of induction to avoid the accumulation of toxic intermediates as NO (see below). A recent bioinformatics approach also identified the nir, nor and nos genes as Dnr-dependent in P. aeruginosa (Rodionov et al., 2005).
The three remaining Dnr-dependent genes seem to be indirectly influenced by Dnr, as no obvious Anr box was found in their putative promoter regions (Tables S1 and S4). These genes encode a C4-dicarboxylate transport protein (PA1183), a hypothetical protein (PA3519) and a component of an RND type efflux pump mexG (PA4205).
Due to the Anr-Dnr cascade, we expected Dnr-dependent genes to be downregulated in the Anr mutant. The dctA, mexG and nosZ genes revealed some downregulation in the anr mutant strain, however, above the 0.5-fold cutoff. This response may be caused by the low overall anaerobic induction of the respective genes (between 2.27- and to 2.47-fold induction).
Proteome analysis of the Anr and Dnr regulons reveals delayed nitrate reductase formation
A limited proteome approach was used to control and partly extend the results of the outlined transcriptome studies to the protein level of gene expression. We extracted proteins from wild type P. aeruginosa and the corresponding anr or dnr mutant strains grown under identical conditions as described for the transcriptome analysis, but extended the anaerobic incubation to 6 h for an adequate proteome response. This proteomic analysis retrieved only two anaerobically induced proteins involved in denitrification that were strongly reduced in cell free extracts prepared from both the anr and dnr mutants. Those were NirS, the nitrite reductase and the regulatory protein NirQ (Table S5). The promoter regions of the corresponding genes contain Anr boxes and have been reported to be Dnr-dependent (Arai et al., 1995). Present in both mutant strains, however in lower concentration, were the putative periplasmic molybdate binding protein ModA, the oxygen-dependent coproporphyrinogen III oxidase HemF, involved in haeme biosynthesis, and NarH, encoding the beta chain of the membrane-bound nitrate reductase. Again, Anr boxes were identified in the promoter regions of the corresponding genes. In agreement, the anaerobic expression of the hemF gene has previously been reported to be dependent on both Anr and Dnr (Rompf et al., 1998). The proteome data indicate that the anaerobic expression of the genes encoding NarH and ModA seems to be constitutive at low level, but depends on the dual activity of both Anr and Dnr for a strong activation. Recent reporter gene experiments confirmed the dual Anr and Dnr dependence of anaerobic narK1K2GHJI transcription (Schreiber et al., 2007). However, in our transcriptome analysis we did not observe the anaerobic induction of the corresponding hemF, narH and modA genes. This is likely caused by the different anaerobic incubation periods of 2 h chosen for the transcriptome and 6 h for the proteome analysis. Particularly in the case of the respiratory nitrate reductase NarGHI, a delayed production of these proteins together with an early onset of nirS induction might be essential to prevent the accumulation of toxic intermediates as nitrite and NO. In agreement with the transcriptome analysis the proteome approach detected ArcA and ArcC, both involved in arginine fermentation, to be produced in an Anr-dependent manner. KatA protein formation was also found Anr-dependent in our proteome analysis. In agreement we also found an indication of Anr dependence of the katA gene in our transcriptome analysis. However, due to a high pde value we excluded the 0.36-fold downregulation of the katA gene in our transcriptome analysis. Moreover, we found the PA5496 gene product encoding a subunit of a vitamin B12-dependent two-component class II ribonucleotide reductase, consisting of NrdJb (PA5496) and NrdJa (PA5497) respectively (Torrents et al., 2005). This enzyme is presumed to be involved in DNA repair or replication under low oxygen tension (Torrents et al., 2005). We detected an Anr box in the putative promoter region of PA5497. However, the distance of this Anr box is only 15 bp to the translational start site. Therefore, transcriptional activation via Anr seems unlikely. As we failed to detect the PA5496-encoded protein in the dnr mutant, we propose these genes to be indirectly dependent on Dnr.
Interestingly, the protein pattern of the anr mutant strain revealed two proteins, PA1677 and GltB of increased concentration compared with the wild type P. aeruginosa, indicating gene repression by the Anr regulator. The promoter region of the gene encoding the hypothetical protein PA1677 does not contain any obvious Anr box, suggesting this protein to be indirectly dependent on Anr. However, we detected one Anr box 125 nt upstream of the putative ATG start codon in the promoter region of the second potentially Anr-repressed gene gltB, encoding a putative binding protein component of an ABC sugar transporter.
A position weight matrix-based bioinformatics approach of Anr binding site definition identifies new important residues
To create a position weight matrix for the Anr binding site of P. aeruginosa the annotated DNA binding sites present in the PRODORIC database (http://www.prodoric.de) were collected and aligned according to their transcriptional direction (Münch et al., 2003) and an information theory position weight matrix model (Schneider et al., 1986) was created. This was used to detect Anr binding sites in the promoter regions of the Anr-controlled genes derived from transcriptomics and proteomics experiments. Only Anr boxes from genes with strong Anr-dependent induction were selected. These Anr binding sites were used to extend the existing model to 40 experimentally confirmed Anr binding sites (Table S3).
The sequence logos (Schneider and Stephens, 1990) comprising the underlying model, revealed some new details about the possible function of the Anr box of Anr-dependent genes (Fig. 4). The degree of sequence conservation peaked at the positions 2, 3 and 12, 13 in a cosine wave-like form. Both the spacing of 10 bp and the information content of 2 bits correlate well with a major groove interaction of the dimeric protein (Schneider, 1996).
Bioinformatically predicted Anr boxes
The extended position weight matrix model was further used for a refined search for potential new Anr binding sites throughout the whole genome of P. aeruginosa. We used the information content in bits as scoring function and defined the threshold at a sensitivity of 1.0, which is the score where 100% of all 40 binding sites comprising the position weight matrix are recovered. This approach identified 436 Anr binding sites in a distance of up to 350 bp upstream of a translational start codon. The score derived from the position weight matrix ranged from 14.84 down to 11.78 containing obviously a huge number of false positive binding sites. We analysed if the identified binding sites were located in non-coding intergenic regions and determined 95% of the identified binding sites with a score of > 13.88 to be located in non-coding regions, as has been observed previously (Robison et al., 1998). The ratio of binding sites in non-coding regions decreased dramatically in dependence of the score and reached 33% for Anr boxes with an information score of 11.78. In order to refine our search results and to avoid false positives, we validated the sequences according to their genomic environment. All sites determined in coding regions were excluded. Our predicted Anr regulon therefore contains only genes next to Anr binding sites located in non-coding regions. In the case of divergent promoters we listed both genes. A total of 170 Anr binding sites were identified this way (Table S6). Of these predicted binding sites 40 Anr binding sites were experimentally identified as shown in Table S3. Consequently, 130 new Anr binding sites were identified. Obviously different growth conditions and media are necessary to induce transcription of these putative promoters.
The predicted P. aeruginosa Anr regulon
The 170 Anr binding sites in combination with our experimental data allowed us to deduce some general features of the Anr regulon of P. aeruginosa. We carefully checked at two parameters, the score of the predicted Anr binding site and its position relative to the putative ATG start codon. While the distance of the transcriptional start site to the ATG start codon is variable in different Anr-dependent promoters, there is a minimum distance of the Anr box to the transcriptional start site required for gene activation. As already mentioned, the ideal distance of the Anr box centre to the transcriptional start site is −41.5. The transcriptional start site is located upstream of the ribosome binding sequence, which is 7 bp in length and located between 7–14 nucleotides in front of the ATG start codon. Taking this into account, the minimal distance of the centre of an Anr binding site to the ATG start codon would be around 55.5–62.5 nt. Note that Table S6 lists the distance between Anr binding site and ATG start codon, which is between 48–55 nt. This prediction is in perfect agreement with the smallest distance observed for experimentally confirmed Anr binding sites. The Anr binding site of the rfaD gene is located 54 nt in front of the putative ATG start codon, while the Anr binding site in front of PA0526 is located 57 nt upstream the putative ATG start codon. Interestingly, we also identified 31 Anr binding sites with a smaller distance than 54 nt to the putative ATG start codon. This indicates that there might be some Anr binding sites in position that would indicate a role of Anr as a repressor. However, none of this Anr binding sites were identified experimentally so far. In this context an interesting example is the gene plcH, which contains a highly conserved Anr binding site 24 nt in front of the putative ATG start codon.
The majority of Anr binding sites are located between 60–100 nt upstream of the ATG start codon with a peak of location between 60–70 nt. This is in good agreement with the number and distance of the experimentally confirmed Anr binding sites (see Fig. 5).
The predicted Anr regulon of P. aeruginosa consists still of a huge number of hypothetical proteins of unknown function. This clearly reflects the need for further investigations of the corresponding life condition. The currently predicted Dnr- and Anr-dependent regulon can only serve as a map of the putative members.
The P. aeruginosa Anr regulon contains approximately 170 predicted transcription units as determined by an iterative transcriptome, proteome and bioinformatics approach. In contrast to its E. coli counterpart Fnr the P. aeruginosa oxygen regulator Anr is acting almost exclusively as transcriptional activator. The Dnr regulon consists currently of only four direct members, all devoted to denitrification. Consequently, Anr is the major redox regulator of P. aeruginosa.
The initial response of P. aeruginosa to anaerobiosis is unexpectedly complex. In contrast to our expectation, fermentative pathways as the arginine fermentation pathway and pyruvate fermentation are induced at first to immediately provide energy during the transition to anaerobiosis. The alternative respiratory pathway denitrification is induced afterwards and in a sequential manner. Pseudomonas aeruginosa induces the lower denitrification pathway consisting of the nir, nor and nos genes at first, probably to avoid accumulation of toxic intermediates as NO. The nitrate reductase is induced at a later stage. Moreover, a variety of different stress response genes is produced. Many of these gene products, as the universal stress proteins, have an unknown function.
Bacterial strains and growth conditions
The P. aeruginosa PAO1 wild type, the anr deletion mutant PAO6261 and the dnr mutant RM536, all previously described (Dunn and Holloway, 1971; Arai et al., 1995; Ye et al., 1995), were used in this study. Bacteria were grown in a modified AB minimal medium (Clark and Maaløe, 1967), containing 25 µM FeSO4, 20 mM glucose and 50 mM NaNO3 or LB medium supplemented with 50 mM NaNO3 (Sambrook et al., 1989). The 200 ml aerobic cultures were grown in 1 l Erlenmeyer flasks at 37°C and 300 r.p.m. Cells were harvested in the early exponential phase at an OD578 of 0.3. Anaerobic cultures were grown in 135 ml serum flasks filled to 95%. Control experiments with an oxygen electrode (Mettler-Toledo, Giessen, Germany) or the oxygen-sensitive indicator dye resazurin verified anaerobic conditions. In the anaerobic shift experiments, the aerobic culture was grown to an OD578 of 0.3 and then transferred to a 135 ml sealed serum flask. Control experiments verified that oxygen tension decreased within 3–5 min below the detection limit of an oxygen electrode. The detection of nitrate and nitrite was performed as described previously (Hoffmann et al., 1998).
To prevent changes of the protein pattern after harvest, the culture was mixed with a double volume of ice cold potassium-phosphate buffer (100 mM; pH 7.4) and allowed to cool for 5 min. Cells were centrifuged at 8000 g for 20 min at 4°C, washed twice with the potassium-phosphate buffer and suspended in 1/100 vol. of the original culture volume. Protein concentrations were determined in whole cell suspensions using the BCA protein assay (Sigma, Taufkirchen, Germany). For this assay cells were disrupted by incubation of a 360 µl culture aliquot with 150 µl 2 M NaOH for 1 h at 70°C. We modified a protocol described previously (Hanna et al., 2000) and extracted proteins directly from whole cells with phenol and a subsequent acetone precipitation. The precipitated protein was solubilized in sample buffer consisting of 7 M urea, 2 M thiourea, 4% 3-[(3-cholamidopropyl)-dimethylammonio]-1-propanesulfonate (CHAPS), 50 mM dithioerythritol (DTT), 2% Triton X100 and 2% ampholytes (Bio-Lyte; Biorad, Munich, Germany). Protein concentration was determined in the sample buffer using the 2D Quant kit (Amersham, Freiburg, Germany). The 2D electrophoresis was performed using immobilized pH gradient (IPG) strips of 17 cm length covering various pH ranges (pH 5–8; 3–10; 4–7; 4.5–5.5; 5.5–6.7; Biorad, Munich, Germany). The IPG strips were rehydrated overnight in sample buffer containing 500 µg of protein. IEF was conducted at 20°C under mineral oil in the PROTEAN IEF Cell (Biorad, Munich, Germany) for a total of 110 kVh. Prior to SDS-PAGE, the focused proteins in the IPG strips were reduced for 15 min in an SDS equilibration solution containing 15 mM DTT and alkylated for another 15 min in the same buffer containing 150 mM iodacetamide. After the equilibration step the strips were transferred to 10% SDS-PAGE (25.5 cm× 20.5 cm) gels. Electrophoresis was performed at a constant temperature of 20°C with 2 W per gel for 20 h. The gels were stained with Ruthenium II bathophenanthroline disulfonate chelate (RuBPS) (Rabilloud et al., 2001) or with Coomassie Blue G-250 (Neuhoff et al., 1988). Gels were documented with an FX-Scanner (Biorad, Munich, Germany). Analysis and quantification of differential protein spot patterns were performed by using the Software Z3 (Compugen, Tel Aviv, Israel). Gel spots were excised and treated using a slightly modified method described previously (Shevchenko et al., 2000). Briefly, the gel pieces were washed with water, dehydrated with acetonitrile (ACN), rehydrated with 100 mM NH4HCO3, and prior to trypsin (sequencing grade, Promega) digestion again dehydrated by ACN. Peptides were extracted and collected in four elution steps (each 15 min, 37°C) using 25 mM NH4HCO3, ACN, 5% formic acid and again ACN. Extracted peptides were purified using ZipTip C18-microcolumns (Millipore), following the manufacturer's instructions. Proteins were identified by peptide-mass fingerprint (PMF) as well as post-source decay fragmentation data recorded on a Bruker Ultraflex MALDI-TOF mass spectrometer. PMF data were analysed using an internal MASCOT-server at the Helmholtz Centre Braunschweig (version 1.9; Matrix Science) (Perkins et al., 1999) and the NCBI database (restricted to the taxon P. aeruginosa). Only peptides with a MASCOT rank of 1 were considered significant and used for the combined peptide score. The criteria used to accept protein identifications based on PMF data included the extension of sequence coverage (minimum of 30%), the number of peptides matched (minimum of 5) and the score of probability (minimum of 70 for the Mowse score). Lower-scoring proteins were either verified manually or rejected.
RNA isolation for microarray and Northern blot analyses
Twenty-five millilitres of the cultures were cooled down with crushed ice, harvested and subjected to total RNA isolation by a modified hot phenol method (Aiba et al., 1981). The obtained RNA was treated with RNase-free DNase I and further purified by RNeasy columns (Qiagen, Hilden, Germany). The integrity of total RNA as well as the transcription of anaerobic marker genes in the harvested cells was confirmed by Northern blot analysis. The RNA of the nirS or arcA genes served as anaerobic marker to confirm anaerobic conditions, whereas absence of the nirS transcript in Northern blot experiments of aerobically grown cells confirmed aerobic conditions. For Northern blot analysis the RNA samples were separated electrophoretically in 1% agarose gels containing 5% (660 mM) formaldehyde (Ausubel et al., 1995). Total RNA was then transferred to a positively charged nylon membrane by vacuum transfer and covalently linked by UV light. The nirS probe was generated by PCR using the primer pair nirS-NO 5′-AGT TCA ACG AGG CCA AGC AG-3′ and nirS-T7 5′-CTA ATA CGA CTC ACT ATA GGG AGA-TCG TGC TGG GTG TTG TAG AC-3′. The arcA probe was generated with the primer pair arcA-NO-for 5′-CTG ACC GAG ACC ATC CAG AA-3′ and arcA-NO-rev 5′-CTA ATA CGA CTC ACT ATA GGG AGA-CAG CAG GGT GTT GGT GTA GG-3′. DNA labelling was carried out with the Digoxygenin Labelling Kit (Roche, Mannheim, Germany). Hybridization and detection was carried out following protocols of the manufacturer. The dioxetane derivative CDP-Star (NEB, Frankfurt am Main, Germany) was used as a chemiluminescent substrate for membrane-based detection of alkaline phosphatase conjugates.
Microarray experiment and data analysis
The cDNA synthesis, fragmentation and labelling were performed according to protocols for the Affymetrix Pseudomonas GeneChip (Santa Clara, California) with the following modification: 0.6 U DNase I were used per µg cDNA for fragmentation to yield the desired cDNA size range of 50–200 bases as confirmed by gel electrophoresis. This was verified by separating 200 ng of fragmented cDNA on a 15% acrylamide gel and staining with SYBRgold (Molecular Probes, Eugene, United States). Target hybridization, washing, staining and scanning were performed by the Affymetrix Core Facility at the Helmholtz Centre for Infection Research, Braunschweig. For each growth condition, RNA from three independent cultures was used and two technical replicates were performed.
Raw microarray data were preprocessed with the Bioconductor software framework (Gentleman et al., 2004). Expression values were calculated by the Robust Multichip Average method using quantile normalization, background corrected PM intensities and median polish as summarization method (Bolstad et al., 2003; Irizarry et al., 2003a,b).
MIAME compliant array data were submitted to the GEO database (record identifier: GSE17179) and are accessible via the following link:
In addition, the pdes were computed for the following pairwise comparison of experimental conditions: wild type cells grown under aerobic versus anaerobic conditions, Δanr versus wild type cells (both under anaerobic growth) and Δdnr versus wild type cells (both under anaerobic growth). For this purpose, the preprocessed expression data were analysed by a regularized t-test based on a Bayesian statistical framework using the CyberT algorithm (Baldi and Long, 2001; Hatfield et al., 2003). The R package bayesreg, which implements the CyberT algorithm, was used for computing the posterior probabilities of differential expression (free download under http://cybert.microarray.ics.uci.edu/). The mean gene expression levels of the replicate experiments were ranked in ascending order, a sliding window of 101 genes was used and a confidence value of 10 was chosen as weight. More information on the use of this regularized t-test in the context of determining differentially expressed genes of prokaryotic DNA microarray expression data is published elsewhere (Hung et al., 2002). A combination of information from expression ratios and pdes allowed to identify the genes with the highest relative changes in expression values and the highest probabilities of differential expression at the same time.
In the first comparison we identified genes induced or repressed under anaerobic conditions in the P. aeruginosa wild type PAO1. Here we compared the transcriptome profile of P. aeruginosa PAO1 grown under aerobic conditions up to an OD578 of 0.3 with the transcriptome profile of the PAO1 strain, which was first grown under aerobic conditions up to an OD578 of 0.3 and then shifted to anaerobic conditions by transfer to a sealed serum flask and further incubated for 2 h under anaerobic conditions.
In our second comparison we identified genes, which are regulated differently in the anr mutant strain PAO6261. Here we compared the transcriptome profile of the P. aeruginosa wild type PAO1 with the transcriptome profile of the P. aeruginosa anr mutant strain PAO6261. Both strains were harvested after 2 h incubation under anaerobic conditions.
In our third comparison we identified genes, which are regulated differently in the dnr mutant strain RM536. Here we compared the transcriptome profile of the P. aeruginosa wild type PAO1 with the transcriptome profile of the P. aeruginosa dnr mutant strain RM536. Both strains were harvested after 2 h incubation under anaerobic conditions.
All genes repressed less than 0.5-fold and a pde value above 0.95 were selected for further analysis.
Prediction of Anr binding sites and regulated genes
A model of the Anr binding site was created by computing information vector RSequence(l) using following equation:
where f(b,l) is the frequency of each base b at position l in the aligned binding sites (Schneider et al., 1986). We considered the nucleotide bias present in P. aeruginosa by using a linear correction of noise (Schreiber and Brown, 2002). The position weight matrix m(b,l) was afterwards generated by
which is equivalent to the individual letter size of a sequence logo. Scores were calculated by summing up the individual weights. We used the Virtual Footprint tool for pattern search and the prediction of the potential regulons (Münch et al., 2005). The program takes into account the position weight matrix model and is connected interactively to the PRODORIC database (http://www.prodoric.de/vfp). Sequence logos were created using WebLogo (Crooks et al., 2004).
We are indebted to Dr Dieter Haas (Université de Lausanne, Switzerland) for providing the P. aeruginosa anr mutant and Dr Hiruyuki Arai (Riken-Institute, Tokyo, Japan) for the P. aeruginosa dnr mutant. We thank Tanja Toepfer for technical assistance with the microarray hybridization. The help of Dr Dana Held from our laboratory for the construction of the plasmids pDH11 and pDH12 and the mutated HemN promoter is highly acknowledged. This work was supported by grants from the Deutsche Forschungsgemeinschaft, the German Research Centre for Biotechnology, the German Federal Ministry of Education and Research (BMBF) for the Bioinformatics Competence Centre ‘Intergenomics’ (Grant No. 031U110A/031U210A), the BMBF for the National Genome Research Network (NGFN2-EP, Grant No. 0313398A) and the Fonds der Chemischen Industrie to D.J.