GeneChip® expression analysis of the iron starvation response in Pseudomonas aeruginosa: identification of novel pyoverdine biosynthesis genes


  • Urs A. Ochsner,

    1. Department of Microbiology, Campus Box B-175, University of Colorado Health Sciences Center, 4200 East Ninth Avenue, Denver, CO 80262, USA.
    Search for more papers by this author
  • Paula J. Wilderman,

    1. Department of Microbiology, Campus Box B-175, University of Colorado Health Sciences Center, 4200 East Ninth Avenue, Denver, CO 80262, USA.
    Search for more papers by this author
  • Adriana I. Vasil,

    1. Department of Microbiology, Campus Box B-175, University of Colorado Health Sciences Center, 4200 East Ninth Avenue, Denver, CO 80262, USA.
    Search for more papers by this author
  • Michael L. Vasil

    Corresponding author
    1. Department of Microbiology, Campus Box B-175, University of Colorado Health Sciences Center, 4200 East Ninth Avenue, Denver, CO 80262, USA.
    Search for more papers by this author


Upon iron restriction, the opportunistic pathogen Pseudomonas aeruginosa produces various virulence factors, including siderophores, exotoxin, proteases and haemolysin. The ferric uptake regulator (Fur) plays a central role in this response and also controls other regulatory genes, such as pvdS, which encodes an alternative sigma factor. This circuit leads to a hierarchical cascade of direct and indirect iron regulation. We used the GeneChip® to analyse the global gene expression profiles in response to iron. In iron-starved cells, the expression of 118 genes was increased at least fivefold compared with that in iron-replete cells, whereas the expression of 87 genes was decreased at least fivefold. The GeneChip® data correlated well with results obtained using individual lacZ gene fusions. Strong iron regulation was observed for previously identified genes involved in biosynthesis or uptake of the siderophores pyoverdine and pyochelin, utilization of heterologous siderophores and haem and ferrous iron transport. A low-iron milieu led to increased expression of the genes encoding TonB, alkaline protease, PrpL protease, exotoxin A, as well as fumarase C, Mn-dependent superoxide dismutase SodA, a ferredoxin and ferredoxin reductase and several oxidoreductases and dehydrogenases. Iron-controlled regulatory genes included seven alternative sigma factors and five other transcriptional regulators. Roughly 20% of the iron-regulated genes encoded proteins of unknown function and lacked any conclusive homologies. Under low-iron conditions, expression of 26 genes or operons was reduced in a ΔpvdS mutant compared with wild type, including numerous novel pyoverdine biosynthetic genes. The GeneChip® proved to be a very useful tool for rapid gene expression analysis and identification of novel genes controlled by Fur or PvdS.


Innovative technologies such as high-throughput sequencing, DNA microarrays, proteomics and sophisticated data analysis tools allow a previously unimaginable global perspective to microbial genetics and may advance the development of new therapeutic agents (Debouck and Goodfellow, 1999; Ivanov et al., 2000). Almost half of the currently over 100 microbial genome projects have been finished, and the complete genome sequences of many human pathogens are known. Over the past few years, simple and often self-made DNA microarrays have been used successfully to study various microbiological aspects including host–pathogen interactions (Cummings and Relman, 2000; Ichikawa et al., 2000), cell cycle regulation (Brun, 2001), genetic diversity (Salama et al., 2000) and biofilm development (Schoolnik et al., 2001). DNA microarrays have also been used to identify target genes of global regulators such as integration host factor of Escherichia coli (Arfin et al., 2000), quorum-sensing in E. coli (DeLisa et al., 2001) and Streptococcus pneumoniae (de Saizieu et al., 2000) and two-component regulators of Bacillus subtilis (Ogura et al., 2001). Several studies focused on environmental changes, such as the response of group A Streptococcus to alterations in growth temperature (Smoot et al., 2001) or the effect of isoniazid on resistant Mycobacterium tuberculosis (Wilson et al., 1999). The only published microarray study on the influence of iron on the genome-wide expression profile so far has focused on Pasteurella multocida (Paustian et al., 2001).

Pseudomonas aeruginosa possesses the largest bacterial genome completely sequenced today, with 6264 403 bp representing over 5500 genes ( Stover et al., 2000 ) and, recently, an oligonucleotide array called GeneChip® has become available from Affymetrix. The total 5900 spots on the GeneChip® represent 5769 probes to the P. aeruginosa PAO1 strain, which include 5570 annotated open reading frames (ORFs), rRNA and tRNA genes, as well as 199 intergenic regions. In addition, 117 probes represent genes from non-PAO1 Pseudomonas strains, transposon-related genes and antibiotic resistance genes, whereas 14 probes serve as controls for genes from B. subtilis , Arabidopsis thaliana and Saccharomyces cerevisiae . We used the novel GeneChip® technology to study the global changes in the expression profiles of P. aeruginosa upon iron starvation. To our knowledge, this is the first report on a genome-wide analysis of iron-responsive genes and the first application of the P. aeruginosa GeneChip®.

Pseudomonas aeruginosa is a significant opportunistic pathogen affecting cystic fibrosis (CF) patients and immunocompromised people, and can cause wound infections and pneumonia. Infections with P. aeruginosa are often persistent because of the development of multiple drug resistance and the formation of biofilms ( Hancock and Speert, 2000 ; Hoiby et al., 2001 ). Because iron is essential for growth and survival of P. aeruginosa ( Vasil and Ochsner, 1999; Ratledge and Dover, 2000 ), potential novel drugs may target iron metabolism. The preferred aerobic metabolism requires respiratory enzymes that need iron for their function. However, iron restriction is encountered during infection of a host where virtually all iron is sequestered by host proteins. The powerful iron acquisition systems aimed to overcome iron limitation include the secretion of siderophores, proteases and cytotoxic exotoxin A, as well as haem/haemoglobin utilization systems ( Ochsner et al., 1996 ; 2000a ; Takase et al., 2000 ; Wilderman et al., 2001 ). In contrast, a high-iron milieu during aerobic respiration can lead to the formation of reactive oxygen intermediates, which have deleterious effects on bacterial cells. Therefore, iron metabolism must be carefully balanced in terms of acquisition and storage. The ferric uptake regulator (Fur) plays the central role in the iron starvation response ( Ochsner et al., 1995 ). Interestingly, the fur gene is essential in P. aeruginosa ; however, fur mutants producing altered Fur proteins have been generated ( Barton et al., 1996 ). A large number of Pseudomonas iron-regulated genes (PIGs) have been isolated in a cycle selection procedure based on the binding of Fur to DNA targets ( Ochsner and Vasil, 1996 ). Fur controls several other regulators, including the ‘extracytoplasmic function’ (ECF) sigma factor PvdS, which is required for the production of pyoverdine, exotoxin A and PrpL protease ( Cunliffe et al., 1995 ; Ochsner et al., 1996 ; Wilderman et al., 2001 ). Pyoverdine and PvdS have been associated with virulence in several different models ( Meyer et al., 1996 ; Xiong et al., 2000 ). The PvdS sigma factor has been purified, and binding to several promoters of pvd pyoverdine synthesis genes has been demonstrated ( Wilson et al., 2001 ). Very recently, PvdS has been shown to play a key role in siderophore-mediated signalling, which represents a novel type of bacterial communication that is related in principle to quorum sensing and allows the concerted expression of several known virulence determinants, including exotoxin and protease, in response to iron starvation ( Lamont et al., 2002 ).

This report on the global response to iron starvation is based on two comparisons of expression profiles: (i) cells grown under low-iron versus high-iron conditions; and (ii) wild-type cells versus ΔpvdS cells grown under low-iron conditions. We confirmed the iron-responsive expression of known genes, thereby evaluating the GeneChip®, and we were able to identify many novel iron-regulated genes. In particular, we demonstrate that a number of genes of previously unknown function are regulated by PvdS and are required for pyoverdine biosynthesis. These findings should facilitate the further elucidation of the biochemical pathway leading to pyoverdine production.


Influence of iron on global gene expression

To assess the effect of iron availability on the expression of P. aeruginosa genes, duplicate cultures of wild-type PAO1 cells were grown to stationary phase under iron-limiting or iron-replete conditions. The growth medium was Chelex-treated and dialysed tryptic soy broth supplemented with 50 mM sodium glutamate and 1% glycerol. This medium contains limiting amounts of iron and has been used in many previous studies relating to P. aeruginosa iron metabolism (Liu, 1973; Barton et al., 1996; Ochsner et al., 1996; Vasil et al., 1998). It supports vigorous growth to a final cell density of 4 × 109 cells ml−1 or to roughly 6 × 109 cells ml−1 when supplemented with 200 µM FeCl3 for high-iron conditions. The growth rate is slightly higher under iron-replete conditions, but the cells grown in either low- or high-iron medium enter stationary phase after about 6–8 h; therefore, cell density-dependent regulatory circuits such as quorum sensing are not affected (data not shown). The cultures were grown aerobically at 32°C, which is well within the range of naturally occurring conditions, and we harvested the cells after 18 h to minimize the influence of the early difference in growth rates and to maximize iron starvation under low-iron conditions. The medium had an initial pH of 7.1, which increased to 8.1 and 8.2 at the end of the low-iron and high-iron cultures, respectively; thus, the two conditions were quite comparable. The two sample pairs obtained from cells grown in this low-iron or high-iron medium hybridized to an average of 3978 (69%) of the total 5769 probes portraying the PAO1 genome on the GeneChip®. Interestingly, the number of expressed genes appeared to be somewhat lower in iron-starved cells (3755, 65%) than in cells grown under high-iron conditions (4201, 73%). The gene expression levels were cross-compared, and the average changes were calculated. The number of genes that were regulated by iron, either positively or negatively, is depicted in Fig. 1. In iron-starved cells, the expression of 20 genes was increased 50-fold or more, and 118 genes were at least fivefold upregulated compared with cells grown in iron-rich medium. In contrast, expression of four genes was reduced 50-fold or more, and 83 genes were at least fivefold downregulated. At least 3000 genes were expressed above the detection limit but were unaffected by iron, and ≈ 1800 genes were not significantly active under either condition.

Figure 1.

Gene expression levels in response to iron starvation. The average changes in gene activity between cells grown under low-iron and high-iron conditions were obtained by GeneChip® analysis. The number of genes showing increased or decreased expression and their level of iron regulation are indicated. Genes not expressed under either condition are shown by the lighter grey box. The bars are drawn to scale, such that the enclosed areas are proportional to the number of genes in that group.

Characterization of iron-regulated genes

The P. aeruginosa genes that exhibited the greatest increase in expression upon iron starvation are listed in Table 1. As expected, a strong iron regulation pattern was observed for genes and operons that have been reported to play a role in siderophore-mediated iron acquisition. These included the pyoverdine-related genes PA2386 (pvdA), PA2426 (pvdS), PA2396 (pvdF ), PA2398 (fpvA), PA2424–2425 (pvdL), PA2397 (pvdE), PA2402 (pvdI ), PA2400–2401 (pvdJ ) and PA2399 (pvdD), and the pyochelin -related genes PA4220–4221 (fptA), PA4228–4231 (pchDCBA), PA4222–4226 (pchEFG) and PA4227 (pchR ). Also, the expression of tonB (PA5531) and of several genes involved in haem uptake (PA3407–3408, PA4708–4709, 4710), utilization of heterologous siderophores (PA4158, PA3901, PA4897) and ferrous iron transport (PA4357–4359) was strongly iron regulated. The genes encoding alkaline protease (PA1249) and its secretion apparatus (PA1245–1248), as well as PrpL (PA4175), were expressed at higher levels under iron-limiting conditions compared with iron-replete conditions. The physiological status of iron-starved cells was reflected by the observed induction of several genes that encode enzymes such as fumC and sodA (PA4467–4471), a ferredoxin (PA3530) and ferredoxin reductase (PA3397), cytochrome o ubiquinol oxidase (PA1317–1321), other probable oxidoreductases (PA3768, PA4513, PA5309) and dehydrogenases (PA5150, PA0265, PA5312). Many of these enzymes do not require iron for their function and may thus provide essential cellular functions in the absence of their counterparts that depend on iron as cofactors. Furthermore, a considerable number of regulatory genes were highly iron regulated as well, including seven putative alternative sigma factors of the ECF family together with anti-sigma factors and five other transcriptional regulators. A possible negative regulator is encoded by PA4570, which was 403-fold iron regulated and contained a strong Fur box within its promoter. At least 20 iron-regulated genes or operons were identified that encode putative proteins that lack conclusive homologies to other polypeptides or conserved motifs that would allow deduction of a possible function. Besides the genes listed in Table 1, this group of ‘hypothetical, unclassified, unknown’ (HUU) genes includes PA0026, PA0433, PA1035, PA2761, PA2776, PA4090, PA4792 and PA5473, all of which were iron regulated up to 10-fold.

Table 1. Iron-regulated genes of P. aeruginosa PAO1 with increased expression under iron limitation.
ORF or operonGeneFunction or classFold change
versus high
Fold change
lacZ fusion
low versus high
Fold change
PAO1 low
ΔpvdS low
Fur box or IS box
  1. Operon-forming ORFs are grouped, and the value shown represents the gene with the highest iron regulation. The functions of the encoded proteins are indicated according to the PAO1 genome annotation where possible. ND, not determined; NA, not applicable.

  2. The ratios of the measured β-galactosidase activities are the means of two independent experiments. For clarity, only the significant changes in the comparison of PAO1 and ΔpvdS are indicated. The consensus sequence is GATAATGAT(A/T)ATCATTATC for the Fur box (Ochsner and Vasil, 1996), TAAAT-N16-CGT for the IS (‘iron starvation’) box (Wilson et al., 2001), and lowercase letters indicate non-matching bases.

PA4570 HUU; weak similarity to IclR repressor family403 690 GATAATaAgAATtATTcTC
PA2386pvdA l -Ornithine-N 5 -oxygenase 216 520178 TAAAT-N16-CGT
PA3530 bfd Ferredoxin203ND aATAATGAgAATgATTgTC
PA4220-4221 fptA Ferripyochelin receptor (4221); HUU (4220)182ND cATAATGATAAgCATTATC
PA2426 pvdS ECF sigma factor177 930 GtaAtTGAcAATCATTATC
PA2383-2384 LysR-type regulator (2383); HUU (2384)148 168 12NA
PA0672 hemO ( pig4 ) Haem oxygenase138 250 GAcAtTGAgAtTCAaTAaC
PA2411–2412 Probable thioesterase (2411); HUU (2412)126 347131 TAAtT-N16-CGT
PA4228–4231 pchDCBA Pyochelin biosynthesis124ND NA
PA4467–4471 fumC–sodA Fumarase (4470); (Mn)-superoxide dismutase (4468)1191120 GATAATGAgAtTgATTATt
PA2033–2034 HUU 95  58 GcaAATGcTAATCATTcgC
PA3407–3408 hasRA Haem receptor (3408); haemophore (3407) 70 144 tgTAATaAgAATCAgTcTC
PA2413 Probable aminotransferase 65 153 75 TAAAT-N16-CGT
PA2427 pvdY HUU 61 580 56 TAAAT-N16-CGT
PA4222–4226 pchEFG Pyochelin biosynthesis 55ND NA
PA2451–2452 Similar to E. coli Fes enterochelin esterase 52  74 42 cAAAT-N16-CGT
PA4218–4219 HUU; membrane proteins, adjacent to fptA 49ND NA
PA0471–0472 fiuR ( pig17 ) ECF sigma factor (0472); transmembrane sensor (0471) 46  80 aAaAAcGAgAATtATTcgC
PA1300–1301 ECF Sigma factor (1300); transmembrane sensor (1301) 46 200 GcgAATGATAATacaTATC
PA2396pvdFN5-hydroxyornithine transformylase 45ND 30 TAAAT-N16-CGT
PA3409–3410 pig25 ECF Sigma factor (3410); transmembrane sensor (3409) 40ND ttTtATGATAATgATTtTC
PA2393–2395 Dipeptidase (2393); aminotransferase (2394); reductase (2395) 38 340 31 TAAAT-N16-CGT
PA4227 pchR Transcriptional regulator for pyochelin biosynthesis genes 37ND GgaAATGAgAtTtATTATC
PA4515–4516 piuC ( pig7) HUU 36  25 tccAATGcaAATCAaTATC
PA2398 fpvA Ferripyoverdine receptor 35 212 22 TAAAA
PA2424–2425pvdLNon-ribosomal peptide synthetase (2424); thioesterase (2425) 34  96 26 TAAtT-N16-CGT
PA2397pvdEABC transporter 33ND 34 TAAAT-N16-CGT
PA4357–4359 feoAB Probable ferrous iron transport 32  39 GcTgATtAccATCATgATg
PA2400–2402 pvdJ , pvdI Non-ribosomal peptide synthetases 30  35 26 TAAtT-N16-CGT
PA2467–2468 ECF sigma factor (2468); transmembrane sensor (2467) 30ND GAcAATGcgAtTCATTATC
PA5531 tonB ( pig13 ) TonB protein 24ND ctgAATGATAATaATTATC
PA4895–4896 ECF sigma factor (4896); transmembrane sensor (4895) 20ND GcaAATGAaAActATTATC
PA2399 pvdD Non-ribosomal peptide synthetase 19ND 15 TAAAT
PA4370 HUU 19   5 aAaAATGcaAATCtTTcgC
PA5217 Probable component of ABC iron transporter 18ND acgAATGATAAgaAaTcTC
PA2403–2410 HUU; membrane proteins; ABC transporter (2408, 2409) 15  75 11NA
PA2389–2391 HUU; ABC transporter (2390); outer membrane protein (2391) 14  53 10 TAAAT-N16-CGT
PA2392 HUU 14 640 12 TAAAT-N16-CGT
PA3899-3900 ECF sigma factor (3899); transmembrane sensor (3900) 13ND tAaAATGATAActATTATC
PA4500 Probable component of ABC transporter 13ND NA
PA1245–1248 aprDEF Alkaline protease secretion apparatus 12ND NA
PA4708–4709 phuST Haem binding and transport 12  36 GcTAATGcaAATaATTATC
PA2385 Probable acylase 11ND GATAATGgcgATCtTcATC
PA2531 Probable aminotransferase 11  24 11 TAAAg-N16-CGT
PA3768 Probable metallo-oxidoreductase 11ND NA
PA4513 Probable oxidoreductase 11ND NA
PA4158 fepC Ferric enterobactin transport protein 10ND NA
PA5150 Probable short-chain dehydrogenase 10   5  6 gAAAg-N16-CGT
PA0266 gabT 4-Aminobutyrate aminotransferase  8ND NA
PA0707 toxR ( regA ) Transcriptional regulator ToxR  8 300  6NA
PA1134 HUU  7  33  7 TAAAc-N16-CGT
PA3165 hisC2 Histidinol-phosphate aminotransferase  7ND NA
PA4175 prpL Endoproteinase (protease IV)  7  15 11 TAAAa-N16-CGT
PA4710 phuR Haem receptor  7  15 GATAATtATttgCATTAgC
PA0929–0930 pirR ( pig19 ) Two-component regulatory system  6  40 acgAtTGcTAATCAaTcTt
PA3901 fecA TonB-dependent receptor  6ND NA
PA0265 gabD Succinate-semialdehyde dehydrogenase  5ND NA
PA0500 bioB Biotin synthase  5ND NA
PA0968 tolQRAB Tol system  5ND GAagATGgTAATtAaTtgC
PA2387 pig32 Sigma factor (ECF)  5ND ctTgATGAgAATtATTATa
PA4897 optI Outer membrane receptor  5ND NA
PA5309 Probable oxidoreductase  5ND NA
PA0346 HUU; predicted signal peptide  4   2  6 TgAAT-N16-CGT
PA0818 HUU; predicted signal peptide  4   9  4 TgAAT-N16-CGT
PA1316 Probable MFS transporter  4ND NA
PA1317–1321 cyoABCDE Cytochrome o ubiquinol oxidase  4ND NA
PA3811 hscB Heat shock protein  4ND NA
PA3812 iscA Probable iron-binding protein  4ND NA
PA4633 Probable chemotaxis transducer  4ND NA
PA4833 Similar to haemolysin Hly-III family  4   3  4 TAAAa-N16-CGT
PA4973 thiC Thiamin biosynthesis  4ND NA
PA5312 kauB Probable aldehyde dehydrogenase  4ND NA
PA5313–5314 paaT Probable pyridoxal-dependent aminotransferase  4ND NA
PA1003 LysR-type regulator  3   2  4 TAAAT-N16-CGT
PA1148toxAExotoxin A  3 125  3 TAAAT
PA1249 aprA Alkaline protease  3   3 NA
PA2388 Transmembrane sensor  3ND tAaAtTtATAATaATTcTC
PA3397 fpr Ferredoxin reductase  3ND NA

The GeneChip® data were validated using a mini-library of promoter fusions to the lacZ reporter gene, and the iron response of these fusions is also shown in Table 1. The growth conditions for the lacZ reporter cultures were identical to the conditions under which the cells for GeneChip® expression profiling had been grown. The two methods correlated very well in terms of the observed changes in gene expression, although in a few cases (i.e. toxA, toxR), the GeneChip® detected a substantially lower degree of iron-regulation than the reporter fusions (see Discussion).

Identification of PvdS-regulated genes

Only 27 of the 78 iron-regulated genes or operons shown in Table 1 contain a Fur box-like sequence element in their promoter region. Other regulators besides Fur thus appear to govern the indirect iron control of a large number of genes. For example, pvdS is Fur regulated and encodes an alternative sigma factor, which has been found to be essential for the expression of the pyoverdine biosynthesis genes pvdA, pvdD, pvdE, pvdF, ptxR and pvdY (Wilson et al., 2001), the genes toxA and regA (Ochsner et al., 1996) and prpL protease (Wilderman et al., 2001). To identify additional PvdS-regulated genes, we used the GeneChip® to compare the global expression profile of ΔpvdS mutant cells to PAO1 wild-type cells under iron-limiting conditions. We found 26 genes or operons that were affected by PvdS under these conditions (Table 1). The results were confirmed independently by comparing the activities of specific gene fusions to the lacZ reporter gene in PAO1 wild-type cells and ΔpvdS mutant cells (data not shown). PvdS seems to act exclusively as a positive regulator of gene expression, and activation by PvdS correlated well with the observed iron regulation of these genes, as shown above. With a few exceptions, the PvdS-regulated genes possessed a typical PvdS-binding motif (iron starvation ‘IS’ box) in the promoter region. The IS box participates in sequence-specific recognition by PvdS to enable transcriptional initiation and is composed of a conserved TAAAT element spaced 16 bases from an optional CGT element (Rombel et al., 1995; Wilson et al., 2001). The GeneChip® analysis not only confirmed known PvdS targets, but also identified several additional PvdS-regulated genes, including genes located within a cluster associated with pyoverdine biosynthesis (see below). Also among the novel PvdS-regulated genes is PA4833 encoding a basic 23 kDa membrane protein with > 0% identity to members of the haemolysin Hly-III protein family. PA1003 encodes a LysR-type transcriptional regulator and is located next to the phnAB phenazine biosynthesis operon. Mutants affected in PA1003 exhibited a bright-yellow pigmentation (data not shown). Other PvdS-regulated genes include PA0346 and PA0818, both of which encode possible extracellular proteins of unknown function, and PA2531, PA1134 and PA5150.

Novel pyoverdine biosynthesis genes

The majority of the PvdS-regulated genes were located in a 120 kb region spanning from PA2383 to PA2452 (Fig. 2). Mutants affected in single loci within this pvd cluster were constructed and analysed for fluorescence on low-iron CAA agar, pyoverdine in supernatants and growth on CAA agar containing EDDHA (see Experimental procedures). Several novel genes within the pvd cluster were essential for pyoverdine-mediated iron uptake (Fig. 2). PA2385 encodes a putative 84 kDa acylase and is located downstream of pvdA, and a putative promoter for PA2385 containing a Fur box was detected within the pvdA coding sequence. Mutations in PA2392 or in the operon PA2393–2395 also led to a pyoverdine-negative phenotype, and possible functions of the encoded proteins deduced from sequence similarities are dipeptidase (PA2393), aminotransferase (PA2394) and iron(III)reductase (PA2395). PA2412 and PA2413, both of which were directly regulated by PvdS, were indispensable for pyoverdine synthesis as well and encoded a thioesterase and a probable aminotransferase. The PvdS-dependent operon containing PA2424 (pvdL) and PA2425 has recently been studied in great detail for its role in pyoverdine biosynthesis (D. Mossialos, U. Ochsner, C. Baysse, P. Chablain, J.-P. Pirnay, N. Koedam, et al., submitted). Additional genes within the pvd cluster may also play a role in pyoverdine production. For example, the operon containing PA2389–2391 and PA2427 (pvdY) all encode proteins of unknown function, and the corresponding mutants produced pyoverdine at significantly reduced levels. Mutations in PA2384 or in PA2403 led to phenotypes characterized by intense yellow pigmentation on LB agar, which may result from the accumulation of a chromophore precursor.

Figure 2.

Cluster of PvdS-regulated genes. Selected parts of a 120 kb portion of the PAO1 genome spanning from PA2383 to PA2452 are shown, and the gene names are given where known. Black arrows represent genes that are regulated by PvdS. White arrows represent genes that are controlled directly by Fur and contain Fur boxes (vertical black bars). Promoters are shown as boxes and are shaded if they harbour a PvdS-binding motif. Deletion mutants affected in specific genes within the cluster were generated, and their phenotypes are indicated, including fluorescence on CAA agar, pyoverdine concentration measured in supernatants of stationary phase CAA cultures and ability to grow on CAA agar containing the chelator EDDHA. Under these conditions, the pyoverdine levels of PAO1 wild-type cultures were 0.320 ± 0.04 mg ml −1 .


Iron starvation in P. aeruginosa leads to a distinct stress response characterized by the derepression or induction of genes involved in iron acquisition. The large number of genes affected by iron reflects the importance of this micronutrient for growth and pathogenesis. The growth medium itself had little influence on the global response to iron starvation. The data presented here are based on low- and high-iron D-TSB medium, but the groups of genes that were differentially expressed with respect to iron were very consistent between cells grown in either D-TSB or M9 medium (data not shown). One of the few exceptions was the PA3478–3479 (rhlA–rhlB) rhamnolipid biosynthesis operon, which is controlled by quorum sensing and RpoN (Ochsner and Reiser, 1995), and 80-fold iron regulated in M9 but constitutive in D-TSB. In contrast, the group of genes that were most active under high-iron conditions was much less conserved in cells grown in D-TSB versus M9 medium and will be discussed elsewhere.

The iron-regulated genes identified by GeneChip® analysis shown in Table 1 include the majority but not all Fur-regulated genes that we had isolated previously by cycle selection (Ochsner and Vasil, 1996; Vasil and Ochsner, 1999). Some of the genes that contain an ex­perimentally confirmed Fur box but were not expressed according to the GeneChip® data encode outer membrane receptors for siderophores produced by other bacterial or fungal species (Meyer, 1992). P. aeruginosa has evolved specific regulatory systems to allow expression of the receptor genes only in the presence of their cognate siderophore. This was first demonstrated for the ferrienterobactin receptor gene (pfeA), which is not only repressed by Fur in the presence of iron, but is also subject to positive control by the enterobactin-responsive PfeR–PfeS two-component regulatory system (Dean et al., 1996). Thus, the absence of pfeA transcripts in cells grown in D-TSB is not surprising, as this medium did not contain enterobactin. At least five additional Fur-regulated receptor genes (PA1910, PA0470, PA0931, PA2466, PA0151) are located next to a regulatory locus encoding either a two-component regulatory system or a sigma factor and transmembrane sensor. Also, these receptor genes were not expressed under the growth conditions used and may thus be inducible by their cognate siderophores. 

A critical evaluation of the GeneChip® results showed that the data obtained correlated very well with the measured iron regulation based on gene fusions to the lacZ reporter gene. Exceptions included toxA in which the iron regulation was only threefold using the GeneChip®, but 125-fold using a toxA–lacZ reporter gene. The two methods are quite different, in that the microarray data represent the current RNA pool, whereas the reporter gene constructs yield the total accumulation of β-galactosidase. It is possible that mRNAs with short half-lives may be under-represented; however, the toxA mRNA is quite stable with a half-life of 8–10 min (Lory, 1986). Earlier studies on toxA expression were based on Northern blot analysis or RNase protection assays and indicated that toxA transcripts were absent when the bacteria were grown in media containing iron at concentrations of 10 µM or higher (Lory, 1986; Ochsner et al., 1996). The failure of the GeneChip® to detect the strong iron regulation for toxA is more likely to be caused by a high background signal for particular probe pairs. For every gene, the P. aeruginosa GeneChip® contains at least 10 25-mer ‘perfect match probes’ and so-called ‘mismatch probes’ with a single mismatch at the centre base. The average differences of the hybridization signals are then calculated for each of these probe pairs. To investigate the reasons for the poor output relating to toxA iron regulation, we analysed the 13 probe pairs for toxA in greater detail (Fig. 3A). In several cases (probe pairs 5, 6 and 11), the intensities of the low-iron and high-iron signals were virtually identical. Moreover, in one case (probe pair 9), the mismatch probe gave excess hybridization signals over the perfect match counterpart, and this was observed for the low- or high-iron GeneChips. Removal of these uninformative probe pairs using the microarray suite tool ‘Probe Mask’ resulted in a more realistic iron regulation. Also, data obtained for individual genes that form an operon were often somewhat inconsistent and showed large fluctuations, although these genes were obviously transcribed from the same mRNA. A good example are the three genes PA2393, PA2394 and PA2395, which form an operon involved in pyoverdine synthesis as described above (Fig. 3B). Similarly, the five-gene operon PA4467–4471 (Hassett et al., 1997) that contains fumC (PA4470) and sodA (PA4468) resulted in a gradual decline of the observed iron regulation for each subsequent gene of the operon (Fig. 3B). However, the iron regulation pattern also remained sig­nificant (> fivefold) for the last gene of this operon. The fluctuations and decreasing intensities of the hybridization signals for genes located in a large operon are more likely to be an effect of natural RNA degradation from the 3′ end than a limitation of the GeneChip®. Thus, the overall performance of the GeneChip® and of the data analysis software was excellent.

Figure 3.

GeneChip® probe specificity (A) and variability within operons (B).

A. The location of the 13 GeneChip® probe pairs for the toxA gene are shown schematically as arrows, and the hybridization signals of target samples from low- or high-iron conditions to the perfect match or mismatch probes are indicated as black and grey bars respectively. Non-specific hybridization of a probe pair to target cDNA is marked by an asterisk, and excess hybridization to samples from either growth condition is indicated by vertical arrows.

B. GeneChip® expression profiling of multigene operons. A single promoter that drives expression of the polycistronic mRNA is shown by a black box. The iron regulation detected by two representative GeneChip® pairs illustrates the declining degree of regulation for each gene along the operon.

The GeneChip® technology allowed a rapid analysis of the expression profiles in response to iron. Numerous novel iron-regulated genes were identified, including genes encoding a possible Fur-dependent repressor (PA4570), a PvdS-regulated putative homologue of the haemolysin-III family (PA4833) and many proteins of unknown function. A large group of PvdS-regulated genes located in the pvd region were found to be essential for pyoverdine-mediated iron acquisition. Pyoverdine is composed of a peptide moiety of diverse structure among different Pseudomonads, and a conserved fluorescent chromophore, which can carry a variable acyl chain (Meyer, 2000). The peptide part contains unusual amino acids such as d-serine and N 5-formyl-N 5-hydroxyornithine and is produced by non-ribosomal peptide synthetases. The enzymatic pathway leading to the dihydroxyquinoline chromophore is poorly understood. Several investigative groups are actively contributing to the elucidation of the complex biosynthetic pathway for pyoverdine (Lehoux et al., 2000; Meyer, 2000; D. Mossialos, U. Ochsner, C. Baysse, P. Chablain, J.-P. Pirnay, N. Koedam, et al., Submitted), and our discovery of numerous novel genes with a previously unknown role in siderophore synthesis will support these efforts. Pyoverdine has gained recent attention for several reasons, including (i) its potential in the so-called siderotyping (Meyer, 2000); (ii) its use as a vehicle for the delivery of coupled antimicrobial agents through the cell envelope (Kinzel et al., 1998; Hennard et al., 2001); and (iii) its role in the siderophore-mediated signalling that regulates the expression of several virulence determinants including exotoxin A and protease (Lamont et al., 2002).

Experimental procedures

Bacterial strains, plasmids, media and chemicals

Pseudomonas aeruginosa PAO1 and an isogenic Δ pvdS ::Gm mutant ( Ochsner et al., 1996 ) were used for expression analysis. E. coli strains DH5α and SM10 were used as hosts for plasmid cloning. Luria–Bertani (LB) was used for strain maintenance and contained 1.5% agar (Difco) in solid media. Low-iron medium D-TSB was tryptic soy broth treated with Chelex-100 resin (Bio-Rad), dialysed and supplemented with 50 mM glutamate and 1% glycerol, and 200 µM FeCl 3 was added for high-iron medium. Alternatively, M9 minimal medium containing 0.4% glucose was used ( Sambrook et al., 1989 ), without or with 200 µM FeCl 3 . Casamino acids (CAA) medium was used for pyoverdine detection, and non-fluorescent mutants were also cultivated in CAA medium containing 0.5 g l −1 ethylenediamine di( o -hydroxy)phenylacetic acid (EDDHA). Cultures were grown aerobically for 18 h at 32°C and 260 r.p.m. Where needed, antibiotics were added at the following concentrations: for E. coli , ampicillin (100 µg ml −1 ), gentamicin (15 µg ml −1 ), tetracycline (15 µg ml −1 ); for P. aeruginosa , carbenicillin (750 µg ml −1 ), gentamicin (75 µg ml −1 ), tetracycline (150 µg ml −1 ).

RNA isolation, generation of cDNA probes and data analysis

Total RNA was isolated from P. aeruginosa by the hot phenol method followed by DNase I treatment, and the integrity of total RNA was confirmed by RNase protection analysis using a riboprobe specific for the constitutively expressed omlA gene as described in detail elsewhere (Barton et al., 1996; Ochsner et al., 1999). The GeneChip® probes were prepared according to the protocol supplied by the manufacturer (Affymetrix) with a few modifications. Briefly total RNA was purified further through RNeasy spin columns (Qiagen). The following reactions were performed in a GeneMate thermocycler (ISC Bioexpress). Control spike transcripts (130 pM) were added to 10 µg of RNA, and random primers (Invitrogen Life Technologies) were annealed (10 min at 70°C, 10 min at 25°C). First-strand cDNA was synthesized with 25 U µl−1 SuperScript II (Invitrogen Life Technologies) in the presence of 0.5 mM dNTPs, 0.5 U µl−1 SUPERase RNase inhibitor (Ambion) and 10 mM dithiothreitol (DTT; 10 min at 25°C, 60 min at 37°C, 60 min at 42°C, 10 min at 70°C). After RNA removal by alkaline treatment and neutralization, cDNA was purified using QuickSpin columns (Qiagen). For the fragmentation, 3 µg of cDNA and 0.75 U of RQ1 DNase I (Promega) were used (10 min at 37°C, 10 min at 98°C). The desired cDNA size range of 50–200 bases was verified by separating 200 ng of cDNA on a 4–20% acrylamide gel and staining with SYBR gold (Molecular Probes). The fragmented cDNA was then end-labelled with biotin-ddUTP using the Enzo BioArray™ terminal labelling kit (60 min at 37°C). Target hybridization, washing, staining and scanning were performed by the Affymetrix Core Facility using a GeneChip® hybridization oven, a Fluidics station and microarray suite software (Affymetrix).

General genetic procedures

Polymerase chain reaction (PCR) was performed using Taq polymerase and custom-made primers (Invitrogen Life Technologies) in a Perkin-Elmer Cetus thermal cycler, with 30 cycles of denaturing (1 min, 94°C), annealing (1 min, 55°C) and extension (1 min kb−1 DNA, 72°C). The PCR products were purified, cloned into pCRII-2.1 (Invitrogen) and sequenced with Sequenase 2.0 (United States Biochemical) and M13 primers. Transcriptional fusions to lacZ were made by cloning the promoters into pPZTC, which is a modified version of the reporter plasmid pPZ30 (Schweizer, 1991). Isogenic mutant strains affected in specific genes were generated as described in detail elsewhere (Ochsner et al., 2000b). Briefly, a suitable portion of the target gene was amplified by PCR and cloned into pCRII-2.1, followed by replacement of an internal gene fragment with a gentamicin (GmR) resistance cassette obtained from pPS856 (Hoang et al., 1998). The construct was ligated into pEX100T (Schweizer and Hoang, 1995), transformed into E. coli SM10 and used for allelic exchange in a biparental mating with P. aeruginosa PAO1. Gene replacements were verified by PCR across the relevant region.

Biochemical and other analytical methods

β-Galactosidase activities in soluble cell extracts were determined using ONPG (Sigma) as the substrate and expressed as International Units (U mg−1) with a millimolar extinction coefficient for ONPG of 3.1 as described previously (Ochsner et al., 2000b). Fluorescence of bacterial cultures was detected with a UV lamp at 254 nM. Pyoverdine concentrations were calculated from the absorbance at 403 nM of fivefold diluted culture supernatants, and purified pyoverdine was used as a standard.


We thank Cystic Fibrosis Foundation Therapeutics, Inc. for subsidising the Affymetrix GeneChip® during its introductory period. We also thank Gao Bifeng and Todd Woessner for technical assistance, and Iain Lamont for the generous gift of purified pyoverdine. This work was funded by a grant from the National Institutes of Health (AI-15940) to M.L.V. and by a grant from the Cystic Fibrosis Foundation to M.L.V. and P.J.W.