Using DNA microarray analysis, mRNA levels from wild-type Yersinia pestis cells treated with the iron chelator 2,2′-dipyridyl were compared with those supplemented with excessive iron, and subsequent to this, gene expression in the fur mutant was compared with that in the wild-type strain under iron rich conditions. The microarray analysis revealed many iron transport or storage systems that had been induced in response to the iron starvation, which is mediated by the Fur protein, using the iron as a co-repressor. The iron–Fur complex also affected some genes involved in various non-iron functions (ribonucleoside-diphosphate reductase, membrane proteins, electron transport and oxidative defense, etc.). The Fur protein still participated in the regulation of genes involved in broad cellular processes (virulence factors, pesticin activity, haemin storage and many proteins with unknown functions) that were not affected by iron depletion conditions. In addition to its classical negative regulatory activities, the Fur protein activates gene transcription. Using bioinformatics tools, we were able to predict the Y. pestis Fur box sequence that was clearly the over-presented motif in the promoter regions of members of the iron–Fur modulon.
In mammals, iron is bound to Fe3+-binding proteins and haemoproteins which maintain an exceedingly low level of free iron (10−18 M) in order to sustain bacterial growth (Wandersman & Delepelaire, 2004). Successful pathogenic bacteria have evolved a variety of strategies in order to acquire iron from the iron or haem resources of mammalian hosts (summarized in Table S1).
Iron uptake is carefully regulated by bacteria, often mediated by ferric uptake regulator (Fur) machinery (Escolar et al., 1999; Hantke, 2001; Andrews et al., 2003). Under iron-rich conditions, Fur binds the divalent ion, acquires a configuration able to bind target DNA sequences and inhibits the transcription of target genes and operons. Under iron-restricted conditions, Fur does not bind the ion, and the genes which are commonly found in iron-related functions are expressed, ensuring that the bacterial cells can scavenge iron from the extracellular environment.
In Yersinia pestis, five iron acquisition systems have been characterized (Yfe, Yfu, Ybt, Hmu and Has), and all of them are controlled by the Fur protein (Bearden et al., 1998; Thompson et al., 1999; Carniel, 2001; Gong et al., 2001; Rossi et al., 2001; Schubert, 2004). At least two (Ybt and Yfe) of them have proven to be required for the full virulence of Y. pestis (Bearden et al., 1998; Bearden & Perry, 1999). In this study we compared the gene expression profile of Y. pestis under both iron rich and depleted conditions. The global expression pattern of a fur mutant was subsequently tested under iron rich conditions. The gene regulation mediated by the Fur protein that plays a major role in the iron assimilation of Y. pestis was investigated.
Materials and methods
Bacterial strains and mutant construction
Yersinia pestis wild-type (WT) strain 201 was isolated from Microtus brandti in Inner Mongolia, China. It has major phenotypes as F1+ (able to produce fraction 1 antigen or the capsule), VW+ (presence of V antigen), Pst+ (able to produce pesticin) and Pgm+ (pigmentation on Congo-red media). Strain 201 has an LD50 of less than 100 cells in mice, following subcutaneous challenge. Strain 201 belongs to a newly established Y. pestis biovar, microtus (Zhou et al., 2004). Biovar microtus strains are thought to be avirulent to humans, although they are highly lethal to mice (Zhou et al., 2004).
Construction of the Y. pestis fur mutant strain was done as previously described by Datsenko and Wanner (Datsenko & Wanner, 2000). Briefly, plasmid pKD46 was first moved into Y. pestis 201 by electroporation. The fur::Kan mutagenic cassette was PCR amplified from plasmid pKD3 with the primers Fur-CM-F (5′-cagccttaatttgaatcgattgtaacaggactgaatccgctgtaacgcactgagaagc-3′) and Fur-CM-R (5′-gtgcttaaaatctttataagagtaatgcgataaaacgataagattgcagcattacacg-3′) and transformed into strain 201/pKD46. Mutants were selected by plating electroporated cells on agar plates containing kanamycin. Resistant transformants were subsequently cloned into a single colony. Chromosomal integration of the mutagenic cassette was confirmed by PCR and sequencing using oligonucleotides external to the integrated cassette (data not shown). The mutants were incubated overnight at 37°C and then tested for the loss of the temperature-sensitive plasmid pKD46 by looking for penicillin sensitivity. The elimination of the helper plasmid was verified by PCR (data not shown).
A chemically defined TMH medium (Straley & Bowmer, 1986) was used to cultivate the bacteria. Both the WT strain and the fur mutant were grown at 26°C to the middle exponential growth phase (an A620 of about 0.8) in TMH medium containing 10 μM of FeCl3 as the sole iron source. The cell cultures were diluted 1 : 20 in fresh TMH medium and grown for at least 10 generations in the medium at 26°C prior to reaching to the middle exponential growth phase (an A620 of about 0.8).
A comparison of transcriptional profiles was carried out for cells growing at 26 and 37°C, respectively. For cells grown at 37°C, the bacteria were pre-cultivated to the middle exponential phase at 26°C (since Y. pestis grows much more slowly in vitro at 37°C than at 26°C) and then the cultures at 26°C were transformed to 37°C for 1 h prior to the addition of the iron chelator 2,2′-dipyridyl (DP) or FeCl3.
The cultures obtained at 26 or 37°C, were divided and one-half was treated with 100 μM of DP to elicit iron starvation. The other half was supplemented with an additional 40 μM FeCl3 to ensure iron rich conditions. Growth was continued for 30 min at 26 or 37°C and the cells were then harvested for the isolation of total RNA.
DNA microarray analysis
Details of the DNA micrroarray analysis has been described previously (Qiu et al., 2005; Zhou et al., 2005). Briefly, total cellular RNA was isolated and then used to synthesize cDNA in the presence of aminoallyl-dUTP, genome directed primers (GDPs) and random hexamer primers. The aminoallyl-modified cDNA was then labelled with Cy5 or Cy3 dye. Glass slides spotted in duplicate with 4005 PCR amplicons, representing about 95% of the non-redundant annotated genes of Y. pestis CO92 (Parkhill et al., 2001) and 91001 (Song et al., 2004), were used for probe hybridization. The scanning images were processed with GenePix Pro 4.1 software (Axon Instruments). Spots with a background-corrected signal intensity (median), in both channels, of lower than twofold background intensity (median) were rejected from further analysis, and the remaining data points were then normalized by total intensity normalization methods. The normalized log2 ratio of the test/reference signal for each spot was recorded. The averaged log2 ratio for each gene was finally calculated. Significant changes in gene expression were identified using Significance Analysis of Microarrays (SAM) software (Tusher et al., 2001).
Computational analysis of the Fur-binding sites
The 500-bp promoter regions extending upstream from the start codon of every first gene in each indicated transcriptional unit were retrieved with the ‘retrieve-seq’ program (http://rsat.ulb.ac.be/rsat/). The MotifSampler program (Thijs et al., 2002) was used for detection of the over-represented motifs within the promoter regions of a set of Fur-regulated genes. Following this, the consensus sequence logo was built by compiling the potential motif sequences of each gene with the ‘Logo’ algorithm (Schneider & Stephens, 1990).
The ‘fuzznuc’ routine, a part of the Emboss suite of programs (http://emboss.sourceforge.net), was used to match the Fur consensus sequence within the promoter regions, and the coliBASE Pattern Search tool (http://colibase.bham.ac.uk/pattern/index.cgi?help=pattern&frame=pattern) to search the Y. pestis genome for the homologous Fur-binding sites.
The iron–Fur modulon in Yersinia pestis
We first compared the RNA from WT cells supplemented with iron with that treated with DP, and then gene expression in the fur mutant was compared with expression in the WT strain in the presence of supplementary iron. Here the iron–Fur modulon was defined as those genes whose transcription was affected by both DP treatment and the fur mutation. Accordingly, 56 genes were found to be regulated by the iron–Fur complex, and among these, 48 were positively controlled and 8 negatively (Table 1).
The most obvious transcriptional change was detected in the 29 genes involved in iron uptake. The transcription of these genes was induced under both iron starvation as well as in the fur mutant. They can be assigned to four types of iron uptake system. The first includes the 13 genes encoding components of siderophore-based iron transport systems. The second is the hmuRSTUV operon. The haemoprotein-receptor-based Hmu system plays roles in both the utilization of haemin and various haemoproteins in Y. pestis (Thompson et al., 1999). The third type contains the yfeABCD and yfuABC operons. The yfe locus comprises five genes arranged in two distinct operons, yfeABCD and yfeE, the larger of which encodes an ABC transport system involved in the transport of both iron and manganese (Bearden et al., 1998; Bearden & Perry, 1999). The yfu locus consists of three genes, yfuA, B and C (Gong et al., 2001). Both Yfu and Yfe belong to the metal-transporting binding-protein-dependent ABC system. The final group is the TonB–ExbB–ExbD complex (Perry et al., 2003; Postle & Kadner, 2003).
Bacteria possess two types of iron-storage protein, the haem-containing bacterioferritins (Bfr) and the haem-free ferritins encoded by ftnA (Andrews, 1998). bfr is the second gene in an operon and is preceded by bfd. We observed that bfd and bfr were up-regulated under both iron starvation and in the fur mutant, supporting the notion that Bfd acts as a Bfr reductase mediating the release of iron from Bfr under iron restriction (Garg et al., 1996; Quail et al., 1996). However, the transcription of ftnA is repressed following iron starvation, while it is induced in the fur mutant.
Various/unknown functions negatively controlled by Fur
The iron – Fur complex most likely repressed the transcription of the nrdHIEF operon in Y. pestis, an effect which has been observed in E. coli (Vassinova & Kozyrev, 2000; McHugh et al., 2003). In addition, transcription of 13 genes with unknown or unsigned functions appeared to be repressed by the iron–Fur complex. Among them are the mntH gene encoding a manganese transport protein, YPO1528 encoding a homology of ferric iron reductase, and five genes encoding membrane components. Taken together, these genes may represent additional, uncharacterized candidates which play a role under iron-restricted conditions. Remarkably, the induction of the selected component of the cell membrane would presumably affect the permeability of the cell membrane and may play a previously unappreciated role in iron homeostasis.
Functions positively controlled by Fur
The napPFDABC operon, sodB and katA, were shown to be repressed by both DP treatment and fur mutation, indicating that the transcription of these genes was induced by the iron–Fur complex. The six nap genes encode the periplasmic nitrate reductase system. Oxidative stress is potentiated by iron because iron participates, with H2O2, in the Fenton reaction, leading to the formation of a very reactive hydroxyl radical (OH−), which can react with any cellular macromolecule (Touati, 2000). Both iron superoxide dismutase (SodB) and catalase (KatA) contribute towards counteracting this damage. There is increasing evidence for a coordination between the regulation of iron homeostasis and defence against oxidative stress (Touati, 2000).
Additional effects of the fur mutation on gene expression
We observed a broad array of genes (Table S2) whose transcription was affected by the fur mutation but not by the DP treatment. An interesting observation was the negative regulation of an array of virulence-related genes, indicating that Fur acted as a repressor of virulence genes under iron-rich conditions. Among them are the caf1 and psa operons that encode the F1 capsule (Zavialov et al., 2003) and pH6 antigen (Payne et al., 1998), respectively. Both of them (Du et al., 2002; Huang & Lindler, 2004) function as antiphagocytic factors independent of Yops (Bleves & Cornelis, 2000). Additional virulence-related genes include slyA, ail and the ure operon. SlyA is required for the virulence and survival of Salmonella enteritidis in professional macrophages (Stapleton et al., 2002). The ail gene is required for invasion and serum resistance phenotypes in the enteropathogenic Yersinia spp. The production of urease encoded by the ure operon contributes to the pathogenicity of Y. pseudotuberculosis, as it is necessary for oral transmission (De Koning-Ward & Robins-Browne, 1995).
The pgm locus and haemin storage
The pgm locus (YPO1902-1967 on CO92 chromosome) contains the two established virulence-related gene clusters, the high-pathogenicity island (HPI) and the hms (haemin storage) system (Buchrieser et al., 1999). We observed the obviously reduced expression of the cluster YPO1937-1965 in the fur mutant (Fig. 1), indicating a positive regulation of the corresponding genes by the Fur protein under iron-rich conditions. This cluster can be roughly separated into three functional parts: a membrane-related gene cluster, the hms locus that accounts for the hms phenotype and the ast operon responsible for arginine utilization.
The pesticin activity gene (pst) is located on plasmid pPCP1. The psn gene from the pgm locus encodes the pesticin receptor, and the Y. pestis strains with an in-frame deletion in psn are pesticin resistant (Fetherston et al., 1995). The pesticin sensitivity of Y. pestis is affected by the iron status of the cells (Brubaker & Surgalla, 1961), and the fur mutation increases pesticin sensitivity in Y. pestis strains (Staggs et al., 1994). In this study, transcription of pst was enhanced in a fur mutant under iron-rich conditions, while that of psn was repressed.
Prediction of Yersinia pestis Fur box consensus
The Escherichia coli Fur protein, following its association with iron, binds a 19-bp consensus site with the sequence 5′-GATAATGATAATCATTATC-3′ (known as the classic Fur box), and represses the transcription of the downstream genes (Escolar et al., 1999). Genomic transcriptional profiling of the fur mutant enables us to define the iron–Fur modulon of Y. pestis (Table 1). All the gene members of this modulon were examined visually, according to their transcriptional organization in relation to the surrounding genes, to identify putative transcriptional units (defined as a cluster of adjacent genes that have intergenic regions <200 bp in length and were putatively transcribed in the same orientation). Members of the iron–Fur modulon were combined into 33 putative transcriptional units (Table 1). Following this, the MotifSampler program (Thijs et al., 2002) was used to find over-represented motifs in the promoter regions of these transcriptional units. This analysis generated a conserved motif that was thought to be the Fur box consensus in Y. pestis (Fig. 2). The resulting 19 bp consensus sequence is a 9-1-9 inverted repeat (5′-AATGATAATNATTATCATT-3′).
Transcriptome analysis of Fur regulation in Yersinia pestis
In this work we have reported on the global transcriptional responses of Yersinia pestis, grown at both 26 and 37°C, to iron starvation. Complementary microarray experiments were performed to compare the global gene expression patterns of both the WT strain and the fur disruptant grown under iron-rich conditions. During its life cycle, Y. pestis must acclimatize itself to the shift in temperature between the flea blockage (about 26°C) and that of a warm-blooded host (37°C) (Perry & Fetherston, 1997). Cells grown in TMH media at 37°C or 26°C shared highly conserved transcriptional profiles in response to an iron-starved environment (Fig. 3). No growth-temperature-specific iron uptake or storage system was identified in this work. Temperature appears to affect the regulation of Fur on some cellular pathways, but this impact shows no influence on Fur-dependent iron assimilation functions (see Tables 1 and S2).
The iron–Fur modulon was defined as those genes whose transcription was affected by both DP treatment and the fur mutation, as shown in Table 1. The Feo system, encoded by the Fur-dependent feoAB genes in E. coli, is specific for ferrous iron transport (Andrews et al., 2003; Hantke, 2003). In our experiments, transcription of the feoAB genes was not affected by iron depletion (the reason for this may be that Fe3+ was used as the sole iron resource), but was dependent on Fur (Table 1). Table 1 still included two known Fur-dependent iron uptake systems from Y. pestis (Ybt and Has). The ybt operon was DP-inducible but not affected by Fur, while both DP and Fur did not affect the has genes in this study. The six has genes form two operons, hasRADEB and hasF. The haemophore-based Has system is involved in haemoprotein utilization, but does not contribute to haemin utilization in Y. pestis (Rossi et al., 2001). Ybt is a siderophore-based transport system composed of 11 genes organized into four operons (reviewed in references Carniel, 2001; Schubert, 2004). The discrepancy in the results from Ybt and Has is probably related to the growth conditions employed here, which are unlikely to favour the expression of all the Fur-controlled genes.
Overall, the Fur regulatory system in Y. pestis is quite complex. Many iron acquisition or storage systems are regulated in response to iron availability, which is mediated by the Fur protein using iron as a co-factor (Table 1). Since Fur is defined by its repression of iron uptake genes, a fur mutant may have a higher influx of iron than a WT strain under iron-rich conditions, which would affect bacterial growth and protein expression. Indeed, the iron–Fur complex also affected some genes involved in various non-iron functions (Table 1). The Fur protein still participated in the regulation of genes involved in broad cellular processes that were not affected by iron depletion (Table S2). In addition to its classical negative regulatory activities, Fur still activates gene transcription. Y. pestis Fur appears to be a general regulator rather than a specific repressor.
The Biovar microtus strain 201 used in this study is avirulent to humans, but highly lethal to mice (Zhou et al., 2004). Further studies should be conducted in order to investigate the differences or similarities between pathogenic and avirulent Y. pestis strains in relation to their Fur mediated regulation of iron assimilation.
Computational promoter analysis of Fur regulation
Our microarray analysis identified genes which were both directly and indirectly controlled by the Fur regulator. Using bioinformatics tools, we predicted the existence of the Y. pestis Fur box sequence that was clearly the over-presented motif in the promoter regions of members of the iron–Fur modulon. The predicted Fur box may be more accurately represented in Y. pestis than the classic E. coli 19-bp sequence. The existence of homologues of the Y. pestis Fur box could be predicted in the promoter sequences of most of the Fur-regulated genes identified by the microarray analysis. The presence of a sequence with a high similarity to the Fur box is a very good predictor of Fur-specific binding. Further biochemical experiments, such as a gel mobility shift assay and DNase I footprinting analysis, will be needed in order to determine whether these genes are the functional targets of Fur-specific binding.
Three interpretations of the functional pattern of Fur binding to the E. coli Fur box have been proposed: originally, Fur was assumed to recognize two 9-bp inverted repeats, 5′-GATAATGAT-3′, separated by one unmatched A; the same 19-bp sequence can be viewed as a combination of three adjacent repeats, 5′-NATA/TAT-3′ in a head-to-head-to-tail orientation; and that the 19-bp consensus sequence could in fact be recognized by Fur as three repeats of the NATA/TA hexamer rather than as a 19-bp palindrome (Escolar et al., 1999; Andrews et al., 2003). The Bacillus subtilis Fur box was identified as a 7-1-7 inverted repeat of TGATAATNATTATCA (Baichoo et al., 2002). In addition, modelling of the recently determined structure of Pseudomonas aeruginosa Fur on to a DNA duplex also suggested that two dimers bind per 19-bp Fur box (Pohl et al., 2003). These results imply that the primary DNA determinants for Fur binding are likely to be similar to these three organisms and include a core 7-1-7 (or closely related) inverted repeat of TGATAATNATTATCA (Escolar et al., 1999; Baichoo & Helmann, 2002; Baichoo et al., 2002; Andrews et al., 2003; Pohl et al., 2003), which is also the case for Y. pestis, as was shown in this study.
We searched the genome of Y. pestis CO92 for sites matching the predicted 19-bp Y. pestis Fur box consensus sequence using the coliBASE Pattern Search tool, set up with a 4-bp mismatch. A total of 99 genes were found to carry a DNA sequence with sufficient homology to the Fur consensus binding sequence within 500-bp upstream sequences; 28 of these were differentially regulated by Fur as determined by the microarray analysis. In some instances (24 genes), distinct Fur motifs were found more than once on the same intergenic region.
Fur is a member of a family of related regulators including DtxR, PerR and Zur. There are at least two members (Fur and Zur) of this family of regulators encoded on the CO92 genome. There is the high level of similarity between the Zur and Fur binding sites (Baichoo & Helmann, 2002). Therefore, a whole-genome matching of the Fur box sequence may include Fur, as well as Zur, binding sequences.
Financial support for this work came from the National Natural Science Foundation of China (grant no. 30430620). D.Z. and L.Q. contributed equally to this work.