INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

The OxyR transcription factor is found in many prokaryotic organisms (reviewed in reference 17). This LysR-type regulator activates the expression of numerous genes in response to oxidative stress, in particular upon exposure to hydrogen peroxide, and has served as a paradigm for understanding cellular sensing of oxidative stress. Hydrogen peroxide directly activates OxyR through the formation of an intramolecular disulfide bond (22). Oxidized OxyR then activates transcription of antioxidant genes, including katG (encoding hydroperoxidase I), ahpCF (encoding an alkyl hydroperoxide reductase), dps (encoding a nonspecific DNA binding protein), gorA (encoding glutathione reductase), grxA (encoding glutaredoxin I), and oxyS (encoding a small regulatory RNA) (reviewed in reference 17).

Much has been learned about cellular defenses against oxidative stress through studies of the OxyR regulon. Thus, as part of a continuing effort to better define the physiological roles of OxyR, we initiated a computational approach to identify additional OxyR-regulated genes. We used an algorithm based on information theory (11) that uses previously identified OxyR binding site sequences as a model to search through the entire Escherichia coli genome for putative OxyR binding sites. Our computational approach predicted several sites that had not been identified previously by experimental means. One target gene, identified using this approach and experimentally confirmed to be regulated by OxyR, encodes the iron metabolism regulator Fur, demonstrating coordinate regulation between oxidative stress response and iron metabolism (23).

Here we report our computer-directed identification of three more OxyR binding sites: one upstream of dsbG, one upstream of fhuF, and one within yfdI. The dsbG gene encodes a periplasmic chaperone-isomerase that facilitates correct folding of disulfide bond proteins (1, 2, 14, 20). The fhuF gene encodes a 2Fe-2S protein whose likely function is ferric iron reduction, required for iron uptake by the cell (8). The yfdI-encoded gene product shares homology with ligases, but its biochemical function has not been experimentally demonstrated. We confirmed the presence of OxyR binding to the three sites by in vitro DNase I footprinting assays. We also examined expression of the dsbG and fhuF genes and found the regulation of these two genes to be unique.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Computer search program. The initial computational search was carried out as described elsewhere (23). Seven OxyR target sites identified by footprinting, E. coli oxyR (position 163 in GenBank entry JO4553), katG (position 68 in GenBank entry M21516), ahpC (position 116 in GenBank entry D13187), dps (position 202 in GenBank entry X69337), gor (position 60491 in GenBank entry U00039), and grxA (position 207 in GenBank entry M13449); Mu phage mom-1 (position 68 in GenBank entry VO1463); Salmonella orf; and two sites identified by homology, E. coli ahpC (position 116 in GenBank entry D13187) and a second Mu phage mom-2 site (position 59 in GenBank entry VO1463), were used to generate an individual information weight matrix (11). The matrix was scanned across the entire E. coli genomic sequence (3), and the identified sites were sorted by information content. Local regions of the genome surrounding the strongest sites were displayed using the Lister program (version 9.02) (see Fig. 1, 4, and 7) to show coding regions along with sequence walkers representing potential binding sites (12). Detailed Lister maps of the 20 sites with the highest information content are available online (http://www.lecb.ncifcrf.gov/~toms/paper/zheng.storz2001/). A subsequent computational search was carried out using the OxyR target sites at E. coli oxyR, katG, ahpC, dps, gor, and grxA and at Mu phage mom-1 as well as at E. coli fur, dsbG, fhuF-1, fhuF-2, yfdI, trxC, flu, sufA, and hemH (positions 710103, 637851, 4603273, 4603357, 2467337, 2716640, 2069355, 1762663, and 497211 in GenBank entry U00096, respectively) to generate the individual information weight matrix.

Plasmids and strains. The DNA sequence and coordinates are for E. coli from GenBank accession no. U00096 (3). The plasmids used in the study were constructed using fragments PCR amplified from chromosomal DNA. The sequences of all oligonucleotides are listed in Table 1. To generate the dsbG promoter plasmid (pGSO123) used to test for OxyR binding, a 230-bp fragment generated using primers 433 and 434 was cloned into the EcoRI and BamHI sites of pUC8. To generate the ahpC-dsbG promoter plasmid (pGSO124) used to test for RNA polymerase binding in the presence of OxyR, a 430-bp fragment generated using primers 709 and 710 was cloned into the SmaI site of pRS415. To construct the fhuF promoter plasmid (pGSO129) used to test for OxyR binding, a 240-bp fragment generated using primers 473 and 474 was cloned into the EcoRI and BamHI sites of pUC18. To generate the ydfI plasmid (pGSO130) to test for OxyR binding, a 180-bp fragment generated using primers 475 and 476 was cloned into the EcoRI and BamHI sites of pUC18. The sequence of all inserts was verified. The strains used in the study, MC4100 (wild type), GSO47 (MC4100 oxyR::kan), and GSO72 (MC4100 fur::kan), were described previously (23).

Growth conditions. Cultures were grown at 37°C in Luria-Bertani (LB) rich medium or M63 minimal medium supplemented with 2 mg of glucose/ml and 20 µg of vitamin B₁/ml (15). 2,2'-Dipyridyl (0.5 mM) was added to some cultures to achieve iron depletion.

RNA isolation and primer extension assays. Exponential-phase cultures (optical density at 600 nm = 0.2 to 0.5) were split into aliquots: one aliquot was left untreated, and the other aliquot was treated with the indicated amounts of hydrogen peroxide. After 10 min, the cells from 5, 10, or 25 ml of culture were harvested and resuspended in 1 ml of Trizol equilibrated at 4°C (Gibco BRL). All subsequent purification steps were carried out according to the Trizol reagent manual (based on reference 4). RNA samples were subjected to primer extension assays as described elsewhere (21), using primer 709 specific to ahpC, primer 610 specific to dsbG, and primers 704 and 706 specific to fhuF.

DNase I footprinting. The DNase I footprinting assays of purified OxyR binding to the dsbG, fhuF, and yfdI fragments were carried out as described previously (19).

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Initial search for new OxyR binding sites. An initial OxyR binding site model was constructed using nine OxyR target sequences. Seven of these binding sites, upstream of the E. coli oxyR, katG, dps, gorA, and grxA genes; the Salmonella orf gene; and the Mu phage mom gene, were previously confirmed by DNase I footprinting. Two sites, a second Mu phage mom site and a site upstream of the E. coli ahpC gene, were included based on their homology to the confirmed binding sequences. The average information content of the nine sites used in this model is 16.7 ± 1.9 bits. Information content is a measure of the amount of pattern with respect to the model and is quantitated in bits, the choice between two equally likely possibilities. Our initial model was used to search the entire E. coli genome sequence for additional OxyR binding sites. Table 2 lists the 20 sites with the highest information content based on this initial model. The information content of the 20 sites ranges from 11.8 to 26.0 bits, and five of the six E. coli OxyR binding sites used in the model fell within the top 20 sequences. The sixth E. coli site, upstream of gorA, had an information content of 11.2 bits. In a previous study, we showed that OxyR binds and regulates the promoter of the fur gene, one of the sites identified by our search (23). Here we examine OxyR binding to the sites upstream of the dsbG, fhuF, nmpC, and ybaL genes and to sites within the yfdI and ybbW open reading frames (ORFs).

OxyR binding upstream of dsbG. The computer calculation predicted an OxyR binding site centered at position 637851, 54 bp upstream of the dsbG start codon and 238 bp away from the center of a known OxyR binding site proximal to the divergently transcribed ahpC gene (Fig. 1). The predicted site has an information content of 23.1 bits, which is higher than that of all known OxyR sites except the one in the grxA promoter region (26.0 bits). The dsbG site is homologous to the OxyR binding site detected upstream of the putative Salmonella enterica serovar Typhimurium dsbG gene (18) and previously designated Salmonella orf (19). To verify OxyR binding to this E. coli site, we carried out DNase I footprinting assays. We found that oxidized (Fig. 2A) but not reduced (data not shown) OxyR binds to the predicted site with an affinity that is comparable to that for the site in the ahpC promoter.

View larger version (49K): [in this window] [in a new window]   FIG. 1.   Sequence of the ahpC and dsbG promoters. The ahpC and dsbG transcription starts are marked by black arrows, and the starts of the corresponding ORFs are denoted by white arrows. The DNase I footprints for OxyR binding to the top and bottom strands are indicated by the dark gray boxes. The DNase I footprint for RNA polymerase binding adjacent to the dsbG-proximal OxyR binding site on the bottom strand is indicated by the light gray box. The locations of the two predicted oxyR sites are shown by sequence walkers (12), in which the rectangles surrounding the T's at positions 637851 and 638089 indicate the centers of the binding sites. A sequence walker consists of a string of letters in which the height of each letter shows the contribution that the corresponding base would make to the average sequence conservation shown by the sequence logo of all binding sites (10). The sequence walker is given on a scale of bits of information. The scale, from 3 bits up to the maximum conservation at +2 bits, is given by the rectangle surrounding the T. Positively contributing bases are above the zero line, and negatively contributing bases are below the line. By using bits, the heights of all letters can be added together to obtain the information content of a site.

View larger version (85K): [in this window] [in a new window]   FIG. 2.   DNase I footprinting assays of oxidized OxyR binding to the top and bottom strands of the dsbG promoter in the absence of RNA polymerase (A) and to the top strand in the presence of RNA polymerase (B). The regions protected by OxyR are indicated by the brackets in panel A. All samples were run in parallel with Maxam-Gilbert G/A sequencing ladders. (A) For OxyR binding to the top strand relative to the dsbG promoter, the 230-bp EcoRI-BamHI fragment of pGSO123 was labeled with 32P at the EcoRI site. For OxyR binding to the bottom strand relative to the dsbG promoter, the 32P-labeled primer 710 and unlabeled primer 709 were used to PCR amplify a 430-bp fragment containing both the ahpC and dsbG promoter sequences. (B) For OxyR and RNA polymerase binding to the top strand relative to the dsbG promoter, a 250-bp fragment was PCR amplified from pGSO124 using primers 710 and 726. The amplified fragment was digested with EcoRI, labeled with 32P, and then digested with SmaI.

OxyR binding upstream of dsbG activates ahpCF expression. For all previously known OxyR-activated genes, OxyR binds at a site directly upstream of the 35 region of a target gene (19). Thus, the proximity of the OxyR binding site to the dsbG start codon led us to speculate that OxyR binding to this site is required for OxyR activation of dsbG expression in response to oxidative stress. However, multiple primer extension experiments failed to detect a dsbG transcription start in the vicinity of the OxyR binding site. To further elucidate the role of OxyR binding to the site upstream of dsbG, we examined RNA polymerase binding in the presence of OxyR. Previous studies have shown that oxidized OxyR binding leads to RNA polymerase recruitment to the promoter (6). As shown in Fig. 2B, RNA polymerase did not bind to the promoter fragment in the absence of other proteins. RNA polymerase binding was observed in the presence of OxyR. Surprisingly, the binding was to sequences upstream rather than downstream of the OxyR binding site relative to the dsbG start codon. This result suggested that OxyR binding to the site adjacent to dsbG was functioning to activate expression of a transcript encoding ahpC. Indeed, Northern blotting (data not shown) and primer extension assays (Fig. 3) of wild-type and oxyR mutant cells grown in rich medium showed that hydrogen peroxide treatment led to the OxyR-dependent induction of two ahpC mRNAs, one transcript initiating at position 638144 and a second initiating at position 637916. Primer extension assays carried out with a variety of primers throughout the entire ahpC-dsbG inter-ORF region led to the detection of a hydrogen peroxide-induced transcript encoding dsbG initiating at position 638023. This start suggests that there is an approximately 100-bp overlap between the longer ahpC transcript and the dsbG mRNA. Whole-genome expression experiments also show the presence of overlapping ahpC and dsbG mRNAs (C. Rosenow, unpublished data), but the reason for the overlapping transcripts has not been elucidated.

View larger version (71K): [in this window] [in a new window]   FIG. 3.   Primer extension assays of ahpC and dsbG expression in wild-type and oxyR mutant strains grown in LB medium. Exponential-phase cultures were split into two aliquots: one aliquot was left untreated, and the other was treated with 1 mM hydrogen peroxide. The cells were harvested after 10 min, total RNA was isolated, and primer extension assays were carried out with primer 709 specific to ahpC and primer 610 specific to dsbG. The neighboring sequencing reactions were carried out with the same primers.

OxyR binding to the fhuF promoter. The computer search indicated an OxyR binding site centered at coordinate 4603357, 126 bases upstream of the start codon of the fhuF gene (Fig. 4). This site has an information content of 11.8 bits, which is lower than that of all the known OxyR binding sites except the one in the gorA promoter (11.2 bits). However, given our previous findings that OxyR regulates the expression of the iron metabolism regulator Fur, possible OxyR regulation of a putative iron reductase was of interest. Thus, we carried out DNase I footprinting experiments to test for OxyR binding to the promoter region of fhuF. As shown in Fig. 5, the site predicted by the computational search (fhuF-2) indeed is protected from DNase I digestion by the oxidized OxyR protein. Interestingly, we observed another oxidized OxyR binding site (fhuF-1) of slightly higher affinity centered at coordinate 4603273, 42 bases upstream of the fhuF start codon. Since the second site had an information content of 4.8 bits, OxyR binding to this position was an indication that the initial binding site model could be improved.

View larger version (41K): [in this window] [in a new window]   FIG. 4.   Sequence of the fhuF promoter. The fhuF transcription start is marked by the black arrow, and the start of the FhuF ORF is denoted by the white arrow. The DNase I footprints for OxyR binding to the top and bottom strands are indicated by the dark gray boxes. The DNase I footprints for Fur binding to the top strands (K. A. Lewis, B. Doan, M. Zheng, G. Storz, and T. D. Schneider, unpublished data) are indicated by the light gray boxes. The locations of the two predicted oxyR sites are shown by sequence walkers (12), in which the rectangles surrounding the A at position 4603357 and the T at position 4603273 indicate the centers of the binding sites. The site with an information content of 11.8 bits is designated fhuF-2, and the site with an information content of 4.8 bits is designated fhuF-1.

View larger version (69K): [in this window] [in a new window]   FIG. 5.   DNase I footprinting assays of oxidized OxyR binding to the top and bottom strands of the fhuF promoter. The regions protected by OxyR on both strands are indicated by the brackets. The 240-bp BamHI-EcoRI fragment of pGSO129 was labeled with 32P at either the BamHI site (top strand) or the EcoRI site (bottom strand). The samples were run in parallel with Maxam-Gilbert G/A sequencing ladders.

OxyR and Fur repression of fhuF. To determine whether fhuF is regulated by OxyR, the wild-type and the oxyR and fur mutant strains were grown in the absence and presence of the iron chelator 2,2'-dipyridyl. We then examined fhuF expression without and with hydrogen peroxide treatment by primer extension assays (Fig. 6). These assays showed that, in the absence of oxidative stress, fhuF mRNA levels are strongly repressed by Fur under iron-rich (compare lanes 1 and 3 with lane 5) but not iron-poor (compare lanes 7, 9, and 11) conditions. This repression is in agreement with the strong Fur regulation reported by Stojiljkovic et al. (16). Treatment with hydrogen peroxide led to fhuF repression in the wild-type and the fur mutant strains, but not the oxyR mutant strain, under both iron-rich and iron-depleted conditions (compare lanes 2, 6, 8, and 12 with lanes 4 and 10). These results show that OxyR represses fhuF expression under conditions of oxidative stress. The derepression of fhuF observed when the oxyR strain grown under iron-rich conditions was treated with hydrogen peroxide may be due to reduced Fur binding upon oxidative damage to iron-loaded Fur.

View larger version (51K): [in this window] [in a new window]   FIG. 6.   Primer extension assays of fhuF expression in wild-type, oxyR, and fur strains grown in LB medium without (lanes 1 to 6) and with (lanes 7 to 12) 1 mM 2,2'-dipyridyl. Exponential-phase cultures were split into two aliquots: one aliquot was left untreated, and the other was treated with 1 mM hydrogen peroxide. The cells were harvested after 10 min, total RNA was isolated, and primer extension assays were carried out with primer 706 specific to fhuF. The neighboring sequencing reactions were carried out with the same primer.

OxyR binding to the yfdI gene. In addition to predicting OxyR binding sites upstream of the fur (23), dsbG, and fhuF genes, our computer model predicted OxyR sites centered 9 bp upstream of the nmpC and 81 bp upstream of the ybaL start codons as well as sites within ORFs. To test whether OxyR bound to these sites, we carried out DNase I footprinting experiments. We did not observe strong binding to the nmpC and ybaL promoter regions or to a fragment carrying the predicted site within the ybbW ORF (data not shown). However, we did observe strong binding by oxidized OxyR (Fig. 7 and 8) and weak binding by reduced OxyR (data not shown) to the predicted site within the yfdI ORF. The role of OxyR binding to this site is not clear. We carried out primer extension assays using five different primers but failed to detect the start of an OxyR-regulated transcript on either strand within the vicinity of the binding site. Microarray experiments showed that the expression of a transcript encoded within yfdI is repressed upon treatment with hydrogen peroxide, but this repression is observed in both a wild-type strain and an oxyR mutant strain (24).

View larger version (16K): [in this window] [in a new window]   FIG. 7.   Sequence of the yfdI gene. The YfdI ORF, which extends from position 2467151 to 2468482, is indicated by the white arrow. The DNase I footprints for OxyR binding to the top and bottom strands are indicated by the dark gray boxes. The location of the predicted oxyR site is shown by a sequence walker (12), in which the rectangle surrounding the A at position 2467330 indicates the center of the binding site.

View larger version (72K): [in this window] [in a new window]   FIG. 8.   DNase I footprinting assays of oxidized OxyR binding to the top and bottom strands of the yfdI gene. The regions protected by OxyR on both strands are indicated by the brackets. The 180-bp BamHI-EcoRI fragment of pGSO130 was labeled with 32P at either the BamHI site (top strand) or the EcoRI site (bottom strand). The samples were run in parallel with Maxam-Gilbert G/A sequencing ladders.

Improved search for new OxyR binding sites. Seven previously identified sites (oxyR, katG, ahpC, dps, gorA, mom, and grxA), five OxyR binding sites identified in the initial computational search (fur, dsbG, fhuF-1, fhuF-2, and yfdI), two sites identified by our DNA microarray studies (sufA and hemH) (24), and two sites identified in other studies (flu and trxC) (9; M. Zheng and G. Storz, unpublished data) were used to generate a new model to search the complete E. coli genome. A direct comparison of the initial and the improved models is given online (http://www.lecb.ncifcrf.gov/~toms/paper/zheng.storz2001/), and the 20 sites with the highest information content (ranging from 15.1 to 29.3 bits) based on the second model are listed in Table 2. Although the overall compositions of the two models were similar, a few differences were noted: additional bases were accommodated at some positions (such as a G at position 5 relative to the center of the binding site), and there was some focusing on specific residues at other positions (such as on A at position 4). Interestingly, using the second model, four of the known OxyR binding sites, at the dps, trxC, hemH, and gor promoters, were no longer among the 20 sites with the highest information content. The gorA gene does not show strong OxyR regulation (7), but dps, trxC, and hemH are clearly induced by OxyR in response to oxidative stress (9, 17, 24). It also is noteworthy that several of the top 20 sites identified by the improved search are within ORFs.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

Use of a computational approach based on information theory to identify OxyR binding sites. As the complete sequences of more and more genomes become available, the extraction of biological information by computational analysis of sequence data is quickly becoming an indispensable tool in biological research. While searches for gene homology are routine, searches for DNA binding sites are less common, in part because the sequences recognized by a DNA binding protein are short and often vary considerably from a simple consensus sequence. A quantitative model based on information theory was constructed for describing the DNA recognition sequences for the OxyR transcription factor. This model was used to search for additional OxyR target sequences that were then experimentally tested for OxyR binding in vitro and OxyR-dependent expression in vivo.

Identification of OxyR-regulated genes. Our current search has led to the identification of a new OxyR-activated gene, dsbG, and a new OxyR-repressed gene, fhuF. The periplasmic DsbG protein has been shown to be a chaperone and disulfide bond isomerase involved in the correct folding of disulfide-containing proteins (2, 14). Periplasmic disulfide bond formation is a highly controlled process involving a series of Dsb proteins (5). The rate of disulfide bond formation is likely to increase upon exposure to hydrogen peroxide. Thus, the induction of disulfide bond isomerases may be critical in a defense against peroxide stress. In this context, it would be interesting to test whether expression of other E. coli Dsb proteins is modulated by oxidative stress. The FhuF protein has been suggested previously to be a ferric ion reductase that reduces siderophore-bound ferric ion upon import into the cytoplasm (8). Repression of fhuF transcription by OxyR may lead to decreased levels of FhuF protein which in turn would slow iron uptake, thus providing another mechanism for the cell to minimize the escalation of oxidative damage by the Fenton reaction.

Noncanonical modes of regulation by OxyR. For most of the OxyR-activated genes, OxyR binding sites are located upstream of the 35 region (19). It is likely that OxyR binding to these sequences allows for RNA polymerase recruitment to the promoters. Our discovery of a binding site proximal to the dsbG gene led us to suggest that OxyR binding to this site would lead to OxyR regulation of dsbG. Unexpectedly, we found that OxyR binding to the dsbG-proximal site regulates the expression of the divergent ahpC transcript, while OxyR binding to the ahpC-proximal site regulates expression of the dsbG gene. These findings illustrate how predictions about promoters and regulation solely based on sequence analysis may be misleading and reinforce the need to experimentally test computer-based predictions. The respective starts of the dsbG and ahpC transcripts indicate that the two transcripts overlap by over 100 nucleotides. The reason for the overlap is not known, but pairing between transcripts may allow for an additional level of regulation. Whole-genome expression experiments indicate that more than 150 E. coli RNAs overlap with transcripts on the opposing strand (C. Rosenow, unpublished data). Thus, the overlap between the 5' ends of the dsbG and ahpC mRNAs may represent a more widespread phenomenon.