Expression of the soxR Gene of Pseudomonas aeruginosa Is Inducible during Infection of Burn Wounds in Mice and Is Required To Cause Efficient Bacteremia

Abstract

Burn wounds are prone to infection by Pseudomonas aeruginosa, which is an opportunistic pathogen causing various human diseases. During infection, the bacterium senses environmental changes and regulates the expression of genes appropriate for survival. A purine-auxotrophic mutant of P. aeruginosa was unable to replicate efficiently on burn wounds, suggesting that burn wounds are purine-deficient environments. An in vivo expression technology based on purEK gene expression was applied to the burned mouse infection model to isolate P. aeruginosa genes that are specifically induced during infection. Four such in vivo-inducible (ivi) genetic loci were identified, including the gene for a superoxide response regulator (soxR), the gene for a malate synthase G homologue (glcG), an antisense transcript of a putative regulator responding to copper (copR), and an uncharacterized genetic locus. SoxR of Escherichia coli is known to regulate genes involved in protecting the bacterium against oxidative stress. The expression of soxR was proven to be highly inducible during the infection of burned mice and also inducible by treatment with paraquat, which is a redox-cycling reagent generating intracellular superoxide. The SoxR protein functions as an autorepressor in the absence of paraquat, whereas in the presence of paraquat, this autorepression is diminished. Furthermore, a soxR null mutant was shown to be much more sensitive than wild-type P. aeruginosa to macrophage-mediated killing. In support of this observation, a soxR null mutant exhibited a significant delay in causing systemic infections in the burned mice. Since most mortality in burn patients is caused by systemic infection, the defect in the ability to cause efficient bacteremia in burned mice suggests an important role of the soxR gene in the infection of burn wounds.

In the United States, more than one million people suffer from thermal injury every year, and 60 to 80% of them require medical attention in hospitals or major burn centers (44). About 5,000 of those treated patients die each year, despite advances made in medical treatment (34). Burn injury results in a loss of the normal skin barrier and suppresses the immune system. These pathophysiological alterations make burn patients highly susceptible to several bacterial pathogens, such as Pseudomonas aeruginosa, Streptococcus pyogenes, and Staphylococcus aureus (1, 48). Infecting bacteria can easily penetrate into the subcutaneous soft tissue and proliferate aggressively, causing high mortality due to bacteremia and septic shock (39). The infection of burn wound tissue by bacterial pathogens also contributes to slower wound healing, loss of skin grafts, and severe scar formation (14, 30).

P. aeruginosa is an opportunistic pathogen that not only poses a threat to burn patients but also causes significant mortality and morbidity in cystic fibrosis patients and immunocompromised patients (11). Infection with Pseudomonas from contaminated hospital environments results in severe, life-threatening complications (7). P. aeruginosa has many virulence factors that contribute to infection, penetration, and survival against the host defense systems (7). In addition, P. aeruginosa is ubiquitous throughout the environment due to its great nutritional versatility, resulting in contamination of hospital equipment such as surgical and catheterization equipment (4). Numerous virulence factors, nutritional versatility, and resistance to many commonly used antibiotics make it difficult to eradicate the microorganism from hospital environments. The routine administration of systemic antibiotics generally does not prevent wound colonization, because burn eschar is relatively avascular and systemic antibiotics fail to achieve bactericidal levels in burn wounds unless the antibiotics are used at high dosages (5). Moreover, overdosing with commonly used antibiotics could select for the emergence of “superbug” strains exhibiting high antibiotic resistance. Therefore, the development of new therapeutic strategies for the control of P. aeruginosa infection in burn patients is imperative.

At the initial stage of infection by P. aeruginosa, the pathogen must colonize and proliferate in a new environment that is totally different from its natural reservoir. The new host environment renders several stresses, such as limitation of certain nutrients, thermal stress, osmotic stress, and oxidative stress. To adapt to the new environment, bacteria change the pattern of gene expression, turning on appropriate genes while turning off unnecessary and/or deleterious genes. Studies of such regulated genes could provide valuable clues for the development of molecular therapeutic strategies to control P. aeruginosa infections (41). To isolate P. aeruginosa genes specifically induced in host environments, a genetic selection system called in vivo expression technology (IVET), first described for Salmonella typhimurium (29), has been successfully applied to P. aeruginosa (49). In this study, we have extended the application of the IVET selection system to the burned mouse infection model and have identified four genetic loci that are specifically induced in burned mouse tissue. We focused on one of the loci, the soxR gene, which encodes a putative transcriptional regulator functioning as an autorepressor. Mutational analysis demonstrated that this locus is required for efficient dissemination into deeper organs, indicating that soxR is an important virulence factor.

MATERIALS AND METHODS

Bacterial strains and plasmids.

The bacterial strains and plasmids used in this study are listed in Table 1. The chromosomal cointegrate bank used for IVET selection has previously been described (49), and the cosmid clone bank of PAK chromosomal DNA was also described previously (25). The soxR::lacZ fusion was generated by introducing the soxR promoter region from pSF21-7 (BamHI-KpnI fragment) into the promoterless lacZ fusion vector pDN19lacΩ (46), resulting in pHW9802.

TABLE 1.

Strains and plasmids used in this study

Strain or plasmid	Descriptiona	Reference or source
P. aeruginosa strains
PAK	Wild-type clinical isolate	13
PAK-AR2	PAK with the purEK genes deleted; Spr Smr	49
PAK-soxR::Ω	PAK with chromosomal disruption of the soxR locus; Spr Smr	This study
SF21	Strain isolated by IVET selection with surface infection of burned mice; Spr Smr Apr Tcr	This study
Plasmids
pDN19lacΩ	Promoterless lacZ fusion vector; Spr Smr Tcr	46
pHW9704	soxR clone in pTZ18R to construct pHW9706; Apr	This study
pHW9706	soxR gene disrupted by insertion of Ω cassette in pHW9704; Apr Spr Smr	This study
pHW9708	Intact soxR clone in pTZ18R for sequencing analysis; Apr	This study
pHW9802	soxR promoter fused to lacZ in pDN19lacΩ; Spr Smr Tcr	This study
pHW9808	soxR gene fused to His6 in a fusion vector pQE32; Apr	This study
pHW9811	pbpC structural gene fused to lacZ in pDN19lacΩ; Spr Smr Tcr	This study
pHW9812	Intact soxR clone in pDN19lacΩ; Spr Smr Tcr	This study
pQE32	His6-tagged fusion vector; Apr	Qiagen
pSF21-7	Plasmid rescued from the chromosome of SF21 by HindIII digestion; Apr	This study
pSJ9443	IVET selection vector; Apr Tcr	49

^aSp^r, spectinomycin resistance marker; Sm^r, streptomycin resistance marker; Ap^r, ampicillin resistance marker; Tc^r, tetracycline resistance marker. 

Escherichia coli and P. aeruginosa were grown in Luria agar (L agar) or L broth at 37°C. Minimal medium A (MinA) (13) was also used for the growth of P. aeruginosa. Antibiotics were used at final concentrations of 100 μg of ampicillin/ml, 50 μg of spectinomycin/ml, 25 μg of streptomycin, and 20 μg of tetracycline/ml for E. coli or at 150 μg of carbenicillin/ml, 200 μg of spectinomycin/ml, 200 μg of streptomycin/ml, and 100 μg of tetracycline/ml for P. aeruginosa. Adenine was added at 5 μg/ml as a limited supplement to support the growth of purEK-defective strains on MinA.

Generation of a soxR null mutant strain (PAK-soxR::Ω).

The soxR gene in pHW9704 was disrupted by inserting a 2-kb Ω cassette that encodes resistance to spectinomycin and streptomycin. The resulting plasmid, pHW9706, was linearized by SspI digestion and electroporated into the wild-type strain PAK for marker exchange. Transformants were initially selected on L agar containing spectinomycin and streptomycin, and the resulting colonies were further tested for sensitivity to carbenicillin. Total chromosomal DNAs from carbenicillin-sensitive strains were isolated and analyzed by Southern hybridization to confirm the double crossover of the disrupted soxR gene into the chromosome.

Burned mouse model.

Burned mice were generated by utilizing the procedure described previously by Collins and Roby (10) with modifications. One day before the experiment, female CF-1 mice weighing 22 to 24 g were anesthetized by intramuscular injection of ketamine hydrochloride (1 mg per mouse). The dorsal sides of the mice were shaved and then treated with Brush-on Facial Hair Remover (Del Laboratories, Inc., Garmingdale, N.Y.) to completely remove short hairs in the shaved area. On the next day, the mice were anesthetized again, and approximately 6% of the total body surface area (1.5 × 3.0 cm²) was subjected to a 4-s burning over a grid-top gas burner (17). Immediately after the burning, the mice were administered 0.5 ml of a 0.15 M NaCl solution by intraperitoneal injection. The burned mice were inoculated on the surfaces of the burn wounds with 10 μl of bacterial suspension and individually housed in sterile cages. To recover bacterial cells, the mice were sacrificed by cervical dislocation, and the burn wound skin, spleen, and liver were homogenized separately. The homogenates were serially diluted 10-fold and plated on appropriate media to count the number of viable bacterial cells.

In vitro induction assay of the soxR gene.

Overnight cultures of PAK strains carrying either pDN19lacΩ, pHW9802, or pHW9811 were diluted 100-fold into fresh L broth containing appropriate antibiotics and shaken at 37°C until the optical density at 600 nm reached 0.1. Inducing agents were then added to various final concentrations, and samples were shaken vigorously (300 rpm) at 37°C for 3 h before β-galactosidase activities were measured (31). Methyl viologen hydrate (paraquat; Acros) as a redox-cycling agent was used to generate intracellular superoxide (50). For extracellular superoxide experiments, bovine milk xanthine oxidase (Boehringer Mannheim) was used with hypoxanthine (Sigma) as a substrate (35).

Test of survival of P. aeruginosa within activated macrophages.

The in vitro survival of P. aeruginosa within activated macrophages was tested as described previously (35). Female CF-1 mice were injected intraperitoneally with 1.5 ml of sterile thioglycolate 4 days before harvesting of cells by peritoneal lavage with ice-cold RPMI (Cellgro) medium containing 2% fetal calf serum and 1 U of heparin per ml. Harvested cells were washed twice at 4°C with ice-cold RPMI medium containing 2% fetal calf serum and resuspended in SMEM (minimum Eagle’s medium containing 2 g of NaHCO₃ per liter, 3.5 g of glucose per liter, 0.11 g of sodium pyruvate per liter, 15 mM HEPES, and 5% fetal calf serum). After counting of the number of viable cells by trypan blue staining, 2 × 10⁶ cells were placed in each well of 24-well tissue culture plates and incubated in a 5% CO₂ incubator at 37°C for 2 h. The wells were then washed four times with phosphate-buffered saline (PBS) before being inoculated with bacteria. Bacterial cells were grown to log phase in L broth and opsonized with 10% mouse serum in SMEM for 30 min. One milliliter of the opsonized bacteria (2 × 10⁷ cells/ml in SMEM) was added to each well containing adherent viable macrophage cells. The plates were centrifuged at 200 × g for 10 min and incubated in a 5% CO₂ incubator at 37°C for 1 h to allow for phagocytosis. The wells were washed four times with PBS, and 1 ml of SMEM containing 200 μg of gentamicin was added before the plates were returned to the incubator. Samples were washed twice with PBS at the indicated time, and 0.5 ml of 1% deoxycholate was added to lyse macrophages, followed by washing with 0.5 ml of PBS. The two 0.5-ml solutions were combined and plated on L agar containing appropriate antibiotics to count the number of surviving bacteria.

Recombinant DNA methods.

Standard methods were used for plasmid DNA preparation, restriction enzyme digestion, and cloning (42). DNA sequence analysis was performed by PCR-mediated Taq DyeDeoxy Terminator Cycle sequencing with an Applied Biosystems model 373A DNA sequencer. To sequence the DNA fragments upstream of the purEK operon, a purEK primer complementary to the 5′ end of the purEK gene was used (49). DNA restriction enzyme site and open reading frame analyses were conducted by using the DNA Strider program. BLAST and Genetics Computer Group programs were used to search for and analyze DNA and amino acid sequence homologies. Southern hybridizations were carried out by using the ECL labeling and detection kit from Amersham.

RESULTS

Purines are limited in the tissues of mouse burn wounds.

In order to determine whether the purEK-based IVET selection system is applicable to the burned mouse infection model, the ability of a purEK deletion strain, PAK-AR2, to replicate in burn wounds was compared to that of the wild-type PAK. Each burned mouse was infected on the surface of the wound with 10³ cells of a mixture consisting of PAK and PAK-AR2 in a ratio of 1:17. At 1, 2, and 7 days after inoculation, burn wound tissues were homogenized and plated to determine the number of viable cells of each bacterium; PAK was plated on MinA, and PAK-AR2 was plated on L agar containing spectinomycin and streptomycin.

As shown in Table 2, the number of PAK cells increased 100- and 1,000-fold over the number of PAK-AR2 cells after 24 and 48 h, respectively. After day 7, PAK reached more than 10⁶ cells/ml, whereas none of the PAK-AR2 survived in the burn wound tissue. These results indicate that burn wounds do not contain large enough amounts of purine metabolites to support the growth of purEK deletion strain PAK-AR2, suggesting that the IVET selection system is applicable to the burned mouse model. Since we have an IVET bank that consists of at least 10⁴ independent chromosomal cointegrates, a 7-day incubation will provide sufficient time to allow replication of PAK in burn wounds while eliminating the slower- or nonreplicating cells like PAK-AR2.

TABLE 2.

Replication rates of wild-type PAK and purEK deletion strain PAK-AR2 in burn wounds

Incubation time	PAK/PAK-AR2 ratio	Fold increase
0 h	1:17	1
24 h	10:1	1.7 × 102
48 h	102:1	1.7 × 103
7 days	106:0	>1.7 × 107

Isolation of in vivo-inducible (ivi) genes by using the burned mouse infection model.

The IVET system is based on random integration of a promoterless purEK operon, encoding an enzyme required for purine biosynthesis, into the bacterial chromosome of a P. aeruginosa strain containing a chromosomal deletion of the purEK region. Since the purEK deletion strain is unable to grow in the absence of purine, growth of the bacteria in environments lacking purine is possible only if the promoterless purEK is inserted downstream of an active promoter. To distinguish promoters that are specifically induced in vivo from those expressed constitutively, bacteria recovered from in vivo selection are screened for growth on minimal medium supplemented with a limited amount of purine. Large colonies represent purine-independent growth, and therefore the purEK genes are under the control of constitutively expressed promoters in those bacteria. Small-colony variants represent bacteria that require purine for growth, and therefore the purEK genes are transcribed by promoters that are active in vivo and inactive on minimal medium (in vitro). To isolate P. aeruginosa genes specifically induced during the infection of burn wounds, 10⁶ cells of the IVET bank were inoculated onto the surface of each burn wound site of four mice. After 7 days of incubation, bacterial cells were recovered from the wound tissue by plating the tissue homogenates on MinA containing appropriate antibiotics and 5 μg of adenine per ml to support limited growth of non-purEK-expressing cells. Twenty-four small colonies, which should theoretically have promoters that are active in vivo but inactive in vitro, were randomly picked. Total chromosomal DNAs from the 24 isolates were isolated and digested with either EcoRI or HindIII, and Southern hybridization was conducted with a 1.7-kb purEK gene from pSJ9443 as a probe. Four different patterns of Southern hybridization were observed; 21 isolates had identical hybridization patterns (named sf17) and 3 isolates had distinct patterns (named sf2, sf7, and sf21) (data not shown), implying that we isolated four possible ivi loci.

To further characterize the four genetic loci, DNA fragments upstream of the promoterless purEK were isolated by an “EcoRI rescue” method which involves EcoRI digestion of the chromosomal DNA, self-ligation, transformation into E. coli, and selection for ampicillin resistance (49). The rescued plasmids were used to sequence DNA upstream of the purEK gene. All six possible reading frames of the obtained 500-bp sequences were used to search for their homologues in the GenBank database (National Center for Biotechnology Information and National Institutes of Health) with the BLASTX program. The search results (Table 3) indicated that the genetic locus presented by the 21 identical patterns (sf17) had a high degree of sequence similarity to the copR gene of Pseudomonas syringae (32), but in an antisense direction. The other three loci, sf21, sf7, and sf2, were homologues to the soxR gene of P. aeruginosa PAO1 (53), the glcG gene of E. coli (33), and an open reading frame with no homology to any known genes, respectively. Both soxR and copR are members of two-component regulatory systems; these systems are involved in protection of bacteria against oxidative stress and large amounts of toxic copper, respectively. The glcG gene in E. coli encodes malate synthase G (MSG), which is involved in the glyoxylate cycle. The soxR gene was chosen for further study.

TABLE 3.

P. aeruginosa loci identified to be inducible during infection of burned mice

Locus	Similar protein in databases (GenBank no.)	Function	BLASTX P value
sf21	P. aeruginosa SoxR (X95517)	Transcriptional activator	2.0 × 10−58
sf7	E. coli GlcG (L43490)	MSG	7.1 × 10−40
sf17	P. syringae CopR (L05176)	Transcriptional activator	1.2 × 10−8
sf2	None

The soxR gene is highly and specifically induced during infection of burn wounds.

To verify the inducibility of the soxR gene under in vivo conditions, the relative replication rate of the original isolate SF21 compared to that of the purEK deletion strain PAK-AR2 was tested both under in vivo (burn wound) and in vitro (MinA) growth conditions. Strain SF21 has a promoterless purEK operon fused downstream of the soxR promoter in the chromosome of PAK-AR2. Since purine is limited under both growth conditions, the relative replication rate of the SF21 over that of PAK-AR2 reflects the strength of the soxR promoter. For in vivo inducibility tests, overnight cultures of PAK-AR2 and SF21 were mixed in a 1:1 ratio, and 10 μl of the bacterial mixture (200 bacterial cells) was inoculated onto the burn wound surfaces of nine mice. After 1, 3, and 7 days, burn wound tissues of three mice were homogenized separately and plated on L agar to count cells of both bacterial strains and on L agar containing carbenicillin and tetracycline to count viable SF21 cells. For in vitro control tests, PAK-AR2 and SF21 were mixed in a 1:1 ratio and inoculated into MinA at a bacterial density of 2 × 10³ cells per ml. Viable bacterial cells were counted in a similar manner after 1, 3, and 7 days of incubation at 37°C.

Under in vitro growth conditions, the viability of SF21 and PAK-AR2 decreased gradually over the 7-day period, with no significant changes in the ratio of the two bacterial strains (Fig. 1A), indicating that the soxR promoter is inactive under the in vitro growth conditions. Under in vivo growth conditions (Fig. 1B), SF21 and PAK-AR2 replicated at similar rates until day 3, possibly by utilizing residual purines. However, by day 7, viable cells of SF21 increased to almost 10⁷ per mouse whereas no PAK-AR2 was recovered, demonstrating that the soxR promoter is highly active under the in vivo growth conditions. The effect of the soxR promoter may not have been obvious in the initial 3 days due to the presence of residual purines. However, by day 7 residual purines may have been depleted, thereby accounting for the dramatic increase in the number of SF21 cells.

FIG. 1.

Comparison of growth rates of the IVET isolate SF21 and the purEK deletion strain PAK-AR2 in vitro (A) and in vivo (B). For in vitro growth (A), a mixture of the two bacteria was inoculated into MinA and grown at 37°C for 1, 3, and 7 days. Bacterial cell densities were determined by colony counting on appropriate media. For in vivo growth (B), a mixture of the two bacteria was inoculated on the surfaces of mouse burn wounds. Both bacterial strains were recovered from burn wounds after 1, 3, and 7 days of incubation to determine the number of viable cells.

Sequence analysis of the soxR gene.

To isolate the intact soxR gene, a cosmid clone bank of PAK chromosomal DNA was screened by colony hybridization with the partial soxR gene sequence isolated from pSF21-7 as a probe. Several positive colonies were isolated and confirmed by Southern hybridization (data not shown). A subclone containing the intact soxR gene, pWH9708, was sequenced in both directions.

The soxR gene of P. aeruginosa PAK encodes a 156-amino-acid protein with a predicted molecular mass of 17 kDa. Indeed, when the soxR open reading frame was fused behind a His₆ tag, an expected 17-kDa fusion protein was observed (Fig. 2). The soxR sequence of PAK varies from that of strain PAO1 by only one base, without affecting the amino acid sequence. The amino acid sequence of SoxR is 58% identical to that of the E. coli SoxR, and both proteins possess a 20-residue helix-turn-helix motif near the N terminus which is predicted to have a DNA-binding function (Fig. 3). Both SoxR proteins also contain a region with four cysteine residues (CX₂CXCX₅C), a metal-complexing redox center that may detect superoxide (52).

FIG. 2.

Overexpression of the SoxR protein. E. coli harboring either the fusion vector pQE32 (lane 2) or His-SoxR fusion construct pHW9808 (lane 3) was induced with 1 mM IPTG (isopropyl-β-d-thiogalactopyranoside) for 3 h, and the total cellular proteins were separated on a sodium dodecyl sulfate–12% polyacrylamide gel, followed by Coomassie blue staining. The His-SoxR fusion protein is marked by an arrow. Lane 1, protein molecular mass standards (high range; Gibco BRL).

FIG. 3.

Comparison of the SoxR amino acid sequences of E. coli, S. typhimurium, and P. aeruginosa PAK. The helix-turn-helix motif (boldface) in the N terminus (with a possible DNA-binding function) and a region with four cysteines (CX₂CXCX₅C) (dots) in the C terminus (a putative metal binding site) are marked.

Regulation of the soxR gene.

The SoxR protein of E. coli is known to function as a sensor to respond to intracellular superoxide. Activated SoxR stimulates the expression of SoxS, a transcriptional activator of several target genes such as the Mn-containing superoxide dismutase gene (sodA), the endonuclease IV gene (nfo), and the glucose-6-phosphate dehydrogenase gene (zwf) (15). To test whether the soxR gene of P. aeruginosa is also activated by superoxide, a soxR::lacZ fusion construct, pHW9802, was utilized. In addition, to test whether gene expression from the soxR promoter affects the expression of downstream genes, pbpC, located downstream of soxR, was also fused to lacZ, resulting in pHW9811. PAK containing either the promoterless lacZ fusion vector pDN19lacΩ, pHW9802, or pHW9811 was incubated with different concentrations of paraquat, and the β-galactosidase activities were measured.

As shown in Fig. 4, the β-galactosidase activity in strain PAK(pHW9802) was increased sevenfold in response to 1 mM paraquat, whereas PAK(pHW9811) and the negative control PAK(pDN19lacΩ) were not affected by the same treatment. These results suggest that the presence of intracellular superoxide generated by paraquat (26) stimulates the expression of soxR but that its expression does not affect the expression of the downstream pbpC gene, indicating that it is not controlled by the same promoter. Unlike the positive effect of intracellular superoxide, expression of the soxR gene in P. aeruginosa was not affected by either extracellular superoxide or hydrogen peroxide (data not shown). Interestingly, when the soxR::lacZ fusion construct was introduced into a soxR mutant background, the resulting strain, PAK-soxR::Ω(pHW9802), gave fivefold-higher β-galactosidase activity than the strain with the construct in the wild-type PAK background in the absence of paraquat (Fig. 5). These results indicate that the SoxR protein of P. aeruginosa functions as an autorepressor in the absence of an inducer (paraquat), while the inducer eliminates the autorepressor function.

FIG. 4.

Induction of the soxR gene by treatment with paraquat (PQ). PAK strains containing pDN19lacΩ (vector control), pHW9802 (soxR::lacZ fusion), or pHW9811 (pbpC::lacZ fusion) were grown in L broth at 37°C in the presence of 0 to 1.0 mM PQ. Samples were taken after 3 h of induction to measure β-galactosidase activities. Average values from three repeated tests are shown.

FIG. 5.

Expression of the soxR gene in the presence and absence of a functional soxR gene. The soxR::lacZ fusion construct (pHW9802) and a lacZ fusion vector (pDN19lacΩ) were each introduced into wild-type PAK or a soxR null mutant, PAK-soxR::Ω (soxR). Overnight cultures of the bacteria were used to measure β-galactosidase activities. Average values from three repeated tests are shown.

The soxR null mutant strain of P. aeruginosa causes a delayed systemic infection in burned mice.

To study the role of SoxR in the infection of burn wounds, a soxR mutant strain was generated by insertion of the Ω fragment (see Materials and Methods). Since soxR and its immediately downstream gene, pbpC, are not in an operon structure, insertional inactivation of the soxR gene should not have a polar effect on downstream genes. The soxR mutant strain PAK-soxR::Ω was compared to the wild-type PAK for the abilities to colonize in burn wounds and to cause bacteremia. The two bacterial strains were individually inoculated onto the surfaces of burn wounds with a dose of 2 × 10³ cells per mouse, and bacteria were recovered from burn wound skin, spleen, or liver after 1, 3, and 7 days of incubation. Recovered bacteria were serially diluted and plated on MinA to count the number of viable PAK cells and on L agar containing spectinomycin and streptomycin to count viable PAK-soxR::Ω cells.

As shown in Fig. 6, at 1 day postinfection, both strains increased almost 10⁵-fold over the initial inoculum in the burned skin, but no bacteria were detected in spleen or liver. After 3 days of incubation, the number of PAK cells in the burned skin was not significantly different from the number of soxR mutant cells. However, a significant difference was observed in the spleen and liver, where 10³ and 10⁴ PAK cells, respectively, were found, whereas no soxR mutant cells were detected (P < 0.01). By day 7, the number of soxR mutant cells remained similar to the number of PAK cells on the skin, but the soxR mutant started to appear in spleen and liver in numbers similar to those seen for PAK infection at day 3. Since the growth rates of the two strains under in vitro conditions (in L broth or MinA) showed no significant difference (data not shown), the delayed systemic infection by the soxR null mutant indicates that the soxR gene plays an important role in systemic infection.

FIG. 6.

Effect of soxR on virulence of P. aeruginosa. Burned mice were infected with either PAK or the soxR mutant PAK-soxR::Ω, and the bacterial cells were recovered from burned skin tissues (A), spleens (B), or livers (C) after 1, 3, and 7 days of incubation. Three mice were used for each strain at each time point. This experiment was performed twice. Error bars indicate standard deviations.

The soxR null mutant is more sensitive to macrophage-mediated killing than wild-type PAK.

The observation that the disruption of soxR causes a delay in systemic infection indicates that soxR plays a significant role during host infection. One possible explanation for this could be a correlation with activated immune cells such as macrophages during systemic infection. To determine the sensitivity of bacterial cells to macrophage-mediated killing, three bacterial strains, i.e., wild-type PAK, the soxR null mutant, and the mutant complemented by an intact soxR gene (pHW9812), were incubated with activated macrophages in vitro, and intracellular bacteria were recovered at various times to determine the number of surviving bacterial cells as described in Materials and Methods.

As shown in Fig. 7, the soxR mutant is much more sensitive to macrophage-mediated killing than wild-type PAK. The number of soxR mutant cells decreased rapidly after 8 h of cocultivation, and none survived by 16 h of cocultivation. Furthermore, when the soxR null mutant was complemented by an intact soxR gene (pHW9812), the resulting strain was more resistant than wild-type PAK to macrophage-mediated killing, presumably due to a multi-copy-number effect of the soxR gene. These results demonstrate that the soxR gene plays an important role in protecting bacteria against macrophage-mediated killing, and it may account for the delayed systemic infection caused by the soxR null mutant during burned mouse infection.

FIG. 7.

Effect of soxR on survival of P. aeruginosa within activated macrophages. Activated macrophages were isolated by peritoneal lavage 4 days after injection of thioglycolate. Bacterial cells were opsonized with mouse serum and were incubated with macrophages for phagocytosis. At the indicated times, bacterial cells were recovered by lysis of macrophages with deoxycholate solution and plated on L agar containing appropriate antibiotics. This experiment was repeated in triplicate. Wild-type P. aeruginosa PAK, a soxR null mutant strain, and the soxR null mutant strain complemented by a plasmid (pHW9812) containing the intact soxR gene were used. Error bars indicate standard deviations.

DISCUSSION

In most studies using the burned mouse model, bacteria were injected subcutaneously, allowing the bacteria to proliferate quickly and spread systemically to cause septic shock, resulting in a P. aeruginosa 50% lethal dose of as low 10² to 10³ cells. However, natural infections occur through the surfaces of burn wounds, where the bacterial pathogens colonize and penetrate the burn eschar depending on the bacterial invasiveness, local wound factors, and degree of immunosuppression (20). For the purpose of the IVET selection, the mice must be alive for a certain length of time after being infected by at least 10⁴ IVET bank cells so that the bacteria have sufficient time to replicate, eliminating slow- and nonreplicating cells. To accomplish this, we used a burned mouse infection model which gives only 6% total body surface burn, whereas 15 to 30% total body surface burn is usually used (23). In addition, to mimic the natural route of infection, the IVET bank cells were inoculated on the surfaces of burn wounds instead of by subcutaneous injection. Therefore, our selection procedure should preferentially isolate genes relevant to natural infections of burn wounds.

Burn wounds are highly nutritional environments, capable of supporting the growth of pathogens. Burn injury causes a change in levels of hormones such as catecholamine, cortisol, glucagon, and insulin (51), which promote proteolysis and lipolysis (18), releasing large amounts of amino acids (12), glycerol, and free fatty acids into circulation. Especially, high levels of cortisol lead to excessive protein catabolism, causing loss of tissue protein and a high concentration of free amino acids (hyperaminoacidaemia) (40). The purEK deletion strain PAK-AR2 was able to replicate slowly in the burn wounds within the first 3 days of inoculation, probably due to residual purines available for the bacterium to utilize. However, PAK-AR2 is unable to replicate after 3 days, implying that either the available purines are being used up by the replicating bacteria or the host tissue stops releasing them.

We have identified four different in vivo-inducible (ivi) genes, including glcG, copR, soxR, and an unknown gene. The glcG gene of E. coli encodes MSG, which is one of two isoenzymes (33). Malate synthase condenses glyoxylate with acetyl coenzyme A to yield malate in the glyoxylate cycle. The malate is then oxidized to oxaloacetate, which can condense with another acetyl coenzyme A to start the tricarboxylic acid cycle. Glyoxylate is generally produced from two chemical sources: glycolate and acetate. When E. coli grows on glycolate as a sole carbon source, MSG accounts for almost 100% of the intracellular malate synthase activity (47). Glycolate is a two-carbon molecule which stimulates collagen production and provides anti-inflammatory activity with antioxidant properties during the wound healing process (37). It is likely that burn wounds generate high levels of glycolate to assist the wound healing process, which in turn induces the expression of P. aeruginosa glcG.

The other identified gene, copR, together with copS forms a two-component regulatory system involved in bacterial defense against toxic concentrations of copper (32). The putative transmembrane protein CopS senses high levels of free Cu²⁺ ions in the periplasm and phosphorylates the cytoplasmic CopR protein, which induces the expression of the cop operon (8). In our isolates, the copR gene was fused in an opposite direction to the promoterless purEK operon, indicating the presence of an antisense RNA-mediated copR regulation. This suggests that burn wounds are low-copper environments; therefore, copR expression needs to be down regulated. This is reasonable, because burn injury causes a reduction in tissue copper levels such that 20 to 40% of the body copper content is lost within 7 days after burn injury (3). The decrease in copper might be due to disturbances in the synthesis of ceruloplasmin, which is a carrier protein of copper (43). Therefore, a dramatic decrease of copper in burn wounds would generate copper-deficient environments that suppress the expression of copR in P. aeruginosa.

Both prokaryotic and eukaryotic cells have developed defense systems against oxidative stresses (19), because reactive oxygen species, including superoxide (O₂.⁻), hydrogen peroxide (H₂O₂), and hydroxyl radical (ȮH), are easily generated by normal aerobic metabolism (45) and by exposure to environmental agents such as radiation, metals, or redox-cycling agents such as plambagin, menadione, and paraquat (26). Superoxide is a relatively weak reactive oxygen species, which can cause limited damage to cell membranes and also inactivate superoxide-sensitive enzymes such as aconitase and fumarase. However, the main toxicity of superoxide is caused by its derivative hydroxyl radical, which is highly reactive and directly oxidizes critical cellular targets, including DNA, lipids, and proteins. To block the conversion of superoxide to hydroxyl radical and to repair damaged cellular components, bacteria utilize the soxRS regulon, which is activated by oxidative stress (26). The soxRS system of E. coli operates in two stages of transcriptional activation: a redox signal activates SoxR as a potent transcription activator for the soxS gene (36), whose product then triggers transcription of ∼10 regulon genes (2). The products of the soxRS regulon include Mn-containing superoxide dismutase; the oxidative DNA repair enzyme endonuclease IV; glucose-6-phosphate dehydrogenase, which produces NADPH; a micF-encoded antisense RNA which blocks expression of the outer membrane porin OmpF to restrict the entry of compounds into cells (9); the redox-resistant isoenzymes fumarase C and aconitase A (28); and an NADPH:ferredoxin oxidoreductase (28). Disruption of the soxRS regulon renders E. coli sensitive to both oxidative stress and macrophage-mediated killing (38).

Burn injury causes several broad systemic responses, such as activation of macrophages and neutrophils, production of cytokines and inflammatory mediators, and stimulation of the metabolic response (16), all of which can lead to an increased production of superoxide. However, superoxide is known to be a weak and unstable oxidant, which does not easily cross the cell membrane. Therefore, in order for induction of the expression of soxR to occur, bacteria must be exposed to high levels of superoxide or certain redox-cycling materials that can readily penetrate the cell membrane and generate intracellular superoxide. To date, we do not know of any natural redox-cycling agents present in burn wounds. However, certain redox-cycling agents, such as paraquat (21) and CuSO₄ (27), are known to directly generate a bactericidal flux of superoxide when present in the bacterial cytoplasm, whereas other agents, such as plambagin and menadione, directly inhibit respiratory NADH dehydrogenase to cause an intracellular redox imbalance (24). Significant changes in the NAD(P)⁺/NAD(P)H ratio are also known to induce the E. coli soxRS system (28).

Attempts to obtain a 50% lethal dose for PAK and the soxR mutant by using surface infection of the burned mice were not successful. However, a significant difference was observed when the two bacterial strains were used to individually infect the burned mice. On the burn wound skin, there is no difference in the growth rates of the soxR mutant and the wild-type PAK strain. However, the soxR mutant was found in the liver and spleen much later than the wild-type PAK. We have tested a number of possibilities that could account for the delayed systemic infection by the soxR mutant. First, the effect of a soxR mutant on bacterial motility was tested. Since the SoxR protein functions as a sensor of oxidative stress, a soxR mutant might be defective in chemotaxis, resulting in an inability to move away from the source of oxidative stress (chemorepellant) and resulting in slower spreading. However, in vitro bacterial motility tests in the presence or absence of paraquat failed to detect any significant difference between the two strains. Second, the possibility of the soxR mutant being more sensitive to serum was tested. Sera prepared from mice 3 days after burn injury were used to conduct a serum bactericidal assay (6). Again, no difference in the sensitivities of the two bacterial strains to the serum-mediated killing was observed. Finally, our in vitro macrophage-mediated killing assay indicated that the soxR mutant strain is much more sensitive to killing by activated macrophages than the wild-type PAK strain. This could explain, at least partially, the delayed dissemination of the soxR mutant in burned mice.

This study showed a successful application of the IVET system to the burned mouse infection model and demonstrated that soxR is specifically induced during the infection of burn wounds. The soxR gene is a member of soxRS two-component regulatory system, which functions as a bacterial defense system against oxidative stress, indicating that burn wounds are abundant in oxidative agents. Indeed, burn injury is known to generate high oxidative stress via the increased metabolic response and activation of immune cells such as macrophages and neutrophils. Encountering the high oxidative stress during infection of burn wounds, P. aeruginosa seems to increase the expression of a soxR-controlled defense system. SoxR of P. aeruginosa has been shown to function as a strong autorepressor in the absence of inducing signals, and this autorepression is presumably diminished when P. aeruginosa senses oxidative stress during the infection of burn wounds. A recent report from Demple’s group (22) demonstrated that E. coli SoxR also functions as an autorepressor. In the presence of inducing signals, the SoxR protein of P. aeruginosa, like that of E. coli, may further become a transcriptional activator of soxS, which in turn activates downstream effector genes. However, we have not yet found the soxS homologue in the almost complete genome sequence of the P. aeruginosa by sequence homology searching. An active search for a functional homologue of the soxS gene is under way.