Role for σ³⁸ in Prolonged Survival of Escherichia coli in Drosophila melanogaster Free

Single-cell bacteria are directly exposed to environment in nature, and they carry sophisticated genetic systems for adaptation by altering the pattern of genome expression (1, 2). After complete sequencing of the genome, the whole set of ∼4600 genes has been predicted to exist in the model prokaryote Escherichia coli K-12 (3, 4). Even for the best-characterized model organism E. coli, the functions of approximately one third of its genes remain unidentified or unpredicted because those genes are not expressed under laboratory culture conditions. Under stressful conditions in nature such as within host animals, however, these uncharacterized silent genes could be expressed for adaptation. The pattern of genome expression is controlled mainly at the step of transcription. The DNA-dependent RNA polymerase of E. coli is a multisubunit holoenzyme with the subunit composition α₂ββ′ω and one of seven species of the promoter-recognizing subunit σ (5–7). Among seven σ factors, σ⁷⁰ is responsible for the transcription of most genes expressed in the exponential phase of cell growth, whereas it is replaced by rpoS-encoded σ³⁸, another housekeeping σ factor, upon entry into the stationary phase (5, 8). The function of the holoenzyme of RNA polymerase is further modulated through interaction with a set of ∼300 transcription factors (1, 2). Therefore, the decision regarding gene utilization is executed by both σ factors and transcription factors.

The invasion of microbial pathogens evokes immune responses in host organisms. The frontline of host immunity is various reactions that compose innate immunity (9–11). The mechanism and meaning of innate immune responses have been studied in detail using fruit fly Drosophila melanogaster, a genetically tractable model animal equipped with only this type of immunity. Previous studies have revealed that substances located at the surface of pathogens are recognized by soluble and cell-bound receptors present in the hemolymph and the surface of immune cells called hemocytes, respectively, resulting in the induction of antimicrobial peptide production in the fat body and a variety of cellular immune reactions (12–17). This mechanism of immunity against invading pathogens is fundamentally conserved beyond species including humans (18, 19). On the other hand, microbial pathogens possess strategies to avoid or resist host immunity (20–26). Therefore, it is reasonable to presume that microbial pathogens change gene expression patterns when they are exposed to assaultive environments, such as immune reactions in the host. However, it is still unknown what host substances are responsible for triggering a change in gene expression and how microbes respond to such stimuli.

σ³⁸, which is responsible for the expression of E. coli genes in the stationary phase of cell growth, also plays an important role under various stressful conditions such as oxidation, acid shock, temperature shock, osmotic shock, and starvation (5, 8, 27–29). Therefore, this σ factor is a strong candidate for the component responsible for altering gene expression in bacteria exposed to host environments. σ³⁸ has been shown to be necessary for the pathogenicity expression of various species of bacteria (30, 31) and the drug resistance of E. coli (28, 29); however, how σ³⁸ participates in such behaviors of bacteria is unknown. In this study, we determined the role for σ³⁸ in the infectious properties of E. coli using Drosophila as a model host.

The following lines of Drosophila were used in this study: Oregon R (Kyorin-Fly, Kyorin University, Tokyo, Japan) as a wild-type fly line, w¹¹¹⁸, imd¹ (32) (a gift from B. Lemaitre), Rel^E20 (33) (Bloomington Drosophila Stock Center, Indiana University, Bloomington, IN; stock number 9457), UAS-rpr (Bloomington Drosophila Stock Center; stock number 5824), pxn-GAL4 8.1 (a gift from M. J. Galko), PGRP-LE¹¹² (34) (a gift from S. Kurata), and Atg1^Δ3d (35) (a gift from T. P. Neufeld). Some of the fly lines were used after changing balancers. The E. coli K-12 strain BW25113 and its derivatives, JW5437 (rpoS-deficient mutant), JW1721 (katE-deficient mutant), and JW3914 (katG-deficient mutant) were obtained from Keio Collection, a library of E. coli with deletions in the open-reading frame of individual genes (36) (National BioResource Project: National Institute of Genetics, Mishima, Japan). Bacteria were cultured at 37°C with Luria-Bertani medium supplemented with antibiotics when required, harvested at the stationary phase of cell growth, washed with ice-cold PBS, and used in the experiments. To prepare dead bacteria, E. coli was washed with PBS, resuspended in PBS, and incubated at 100°C for 30 min. The plasmids used to express E. coli genes were obtained from ASKA Clone (National BioResource Project), a library of E. coli genes cloned into the plasmid pCA24N (37), in which the isopropyl β-d-1-thiogalactopyranoside–inducible promoter T5-lac drives transcription. Hemocytes were isolated from third-instar wandering larvae according to the standard procedure (38) with modifications (39). The cell line l(2)mbn, established from the hemocytes of Drosophila larvae, was maintained at 25°C with Schneider’s Drosophila medium (Life Technologies Japan, Tokyo, Japan) containing 10% (v/v) heat-inactivated FBS as described previously (39).

Male adult flies were infected with E. coli by a microinjection into the hemocoel, as described previously (39). Flies (3–7 d after exclusion, 15–20 flies per vial, 1–3 vials in each experiment), were anesthetized with CO₂ and injected in the abdomen with given numbers of E. coli as a 50-nl suspension in PBS, except for the experiments with 6 × 10⁶ bacteria that were suspended in 150 nl of PBS, with the aid of a nitrogen gas–operated microinjector (Narishige, Tokyo, Japan). The size of E. coli inoculum is indicated in each figure or figure legend. Flies that received the injections were maintained at 29°C until they were subjected to the analyses. Under the conditions adopted in this study, the abdominal injection of 150 nl of PBS did not influence the life of flies at least for 7 d (Supplemental Fig. 1A). To determine CFU of bacteria existing in flies, viable flies after infection (five flies from each vial, 3–4 vials in each experiment) were homogenized in PBS using a plastic pestle. The resulting homogenates were plated onto agar-solidified Luria-Bertani medium and maintained at 37°C overnight, and the number of colonies was determined and expressed as CFUs recovered from one fly.

The level of phagocytosis of E. coli and the rate of killing of engulfed bacteria were determined essentially based on the reported procedures with Staphylococcus aureus (39). To determine the level of phagocytosis in vitro, E. coli surface-labeled with FITC was mixed with larval hemocytes at a phagocyte:bacteria ratio of 1:500 and incubated at 25°C for 10 min. The mixtures were then supplemented with trypan blue to quench fluorescence derived from unengulfed bacteria and examined by fluorescence microscopy for the detection of hemocytes containing fluorescent bacteria. In an assay for phagocytosis in vivo, adult flies were injected abdominally with FITC-labeled E. coli (2 × 10⁶ per fly) as described above and examined by fluorescence microscopy after 2 h to detect the clusters of fluorescent hemocytes that had engulfed bacteria. To determine the viability of bacteria in phagocytes, 1(2)mbn cells were incubated with unlabeled bacteria at 25°C for 10 min, washed with serum-free medium to remove unengulfed bacteria, and further maintained. Cells were then lysed with water, and the lysates were subjected to a colony-forming assay as described above for the experiment with fly homogenates.

Bacteria, adult flies with and without E. coli injection, and l(2)mbn cells before and after incubation with E. coli were lysed with buffer consisting of 10 mM Tris-HCl (pH 7.5), 1% (v/v) Nonidet P-40, 0.1% (w/v) sodium deoxycholate, 0.1% (w/v) SDS, 0.15 M NaCl, 1 mM EDTA, and 1% (v/v) protease inhibitor mixture (Nacalai Tesque, Kyoto, Japan). Proteins contained in the lysates were separated by SDS-PAGE followed by Western blotting with the anti-σ³⁸ rabbit antiserum (40), anti-RNA polymerase α-subunit rabbit antiserum (40), or anti-(His)₆ Ab (sc-804; Santa Cruz Biotechnology, Santa Cruz, CA), and signals were visualized by a chemiluminescence reaction. The level of mRNA was determined by RT-PCR as described previously (39). Total RNA was prepared from bacteria, flies, and cultured cells using TRIsol (Life Technologies Japan) and used as a template for RT with random primers. The resulting cDNA was subjected to semiquantitative PCR or quantitative PCR using Thunderbird SYBR qPCR Mix (Toyobo, Osaka, Japan) and Mx3005p (Agilent, CA). The DNA oligomers used as primers in PCR were as follows: 5′-ATGCAGGGTTCTGTGACAGA-3′ (forward) and 5′-AGAATACGGCGCAGTGCGTT-3′ (reverse) for the mRNA of RNA polymerase α-subunit; 5′-GATGAGAACGGAGTTGAGGT-3′ (forward) and 5′-ACGCAAGTTACTCTCGATCAT-3′ (reverse) for σ³⁸ mRNA; 5′-CGTGAGAACCTTTTCCAATATGATG-3′ (forward) and 5′-TCCCAGGACCACCAGCAT-3′ (reverse) for diptericin mRNA; 5′-GTGGTGGGTVAGGTTTTCGC-3′ (forward) and 5′-TGTCCGTTGATGTGGGAGTA-3′ (reverse) for attacin mRNA, 5′-GACGCTTCAAGGGACAGTATCTG-3′ (forward) and 5′–AAACGCGGTTCTGCATGAG–3′ (reverse) for ribosomal protein 49 mRNA; 5′-ATGAGCACGTCAGACGATATC-3′ (forward) and 5′-AAGTGGGTGATTTTCTCGCG-3′ (reverse) for KatE mRNA; and 5′-CGTGCCAGGCCATACGAAT-3′ (forward) and 5′-TCCATCGATTGAACGGTAAGT-3′ (reverse) for KatG mRNA. In the quantitative RT-PCR analysis, the mRNA levels of the bacterial and fly proteins were shown relative to those of RNA polymerase α-subunit and ribosomal protein 49, respectively, which were analyzed as internal controls.

Results from quantitative analyses are expressed as the mean ± SD of the data from at least three independent experiments, unless otherwise stated in the corresponding figure legends. Other data are representative of at least two independent experiments that yielded similar results, except for the experiment that was done once (Fig. 7B). The number of experiments replicated is shown in the corresponding figure legends. The data in the fly survival assay were statistically analyzed by the two-tailed Student t test or the log-rank test (Kaplan–Meier method) as indicated in figure legends, and all other data were by the two-tailed Student t test. Any p values < 0.05 were considered significant, and the data significantly different from each other are shown with asterisks or p values in the figures.

As a first step toward understanding the role for stress-responsive σ³⁸ in the infectious properties of E. coli, we determined a change in the expression of this σ factor in E. coli injected into the abdomen of adult flies. When the level of σ³⁸ relative to that of the α subunit of E. coli RNA polymerase, an internal control, was determined by Western blotting, we found that σ³⁸ increased within 5 min after injection and returned to its original level after 30 min (Fig. 1A). The signal with migration similar to σ³⁸ found in the lysates of uninfected flies was thought to be nonspecific because no positive signal was detectable in the analysis of fly RNA for σ³⁸ mRNA (data not shown). An analysis in RT-mediated quantitative PCR revealed that the mRNA of σ³⁸ showed an increase after injection into adult flies (Fig. 1B), indicating the augmented expression of σ³⁸-encoding rpoS at the step of transcription. These results suggested that E. coli responded to environments in the hemocoel of adult flies and stimulated the expression of rpoS.

We next examined whether σ³⁸ influenced the survival and pathogenicity expression of E. coli in Drosophila. The rate of growth of rpoS-deficient E. coli was almost equal to that of its parental strain under laboratory culture conditions (Fig. 2A). However, the mutant and parental strains were markedly different when the level of live bacteria residing in Drosophila was determined by an assay for colony formation: the σ³⁸-lacking E. coli strain decreased more rapidly than the parental strain (Fig. 2B). We then analyzed the pathogenic effect of E. coli by determining that the death of flies after the abdominal injection with bacteria. E. coli is generally considered as a bacterium that is not pathogenic to Drosophila, but adult flies died within a few days when they carried a burden of large amounts, higher than 2 million cells in a fly, in the hemocoel (Fig. 2C, left panel). In contrast, the death of flies was not significant when heat-killed bacteria were injected (Fig. 2C, right panel) indicating that the presence of live bacteria in large quantities was the cause of fly deaths. We then compared the pathogenic effect between the rpoS-deficient and parental strains. The results showed that the lack of σ³⁸ made E. coli less pathogenic to flies: more than 70% of flies infected with the mutant strain were alive, even after 7 d in contrast to a ratio of live flies of near 50% at day 4 with the parental strain (Fig. 2D). To confirm the findings described above, we next conducted a “rescue” experiment. The σ³⁸-lacking mutant was transfected with a plasmid containing intact rpoS to be forcedly expressed or an empty plasmid (Fig. 3A), and was used to infect adult flies. We found that the forced expression of rpoS in the mutant bacteria recovered the defects caused by the lack of σ³⁸: rpoS-expressing E. coli survived longer in flies (Fig. 3B) and showed a greater pathogenic effect (Fig. 3C) than the same mutant strain harboring an empty vector. Note that we determined the level of colony-formable bacteria in the time course of hours, not days, because the rpoS mutant seemed to disappear rapidly in flies after the transformation with plasmid. These results indicated that σ³⁸ is required for E. coli to survive in Drosophila. We speculated that E. coli with the aid of σ³⁸ can grow well or avoid killing in the hemocoel of Drosophila, which can allow the bacterium to survive longer resulting in the early death of the host.

The loss of σ³⁸ decreased the level of live E. coli in flies. We thus tested the possibility that the rpoS-deficient E. coli was more sensitive to host immunity than the parental strain. There are two distinct immune responses against invading Gram-negative bacteria in Drosophila, the Imd pathway-mediated production of antimicrobial peptides (a humoral response) and the phagocytic elimination of bacteria by hemocytes (a cellular response) (12–17). We first confirmed the occurrence of a humoral response in flies infected with E. coli. When RNA prepared from adult flies was analyzed with RT-mediated quantitative PCR, there was an increase in the level of signals derived from the mRNA of attacin and diptericin, Imd pathway–inducible antimicrobial peptides, after the abdominal injection of E. coli (Fig. 4A). We then asked whether this response is involved in the action of σ³⁸ using imd¹, a hypomorphic mutant for the Imd pathway (32). This mutant fly line died earlier than the wild-type fly after the abdominal injection with the parental E. coli strain (Fig. 4B, left panel) confirming compromised immunity in imd¹. We then investigated whether the phenotypes of the rpoS-deficient E. coli seen with wild-type flies were also observed with imd¹. As anticipated, CFU of the parental bacteria increased by nearly 10-fold in a day when injected into imd¹ (Fig. 4B, middle panel). In contrast, there was no increase in CFU when the σ³⁸-lacking E. coli was similarly analyzed. Next, the lack of rpoS expression made E. coli temperate against imd¹ flies (Fig. 4B, right panel), as observed with wild-type flies (Fig. 2D). In addition, we performed similar experiments using the fly line Rel^E20, a null mutant for Relish coding for the NF-κB subunit Relish that functions downstream of Imd (33, 41), and found that the data resembled those obtained with imd¹ (Fig. 4C). These results indicated that the Imd pathway has nothing to do with the σ³⁸-dependent survival and pathogenicity of E. coli in Drosophila.

We next tested the involvement of a cellular immune reaction in the action of σ³⁸. For this purpose, we intended to inhibit phagocytosis by injecting latex beads (2 μm in diameter; Life Technologies, Tokyo, Japan) into the hemocoel of adult flies 3–12 h ahead of the infection with bacteria, according to the procedures reported previously (42). We first confirmed that the administration of latex beads weakened immunity in Drosophila by determining the rate of fly deaths after infection with the parental E. coli strain. The presence of latex beads did not influence fly survival without infection (Fig. 5A, left panel), but made flies die earlier than those that had received vehicle alone after the abdominal injection with the parental E. coli (Fig. 5A, right panel), which suggested compromised phagocytosis by the administration of beads. In these flies, the σ³⁸-lacking E. coli was not eliminated more rapidly (Fig. 5B, left panel) and not less pathogenic (Fig. 5B, right panel) than the parental strain. To confirm this, we inhibited phagocytosis by inducing apoptosis in hemocytes. For this purpose, the proapoptotic protein Reaper was specifically expressed in hemocytes using the GAL4-UAS system, and those flies were used as the host for bacterial infection. The induction of apoptosis itself did not kill flies (Fig. 5C, left panel), but made flies vulnerable to infection with E. coli (Fig. 5C, right panel). In addition, the level of phagocytosis of E. coli injected into the hemocoel was significantly reduced in those flies (data not shown). As observed with the flies injected with latex beads, the results with the apoptosis-induced flies indicated that a change in the level of colony-formable bacteria did not differ between the rpoS mutant and parental E. coli (Fig. 5D, left panel), and that the rpoS mutant E. coli was as pathogenic as the parental strain (Fig. 5D, right panel). These results indicated that the intactness of phagocytosis by hemocytes was required for σ³⁸ to contribute to the prolonged survival and pathogenicity expression of E. coli. The results also suggested that σ³⁸ exerted this function by preventing bacteria from being eliminated by phagocytosis. A decrease in the pathogenic effect of E. coli after the loss of σ³⁸ was evident in the fly lines PGRP-LE¹¹² and Atg1^Δ3d (Supplemental Fig. 1B) an in mutant flies deficient in the autophagy of Gram-negative bacteria in hemocytes (34), excluding the involvement of the autophagic killing of engulfed bacteria (43–45) in this action of σ³⁸.

A simple explanation for this phenomenon is that the rate of phagocytosis of E. coli was raised upon the loss of σ³⁸; however, this was not the case because hemocytes isolated from third-instar larvae, which were used as a surrogate for adult hemocytes, effectively phagocytosed E. coli in vitro regardless of the presence of σ³⁸ (Fig. 6A). We next tested the possibility that the rpoS-deficient E. coli was killed more rapidly than the parental strain in phagocytes after engulfment. Because larval hemocytes lysed within 1 h in primary culture for unknown reasons, we used l(2)mbn cells, a cell line established from larval hemocytes, for this experiment. l(2)mbn cells were incubated with E. coli, washed to remove unengulfed bacteria, and lysed immediately or after 1.5 h in culture. The resulting lysates were analyzed for the level of colony-formable bacteria. The results showed that CFUs of σ³⁸-lacking E. coli was reduced to <30% during a 1.5-h incubation, whereas almost 60% of the original CFUs remained with the parental strain (Fig. 6B), indicating the requirement of σ³⁸ for the E. coli resistance to phagocytic killing. We next examined whether the induction of rpoS expression in flies depends on the phagocytosis by hemocytes. For this purpose, l(2)mbn cells and E. coli were coincubated, and hemocytes and unengulfed bacteria were recovered separately and analyzed for the expression level of σ³⁸ and α subunit. We found an increased level of σ³⁸ in the lysate of hemocytes that supposedly had phagocytosed E. coli, whereas this was not true for E. coli that were incubated with hemocytes but remained unengulfed (Fig. 6C, left panel). Augmented expression of σ³⁸ after engulfment was evident at the level of mRNA as well (Fig. 6C, right panel). Furthermore, the inhibition of phagocytosis in flies by the preinjection of latex beads severely retarded the increase of σ³⁸ in E. coli after the abdominal injection (Fig. 6D). These results collectively suggested that the expression of rpoS was enhanced after phagocytosis resulting in the inhibition of killing of engulfed E. coli in hemocytes.

σ Factors, as a subunit of RNA polymerase, recognize promoters and determine which genes are to be transcribed. Therefore, we searched for genes that were induced by σ³⁸ and played roles in the σ³⁸ control of the survival and pathogenic effect of E. coli in Drosophila. Several hundred genes are transcribed in a manner dependent on σ³⁸, called the rpoS regulon (5, 8, 46). Among them, we focused on two genes, katE and katG, because they code for catalase and catalase-peroxidase, respectively—enzymes that detoxify the reactive oxygen species responsible for killing internalized bacteria in phagocytes. We first examined whether the expression of katE and katG was enhanced when E. coli was exposed to host environments. KatE and KatG mRNA levels were determined in the parental E. coli strain before and after injection into the hemocoel of wild-type Drosophila. The results showed that the mRNA level of both KatE and KatG was raised in adult flies (Fig. 7A). The increase of KatG mRNA was reduced when the rpoS mutant strain was similarly analyzed (Fig. 7A), confirming the dependence of katG expression on σ³⁸. A similar observation was made for KatE mRNA; however, a fluctuation in the data was too large to draw a conclusion (data not shown). We then determined the infectious properties of E. coli lacking KatE and KatG. We found that both mutant bacteria recapitulated the behavior of the rpoS mutant: they grew in vitro as actively as the parental E. coli (Fig. 7B), and they were more rapidly eliminated (Fig. 7C) and less pathogenic (Fig. 7D) in flies than the parental strain was. Together, katE and katG are most likely involved in the action of σ³⁸ to maintain the survival and pathogenicity expression of E. coli in Drosophila.

To confirm the role for katE and katG, we examined whether the phenotype of the rpoS mutant reverted by the forced expression of these genes. The σ³⁸-lacking E. coli was transfected with a plasmid that expressed KatE or KatG as a protein fused to the His-tag, and their whole-cell lysates were analyzed for the presence of these proteins. The results of Western blotting showed the presence of signals with the expected molecular masses of His-KatE and His-KatG (Fig. 8A). We then used these E. coli to infect adult flies to determine whether lowered survival and pathogenicity due to the lack of σ³⁸ were recovered. The results showed that the expression of either katE or katG raised both survival (Fig. 8B) and pathogenicity (Fig. 8C) of the rpoS mutant in adult flies. In contrast, the forced expression of katE or katG in the rpoS mutant did not change the level of phagocytosis by larval hemocytes in vitro (Fig. 8D). Moreover, an increase in bacterial pathogenicity by the expression of katE and katG was not observed when phagocytosis-compromised flies were used as hosts (Fig. 8E). Finally, the killing of the rpoS mutant in l(2)mbn cells after engulfment was recovered by the expression of katE and katG (Fig. 8F). The above-described results collectively suggested that KatE and KatG are responsible for the σ³⁸ control of E. coli survival and pathogenicity in Drosophila.

Although the involvement of σ³⁸ in the pathogenicity expression in host organisms and the response to stressful conditions of E. coli has been recognized for mammals (30, 31), it remained to be solved how this σ factor functions under such circumstances. In the current study, we found that σ³⁸ is required for the survival and pathogenicity expression of E. coli in fruit fly Drosophila, a genetically tractable model host. Furthermore, the forced expression of katE and katG, known target genes of σ³⁸, complemented a mutation in σ³⁸-encoding rpoS. The effect of σ³⁸ became less clear when phagocytosis-deficient flies were used as hosts; however, the susceptibility of E. coli to phagocytosis by Drosophila hemocytes did not change with the loss of σ³⁸. Instead, the E. coli lacking σ³⁸ was killed more rapidly than the parental strain in phagocytes after engulfment. katE and katG code for catalase and catalase-peroxidase, respectively, that detoxify reactive oxygen species. We thus interpret the observations as that σ³⁸ contributes to the prolonged survival of E. coli in Drosophila by enhancing the transcription of katE and katG for protecting engulfed bacteria from killing in phagocytes, although this scenario needs to be confirmed in vivo. We did not observe an effective growth of E. coli in adult flies under the conditions we adopted in this study, suggesting that the pathogenic effect was demonstrated not by the virulence of σ³⁸-expressing E. coli , but through reduced endurance of adult flies because of the prolonged existence of bacterial burden.

The cellular level of σ³⁸ is controlled at the steps of transcription, translation, and protein degradation (5, 8, 47). The increase in the mRNA level of σ³⁸ in adult flies indicates the involvement of a transcriptional control. We gained evidence for the involvement of phagocytosis in the induction of rpoS expression, and it was most likely that the transcription of rpoS is stimulated when E. coli is engulfed by hemocytes. The transcription of rpoS is enhanced when E. coli encounters conditions unfavorable for growth, possibly as a part of the quorum-sensing system (8). In fact, a quorum-sensing transcription factor of Pseudomonas aeruginosa has been shown to be involved in the mitigation of immune reactions in Drosophila (48). The level of rpoS transcription is under positive and negative regulation: stimulation by ppGpp and polyphosphate, and inhibition by cAMP and UDP-glucose (5, 8). Determining quantitative changes in the above-mentioned components of E. coli before and after engulfment should be important to clarify the mechanism for rpoS induction in phagocytes. In addition, the specific degradation of σ³⁸ by a protease called ClpP is known (49), and this could explain why a high level of σ³⁸ lasted only for a short period. The involvement of σ factor in the bacterial response to host environments has also been shown for adherent-invasive E. coli, which is a pathogen that is causative of inflammatory bowel disease. This E. coli induces the production of flagellum, which helps bacteria adhere to cell lines derived from human intestines, with the aid of σ²⁴ (50). In this example, σ²⁴ undergoes alterations, most probably not in abundance but in subcellular localization, through the degradation of a specific anti-sigma factor that sequesters σ²⁴ to the inner membrane (51). As a result, rapid killing of σ³⁸-lacking E. coli after engulfment could be partly explained by the action of σ³⁸ to raise bacterial adherence to the surface of hemocytes.

Bacteria appear to be equipped with strategies to resist killing in phagocytes (20, 26, 52–54). Salmonella avoid NADPH oxidase-dependent killing by inhibiting the delivery of the oxidase to phagosomes; Listeria escape from phagosomes to the cytoplasm by disrupting phagosomal membranes; Mycobacteria inhibit the acidification of phagosomes; and Legionella, Coxiella, Chlamydia, and Leishmania retard the maturation of phagosomes. Our proposal that E. coli avoids killing within phagosomes/phagolysosomes by expressing enzymes to detoxify reactive oxygen species should add another example to the list of mechanisms by which bacteria survive in the phagocytes of host organisms. However, further study is needed to delineate this new mechanism. In particular, it will be necessary to identify and characterize factors in hemocytes that trigger the transcription of rpoS and bacterial substances that respond to such host stimuli. We also need to verify the conservation of such a behavior in E. coli for other bacterial species and in vertebrate animals including humans.

We thank the National BioResource Project for bacterial strains and plasmids; Michael J. Galko, Shoichiro Kurata, Bruno Lemaitre, Thomas P. Neufeld, Kyorin-Fly, and Bloomington Drosophila Stock Center for fly lines; and Ayumi Kuroda and Takahiro Ito for contributing to the RT-PCR analysis. We also acknowledge the use of FlyBase and RegulonDB.

The authors have no financial conflict of interest.