Pseudomonas aeruginosa RhlR is required to neutralize the cellular immune response in a Drosophila melanogaster oral infection model

Abstract

An in-depth mechanistic understanding of microbial infection necessitates a molecular dissection of host–pathogen relationships. Both Drosophila melanogaster and Pseudomonas aeruginosa have been intensively studied. Here, we analyze the infection of D. melanogaster by P. aeruginosa by using mutants in both host and pathogen. We show that orally ingested P. aeruginosa crosses the intestinal barrier and then proliferates in the hemolymph, thereby causing the infected flies to die of bacteremia. Host defenses against ingested P. aeruginosa included an immune deficiency (IMD) response in the intestinal epithelium, systemic Toll and IMD pathway responses, and a cellular immune response controlling bacteria in the hemocoel. Although the observed cellular and intestinal immune responses appeared to act throughout the course of the infection, there was a late onset of the systemic IMD and Toll responses. In this oral infection model, P. aeruginosa PA14 did not require its type III secretion system or other well-studied virulence factors such as the two-component response regulator GacA or the protease AprA for virulence. In contrast, the quorum-sensing transcription factor RhlR, but surprisingly not LasR, played a key role in counteracting the cellular immune response against PA14, possibly at an early stage when only a few bacteria are present in the hemocoel. These results illustrate the power of studying infection from the dual perspective of host and pathogen by revealing that RhlR plays a more complex role during pathogenesis than previously appreciated.

Thanks to its powerful genetics and the lack of an adaptive immune response, the fruit fly Drosophila melanogaster is an ideal host in which to study many evolutionarily conserved features of host–pathogen relationships (1). The Drosophila host defense response in a septic injury model (in which pathogen cells are introduced directly into the body cavity) relies on the rapid activation of immune defenses, including coagulation and melanization, phagocytosis of invading microorganisms mediated by hemocytes, and a potent systemic humoral response involving the production of antimicrobial peptides by the fat body, the insect equivalent of the mammalian liver (1, 2). In the case of bacterial infections, pattern recognition receptors (PRRs) discriminate between two types of peptidoglycan (PGN). Diaminopimelic acid-type PGN triggers the immune deficiency (IMD) pathway. The antibacterial action of the IMD pathway is mediated in part by antimicrobial peptides (AMPs), including diptericin, which is active against Gram(−) bacteria. In contrast, lysine-type PGN, which is found in some Gram(+) bacteria, leads to the systemic activation of the Toll pathway that functions in parallel to the IMD pathway to activate the expression of a partially overlapping set of immune effectors, including the AMP drosomycin.

The Drosophila defense against infection is not limited to immunity in the body cavity (hemocoel). For example, intestinal infection models have revealed a role for the IMD pathway in barrier epithelia, including the midgut epithelium (3–6). Previously, we developed a Drosophila oral infection model with the potent entomopathogenic bacterium Serratia marcescens, which is able to cross the intestinal barrier (4). Interestingly, S. marcescens loses virulence in the hemocoel and is controlled by phagocytosis. We identified about 900 Drosophila genes that may be involved in defense against ingested S. marcescens (7). The well-studied human opportunistic pathogen Pseudomonas aeruginosa is also a potent Drosophila pathogen (8). Because of the extensive genetic tools available for P. aeruginosa, including a nonredundant transposon mutant library (9), and the multifaceted nature of P. aeruginosa virulence (10, 11), we and others have used Drosophila–P. aeruginosa oral infection models to study evolutionarily conserved mechanisms underlying infectious disease (12–16; reviewed in ref. 17). Here, we address pathogenesis from the dual perspective of host and pathogen by using mutants in both Drosophila and P. aeruginosa. We find with our infection protocol that ingested P. aeruginosa strain PA14 traverses the gut barrier and kills its host through a systemic infection. The P. aeruginosa quorum-sensing regulator RhlR is required for virulence and may allow P. aeruginosa to circumvent the hemocyte-mediated cellular immune response.

Results

Ingested P. aeruginosa Kills Flies by Bacteremia in the Hemocoel.

We monitored the survival of adult flies fed on a sugar solution (supplemented, or not, with bacterial growth medium) containing P. aeruginosa strain PA14. The severity of infection ranged from asymptomatic (sucrose-only solution) to severe (sucrose solution supplemented with bacterial growth medium) (Fig. 1 A and B and Fig. S1A). Typically, flies died more slowly (after about 8 d) when P. aeruginosa was ingested than after direct inoculation into the hemocoel in the septic injury model (48 h; Fig. S2A) (8, 14, 18, 19). P. aeruginosa PA14 did not appear to persistently colonize the fly intestine despite the presence of a stable steady-state number of viable bacteria in the intestine when flies were continuously feeding on the pathogen (Fig. S2B). Indeed, flies that were fed on P. aeruginosa for up to 3 d did not succumb to the infection when transferred to vials containing only a sterile sucrose solution (Fig. 1C) and actually cleared the bacteria from the gut. In contrast, flies transferred to the sterile sucrose medium after feeding on PA14 for 4 d died with similarly rapid kinetics as flies fed continuously on the pathogen, except that killing occurred about a day later, even though the pathogen was cleared from the digestive tract (Fig. 1C and Fig. S2B). In our experimental setting, we observed no significant degradation of the intestinal epithelium or an increase in intestinal stem cell proliferation, even at late stages of infection (Fig. S3 and SI Results and SI Discussion).

Fig. 1.

Systemic and cellular immune responses contribute to host defense against orally ingested P. aeruginosa PA14. (A and B) Survival following PA14 oral infection. IMD pathway mutants [imd (P = 0.0003, n = 8), key (P = 0.00005, n = 22)] and Toll pathway mutants [MyD88 (P = 0.0001, n = 22), spätzle (spz) (P = 0.01, n = 4)] succumbed faster to the infection than wild-type (wt) flies (A). Flies defective for phagocytosis [eater (P = 0.01, n = 3); latex bead-injected flies: wtΔphag (P = 8 × 10⁻⁷, n = 9)] also died faster than wild type (B). (C) Flies were either fed continuously or fed for the indicated period on the bacterial solution and then fed on a sterile sucrose solution that was changed daily; survival data are shown. At least 4 consecutive days of feeding were required to develop a lethal infection. (D) Bacterial titers measured in the hemolymph collected from batches of 10 flies in seven independent experiments are shown on a logarithmic scale. The values shown correspond to the bacterial titer per fly. Error bars are ±SD.

P. aeruginosa PA14 was able to cross the intestinal epithelium, although bacteria were barely detectable in the hemolymph after the first day of feeding on PA14 unless phagocytosis was blocked (Fig. 1D). Afterward, the titer of PA14 slowly increased in the hemolymph of wild-type flies or behaved somewhat erratically in flies with impaired host defense (see below) during the first 3–4 d while remaining at an absolute level of fewer than 100 bacteria per fly in the hemolymph. When we injected a similar number of bacteria in the body cavity (septic injury model), 50% of the flies succumbed within 48 h (Fig. S2A). In contrast, a similarly high number of bacteria were found in the hemolymph of orally infected wild-type flies around day 4, yet flies succumbed only starting from days 7–8 (Fig. 1 A and D). Thus, P. aeruginosa in the hemocoel appeared to be initially less virulent than in the septic injury model. Likewise, S. marcescens, another Gram(−) opportunistic pathogen able to orally infect Drosophila, is much more virulent following a septic injury. S. marcescens also crosses the intestinal barrier in an oral infection model, but in contrast to P. aeruginosa, does not proliferate in the hemocoel (4). Taken together with the lack of persistent colonization of the intestine by P. aeruginosa PA14, the steadily increasing bacterial titer in the hemolymph suggests that orally infected flies die from bacteremia.

Both a Systemic and a Cellular Immune Response Are Required in the Host Defense Against Ingested P. aeruginosa.

Next, we investigated the different facets of host defense in the P. aeruginosa oral infection model. As expected, IMD pathway mutants succumbed significantly earlier than wild-type flies of the same genetic background (Fig. 1A). We observed the induction of Diptericin reporter transgenes in the distal proventriculus and proximal midgut from day 1 onward (the proventriculus is the valve-like structure that connects the foregut to the midgut; Fig. 2 B–D and Fig. S4). In contrast to its expression in the proventriculus, the expression of the Diptericin-LacZ reporter was induced only from day 5 onward in fat body lobules in which the systemic immune response takes place (Fig. 2 E–G). The bulk of endogenous Diptericin mRNA started accumulating at days 4–5, suggesting that the majority of Diptericin expression occurs in the fat body (Fig. 2A). As expected, Diptericin expression was not induced in flies in which the IMD pathway gene kenny (key) is mutated, but was induced in MyD88 mutant flies in which the Toll pathway is abrogated (Fig. 2A) (20, 21).

Fig. 2.

An early-activated local IMD response and a late systemic IMD response both contribute to host defense against orally ingested P. aeruginosa PA14. (A) qRT-PCR analysis of the induction of Diptericin, a classic IMD pathway readout, in infected flies. Results are expressed as a percentage of the induction measured 6 h after a septic injury challenge with E. coli. P values (*) refer to the comparison between infected and noninfected flies of the same genotype: *P < 0.05; **P < 0.01; ***P < 0.001; n = 7. Other P values (°) refer to the comparison between mutant and wild-type flies at the same day of infection: °P < 0.05; n = 7. (B–G) β-Galactosidase staining of Diptericin-LacZ flies. Diptericin is induced in the proventriculus (arrows) throughout the infection (B–D), whereas systemic Diptericin induction in the fat body (arrowheads) of the fly occurs in later stages of the infection (E–G). (H) Rescue of the imd PA14 susceptibility phenotype by overexpression of a UAS-imd⁺ transgene (>IMD) with a gut (NPG4G80), a hemocyte (hmlG4G80), or a fat body (ylkG4)-specific driver as documented by the average time it takes to kill 50% of the flies (LT50). Note that AMPs synthesized in hemocytes and the fat body are secreted into the hemocoel. In this series of experiments, wild-type flies succumbed somewhat earlier than usual. P values computed by comparison with imd mutant flies: *P < 0.05; **P < 0.01; ***P < 0.001; n = 5. Error bars are ±SD. ns: not significant. (Scale bars: 500 μm.)

To determine which tissues are functionally relevant to the IMD defense against ingested PA14, we expressed a transgenic wild-type copy of imd either in the midgut, in hemocytes, or in the fat body of an imd mutant (rescue by overexpression) or of wild-type flies (overexpression) using the UAS-Gal4 expression system (22). To assess the degree of susceptibility to infection, for each survival experiment we computed the time required to kill 50% of the flies [median lethal time 50 (LT50)]. The overexpression of the imd transgene in a wild-type background did not significantly enhance protection against PA14. In contrast, the imd susceptibility phenotype was rescued by overexpressing the wild-type gene in the midgut, hemocytes, or fat body, suggesting that the IMD pathway can control defense responses in at least three different immune tissues (Fig. 2H and SI Discussion).

Typically, the Toll pathway is not strongly activated by Gram(−) bacteria (2), yet P. aeruginosa [a Gram(−) bacterium] has been shown to induce systemically the Toll pathway (18). Consistent with the latter study, Toll pathway mutant flies such as spätzle and MyD88 were more sensitive to oral P. aeruginosa infection (Fig. 1A). Accordingly, a Drosomycin-GFP reporter transgene (Toll pathway readout) was expressed only in the fat body from day 5 onward (Fig. 3 B and C). Similarly to Diptericin, the expression of endogenous Drosomycin as measured by qRT-PCR also became significant only from day 5 onward (Fig. 3A). The overexpression of a wild-type copy of MyD88 in hemocytes but not in the midgut was sufficient to rescue the P. aeruginosa susceptibility phenotype of a MyD88 mutant (Fig. 3D). Similar to imd overexpression, transgene-mediated activation of the Toll pathway using UAS-Toll^10b (encoding a constitutively active form of the receptor) or UAS-MyD88⁺ in the midgut or the hemocytes before P. aeruginosa ingestion did not provide enhanced protection against the infection in a wild-type background (Fig. 3D). Taken together, the expression data and the genetic experiments suggest that in the late stages of the infection process the Toll pathway acts through the systemic immune response to impede P. aeruginosa infection.

Fig. 3.

Late Toll pathway activation contributes to systemic host defense against orally ingested P. aeruginosa PA14. (A) qRT-PCR analysis of the induction of Drosomycin, a classical readout of Toll pathway activation, in infected flies. Results are expressed as a percentage of the induction measured 24 h after a septic injury challenge with M. luteus. P values (*) refer to the comparison between infected and noninfected flies of the same genotype: *P < 0.05; **P < 0.01; ***P < 0.001; n = 6. No significant difference was observed between wild-type (wt) and key flies with respect to Drosomycin expression levels. (B and C) Drosomycin-GFP reporter induction in the fat body upon infection. (D) Overactivation of the Toll pathway and rescue of the MyD88 susceptibility phenotype by overexpression of a UAS-MyD88⁺ transgene (>MyD88) with a gut (NPG4G80) or a hemocyte (hmlG4 or hmlG4G80)-specific driver, as documented by the average time it takes to kill 50% of the flies (LT50). Rescue was observed by overactivation of the Toll pathway in hemocytes, but not in the gut. Note that AMPs synthesized in hemocytes are secreted into the hemocoel. The UAS-Toll^10B transgene (>Toll10B) expresses a gene encoding a constitutively active form of the Toll receptor. In this series of experiments, wild-type flies succumbed somewhat earlier than usual. P values are compared with MyD88 mutant flies: *P < 0.05; **P < 0.01; n = 5. Error bars are ±SD. ns: not significant. (Scale bar: 500 μm.)

A cellular immune response constitutes an important arm of host defense in several infection models, likely through phagocytosis (13, 14, 23–25). We therefore asked whether phagocytes play an important role in our P. aeruginosa PA14 feeding model. We impaired the cellular response either by injecting nondegradable latex beads (26) or by using mutant flies deficient for the phagocytic receptor Eater (23). In both cases, we observed a significantly reduced resistance to ingested P. aeruginosa (Fig. 1B), even under conditions in which ingested P. aeruginosa does not kill wild-type flies (sucrose only, Fig. S1 B and C). Flies in which the cellular response was blocked by latex bead injection displayed a higher bacterial titer than wild-type flies (Figs. 1D and 4B). We therefore investigated the possibility that PA14 impairs the phagocytic machinery of hemocytes. Even during the final phase of the infection, however, hemocytes were still able to ingest fluorescein-labeled Escherichia coli, suggesting that hemocytes are present and not impaired in their ability to phagocytose bacterial particles (Fig. S5).

Fig. 4.

RhlR, but not LasR, is required to counteract the cellular immune response against P. aeruginosa PA14. (A) Survival experiments in wild-type Drosophila to analyze virulence of P. aeruginosa mutants in known virulence factors. The average time that it takes to kill 50% of flies (LT50) is plotted. Two rhlR transposon insertion mutants [37943 (referred to as rhlR) and 34255] and a deletion (ΔrhlR) displayed the same attenuated virulence phenotype, whereas other mutants were not significantly less virulent than wild-type (wt) PA14. pscD is a deletion mutation that affects the secretion machinery and thus prevents the secretion of all T3SS effectors, including ExoT. exoT mutant bacteria were tested in independent experiments using flies of a different genetic background and also did not show a phenotype (n = 3). *P < 0.05; **P < 0.01; ***P < 0.001; n = 3 or 4 depending on the mutant tested. (B–F) In the Insets, the genotype of the host [wild-type (wt) or mutant flies] and the genotype of the pathogen (PA14 refers to wild-type PA14) are indicated. (B) Bacterial counts per fly measured in the hemolymph collected from PA14 and rhlR infected wild-type (wt) and latex bead-injected flies (wtΔphag) expressed on a logarithmic scale (n = 3). (C) Survival experiments using wild-type PA14 and rhlR mutant bacteria. rhlR mutants are less virulent (P values PA14 vs. rhlR in wild-type flies: P = 0.0017, n = 7; key flies: P = 0.0020, n = 6; MyD88 flies: P = 0.0001, n = 7). (D) rhlR mutant bacteria killed phagocytosis-deficient latex bead-injected flies as rapidly as wild-type bacteria (P > 0.05, n = 6). (E) Survival experiments using wild-type PA14 and lasR mutant bacteria. lasR mutants are as virulent as PA14 in wild-type flies and less virulent in key and MyD88 flies (P values PA14 vs. lasR in wild-type flies: P = 0.33, n = 5; key flies: P = 0.027, n = 2; MyD88 flies: P = 0.025, n = 3). (F) lasR mutant bacteria killed phagocytosis-deficient latex bead-injected flies somewhat less rapidly than wild-type bacteria (P = 0.013, n = 4). Error bars in A and B are ±SD.

Taken together, our data suggest that different host defenses become relevant at distinct stages of the infection.

RhlR, but Not the LasR Acylhomoserine Lactone Quorum-Sensing Transcription Factor, Is Required for the Virulence of Orally Ingested P. aeruginosa PA14.

To determine which bacterial factors influence the virulence of PA14 in the oral infection model, we challenged wild-type flies with bacteria defective for the type III secretion system (T3SS) (pscD) or one of its effectors (exoT); the GacA virulence regulator; the AprA alkaline protease, which is important for Pseudomonas entomophila virulence in an oral infection model of Drosophila (6); the LasR N-3-(oxododecanoyl)homoserine lactone quorum-sensing regulator; the LasB elastase; the MvfR quinoline quorum-sensing regulator; the PhoB low-phosphate response regulator; and PA14 deficient for the formation of biofilm (pelA). All of these bacterial mutants displayed normal virulence in wild-type flies (Fig. 4A and Fig. S6C). In contrast, several independent PA14 rhlR mutants, which are deficient for the C4-acylhomoserine lactone-dependent quorum-sensing regulator RhlR, a second acylhomoserine lactone quorum-sensing system in P. aeruginosa (27), were severely impaired in virulence and killed the flies 3–4 d later than wild-type PA14 (Fig. 4 A, C, and D and Fig. S6 A and B). A lasR-rhlR double mutant behaved like the single rhlR mutant in wild-type flies (Fig. S6C). Interestingly, flies infected with rhlR mutants did not succumb as synchronously as flies infected with wild-type PA14. The survival curve was significantly shallower as quantified by the Hill coefficient, which measures the steepness of a sigmoid curve (Fig. S7). Consistent with their reduced ability to kill flies, the titer of the rhlR mutants in the hemolymph was lower than the titer of PA14 and reached a maximum of around 100 bacteria per fly at day 6 of infection (Fig. 4B). Thus, rhlR mutant bacteria appear to be cleared more efficiently from the hemolymph. In keeping with these data, the systemic immune response was hardly induced, as measured by the accumulation of Drosomycin and Diptericin mRNAs (Fig. S8A).

RhIR Is Required to Circumvent the Cellular Immune Defense of P. aeruginosa Orally Infected Flies.

To distinguish the possibilities that RhlR is required to counteract or elude either the systemic humoral immune response or the cellular arm of host defense, or both, we first infected wild-type, key, or MyD88 mutant flies with either wild-type P. aeruginosa or an isogenic rhlR mutant. We found that rhlR mutant bacteria killed key or MyD88 mutant flies with the same low virulence as wild-type flies (Fig. 4C).However, the titer of PA14 rhlR was somewhat higher in the key or MyD88 mutants than in wild-type flies, presumably because the bacteria were not cleared as efficiently from the hemolymph (Fig. S8B). As expected, PA14 rhlR mutants did not induce Diptericin in key mutants, but Drosomycin was induced to somewhat lower levels by PA14 rhlR in the key mutant than in wild-type flies (Fig. S8A). Conversely, PA14 rhlR did not induce Drosomycin in MyD88 mutants, and Diptericin was induced to significantly lower levels than those measured after an oral challenge with wild-type P. aeruginosa (Fig. S8A).

Next, we impaired the cellular immune response either by injecting latex beads before feeding the flies on rhlR mutant bacteria or by using eater mutants (sucrose-only conditions, as described above). In striking contrast to MyD88 or key mutants, phagocytosis-deficient flies succumbed almost as rapidly as controls fed with wild-type P. aeruginosa (Fig. 4D and Fig. S1B). In other words, the rhlR mutant is highly virulent when the cellular immune response is impaired. Moreover, the rhlR bacterial titer measured in latex bead-treated flies was as high as that of PA14 in wild-type flies (Fig. 4B). These results suggest that RhlR's role in virulence in this infection model is to circumvent the cellular arm of immunity. Interestingly, whether the rhlR mutant bacteria were virulent or not in immunodeficient flies, we noted that the slopes of survival curves were shallower than with wild-type P. aeruginosa, indicating that the requirement for RhlR in synchronizing the rate of death among infected flies is independent of host defenses (Fig. 4 C and D and Fig. S7). We also assessed the role of RhlR in a septic injury model. Consistent with the oral infection model, rhlR mutants were significantly less virulent in wild-type flies, but not in phagocytosis-deficient, latex bead-injected flies (Fig. S9), further supporting the idea that RhlR is involved in counteracting hemocyte-mediated defense responses. To determine how RhlR controls virulence in our oral infection model, we tested several known direct transcriptional targets of RhlR. We found that rhlA, rhlB, and phenazine (phzH, phzM, phzS, and Δphz1/2) mutants were as virulent as wild-type PA14 in both wild-type or immunosuppressed flies (Fig. S6C). These data indicate that RhlR exerts its effects through other, yet unknown, effectors.

Unexpectedly, we found that LasR was required for virulence in flies defective for key, MyD88, and phagocytosis (Fig. 4 E and F and Fig. S1C). Thus, in contrast to RhlR, LasR contributes to virulence only in immunocompromised, but not in wild-type, flies.

Discussion

Fly Model of Generalized Bacteremia Following Gastrointestinal Infection.

Drosophila has been widely used as a model host to study P. aeruginosa pathogenesis (12, 14–19, 28, 29). Here, we used an oral infection model to investigate in detail the interplay between bacterial virulence mechanisms and the host response by using both host and pathogen mutants defective in immunity or virulence, respectively. In previously described Drosophila oral infection models using P. aeruginosa, the actual cause of death has rarely been investigated (17). Unlike a previous study (15), we did not observe extensive damage to intestinal epithelial cells in our experimental setting (Fig. S3 and see SI Results and SI Discussion for further discussion). Thus, it is unlikely that flies succumb to intestinal damage. Rather, our results show that some ingested P. aeruginosa cross the digestive tract (a conclusion that can also be drawn from the study reported in ref. 14) and cause a systemic infection as evidenced by the high bacterial titer measured in the hemolymph before death (Fig. 1), akin to human infections caused by foodborne pathogens (30, 31). In support of this conclusion is our finding that flies succumb to bacteremia when they are fed on P. aeruginosa for 4 d and are then transferred to a sterile feeding solution. The flies die, even though the bacteria are quickly cleared from the gut after being transferred (Fig. 1C and Fig. S2). Thus, this oral infection model in Drosophila provides a paradigm in which to study intestinal infections that can lead to bacteremia.

Two Phases of Infection and a Switch in the Virulence Program Controlled by the RhlR Virulence Regulator?

Several studies with Drosophila and other insects have shown that very low numbers of P. aeruginosa cells (as few as 1–10) introduced into the body cavity by microinjection or pricking are able to rapidly multiply, causing a lethal bacteremia over the course of about 2 d (14, 18, 19, 32). The behavior of P. aeruginosa in our oral infection model is markedly different. The bacterial titer in the hemolymph remains low during the first phase of the infection (Fig. 1D and Fig. S8B). This may reflect a low-virulence state of the bacteria that cross the gut barrier as described previously for S. marcescens (4), or it may reflect the ability of humoral or cellular immune defenses to initially cope with the invading bacteria, or both. Indeed, a systemic immune response is significantly induced only at day 5 of feeding, when the bacterial titer in the hemolymph has increased significantly (Figs. 2 and 3). We note that a systemic immune response is induced earlier in the infection in immunodeficient flies (Figs. 2 and 3), in which case the bacterial titer also increases more rapidly (Fig. 1D).

Because PGN is not exposed on the surface of Gram(−) bacteria, they may not be detected by Drosophila’s PRRs unless the bacteria proliferate and release small PGN fragments generated during cell-wall remodeling (4). One explanation of our data showing that a systemic immune response occurs only after 5 d of feeding (Fig. 2 A and E–G and Fig. 3 A–C) is that the P. aeruginosa cells that initially cross the epithelial barrier into the hemolymph are in a relatively avirulent state but eventually switch to a high state of virulence (Figs. 1D and 4B). Alternatively, or concomitantly, the late onset of systemic immunity may reflect the gradual influx of bacteria through the gut into the hemolymph until they reach sufficiently high numbers to overcome local and phagocytic defenses. Finally, it is possible that the P. aeruginosa cells in the intestine or the few that translocate into the hemocoel actively suppress the systemic immune response as has been observed in a septic injury model with P. aeruginosa PA14 (29). This latter hypothesis may appear somewhat unlikely given the low number of bacteria retrieved from the hemolymph during the early phase of the infection. Because hemocytes are phagocytically active throughout the course of the infection and because the systemic immune response is not activated in the first phase of the infection, the cellular response may be the main active defense during the early phase in the hemocoel.

In summary, there seem to be two phases in the infection. In the early phase, bacteria cross the gut barrier and are most likely controlled efficiently in the hemocoel by phagocytes. In the late phase, P. aeruginosa PA14 is able to resist at least partially the cellular immune response through RhlR and then starts proliferating rapidly in the hemolymph, thus activating a systemic immune response, which in turn slows the infection process.

Genetic Analysis of Host–Pathogen Interactions Yields Insights into the in Vivo Roles of P. aeruginosa Quorum-Sensing Virulence Regulators.

By using genetic mutants in both partners of an infectious host–pathogen relationship, we have been able to reveal unexpected in vivo roles for two global regulators of P. aeruginosa virulence: the transcription factors RhlR and LasR. RhlR is the major regulator of C4-homoserine lactone quorum-sensing, one of three quorum-sensing systems in P. aeruginosa (27, 33). Our data show that RhlR plays a key role in the oral infection model as rhlR mutants display strongly attenuated virulence (Fig. 4C). However, RhlR is unlikely to be required for passage through the intestinal barrier because rhlR mutants can kill phagocytosis-deficient flies as rapidly as wild-type PA14 (Fig. 4D and Fig. S1B).

It has been proposed that P. aeruginosa partially inhibits the systemic AMP response induced in the septic injury model (29). We do not think that RhlR is responsible for this virulence function: rhlR mutants also displayed an attenuated virulence phenotype in both IMD and Toll pathway mutant flies (key and MyD88 mutants; Fig. 4C), arguing that NF-κB–independent defense mechanisms contain the infection in these cases. Furthermore, we did not detect enhanced induction of AMP gene expression when wild-type flies were infected with rhlR mutant bacteria (Fig. S8A), suggesting that RhlR is not involved in suppressing the AMP responses. Rather, the decreased induction of the AMP genes likely reflects the reduced ability of rhlR bacteria to proliferate in vivo (Fig. 4B).

In contrast to the results obtained with IMD and Toll pathway mutants, RhlR is dispensable when the cellular immune response is impaired either by the prior injection of latex beads or in eater mutants (Fig. 4D and Fig. S1B). These data suggest that an essential in vivo function of RhlR is to circumvent phagocytosis of P. aeruginosa by professional phagocytes. However, in contrast to the P. aeruginosa toxins secreted by the type III secretion system such as ExoS, RhlR is not required to impair the general phagocytic activity of hemocytes because killed E. coli appear to be ingested normally (Fig. S5). In light of the three explanations delineated above for why P. aeruginosa initially fails to proliferate in the hemolymph and fails to activate a systemic immune response, it is possible that RhlR is required at a critical period during the infection to protect P. aeruginosa bacteria from phagocytic clearance by an unknown mechanism. Interestingly, RhlR function appears to be required at a relatively early time in the infection process when the measured bacterial titer in the hemolymph is rather low (about 100/fly) (compare Fig. 4B to 4D). Quorum sensing should not be activated at this low bacterial concentration.

When either wild-type flies or flies with an impaired immune function were infected with a P. aeruginosa rhlR mutant, they exhibited a shallow survival curve. In contrast, flies that had ingested wild-type PA14 died in a more synchronized manner. Thus, RhlR seems to play an important role in the coordinated onset of pathology in the population of infected flies as a whole, which may be related to its classic role in quorum sensing. In the absence of RhlR, bacteria may behave in a more erratic manner because of a lack of coordination of bacterial virulence properties through quorum sensing.

The RhlR target genes—phenazines, RhlA, and RhlB—have been shown to be important for virulence in several model hosts from plants to mammals (34). Here, we find that these genes are not required for virulence in our Drosophila oral infection model. Thus, RhlR may circumvent the cellular immune response through other as-yet-uncharacterized target genes.

In contrast to RhlR, the transcriptional regulator LasR, which controls 3-oxo-C12-homoserine lactones, is not required for virulence in wild-type flies (Fig. 4 A, E, and F). lasR mutants display a phenotype that is the opposite of rhlR: lasR mutants, unlike rhlR mutants, are attenuated in phagocytosis-impaired flies (Fig. 4 D and F and Fig. S1 B and C). Classically, the two acylhomoserine lactone quorum-sensing systems of P. aeruginosa are thought to function in a hierarchical order, with the LasR system on top of the RhlR regulon (27). However, it was recently shown that RhlR can control the expression of LasR-specific factors independently of LasR and conversely (33). Furthermore, quorum-sensing systems themselves are under environmental control (35). Thus, our study underscores the necessity to functionally dissect the role of virulence factors in vivo in both immunocompetent and immunocompromised hosts to obtain insights into their complex regulatory roles in pathogenesis. Indeed, quorum-sensing defective strains have been isolated from patients (36 and references therein). The finding that LasR is required for virulence in immunocompromised flies reveals a subtler LasR function that may be masked in wild-type flies. This is a reminder that bacterial screens for avirulent mutants in host-sensitized backgrounds are likely to yield insights that may not be gained using only wild-type host organisms. This and many other features of this study highlight the usefulness of model organisms in studying infectious disease.

Materials and Methods

Bacterial Strains and Growth Conditions.

We used the following strains: wild-type strain—P. aeruginosa PA14 (37); rhlR—two transposon insertions, GID3229 ID#37943 (referred to as rhlR) and GID3229 ID#34255 [referred to as 34255 (rhlR)] (9), and two independent sets of deletion mutants that are further described in SI Materials and Methods together with lasR, rhlA, rhlB, pelA (38), and phz1/2 (39) deletions. For deletion mutants, we used ΔpscD; ΔexoT (40), and for insertions we used gacA (37), aprA GID865 ID#23768, lasB GID759 ID#45691, phoB GID3473 ID#48234, rhlA GID2578 ID#23291, rhlB1 GID1159 ID#28984, rhlB2 GID1159 ID#27130, pelA GID86 ID#26187, phzH GID443 ID#39981, phzM GID2109 ID#40343, phzS GID1461 ID#44099 (9), and lasR-RhlR (33). All bacteria were grown in brain–heart infusion broth (BHB) overnight with shaking at 37 °C. We observed similar survival curves of infected flies when PA14 was grown and incubated with Luria broth (LB). E. coli and Micrococcus luteus for qRT-PCR controls were grown in LB overnight with shaking at 37 °C.

Survival Experiments.

An overnight culture of bacteria was centrifuged (4,000 × g, 10 min, 4 °C) and diluted in fresh BHB to obtain a solution of OD₆₀₀ = 2.5. This solution was then diluted 10 times with a sterile 50-mM sucrose solution to OD₆₀₀ = 0.25. Two absorbant pads (Millipore AP1003700) were placed at the bottom of clean medium-size vials (3.5-cm diameter), and 2 mL of bacterial solution was added to the filters before the introduction of about 20 flies, which had been feeding on a 50-mM sucrose solution for 2 d at 25 °C. Survival experiments were performed at 25 °C and 50% humidity, and the number of surviving flies was monitored. For overexpression/rescue experiments, flies were incubated at 29 °C for 48 h (on fly food) before infection to inactivate Gal80 and allow for strong Gal4 activity (7).

For experiments using the oral infection model under conditions in which wild-type flies are not killed (Fig. S1), bacteria were centrifuged and washed in PBS. The pellet was then diluted with 5% sterile sucrose solution to an OD₆₀₀ of 0.1, and 7-mL aliquots of this medium were pipetted onto sterile cotton balls placed at the bottom of empty fly culture vials. Further descriptions of materials and methods are found in SI Materials and Methods.