Mutations That Impact the Enteropathogenic Escherichia coli Cpx Envelope Stress Response Attenuate Virulence in Galleria mellonella

Abstract

In this paper, we show that the larvae of the greater wax moth, Galleria mellonella, can be used as a model to study enteropathogenic Escherichia coli (EPEC) virulence. G. mellonella larvae are killed after infection with EPEC type strain E2348/69 but not by an attenuated derivative that expresses diminished levels of the major virulence determinants or by a mutant specifically defective in type III secretion (T3S). Infecting EPEC inhabit the larval hemocoel only briefly and then become localized to melanized capsules, where they remain extracellular. Previously, it was shown that mutations affecting the Cpx envelope stress response lead to diminished expression of the bundle-forming pilus (BFP) and the type III secretion system (T3SS). We demonstrate that mutations that activate the Cpx pathway have a dramatic effect on the ability of the bacterium to establish a lethal infection, and this is correlated with an inability to grow in vivo. Infection with all E. coli strains led to increased expression of the antimicrobial peptides (AMPs) gloverin and cecropin, although strain- and AMP-specific differences were observed, suggesting that the G. mellonella host perceives attenuated strains and Cpx mutants in unique manners. Overall, this study shows that G. mellonella is an economical, alternative infection model for the preliminary study of EPEC host-pathogen interactions, and that induction of the Cpx envelope stress response leads to defects in virulence.

INTRODUCTION

Enteropathogenic Escherichia coli (EPEC) causes an infection of the lower intestine and severe watery diarrhea, resulting in several hundred thousand deaths each year (11, 51, 63). Typical EPEC strains harbor the EAF (EPEC adherence factor) plasmid that carries genes necessary for the production of bundle-forming pili (BFP), which connect bacteria within microcolonies and lead to the characteristic localized adherence pattern in which clumps of bacteria adhere to intestinal epithelial cells (21, 68). The type IV BFP is likely the predominant adhesin and is an important pathogenicity factor in EPEC (6). Following initial adherence, a type III secretion system (T3SS) translocates virulence effector proteins from the bacterial cytosol directly into the host cell cytoplasm, where they facilitate disease development (3, 12, 14). The last step in infection is intimate adherence, which leads to a localized effacement of absorptive microvilli and the accumulation of host cytoskeletal proteins just beneath the attached bacteria. This leads to the formation of the typical attaching-and-effacing (A/E) lesions characteristic of EPEC infections (65). Most of the genes that are necessary for the formation of the A/E lesions, including those encoding the T3SS and its effectors, are located on a chromosomal pathogenicity island called the locus of enterocyte effacement (LEE) (46).

Expression of the LEE genes is regulated by many factors (5). Of key importance are the EAF plasmid-encoded PerA, PerB, and PerC regulators, which lead to activation of expression of Ler, which is encoded by the LEE (47). Ler, in turn, is thought to be the key regulator of LEE gene expression (4, 7, 20). A number of chromosomally encoded signal transduction proteins, separate from the EAF plasmid or LEE pathogenicity island, have also been demonstrated to regulate EPEC virulence determinant expression (5). One of these is the Cpx envelope stress response (42, 58, 73). This three-component regulatory system is comprised of the membrane-bound histidine kinase CpxA, the cytoplasmic response regulator CpxR, and the small periplasmic protein CpxP, which, in the absence of stress, inhibits the autokinase activity of CpxA (19, 57). Prevailing evidence suggests that the physiological function of the Cpx response is to sense and mediate adaptation to protein misfolding in the envelope. It was recently demonstrated that adaptation involves shutting down production of envelope-localized protein complexes, including virulence determinants (42, 72, 73), as well as activating the expression of protein folding catalysts and degrading factors (DsbA, PipA, PipD, and DegP) (17, 18, 43).

It has long been thought that typical EPEC serotypes cannot infect animals, including mice, suggesting that humans are the only living reservoir (50, 51). However, a study by Savkovic et al. (64) showed that a specific type of mouse (C57BL/6J) can be infected by EPEC and demonstrates some symptoms of disease consistent with those observed in humans, although in much milder forms. Another animal model for the study of human EPEC infections is rabbit, as virulence mechanisms and resulting pathogenicity observed during natural infections of rabbits with O103:K-:H2 EPEC strains are similar to those of human EPEC infections (15, 38, 60). While it is a good model organism to study virulence of EPEC, the ethical considerations as well as housing and food costs have hindered the more widespread use of this model. Other studies of EPEC virulence have been conducted in humans already showing clinical symptoms (9, 25). In addition, a cell culture model of human small intestinal epithelia (Caco-2 cells) (45) has been used to study EPEC pathogenesis. Lastly, the related microbe Citrobacter rodentium has been used as a surrogate to study EPEC disease causation, as it can easily infect mice and shows infection characteristics similar to those of EPEC (13, 41, 53). Although ultimately it is desirable to study important virulence genes using animal models closely related to humans, these models are expensive and time-consuming and are not practical (for most investigators) for the purposes of large-scale genetic screens aimed at identifying new genes affecting host-pathogen interactions.

In this study, we investigated if the larva of the greater wax moth Galleria mellonella could be used as a simple, invertebrate model system to study EPEC infections. Its amenability to infection and its ability to mount an innate immune response makes the larva an attractive model organism for the study of microbial pathogenicity (22). Another advantage of this organism is its large size, which enables precise injections of pathogens and the easy collection of hemolymph and tissue (71). Most importantly, several previous studies showed that G. mellonella is a useful and reliable model for studying the virulence of different pathogenic bacteria, e.g., Burkholderia species (66, 74), Acinetobacter baumannii (55), Listeria species (49), Pseudomonas aeruginosa (48), and Enterococcus faecalis (77). These studies indicate that the ability of these pathogens to kill G. mellonella larvae is directly correlated with the pathogenic potential observed in mammalian model organisms, justifying the use of G. mellonella as a simple infection model for initial broad-scale screening to identify particular bacterial mutants that may be desirable to characterize in detail in more expensive mammalian models. In this study, we show that G. mellonella can be used as an invertebrate model to study enteropathogenic E. coli (EPEC) virulence. We used the G. mellonella infection model to show that mutations that impact the Cpx envelope stress response significantly alter the ability of EPEC to infect G. mellonella.

MATERIALS AND METHODS

Strains and larvae.

Last-instar larvae of Galleria mellonella were obtained from Recorp Inc. (Georgetown, Ontario, Canada) and stored in wood chips at 4°C. All larvae were used within 1 week. All bacterial strains used in this study are listed in Table 1. Bacteria were routinely cultured in Luria-Bertani (LB) medium overnight at 30 or 37°C. To activate virulence factors, strains were transferred (1:50 dilution) into Dulbecco's modified Eagle medium (DMEM-F12) supplemented with 0.1 M Tris-HCl, pH 7.5, at 37°C with shaking until an optical density at 600 nm (OD₆₀₀) of 0.6 was reached. For infections, 1 ml of each culture was harvested by centrifugation and washed three times in 1 ml of 10 mM MgSO₄ · 7H₂O buffer supplemented with strain-specific antibiotics. The final, washed bacterial pellet was resuspended in 1 ml 10 mM MgSO₄ · 7H₂O buffer supplemented with strain-specific antibiotics prior to infection.

Table 1.

Strains used in this study

Strain	Characteristic(s)	Antibiotic resistance	Growth temp (°C)	Reference or source
E2348/69	Wild-type EPEC O127:H6	Streptomycin	37	39
JPN15a	E2348/69 lacking the EAF plasmid	Chloramphenicol	37	33
ALN88	E2348/69 ΔcpxR	Kanamycin	37	52
ALN195	E2348/69 ΔcpxA24	Amikacin	30	42
SV235	E2348/69 ΔbfpA-L	Streptomycin	37	S. L. Vogt and T. L. Raivio, unpublished data
SL189	E2348/69 ΔescN	Streptomycin	37	37
E2348/69	E2348/69 harboring plasmid pxG-1	Chloramphenicol	37	69

^aStrain JPN15 contains the empty plasmid pAC24N (34) to confer resistance to chloramphenicol for the purpose of discriminating JPN15 from the normal flora of G. mellonella.

Infection and hemolymph extraction.

Larvae were infected with 5 μl of selected strains by injection into the hemolymph via the hindmost left proleg. For all experiments, larvae were incubated at 30°C in standard petri dishes for up to 72 h. To collect hemolymph, insects were washed once in 70% ethanol and twice in 10 mM MgSO₄ · 7H₂O to minimize surface contaminants. The abdomen of the larva was pricked with a sterile needle, and the hemolymph was collected with a pipette and transferred into a prechilled tube.

Killing assays and LD₅₀ tests.

All bacterial strains were grown for a brief period in defined DMEM at 37°C prior to infection to induce virulence determinant production. Serial dilutions of each strain tested were prepared (0, 10¹, 10², 10³, 10⁴, and 10⁵ cells/5 μl) in 10 mM MgSO₄ · 7H₂O buffer (supplemented with strain-specific antibiotics) and used for infections. Ten larvae were injected with 5 μl at each dilution, and experiments were conducted at least in triplicate. Control larvae were injected with 10 mM MgSO₄ · 7H₂O buffer to measure any lethal effects of the injection process. No more than one control organism died during any given trial. All infections were carried out at 30°C, since the cpxA* mutant ALN195 does not grow well at 37°C. Once induced, it has been demonstrated that virulence factor production remains elevated even when the cells are kept at 30°C afterwards (61). All organisms were monitored for survival for up to 72 h. To investigate if heat-killed E2348/69 cells are able to kill G. mellonella in a similar fashion, E2348/69 cells were heat inactivated for 30 min at 75°C as previously described (56). Five μl of every dilution injected was spotted onto LB plates and incubated overnight at 37°C to ensure that no viable organisms remained. All organisms were monitored for survival for up to 72 h. Larvae were considered dead when they did not show any response to touch. Data from at least three independent experiments were combined, and the 50% lethal doses (LD₅₀) for each strain were calculated as previously described (31). Significant differences were compared using a Student's t test.

Dissection and microscopy.

To investigate where the bacteria localize during infection, 10 larvae were infected with approximately 10⁴ cells of E2348/69 containing a constitutively upregulated green fluorescent protein (GFP) (Table 1) and incubated for 18 h at 30°C. Uninfected larvae were used as negative controls. Larvae showing advanced signs of melanization were selected and dissected (see Fig. 3B). Larvae were put on ice for 10 min to slow movement and pinned onto black agar plates, and an incision was made at the anterior end and extended along the dorsal side toward the posterior end. Larval skin was pinned down and pictures were taken using an Olympus E-420 camera mounted on a dissecting microscope (×6 magnification) (see Fig. 3A and B). Samples of melanotic material were mounted onto a microscopy slide and examined with a Reichert-Jung Polyvar microscope (×60 magnification) (see Fig. 4). Pictures were overlaid using the freeware software GIMP 2.6.12 (www.gimp.org).

Fig 3.

Dissection of an uninfected (A) and an infected (B) Galleria mellonella larva. Size bar, 10 mm. Arrows indicate zones of melanotic tissue.

Fig 4.

Melanotic flex extracted from an infected larvae. Green fluorescence shows E2348/69 cells, and blue fluorescence (DAPI) shows hemocytes. Size bar, 2 μm.

Bacterial load test.

To examine bacterial growth during infection, bacterial load tests were performed with all strains listed in Table 1. After 0, 4, 18, 24, and 48 h of infection, 10 larvae (in triplicate) were analyzed. We first extracted hemolymph from the living larvae as previously described. The hemolymph-drained larvae were then macerated using a mortar and pestle to obtain whole larval macerate. Dead, melanized larvae were also macerated with mortar and pestle. Ten microliters of hemolymph and 100 mg of larval macerate were used for serial dilutions and plated onto MacConkey plates containing strain-appropriate antibiotics. Plates were incubated for 24 h at 30 or 37°C as indicated. Commensal flora in the macerates of uninfected larvae were shown not to grow under these conditions (data not shown).

G. mellonella immune gene expression and bacterial AMP sensitivity.

To investigate the induction of the G. mellonella innate immune response after infection, 25 larvae were infected with approximately 10³ cells of each individual strain and incubated for 10 min and 4, 16, 24, and 48 h. Twenty-five larvae injected with 10 mM MgSO₄ · 7H₂O were used as injury controls. At selected time points, larvae were chilled on ice for 10 min, the rear ends of the larvae were cut off, and the insides (containing hemolymph, fat bodies, and internal organs) were squeezed into a chilled 15-ml falcon tube. The weight was determined, and 1 ml TRIzol reagent (Ambion) was added for every 100 mg of sample. RNA was extracted as recommended by the manufacturer (Ambion). Three micrograms of RNA was subjected to a DNase digest (1 U/μg) at 37°C for 45 min to remove residual DNA, and the resulting RNA was reverse transcribed with SuperScript II reverse transcriptase (Invitrogen) with 1 μg RNA according to the manufacturer's protocol. Samples treated the same way, but with no reverse transcriptase added, were used as a control to detect any contaminating DNA. Expression of genes encoding the ribosomal protein S7e (housekeeping) and the antimicrobial peptides antimicrobial peptides (AMPs) cecropin and gloverin were evaluated. Primers employed for S7e and cecropin have been previously described (76). The primer sequences for gloverin (forward, 5′-ACTCACATGCCAGTTGACTT-3′; reverse, 5′-TGGGATACACTTATCG-CTTC-3′) were designed using transcriptomic data from G. mellonella recently published by Vogel et al. (71). Quantitative analyses of gene expression were done using the comparative threshold cycle (ΔΔC_T) method as previously described (40). To investigate the sensitivity of the tested bacterial strains to AMPs, the MIC of cecropin B (Sigma) was evaluated as previously described (70). Experiments were conducted in triplicate, and the lowest concentration of cecropin B that reduced the bacterial growth by ≥95% was interpreted as the MIC (70).

RESULTS

EPEC, but not an attenuated derivative, kills G. mellonella larvae.

To examine the feasibility of using G. mellonella to study EPEC virulence, we examined the ability of EPEC type strain E2348/69 to kill larvae (Fig. 1A). Various concentrations of E2348/69 were inoculated into the hemocoel of last-instar G. mellonella larvae, and survival was scored over time. At low inoculums (<10³ CFU/5 μl), E2348/69 did not kill G. mellonella during a 72-h infection period (Fig. 1A). Higher concentrations of E2348/69 (≥10³ CFU) led to a significant decrease in survival following 72 h of incubation, with 10⁵ CFU causing death after 48 h (Fig. 1A). We determined the LD₅₀s after 48 h and found that 2.57 ×10³ cells (P < 0.06) of E2348/69 were necessary to kill 50% of the infected larvae. These data indicate that EPEC is capable of killing G. mellonella.

Fig 1.

Effect of different inocula of wild-type EPEC strain E2348/69 (A), JPN15 (E2348/69 lacking the EAF plasmid) (B), ALN88 (E2348/69 ΔcpxR) (C), ALN 195 (E2348/69 ΔcpxA24) (D), E2348/68 ΔescN (E), and E2348/69 ΔbfpA-L (F) strains on G. mellonella larva survival. Ten larvae were injected with bacteria (from 0 to 10⁵ cells/5 μl) and scored for survival after the indicated periods of time. Results are reported as the means from at least three independent analyses. After 72 h, a significant difference (P < 0.05) was observed (compared to E2348/69) for JPN15, ALN195, and E2348/68 ΔescN strains. No significant difference was observed for ALN88 and E2348/69 ΔbfpA-L strains.

As a further control to demonstrate that the G. mellonella larvae were being killed due to infection with EPEC, we attempted to rescue the larvae by treatment with antibiotics to which E2348/69 is susceptible. Twenty larvae were infected with 10⁴ CFU of E2348/69. Ten of these larvae were injected with 100 μg kanamycin after 4 h via the hindmost proleg, and survival was monitored for 72 h. Larvae treated with the antibiotic showed a significant increase in survival at 48 h relative to those that were untreated (Fig. 2). Following 72 h of incubation, only 7% (±7%) of untreated larvae were alive compared to 90% (±8%) after kanamycin treatment. This experiment confirms that larval death was indeed due to bacterial infection.

Fig 2.

Survival of G. mellonella infected with 10⁴ cells of wild-type E2348/69 for up to 72 h following treatment with 100 μg kanamycin after 4 h of infection. Results represent the means from three independent experiments (P < 0.001).

To begin to understand how EPEC may be causing larval death, we performed infections with GFP-labeled E2348/69 (see Materials and Methods) and then examined hemolymph and dissected larvae using fluorescence microscopy. At 18 h postinfection, we did not detect intracellular or extracellular bacteria within the hemolymph. Dissecting larvae with advanced signs of melanization revealed many melanotic nodules (Fig. 3B). Fluorescence microscopy of the melanotic material revealed a dense load of GFP-labeled E2348/69 bacteria (Fig. 4). These observations suggest that EPEC does not remain within the hemolymph but rather is rapidly encased in melanotic nodules by the G. mellonella larvae. Activation of the humoral arm of the G. mellonella innate immune response by induction of the prophenoloxidase enzyme that leads to melanization is a common feature of many microbial infections of this organism (26, 35).

EPEC growth and virulence factors are required to kill G. mellonella.

To investigate the mechanism of killing employed by EPEC, we infected G. mellonella larvae with serial dilutions of heat-inactivated E2348/69 (0 to 10⁵ cells). We found that heat-killed E2348/69 is not capable of killing the larvae (data not shown). Further, a nonpathogenic E. coli K-12 strain was also unable to kill G. mellonella (data not shown). These data indicate that bacterial growth within the larvae is necessary for killing, and that the pathogenic E2348/69 strain possesses some key attribute(s) not present in the laboratory K-12 strain that are necessary for virulence.

To determine what this attribute is, we tested the ability of JPN15, an attenuated derivative of E2348/69 (32), to kill G. mellonella larvae. JPN15 lacks the EAF plasmid, which encodes the BFP as well as the regulators PerA, PerB, and PerC. Since PerC is involved in activating the expression of the T3S genes on the LEE, JPN15 is not only devoid of BFP but is also defective for T3S (7). Killing assays showed that even at a high inoculum (10⁵ CFU), JPN15 was unable to kill all of the infected larvae (Fig. 1B). Further, the LD₅₀ of JPN15 was 1.2 × 10⁸ (P < 0.06), approximately 5 logs higher than that of wild-type strain E2348/69 (compare Fig. 1A to B). To investigate whether the BFP or the T3S system is responsible for this phenotype, we performed sensitivity assays with E2348/69 ΔescN and E2348/69 ΔbfpA-L mutants. The ΔbfpA-L strain is completely BFP deficient and was able to kill the larvae in a manner almost identical to that of E2348/69 (Fig. 1F). The LD₅₀ of the E2348/69 ΔbfpA-L strain was 4.9 × 10³ and is not statistically different from that for E2348/69. The EscN protein acts as an ATPase for the T3SS, and it has been shown previously that the deletion of this protein leads to a complete loss of secretion competence (1). The ΔescN mutation had a significant impact on the ability of EPEC to kill the larvae (Fig. 1E), with an LD₅₀ of 1.82 × 10⁵ (P < 0.02), approximately 2 logs higher than that of E2348/69. Thus, the ability of EPEC to kill G. mellonella larvae is correlated with the presence of the T3SS but not the adhesive BFP.

EPEC virulence is affected by mutations that alter the Cpx envelope stress response.

Inactivation of the Cpx response in EPEC leads to diminished expression of BFP, while constitutive Cpx induction leads to the removal of all major virulence determinants from the envelope, including the BFP and T3SS (42, 72). Accordingly, we predicted that mutations affecting the Cpx response should diminish virulence. To test this prediction, we measured the ability of a cpxR mutant lacking the Cpx response (ALN88) as well as one carrying a cpxA24 allele which constitutively activates the Cpx response (ALN195) to kill G. mellonella larvae. The cpxR mutant showed a slight attenuation in virulence (LD₅₀ = 4.17 × 10⁴; P < 0.04) (Fig. 1C). In contrast, the cpxA24 mutation rendered the bacteria almost completely avirulent (Fig. 1D). Even when high cells numbers (10⁵ cells) of ALN195 were used to infect, only a few larvae died, and the LD₅₀ (2.5 × 10¹⁰; P < 0.05) was 7-fold higher than that of E2348/69 (Fig. 1D).

Constitutive activation of the Cpx response causes a severe in vivo growth phenotype.

To begin to examine the nature of the virulence defects of the attenuated JPN15 strain and the Cpx mutants, we measured bacterial growth within the larvae for 48 h after infection (Fig. 5). We examined bacterial growth in extracted hemolymph together with whole, macerated live, and macerated dead larvae at various time points after infection (Fig. 5). Growth of the wild-type strain E2348/69 initially increased in the hemolymph, followed by a decrease over the course of infection. Conversely, bacterial numbers in whole, live larvae steadily increased throughout infection (Fig. 5A). Dead larvae contained the highest numbers of recovered bacteria (data not shown). The attenuated strain JPN15 and the cpxR mutant exhibited growth phenotypes similar to that of E2348/69, which suggests that the virulence phenotypes of these strains are not due to altered growth in the host (compare Fig. 5A to B and C). In sharp contrast, the cpxA24 mutant failed to grow at all after the initial inoculation and could not be detected in hemolymph or live whole larvae (Fig. 5D). This finding suggests that the virulence phenotype of this mutant is not due simply to the documented lack of virulence factor production but also to an inability to survive and proliferate within the G. mellonella host.

Fig 5.

Survival and relative bacterial load of G. mellonella infected with different strains of EPEC. Larvae were infected with approximately 10³ CFU of wild-type E2348/69 (A), JPN15 (E2348/69 lacking the EAF plasmid) (B), ALN88 (E2348/69 ΔcpxR) (C), or ALN 195 (E2348/69 ΔcpxA24) (D) and incubated at 30°C for up to 48 h. Bars show survival (in percent); □, hemolymph; ○, whole larvae. After 48 h, a significant difference was observed for every mutant tested compared to the wild type (P < 0.05) in bacteria recovered from the hemolymph and whole larvae.

Wild-type, attenuated, and Cpx mutant EPEC strains induce antimicrobial peptide production in the host.

We wondered if any of the virulence or growth defects we observed were correlated with alterations in, and/or the sensitivity of the bacterial strains to, the host immune response. Bacterial infection is known to lead to increased production of AMPs in insect hosts (28). Thus, we measured transcript levels of the genes encoding the AMPs cecropin and gloverin at different time points following infection with a known quantity of bacteria (Fig. 6A and B). Both antimicrobial peptides are known to be active against Gram-negative bacteria and have been shown to be a prominent part of the G. mellonella immune response when challenged with lipopolysaccharide (LPS) (71). Although differences in the magnitude of the response varied between bacterial strains, we found that infection with E2348/69, JPN15, the cpxR mutant, and the cpxA24 mutants all lead to large increases in the expression of both cecropin and gloverin shortly after infection that persisted to various levels over the course of the experiment (Fig. 6A and B). These data suggest that while there are differences between strains in terms of the host response, all of the bacteria used in this study are recognized by the G. mellonella host and an immune response is mounted. We also investigated whether the virulence and/or growth phenotypes we observed were due to differences in sensitivity to the host immune response. To investigate this, we determined the MIC of cecropin required to kill E2348/69, JPN15, the cpxR mutant, and cpxA24 strains. While E2348/69 and JPN15 had a MIC of 45 μg/ml (±2.5 μg/ml), the cpxR mutant had a MIC of 35 μg/ml (±2.5 μg/ml). The cpxA24 mutant was approximately three times more sensitive to cecropin, with a MIC of 15 μg/ml (±2.5 μg/ml).

Fig 6.

Quantitative real-time PCR analysis of cecropin (A) and gloverin (B) transcripts from G. mellonella at 4, 18, 24, and 48 h after infection with 10³ CFU of the indicated bacterial strains. Values were normalized to that of the housekeeping gene S7e. The y axis shows the relative quantities of the genes encoding cecropin (A) or gloverin (B). The values represent the means from 3 experiments, and the standard deviations are shown. The 0-h time point refers to larvae infected and incubated for 10 min.

DISCUSSION

G. mellonella larvae can be used as an alternative model to assess EPEC virulence defects.

G. mellonella is being used as a simple animal model to study the virulence of a growing list of bacterial pathogens, including P. aeruginosa (31), Campylobacter jejuni (67), Yersinia pseudotuberculosis (10), A. baumannii (55), Burkholderia cepacia (66), and Listeria monocytogenes (49). We show here that EPEC type strain E2348/69 is also able to kill G. mellonella larvae (Fig. 1, 2, and 5). Bacterial numbers capable of killing significant levels of larvae range from 1 (66) to approximately 10⁶ (10). We found the LD₅₀ of E2348/69 to be about 2.5 × 10³, which appears to be somewhat lower than that of Y. pseudotuberculosis, the only other enteric pathogen that has been studied in G. mellonella (10). It is widely accepted that for models such as G. mellonella to be useful measures of virulence, killing should be correlated with the presence of established virulence factors. This has been convincingly demonstrated for a number of pathogens (10, 31, 49, 55), and we show here that an attenuated derivative of E2348/69, JPN15, is avirulent (LD₅₀ = 1.7 × 10⁸) (Fig. 1 and 5) (39, 54). The ability of E2348/69 to kill G. mellonella is also correlated with bacterial growth (Fig. 5 and data not shown), and killing can be prevented by administration of antibiotics to which EPEC is sensitive (Fig. 2). These results suggest the exciting possibility that the G. mellonella model can be used as a model organism to study potential therapeutic agents. Indeed, a number of studies indicate that this is the case for other bacterial pathogens (2, 16, 27, 55).

At this time, we cannot say exactly what the molecular mechanism of pathogenesis is that leads to death of the G. mellonella larvae after infection with EPEC. Our preliminary investigations indicate that EPEC induces an innate immune response (Fig. 3B and 6A and B) that leads to the encapsulation of the microbe within melanotic nodules (Fig. 3B). The formation of such melanotic nodules is one of the first lines of defense for G. mellonella and has been observed previously for infections with live Bacillus cereus and E. coli K-12 strains as well as with heat-killed organisms (23, 59). These nodules are formed in response to nonlethal infections as well (23, 59), where the bacteria are successfully killed and cleared from the organism. Thus, EPEC presumably has some attribute that allows it to proliferate and survive despite being walled off by the host immune response. We did not observe bacteria intracellularly or attached to any larval tissue, and this correlates well with our finding that the BFP adhesins, which have been shown to adhere to N-acetyllactosamine receptors on enterocytes (30), are not required for virulence in the G. mellonella model (Fig. 1F). Larval death was partly dependent, however, on the T3SS. Whether the EPEC T3SS functions in G. mellonella to generate intimate attachment between the microbe and the host cell, leading to the formation of the classic pedestals that are typical of human disease, remains to be definitively determined. It should be noted, however, that secreted bacterial proteins have been implicated in the induction of the prophenoloxidase cascade, leading to melanization in other bacterial infections of G. mellonella (24), thus one possibility is that the T3S effectors of EPEC function in a similar fashion. Altogether, our data support the use of this simple model as a preliminary screening tool that can be used to study the effects of large numbers of mutations on virulence in an affordable manner. Further use of this model should expedite the discovery of new genes that may impact virulence in a variety of pathogens and could eventually serve as new therapeutic targets.

Mutations affecting the Cpx envelope stress response impact virulence.

We have shown that either elimination or activation of the Cpx response leads to reduced virulence factor production, although the effect of inducing the Cpx response is of much larger magnitude and affects more virulence determinants (42, 52, 72). Accordingly, we predicted that mutations affecting the Cpx response should lead to an attenuation of virulence. In fact, the cpxR mutant is slightly attenuated (LD₅₀ = 4.17 × 10⁴) (Fig. 1 and 5), while the cpxA24 mutant exhibits a severe decrease in virulence (LD₅₀ = 2.5 × 10¹⁰) (Fig. 1 and 5). The mild effect of the cpxR mutation on virulence is in line with our observation that its mutation does not greatly impact T3S in vitro (44). Thus, the virulence defect of the cpxR mutant may be due to a lack of expression of other Cpx-regulated genes, which are heavily enriched for encoding envelope proteins (73). Clearly these proteins are not needed for growth in G. mellonella, but perhaps they mediate a different interaction with the host. In this light, it is interesting that the cpxR mutant appears to persist in the hemolymph longer than the wild-type strain (Fig. 5).

In contrast to the cpxR mutant, the cpxA24 mutant exhibits a dramatic in vivo growth phenotype (Fig. 5), suggesting that the diminished virulence of this strain is primarily a result of an inability of the organism to survive within the larvae. Many genes affecting multiple cellular functions are disregulated in the cpxA24 mutant (36), so it is difficult to say at this point why this strain fails to grow in vivo. One possibility is that the cpxA24 mutant is more sensitive to cellular and/or humoral components of the G. mellonella immune response. Although we have not yet investigated this extensively, we find a 3-fold increase in sensitivity of the cpxA24 mutant to cecropin, an AMP demonstrated to be involved in the G. mellonella response to bacterial invaders (71). Interestingly, this finding is in contrast to a recent report by Weatherspoon-Griffin et al. (75), who found that activating the Cpx response via overexpression of the lipoprotein NlpE conferred resistance to some AMPs. While we do not have an explanation for this discrepancy, one possibility is that the stronger, constitutive induction of the Cpx response in the cpxA24 mutant relative to strains overexpressing NlpE has different phenotypic consequences. Regardless of the reason for the growth phenotype, these data suggest that induction of the Cpx response leads not only to a dramatic decrease in virulence factor production but also to severe virulence and in vivo growth phenotypes. These results are consistent with studies of other enteropathogens, including Salmonella enterica serovar Typhimurium and Y. pseudotuberculosis, where it has been demonstrated that Cpx pathway induction leads to virulence defects in tissue culture and mouse models of disease (8, 29, 58, 62). Together, these observations make the Cpx response an attractive target for the development of new therapeutics with the potential to impact a range of enteric pathogens.

Induction of innate immunity.

We measured the innate immune response of G. mellonella after infection by analyzing transcript levels of two AMPs, cecropin and gloverin (Fig. 6A and B). Both AMPs are produced by G. mellonella upon LPS challenge (71). Although all of the strains used in this study resulted in the induction of cecropin and gloverin production, there were differences in the magnitude of the response (Fig. 6A and B). The cpxR mutant appeared to result in the induction of diminished levels of gloverin (Fig. 6B), while infection with the attenuated JPN15 strain and the cpxA24 mutant led to lower levels of cecropin production (Fig. 6A). It is probable that the diminished induction of cecropin by the cpxA24 mutant reflects the inability of this strain to grow in G. mellonella (Fig. 5D). In contrast, the cpxR mutant and JPN15 strains exhibited in vivo growth comparable to that of the wild-type parent strain (Fig. 5). Both of these strains, however, display a different envelope to the host. The Cpx response predominantly impacts the expression of envelope proteins that affect multiple aspects of the envelope, including the cell wall (73), while the JPN15 strain lacks the BFP and makes smaller amounts of the surface-exposed T3SS (6, 7, 33). It has recently been shown that G. mellonella employs, as part of its innate immune arsenal, a number of conserved pattern recognition receptors (PRRs), including at least 3 Toll-like receptors (TLRs) (71). Since many PRRs recognize components of the bacterial envelope, including the cell wall, LPS, and lipoproteins, one possibility is that the JPN15 and cpxR mutant strains are perceived differently by the PRRs of G. mellonella, ultimately affecting the magnitude of the immune response. Interestingly, there appears to be a second peak in production of both cecropin and gloverin at 48 h after infection with the cpxA24 mutant (Fig. 6), despite the fact that we could not detect any bacteria at this or later time points. Although we have no explanation for this observation at this time, we speculate that the initial wave of AMP induction leads to the release of bacterial products that then trigger a second response. Cumulatively, our results demonstrate the feasibility of utilizing G. mellonella as a model to study EPEC virulence and host interaction, and hopefully this report will pave the way for its use by other researchers in the field.