Transcriptional Analysis of a Photorhabdus sp. Variant Reveals Transcriptional Control of Phenotypic Variation and Multifactorial Pathogenicity in Insects^▿

Phenotypic analysis of Photorhabdus variants.

Wild-type and variant traits were assessed as previously described (5) except for the protease assay, which was monitored by a colorimetric method (14). Agglutination assays were assessed as previously described (50) with Photorhabdus bacterial suspensions (10⁸ cells/ml) prepared from exponential-growth or stationary-phase cultures in LB broth.

RNA preparation.

TT01α and VAR* were grown in LB broth for 6 h (5 × 10⁷ CFU/ml; optical density at 540 nm [OD₅₄₀] of 0.5; exponential-growth phase) or 45 h (5 × 10⁸ CFU/ml; OD₅₄₀ of 13 for TT01α and OD₅₄₀ of 8 for VAR*; stationary-growth phase). Total RNA was extracted with TRIzol reagent, according to the manufacturer's instructions (Invitrogen, Carlsbad, CA) and then purified with a High Pure RNA Isolation kit (Roche, Meylan, France), including a DNase I treatment step. The concentration was determined by spectrophotometric measurement of the OD₂₆₀. For each RNA preparation, DNA contamination was assessed by carrying out PCR.

Microarray experiments and analysis.

We synthesized cDNA targets from 10 μg of TT01α and VAR* RNA and labeled them as follows. Each RNA preparation was heated at 70°C for 10 min with random hexamer (0.1 μg/μl; Roche) and cooled at room temperature for 10 min. The RNA was then reverse transcribed by incubation at 42°C for 2 h with SuperScript II (10 U/μl; Invitrogen), dithiothreitol (5 mM; Invitrogen), a mixture of deoxynucleoside triphosphates (1 μl; RPK0147; Amersham/GE Healthcare Europe GmbH, Munich, Germany), and dCTP-Cy3 or dCTP-Cy5 (50 μM; Amersham). The remaining RNA was subjected to alkaline lysis by incubation with 0.2 M NaOH for 10 min at 37°C. The labeled cDNAs were neutralized by adding 625 mM HEPES (pH 8) and immediately purified with a QiaQuick PCR purification kit (Qiagen, Courtaboeuf, France), used according to the manufacturer's recommendations, but with two additional washes in PE buffer (provided in the Qiagen kit). Cy5- and Cy3-labeled cDNAs were mixed, dried under a vacuum, and resuspended in hybridization buffer, as previously described (30).

The Photorhabdus DNA microarray (GEO accession number {"type":"entrez-geo","attrs":{"text":"GPL11042","term_id":"11042"}}GPL11042) and the prehybridization and hybridization procedures have been described elsewhere (30). Each microarray comparison (TT01α versus VAR* growth in the exponential-growth or stationary phase) included six slides, with three dye-swapping replicates. The intensity of the signal for each microarray spot was determined with ArrayVision software (Amersham).

Microarray data were normalized and statistical analyses were carried out with the Spotfire DecisionSite functional genomic tool (http://spotfire.tibco.com/products/decisionsite-functional-genomics.aspx). The raw intensities were normalized to ensure that the experiments were comparable and to remove sources of systematic variation. For a given hybridization, the median intensity of a given variable is adjusted such that it is equal to the median intensity of the control variable (logR − logG = 0, where R and G are the summed intensities for each variable) without the designation of a baseline. As such, the intensity levels of the two channels are mutually adjusted (dye normalization). The trimmed median method was used to calculate a median value for which the effect of outliers was minimized. The trim cutoff was set at 10%, resulting in the exclusion of the top 5% and the bottom 5% of the values from the calculation. The dye-normalized values were used for a global normalization process in which intensities were rescaled with respect to a baseline intensity control.

After normalization, the fold change (FC) was calculated as follows: FC = R if R > 1 and −1/R if R < 1, where R = (median of normalized intensity for VAR*)/(median of normalized intensity for TT01α). We used a Student t test to identify statistically significant differences between groups. The original hypothesis (i.e., null hypothesis) was that the median expression levels of a gene did not differ between the two groups. The null hypothesis was then either rejected or accepted for each gene considered. The results are expressed in terms of P values, indicating the significance of the differences observed. These values correspond to the probability of a type I error leading to the conclusion that two median expression values for a given gene are different when there is, in fact, no difference. A significant difference is considered to exist if the P value obtained is below 0.01. The microarray data are accessible under the GEO accession number {"type":"entrez-geo","attrs":{"text":"GSE24627","term_id":"24627"}}GSE24627.

qRT-PCR.

For the validation of whole-transcriptome data, quantitative reverse transcription-PCR (qRT-PCR) was carried out on the RNA preparations used for Photorhabdus microarray hybridization. cDNAs were synthesized with 1 μg of total RNA using the SuperScript II reverse transcriptase (Invitrogen) and random hexamers (100 ng/μl; Roche Diagnostics). qPCR analyses were performed using a LightCycler FastStart DNA Master^PLUS SYBR green I kit (Roche Diagnostics) with 1 μl of cDNA synthesis mixture and 1 μM concentrations of specific gene primers (Table ​(Table2).2). The enzyme was activated for 10 min at 95°C. The reactions were performed in triplicate at 95°C for 5 s, 61°C for 5 s, and 72°C for 10 s (45 cycles) and monitored in the LightCycler (Roche). Melting curves were analyzed for each reaction, and each curve contained a single peak. The amount of PCR product was calculated from the standard curves obtained from PCR with serially diluted genomic DNA from P. luminescens TT01. All data are presented as a ratio with respect to gyrB, which was used as a control gene. Values are means ± the standard deviation and were compared by using a Student t test.

Sequence analysis.

The sequence annotations of the TT01 genome were from or updated with PhotoScope, the Photorhabdus database of the annotation and comparative analysis platform, MaGe (https://www.genoscope.cns.fr/agc/mage/wwwpkgdb/Login/log.php?pid=13).

Assays of sensitivity to antimicrobial peptides, antibiotics, and toxic compounds.

We assessed the susceptibility of the Photorhabdus variants to antimicrobial peptides (AMPs) or antibiotics, by determining the MICs of colistin, polymyxin B, cecropin A, and chloramphenicol. Stock solutions of colistin methanesulfonate (Sigma) and polymyxin B (Sigma) were prepared in sterile water, to obtain concentrations of 20 and 0.5 mg/ml, respectively. Stock solutions of cecropin A and B (Sigma synthetic molecules C6830 and C1796, respectively) were prepared in 1% acetic acid to obtain a concentration of 0.25 mg/ml. A stock solution of chloramphenicol (Sigma) was prepared in absolute ethanol to obtain a concentration of 20 mg/ml. The antimicrobial molecules were then added directly into the 96-well microtiter plates in 2-fold serial dilutions. Bacterial cells were grown at 28°C overnight in LB broth and were then washed and resuspended in Mueller-Hinton (MH) broth (Biokar, Marnes-la-Coquette, France). A final inoculum equivalent to 10³ CFU was dispensed into the wells of 96-well microtiter plates. The MICs were determined after incubation at 28°C for 48 h. The microtiter plates were read by eye. β-Lactam susceptibility assays were carried out by the agar diffusion method in accordance with the instructions of the Antimicrobial Committee of the French Society for Microbiology. The susceptibilities of these cultures to other toxic compounds were assessed on solid media as follows: an inoculum of Photorhabdus variants equivalent to 10³ CFU was evenly spread onto solidified MH medium containing 1.5% agar (BD Difco, Franklin Lakes, NJ) and allowed to dry at room temperature for 20 min. The various toxic compounds (1.5 mM ZnSO₄, 30% H₂O₂, crystal violet [4 mg/ml], 20% sodium dodecyl sulfate, 25 mM deoxycholate) were diluted in sterile water and sterilized by passage through a filter with 0.2-μm-pore-size pores. Filter paper discs (bioMérieux S.A, Marcy l'Etoile, France) were impregnated with 5 μl of a solution of the toxic compound to be tested and placed in the center of a MH agar plate. Bacterial growth inhibition was evaluated by measuring the diameter of the zone in which no bacterial growth was observed.

DNA manipulation.

Standard DNA manipulations in Escherichia coli were carried out as described by Ausubel et al. (2). We evaluated the functional complementation of VAR* with genes encoding insecticidal toxins (plu1537, pit; plu4092, pirB; and plu2442, xnp2) by constructing low-copy-number plasmids as follows. pit, pirA-pirB, and the operon containing xnp2 (plu2444-plu2443-plu2442-plu2441) were amplified by PCR using the primer pairs XhoI-L-plu1537/PstI-R-plu1537, XhoI-L-plu4093/PstI-R-plu4092, and XhoI-L-plu2444/PstI-R-plu2441, respectively (see Table ​Table2).2). PCR fragments were ligated into pBBR1-MCS hydrolyzed with XhoI and PstI (Roche) to ensure that these fragments were inserted downstream from the P_lac promoter. The resulting plasmids were named pBB-pit, pBB-pirAB, and pBB-xnp2, respectively. All constructs were checked by sequencing (Millegen, Labège, France). Plasmids were introduced into VAR* by mating experiments, as previously described (11). We checked that the P_lac promoter was functional in the VAR* variant by constructing the pBB-P_lac uidA plasmid as follows. The P_lac fragment from pUC19 was inserted upstream from the promoter-less uidA gene in pBB-UIDK (46). The uidA gene encodes the β-glucuronidase enzyme, the specific activity of which was measured as previously described (46).

Pathogenicity assays.

Pathogenicity experiments were performed as previously described (12). Statistical analysis (Wilcoxon test) was performed as previously described (11, 33) using SPSS version 14.0 (SPSS, Inc., Chicago, IL) to compare the mortality patterns.

A modified regulator network in VAR*.

The transcription of 15 regulators or potential regulators (see regulators in Table ​Table3,3, column 1) was found to be modified. For 10 of them, the transcription levels were higher in VAR* than in TT01α: the MrfJ repressor, a putative transcriptional regulator from the LuxR family, three of the 13 Ner-like regulatory proteins encoded in the P. luminescens TT01 genome, the CspE cold-shock protein, the FliZ regulator, the FimB recombinase, and the PqrA and RcsB regulators. Higher levels of FimB and MrfJ proteins may have affects on fimbrial production. This is validated from a genetic perspective, since four genes encoding fimbrial components—a mrfA-like gene (plu0418), mrfA (plu0769), mrfH (plu0777), and a putative adhesin-encoding gene (plu2156)—are more transcribed in VAR* than in TT01α. Moreover, sheep erythrocytes agglutination was only detected with VAR* bacterial cells (Table ​(Table44).

For five molecules, lower levels of transcription were observed in VAR* than in TT01α: the HslJ heat shock protein, three putative transcriptional regulators from the LuxR family, and a putative CI repressor. This large number of regulators with modified transcription patterns suggests that the generation of phenotypic variants in P. luminescens TT01 depends on transcription switching mediated by a network of regulators rather than a single regulator.

Previous regulation analyses have deciphered the role of particular regulatory genes in the generation of Photorhabdus phenotypic variants (22, 40, 53). The AstR-AstS two-component system controls the timing of the phenotypic switch between the wild-type form and phenotypic variants in P. luminescens TT01 (22). Activation of the two-component systems is not dependent on the level of transcription of the genes encoding the proteins involved. Instead, it depends on the phosphorylation status of the cytoplasmic receptor (3). Our approach therefore did not allow the detection of RNA variation for the AstR-AstS or any other two-component system. The gene encoding the HexA/LrhA repressor, a LysR-type transcriptional regulator, is not transcribed in wild-type forms of P. temperata K122 (40). However, our data show no difference in the transcription patterns of the genes encoding HexA/LrhA (Plu3090) or any other LysR-type transcriptional regulator between TT01α and VAR*. Finally, the Ner-like repressor has a more ambiguous role (53). Similar RNA levels for this repressor are found in the wild-type form and phenotypic variants of P. temperata K122. Repression of the wild-type form is observed when the ner gene is harbored on a plasmid and overexpressed in a phenotypic variant. Our data show that three of the 13 ner-like genes present in the P. luminescens TT01 genome are more strongly transcribed in VAR* than in TT01α. Therefore, as in P. temperata K122, ner overexpression is observed in a phenotypic variant of P. luminescens TT01. This overexpression may be mediated by both the redundancy of ner-like genes in the TT01 genome and the higher level of transcription observed in the VAR* variant. The reason for this higher level of transcription is unknown, but it seems likely to be controlled by an upstream global regulator.

Transcriptomic variations of genes supporting phenotypic variations.

The wild-type strain and phenotypic variants of P. luminescens can be distinguished by a set of simple bacteriological tests, such as pigmentation on agar medium, trehalose fermentation, luminescence, the production of intracellular protein inclusions, and antimicrobial and extracellular activities (7, 31). TT01α and VAR* harbor the typical wild-type and phenotypic variant traits, respectively (Table ​(Table4).4). We explored the possibility that some phenotypic variant trait-related genes are regulated at the transcriptional level.

(i) Pigmentation.

On agar media, wild-type forms generate convex, mucoid, orange colonies, whereas phenotypic variants generate flat, nonmucoid colonies without pigmentation (7, 31). The antH gene (plu4187) and plu0947, which are involved in the anthraquinone biosynthesis pathway (10), are less strongly transcribed in VAR* than in TT01α, probably accounting for the loss of pigmentation of VAR*.

(ii) Crystalline inclusion.

One of the characteristic phenotypes of the wild-type bacterium is the presence of intracellular protein inclusions, known as crystals, which are absent from phenotypic variants (4). Crystalline inclusion proteins, CipA and CipB, are regulated transcriptionally in VAR*, since the cipA gene (plu1576), the cipB gene (plu0157), and a cipA-like gene (plu1575) displaying some similarity to cipA (identity, 36%; E-value, 8.4e−06) are less strongly transcribed in VAR* than in TT01α. The role of the crystalline inclusions remains unclear. They may be involved in nutrient storage, including sulfur storage in particular, due to the high methionine content of CipA (4). Moreover, the inactivation of either cipA or cipB through insertion results in mutants displaying variation in phenotypic traits and alteration in interactions with nematodes (4). For these reasons, Bintrim and Ensign speculated that Cip proteins might play a role in the global regulation of phenotypic variation. Since VAR* displayed changes to phenotypic traits but not to interactions with nematodes (D. Clarke, unpublished data), these two sets of phenotypes are probably regulated differently. For example, a small number of cip transcripts may be sufficient for correct interaction with nematodes but not for intracellular crystal formation and the generation of wild-type form traits.

(iii) Antibiotic synthesis.

The phenotypic variant form had lower levels of antimicrobial activity (5). Because classical antibiosis tests are carried out in a Gram-positive bacterium, Micrococcus luteus (Table ​(Table4),4), we extended the antibacterial activity test to other genera of bacteria. The antibacterial activity against the four Gram-positive bacteria observed with TT01α was not observed with VAR* (Table ​(Table5).5). Stilben is the only antimicrobial molecule produced by P. luminescens that has been shown to be active against Gram-positive bacteria (68). None of the genes involved in stilben biosynthesis was transcribed to different levels in VAR* and TT01α. The defect in antibiotic production by VAR* is therefore probably due to posttranscriptional modulation of the stilben biosynthesis pathway or antimicrobial molecules that have yet to be identified. Three kinds of antibiosis patterns were observed against Gram-negative bacteria: (i) no antibacterial activity for either form, (ii) antibacterial activity with TT01α but not with VAR* on both closed and more distantly related bacterial species, and (iii) antibacterial activity with VAR* but not with TT01α on Proteus vulgaris (Table ​(Table5).5). The pyocin-like bacteriocins of P. luminescens can kill related Gram-negative bacteria such as other Photorhabdus strains and E. coli (58). Two integral and two truncated genes encoding pyocin-like bacteriocins (bacteriocins Plu0884 and Plu1894 and bacteriocins Plu0887 and Plu1892, respectively) are less strongly transcribed in VAR* than in TT01α in the exponential growth or in the stationary phase. P. luminescens produces carbapenems of TT01, a class of β-lactam antibiotics with broad-spectrum activity mostly targeting Gram-negative bacteria (21). The cpmB and cpmC genes (plu0183 and plu0184), involved in the biosynthesis of carbapenems are less strongly transcribed in VAR* than in TT01α. The downregulation of both pyocin-like bacteriocins and carbapenems may therefore account for the lower levels of antimicrobial activity against the related E. coli, S. enterica, and Y. enterocolitica strains observed for VAR*, whereas the downregulation of carbapenems may only be sufficient to account for the lower levels of activity against the more distantly related O. anthropi and Pseudomonas sp. strains. For P. vulgaris, against which VAR* displayed a higher level of antibiosis than TT01α, other antimicrobial molecules are probably involved.

(iv) Extracellular protein activities.

Finally, extracellular activities, such as hemolysis and lipase and protease activities, were found to be weaker in the phenotypic variants (5). The activities of two metalloproteases, PrtA and PrtS, were attenuated in the supernatant of the P. temperata K122 phenotypic variant (9, 49). Our data suggest that the transcription of prtA (plu0655) is not modified but that the prtS gene (plu1382) and its downstream gene (plu1381), which probably belongs to the same transcription unit, are less strongly transcribed in VAR* than in TT01α. The regulation of lip-1 (plu3510), encoding a lipase responsible for Tween 80 degradation, is posttranslational in the K122 phenotypic variant of P. temperata (63). Our data confirm this result in P. luminescens TT01α, since no change in transcription was detected for the lip-1 gene. Nevertheless, the uncharacterized lipase genes, plu1517 and plu3159, were less strongly transcribed in VAR*. The low rates of transcription of these two genes may account for poor Tween 20-60 degradation by VAR* (Table ​(Table44).

It is widely accepted that posttranscriptional and posttranslational regulation processes are responsible for the differential phenotypes expressed by the two forms of Photorhabdus (22, 29, 54, 61, 63, 64). In the present study, we show that genes involved in pigmentation, crystalline inclusion, antibiosis, and protease and lipase activities are controlled at the transcriptional level. Interestingly, for the lipase and protease activities, downregulation by both transcriptional and posttranscriptional mechanisms is probably involved in phenotypic variation.

Toxin overproduction in the VAR* variant does not restore virulence.

VAR* is less virulent than ΤΤ01α when injected into Spodoptera littoralis larvae: 50% mortality (LT₅₀) was reached 25 h after injection with ΤΤ01α, but not until 37 h after injection for VAR*, although a larva mortality rate of 100% was nonetheless reached 3 days after infection (Fig. ​(Fig.22).

Open in a separate window

FIG. 2.

Mortality in S. littoralis. Shown is the mortality in S. littoralis infected with the TT01α P. luminescens wild-type form (○), the VAR* phenotypic variant (▪), and VAR* overexpressing toxins or potential toxin genes: the pirAB locus (✳), the pit gene (⧫), and the xnp2 gene (▴). Bacteria obtained at the end of the exponential growth phase were injected into fourth-instar larvae. Mortality values are based on data obtained after the injection of 20 larvae. Note that the virulence of the VAR* variant harboring the pBB-xnp2 plasmid is more attenuated than the VAR* virulence; this is probably due to the energy cost necessary to replicate a plasmid with a large insert (7.3 kb) compared to the 0.5- and 1.7-kb inserts of the pBB-pit and pBB-pirAB plasmids, respectively.

P. luminescens can kill insects by toxemia. The P. luminescens TT01 genome contains a large number of genes thought to encode insecticidal toxins—i.e., the Tc, Mcf, Pir, Txp40, and Pit toxins—and a gene predicted to encode a putative nematicidal protein, the Xnp2 toxin (8, 19, 24, 47, 66). Our transcriptional analyses (microarray hybridization and qRT-PCR) showed that pit (plu1537), pirB (plu4092), and xnp2 (plu2442) were less strongly transcribed in VAR* than in TT01α (Table ​(Table33 and Fig. ​Fig.1).1). Interestingly, plu2443, the second gene of the operon containing xnp2, is also less strongly transcribed in VAR* than in TT01α.

We investigated the possible involvement of a defect in toxin gene transcription in the attenuation of virulence in VAR* by constructing strains overproducing the toxins. This was achieved by placing the pit gene, the pirAB genes and the four genes of the xnp2-containing operon under the control of the P_lac promoter of the pBBR1-MCS1 plasmid and transferring these constructs into VAR*. We first checked that the P_lac promoter was functional in VAR*, by measuring the β-glucuronidase activity of the pBB-P_lacuidA plasmid in VAR* (data not shown). We then carried out pathological assays, by injecting VAR*/pBB-pit, VAR*/pBB-pirAB, and VAR*/pBB-xnp2 into S. littoralis larvae. The overproduction of these three proteins did not restore virulence in VAR* (Fig. ​(Fig.22).

Delayed growth of VAR* in S. littoralis hemolymph.

P. luminescens can also kill insects by bacteremia. We evaluated the role of bacteremia in insect death by assessing the growth of VAR* and TT01α bacteria in S. littoralis hemolymph by injecting each variant into the hemocoel of larvae (5 × 10⁴ CFU/larva) and counting the CFU produced. As for the closely related bacterium Xenorhabdus nematophila, but at a later time point after bacterial injection (59), a partial clearance of circulating bacteria was observed for both TT01α and VAR* (between 6 and 9 h after injection) (Fig. ​(Fig.3).3). Bacterial growth then increased for both TT01α and VAR*. Nevertheless, in contrast to the results obtained for LB broth, in which TT01α and VAR* grew similarly, TT01α grew more rapidly than VAR* in hemolymph (Fig. ​(Fig.3).3). The delayed virulence of VAR* may therefore be due to impaired VAR* bacteremia.

Open in a separate window

FIG. 3.

Bacterial growth after the injection of TT01α (○) and VAR* (▪) into S. littoralis. The graphs show the mean numbers of CFU recovered from the total hemolymph of single larvae (four larvae per time point). Larvae were each injected with 5 × 10⁴ bacteria at time zero. Errors bars indicate the standard errors of the means.

Downregulation of genes involved in drug or stress resistance.

Our transcription profiling study showed that the expression of genes associated with resistance to antimicrobial compounds or the detoxification of toxic compounds was affected in VAR* (Table ​(Table3).3). Five genes were more strongly transcribed in VAR* than in TT01α: mipA, which encodes a protein involved in peptidoglycan synthesis and controlling the growth of the stress-bearing sacculus of Escherichia coli (62); pbgP1, the first gene of an operon encoding the components of an enzymatic pathway involved in arabinose incorporation in the lipid A moiety of lipopolysaccharide and responsible for polymyxin resistance in Salmonella enterica (23, 34, 35); cat, which encodes a chloramphenicol acetyltransferase (13); ccm, which encodes an internal protein, CcmA, influencing cell shape in Proteus mirabilis (37); and uspG, which encodes a universal stress protein that may, in some cases, be linked to resistance to DNA-damaging agents and to respiratory uncouplers (45). Eight genes were found to be less strongly transcribed in VAR* than in TT01α: plu1369, which encodes a putative β-lactamase (15); ompN, encoding a protein of the porin family responsible for the permeability of Gram-negative bacteria to small, polar molecules (55); sodA and katE (plu3068), which encode the manganese-dependent superoxide dismutase SodA and the catalase KatE, respectively, both of which are involved in the detoxification of reactive oxygen species in Escherichia coli (48); and the hcaE, hcaF, hcaB, and hcaD genes, which belong to an operon encoding the components of a pathway involved in the degradation of 3-phenylpropionate and cinnamic acid and participate in oxidative stress resistance in Photorhabdus (16, 17).

We investigated the sensitivity of TT01α and VAR* to various antimicrobial molecules, such as antimicrobial peptides (colistin, cecropin A, and polymyxinB) and antibiotics (ampicillin and chloramphenicol), but both forms were found to be equally resistant to the five molecules (data not shown). We then investigated the sensitivity of the variant to toxic compounds. A significant difference in growth between the variant and the wild type was detected only with hydrogen peroxide (H₂O₂), to which VAR* was more susceptible than TT01α. The lower levels of transcription of the sodA gene, the katE gene and the hca operon are probably responsible for this sensitivity. Interestingly, the difference in inhibition diameters between VAR* and TT01α was similar to that between a P. luminescens TT01 ΔsodA mutant and the corresponding wild-type strain (data not shown). We further investigated the arsenal of genes providing resistance to oxidative stress in the P. luminescens TT01 genome. A careful search of the genome showed that P. luminescens TT01 does not have the other canonical enzymes involved in resistance to oxidative stress, i.e., those encoding SodB, SodC, and KatG (48). Under oxygen-rich conditions resulting from the Fenton reaction, iron is also a source of dangerous radicals (36), and some iron-transporters protect bacterial cells against lethal doses of hydrogen peroxide (20, 56). Interestingly, the transcription patterns of five genes involved in iron uptake were found to be modified in VAR* (Table ​(Table3).3). Dealing with oxidative stress is a relevant challenge for insect pathogens. Indeed, insect hemocytes engaged in phagocytosis generate reactive oxygen intermediates (51). For instance, during encapsulation of the eggs of the wasp parasitoid Leptopilina boulardi, H₂O₂ is one of the principal generated oxidants (52). Therefore, transcriptional alteration of genic arsenal involved in resistance to oxidative stress may explain the altered pathogenicity of VAR*.

We thank Nadège Ginibre for assistance with bacterial growth experiments in the S. littoralis hemolymph.

This study was supported by INRA (grant SPE 2004-1133-2) and by the Ministère de l'Industrie (AAV ASG no. 30; contract A01307).