An insect symbiosis is influenced by bacterium-specific polymorphisms in outer-membrane protein A

Abstract

Beneficial bacterial symbioses are ubiquitous in nature. However, the functional and molecular basis of host tolerance to resident symbiotic microbes, in contrast to resistance to closely related bacteria that are recognized as foreign, remain largely unknown. We used the tsetse fly (Glossina morsitans), which depends on symbiotic flora for fecundity and has limited exposure to foreign microbes, to investigate the tolerance phenomenon exhibited during symbiosis. We examined the potential role of bacterium-specific polymorphisms present in the major bacterial surface protein, outer-membrane protein A (OmpA), on host infection outcomes. Tsetse were successfully superinfected with their mutualistic facultative symbiont, Sodalis glossinidius, whereas infections with Escherichia coli K12 were lethal. In contrast, tsetse were resistant to an E. coli OmpA mutant strain, whereas recombinant Sodalis expressing E. coli OmpA became pathogenic. Profiling of tsetse immunity-related gene expression incriminated peptidoglycan recognition protein (pgrp)-lb as a determinant of the infection outcomes we observed. RNAi-induced knockdown of tsetse pgrp-lb significantly reduced host mortality after infection with otherwise lethal E. coli K12. Our results show that polymorphisms in the exposed loop domains of OmpA represent a microbial adaptation that mediates host tolerance of endogenous symbiotic bacteria.

Bacteria are a highly adaptive and diverse group of organisms with characteristics necessary to inhabit virtually every ecosystem, including habitation within other organisms. Although pathogenic microorganisms that have detrimental effects on animals and plants have been extensively investigated, the greater majority of relationships in nature represent beneficial and sustained “symbiotic” associations (1). For example, the human intestinal microbiome is representative of an enormous yet phylogenetically restricted population of bacteria that provides us with immunological, metabolic, and genetic constituents that we have consequently not been required to evolve (2). To date few experimental studies have been performed to determine the functional and molecular basis of host tolerance to resident microbial flora.

Insects represent a taxon with well documented symbiotic interactions that have allowed these hosts to exploit unusually restricted nutritional resources. Different insect symbioses reflect varying levels of host–microbe integration. Among these are heritable obligate mutualists. These symbionts have been associated with their insect hosts for millions of years and provide metabolic supplementation to their host's nutrient-restricted diets (3). In addition to such highly integrated associations, insects also harbor facultative mutualists. Grouped within the latter designation is Sodalis glossinidius, a tsetse endosymbiont that is closely related to other insect symbionts from Hippoboscid flies (4), grain weevils (5), and lice (6) as well as to important enteric human pathogens including Escherichia coli, Salmonella, and Yersinia. Selective elimination of Sodalis results in reduced tsetse longevity (7). Phylogenetic reconstruction analysis using ftsZ indicated that Sodalis from distant tsetse species have undergone minimal coevolution with their hosts, thus signifying the relatively recent establishment of this symbiosis (8). Sodalis can be cultivated in vitro, and a genetic transformation system has been developed to introduce and express foreign gene products both in vitro and in vivo (9). Furthermore, distant tsetse species can be reconstituted with different Sodalis strains that subsequently mimic natural infections (8).

Sodalis proliferates despite residing in tsetse's hostile gut and hemolymph environment where it can activate and/or be exposed to immune effectors. Analysis of this symbiont's genome sequence indicates that several immunogenic components of its cell membrane are altered, including a truncated lipopolysaccharide (LPS), missing O-antigen and modified outer-membrane protein A (OmpA) (10). Here, we focus on OmpA, which comprises a major component of the outer membrane in the Enterobacteriaceae. This protein functions as both an evasin and target of the mammalian immune system, a modulator of biofilm formation, and as a bacteriophage receptor (11). Additionally, functional studies with several pathogenic bacteria, including Leptospira interrogans (12), Neisseria gonorrhoeae (13), Salmonella enterica (14), and E. coli (15) have demonstrated a direct role for OmpA in virulence phenotypes upon host infection. No information exists on the function of bacterial OmpA as it relates to host-symbiont homeostasis. In the present study, we use the bacteria Sodalis and E. coli (K12) and the Dipteran insects Glossina morsitans morsitans and Drosophila melanogaster to understand the OmpA-mediated mechanisms that influence symbiotic versus pathogenic infection outcomes. We also report on tsetse immunity-related gene expression in response to infection with symbiotic versus foreign bacteria and discuss how these results may contribute to the phenomenon of host tolerance of endogenous flora.

Results and Discussion

E. coli K12 Infections Are Lethal to Tsetse, Whereas Sodalis Proliferates.

Sodalis and E. coli were genetically transformed with a construct that expresses the luciferase reporter gene [pIL; see supporting information (SI) Fig. S1 and Materials and Methods]. Relative light units (RLU) per recombinant (rec) cell in vitro were determined for both recSodalis_pIL and recE. coli_pIL at different time points throughout their log growth phases. recSodalis_pIL produced 8.0 RLU per cell, and recE. coli_pIL produced 2.2 RLU per cell (Fig. S2 A and B, respectively). We next injected 1 × 10³ colony-forming units (CFU) of recSodalis_pIL and recE. coli_pIL directly into tsetse's hemocoel and measured luciferase activity as an indicator of cell growth over time. Bacteria in both infected fly groups exhibited virtually log-phase growth. recSodalis_pIL reached an average peak density of 8 × 10⁴ cells by 14 days after infection (dpi) and had no detrimental effect on host mortality. In contrast, recE. coli_pIL grew rapidly after infection and resulted in the death of all flies by 10 dpi (Fig. 1A). To validate our experimental procedure in a comparative context, we similarly infected Drosophila with recE. coli_pIL. Our results, like those from previous studies (16), demonstrate that Drosophila can survive infection with E. coli K12 (Fig. 1B), whereas tsetse were again susceptible (Fig. 1C). Interestingly, the lethal effect of recE. coli_pIL on tsetse was restricted to systemic infections, because when the same number of live recE. coli_pIL were provided per os in the blood meal, flies were able to clear the infections (Fig. 1D). This finding confirms that tsetse, like other insects (17), exhibit different immune mechanisms in their gut and hemocoel compartments.

Fig. 1.

Host infection outcomes after septic treatment with rec bacteria. (A) Number of cells per fly over time after injecting 1 × 10³ CFU of recSodalis_pIL (black circles) and recE. coli_pIL (red squares) into tsetse's hemocoel. The experiment was repeated at two independent times (n = 25 flies per experiment) for both species of rec bacteria. Rec cell number per fly was determined by dividing RLU expression in vivo by the average number of RLU produced per cell in vitro. All recE. coli_pIL-infected flies perished by 10 dpi. (B and C) Percent survival of female Drosophila (red squares) (B), and female (red squares) and male (red circles) (C) tsetse after systemic infection with E. coli K12 grown in MM media. Systemic infections in tsetse were performed by microinjection with a glass needle, whereas Drosophila were pricked with a tungsten needle dipped a concentrated bacterial solution. Control in B and C are MM media (black squares); WT Sodalis (black circles); heat-killed E. coli K12 (black triangles). n = 50 flies per treatment. (D) Number of recSodalis_pIL and recE. coli_pIL per tsetse fly over time after per os infection with 1,000 CFU of ach bacteria. The experiment was repeated in duplicate (n = 25 flies per experiment) for both species of bacteria. For all experiments, d stands for days after infection.

E. coli's virulence in tsetse depends on bacterium-specific OmpA expression. We speculated that the different tsetse survival outcomes observed upon infection with closely related members of the family Enterobacteriaceae may result from evolutionary adaptations to surface-exposed molecules such as OmpA. However, Western blot analysis using an E. coli anti-OmpA polyclonal antibody confirmed that Sodalis and E. coli express comparable amounts of OmpA (Fig. 2A). To determine whether E. coli OmpA (OmpA_E.coli) may be responsible for this bacterium's lethal effect on tsetse, flies were infected with a transposon-generated mutant strain of E. coli K12 (MG1655; K12ΔOmpA::Tn5Kan-2) (18) that we transformed to express luciferase (recE. coli^ΔOmpA-pIL). Infection outcome with the OmpA mutant strain was different from with WT E. coli. Whereas the majority of recE. coli^ΔOmpA-infected flies survived, those treated with the WT parental strain perished by 12 dpi (Fig. 2B). We next used transcomplementation to express OmpA_E.coli and OmpA_Sodalis in the E. coli^ΔOmpA mutant background (Fig. 2A and Fig. S1 B and C). The OmpA_E.coli-complemented strain (recE. coli^ΔOmpA-pIO_E.coli) exhibited a lethal phenotype in tsetse similar to WT E. coli. In contrast, flies infected with the OmpA_Sodalis-complemented strain (recE. coli^ΔOmpA-pIO_Sodalis) survived similarly to those infected with OmpA mutants (Fig. 2C). Temporal analysis of luciferase expression throughout the infection process indicated that tsetse were able to clear infections with recE. coli^ΔOmpA-pIL and mutants transcomplemented to express OmpA_Sodalis and luciferase (Fig. 2D; using construct pIO_Sodalis-Luc, Fig. S1D). These results suggest that bacterium-specific OmpA molecules are responsible for different infection outcomes in tsetse.

Fig. 2.

OmpA expression and bacterial virulence in tsetse. (A) Western blot analysis of OmpA expression in 1 × 10⁴ CFU of WT Sodalis (Sg) and E. coli K12 (Ec) (Left); OmpA expression in vitro in mutant MG1655 (lane 1), recE. coli^ΔOmpA-pIO_E.coli (lane 2) and recE. coli^ΔOmpA-pIO_Sodalis (lane 3) (Center); OmpA_E.coli expression in recSodalis-pIO_E.coli (pIO) in vitro and in the hemolymph (in vivo) of systemically-infected flies (Right). recSodalis_pIL (pIL) is included as a negative control. The antibody used to probe each blot is indicated below the blot. (B) Systemic infection of tsetse with 1 × 10³ CFU of recE. coli^ΔOmpA-pIL (black squares) and WT parental MG1655 cells (red squares). (C) Survival of tsetse infected with E. coli K12 MG1655ΔOmpA::Tn5Kan-2 transcomplemented to express OmpA_Sodalis (recE.coli^ΔOmpA-pIO_Sodalis; black squares) and OmpA_E.coli (recE.coli^ΔOmpA-pIO_E.coli; red squares). (D) The number of recE. coli^ΔOmpA-pIL (black squares) and recE. coli^ΔOmpA-pIO_Sodais-Luc (black circles) per fly was measured as described in the Fig. 1 legend. (E) Systemic infection of tsetse and Drosophila with recSodalis-pIO_{E coli}. In B–E, all experiments were performed in duplicate, with 25 tsetse per experiment. In E, Drosophila (n = 50) were used as control. For all infections, tsetse received 1 × 10³ CFU of bacteria, whereas Drosophila were pricked with a tungsten needle dipped in a concentrated bacterial solution. d, days after infection.

Sodalis That Express E. coli OmpA are Pathogenic to Tsetse.

To further evaluate the role of OmpA on host infection outcomes, Sodalis were transformed to express E. coli OmpA (using construct pIO_E.coli, Fig. S1B). This gene was engineered so that the recOmpA_E.coli displays a T7 epitope on its fourth outer loop (19). When an anti-T7 antibody was used, the expression of OmpA_E.coli in Sodalis (recSodalis-pIO_E.coli) was confirmed by Western blot analysis in both cultured recSodalis-pIO_E.coli and in hemolymph from recSodalis-pIO_E.coli-infected tsetse (Fig. 2A). Furthermore, localization of recOmpA_E.coli to Sodalis ' outer membrane was confirmed by immunofluorescence (Fig. S3). Tsetse and Drosophila were then systemically infected with OmpA_E.coli-expressing Sodalis, and their mortality rates were subsequently monitored daily. Virtually all tsetse infected with OmpA_E.coli-expressing Sodalis perished by 8 dpi, which was similar to that of flies infected with WT E. coli. In contrast, Drosophila mortality was unaffected by treatment with OmpA_E.coli-expressing Sodalis (Fig. 2E). Given Sodalis ' slower growth rate in comparison with E. coli, we conclude that host death is not simply a consequence of septic shock. Rather, our results show that expression of OmpA from commensal and foreign organisms differentially influences host infection outcomes. We hypothesize that the presence of variations in the structure of OmpA from symbiotic and pathogenic bacteria are determinants of infection phenotype.

Residues Displayed on the External Loops of OmpA Exhibit Polymorphisms.

We aligned OmpA putative transmembrane domains from pathogenic and symbiotic eubacteria. Our alignment included the pathogens E. coli, Shigella flexneri, Salmonella typhimurium, and Yersinia pestis. The symbionts included were Sodalis; CMS, a new lineage from the bloodsucking fly Craterina melbae (4); SOPE, the principle endosymbiont associated with the weevil, Sitophilus oryzae (5); Candidatus Hamiltonella defensa, a facultative symbiont found in many aphid species (20); and Photorhabdus luminescens, which is a mutualistic symbiont of heterorhabditid nematodes (21). Previous phylogenetic analysis indicated that the included symbionts form a distinct lineage related to the enteric pathogens (8). The first 171 aa of OmpA_E.coli form eight membrane-traversing β-barrels connected by four loops (L1–L4) that are exposed to the external environment (22). A previous study using a peptide binding assay revealed that highly-pathogenic domains are located within exposed residues that comprise L1 and L2 of OmpA in E. coli K1 (15). We observed high conservation among the β-barrel structures in all representative bacteria. In contrast, the exposed loop regions, especially loop 1, were divergent in that the symbiont proteins contained amino acid insertions and substitutions that were absent from the pathogens (Fig. 3). The signature motif in loop 1 was conserved among the three closely related pathogens E. coli, Salmonella, and Shigella and also present in Yersinia.

Fig. 3.

Multiple sequence alignment of the first 205 N-terminal putative amino acids of OmpA from Sodalis, SOPE (S. oryzae principal endosymbiont), CMS (C. melbae symbiont), H. defensa, P. luminescens, Y. pestis, E. coli 536 (UPEC), S. typhimurium, and S. flexneri. For SOPE, only the N-terminal 118 aa are included, because the rest of the ORF is interrupted by multiple insertions that generate stop codons in each frame. External loop structures are underlined and designated L1–L4. L1 and L2 were previously proven to encode virulent domains (15). Conserved residues are outlined in black, and substitutions are in gray.

Based on genome comparison studies and phylogenetic analyses, Sodalis and SOPE form sister taxa (8). Interestingly, our results from multiple PCR-amplification experiments indicate that OmpA_SOPE is highly homologous to OmpA_Sodalis but has been rendered a nonfunctional pseudogene. Given the obligate nature of SOPE's association with Sitophilus and its residence within host cells (5), the loss of OmpA function from this symbiont is likely reflective of reductive genome processes. Many intracellular symbionts, including tsetse's obligate mutualist Wigglesworthia glossinidius (23), have, in fact, lost this protein possibly as an extreme adaptation to highly integrated symbioses.

Our alignment also indicates that modifications are present in the exposed loop domains of OmpA from both H. defensa and P. luminescens, even though these microbes exhibit distinctly pathogenic characteristics in their insect secondary hosts. The virulent phenotypes exhibited by these bacteria result from the production of phage-borne toxins, type III secretion system effectors, and immunosuppressive antibiotics (24, 25). However, little is known about the mechanisms that allow H. defensa and P. luminescens to reside as symbionts within aphids and heterorhabditid nematodes, respectively. We speculate that the OmpA polymorphisms these bacteria exhibit may underlie the symbiotic component of their lifecycles. Our OmpA sequence alignment, coupled with our results from functional expression studies, suggest that bacterium-specific variations in host-exposed OmpA domains may be responsible for the differential infection outcomes we observed upon infection of tsetse with Sodalis and E. coli.

Tsetse's Immune Response to Infection with Virulent and Avirulent Microbes.

To address the role of tsetse's immune response in relation to infection outcome, we evaluated the expression profile of a representative sample of tsetse immunity genes. We included antimicrobial peptides (AMPs) (attacin, cecropin, and defensin) and an associated regulator (peptidoglycan recognition protein (pgrp)-lb), a parasite-binding protein (annexin), thioester-containing proteins (tep2 and tep4), and prophenoloxidase-activating enzyme (PoAE). In insects TEPs function as pathogen-specific opsonins that bind to bacteria or parasites and promote their phagocytosis/encapsulation (26), whereas PoAE induces melanization (27). RNA was harvested 4 dpi from flies treated with avirulent (WT Sodalis and E. coli^ΔOmpA) and virulent (WT E. coli and recSodalis-pIO_E.coli) bacteria, and host gene expression levels were measured by using real-time quantitative PCR.

Our relative expression results indicate that the Imd and Toll signal transduction pathways, which control the expression of AMPs (28), are activated in response to avirulent and virulent infection types, as evidenced by the induced expression of attacin, cecropin, and defensin (Table 1). However, when expression levels of these AMPs were compared during avirulent and virulent infections, cecropin transcripts were significantly less abundant in flies that had received virulent bacteria. Insufficient expression of cecropin may explain the lethal effects of WT E. coli and recSodalis-pIO_E.coli on tsetse, because Cecropin exhibits potent Gram-negative bactericidal activity (29). Infection with virulent bacteria also resulted in significantly higher expression of host pgrp-lb than their avirulent counterparts. Given the role of PGRP-LB in Drosophila as the negative regulator of the Imd pathway (30), the induction of pgrp-lb may have prevented activation of the innate immune response in flies infected with virulent bacteria. To test this hypothesis, we used RNA interference to knock down the expression of tsetse pgrp-lb. Using this strategy, we were able to repress transcription of this gene by ≈70% (Fig. 4A) and subsequently reverse the lethal effect of E. coli K12 infection on tsetse (Fig. 4B). Finally, we also observed significantly lower tep 2 and PoAE expression levels in flies infected with E. coli. The lack of phagocytosis, encapsulation, and melanization responses may further enable this bacterium to undergo unimpeded replication, thus becoming lethal to tsetse.

Table 1.

Fold change in the expression of immunity-related genes in tsetse infected with avirulent versus virulent bacteria

Gene	Gene family	GenBank accession no.	Mean fold change ± SEM
			avirulent		virulent
			Sgm-WT	Ec-Δomp	Sgm-pIO	Ec-WT
annexin	Parasite-binding	NP_523370	14.8 ± 3.8	6.7 ± 2.12	1.0 ± 0.4	3.2 ± 1.0
attacin	AMP	AF368909	2.4 ± 0.3	1.5 ± 0.79	2.1 ± 1.1	−1.6 ± 0.07
cecropin	AMP	P83403	18.8 ± 4.6	14.3 ± 2.0	−2.9 ± 0.1	1.3 ± 0.09
defensin	AMP	AAL34112	8.3 ± 1.2	9.5 ± 2.51	3.0 ± 1.5	7.3 ± 0.68
PGRP-LB	AMP regulator	DQ307160	−3.1 ± 0.06	1.1 ± 0.11	2.6 ± 1.1	2.5 ± 0.13
PoAE	Melanization	ABC84592	77.0 ± 2.7	13.8 ± 8.6	1.3 ± 0.08	4.1 ± 1.6
tep2	Phagocytosis	ABC84589	19.6 ± 3.8	11.5 ± 1.24	2.7 ± 0.9	6.0 ± 0.92
tep4	Phagocytosis	ABC84590	4.6 ± 0.39	6.2 ± 1.1	2.9 ± 0.5	4.5 ± 0.75

Ec-WT, WT E. coli-infected flies; Sgm-pIO, OmpA E. coli-expressing Sodalis, Sgm-WT, WT Sodalis-infected flies; Ec-ΔOmpA, OmpA-deficient E. coli-infected flies. Bolded values represent significantly different levels of gene expression between avirulent and virulent strains of the same bacteria. Complete statistical analysis of this data is shown in Table S1.

Fig. 4.

Effect of pgrp-lb knockdown on E. coli infection in tsetse. (A) pgrp-lb expression was reduced by ≈70% in flies injected with PGRP-LB dsRNA compared with controls that received GFP dsRNA. (B) Systemic infection of tsetse with 1 × 10³ WT E. coli 4 days after injection with GFP (black circles) and PGRP-LB dsRNA (black squares). pgrp-lb expression was standardized to host tubulin.

We show that infections with virulent bacteria did not activate host immune responses, whereas infections with avirulent strains induced the expression of immunity-related genes. Although activation of tsetse's immune arsenal eliminated E. coli^ΔOmpA and E. coli^ΔOmpA that express OmpA_Sodalis, WT Sodalis was able to persist, as evidenced by its continuous expression of luciferase. Sodalis ' survival may result from its demonstrated resistance to host AMPs (e.g., Diptericin and Attacin) (31, 32). Alternatively, Sodalis may be protected from elimination by encapsulation or melanization processes by entering into host hemocytes through the use of a type III secretion system (TTSS) (9, 33). In contrast, E. coli is likely cleared from the hemolymph because of its inability to invade and persist within host cells. Further reflective of the homeostatic relationship between tsetse and Sodalis is the fact that expression of immunity-related genes in Sodalis-infected flies had returned to similar levels as those in age-matched uninfected tsetse by 14 dpi (Table S2). Our findings suggest that the fatal outcome of WT E. coli and OmpA_E.coli-expressing Sodalis infections in tsetse may result from the direct inhibitory effects of E. coli OmpA on host immune activation. Alternatively, these bacteria may escape host immune recognition through an indirect OmpA_E.coli-mediated process such as differential regulation of biofilm formation or alteration of bacterial gene expression associated with cellular adhesion and invasion processes.

Conclusions

Recent studies demonstrate that several insects can exhibit adaptive immunity in that they respond specifically, and in a lasting manner, after prior exposure to specific pathogens (34, 35). A comparable mechanism may account for tsetse's initial hyperimmune response to superinfection with its native symbiont Sodalis versus its lack of a robust immune response to E. coli. Our results that demonstrate tsetse's susceptibility to infection with normally nonpathogenic E. coli K12 may reflect the unique ecological aspects of this fly's lifestyle. Tsetse feeds exclusively on sterile vertebrate blood and exhibits an unusual viviparous reproductive strategy where intrauterine larvae are nourished with secretions from female milk glands. This mode of reproduction restricts the exposure of immature stages to a range of environmental microbes during a crucial period of immune system development. In contrast Drosophila, which is resistant to E. coli infection, reproduces on and consumes organic material undergoing decomposition by a wide variety of bacteria. Throughout evolution, tsetse's limited contact with microbial fauna may have compromised the development of a robust systemic immune system capable of eliminating a wide range of foreign microbes, including E. coli. It remains to be seen whether other insect taxa that reside in “sterile” ecological niches and rely on symbiotic bacteria for their survival also exhibit compromised immunity when challenged with foreign microbes.

Materials and Methods

Insects and Bacterial Cultures.

G. m. morsitans were maintained in Yale's insectary at 24°C with 50–55% relative humidity. These flies received defibrinated bovine blood every 48 h through an artificial membrane feeding system (36). D. melanogaster (strain Oregon^R) were treated and maintained at 25°C with 40–50% relative humidity.

S. glossinidius were isolated from surface-sterilized G. m. morsitans pupae and cultured on Aedes albopictus C6/36 cells as described (37). Sodalis, which has a doubling time of ≈24 h, were subsequently maintained cell-free in vitro at 25°C in Mitsuhashi–Maramorosch (MM) medium (1 mM CaCl2, 0.2 mM MgCl2, 2.7 mM KCl, 120 mM NaCl, 1.4 mM NaHCO3, 1.3 mM NaH₂PO₄, 22 mM D (+) glucose, 6.5 g/liter lactalbumin hydrolysate, and 5.0 g/liter yeast extract) supplemented with 5% heat-inactivated FBS (8). E. coli K12 strains (including mutant and WT MG1655) were grown in LB media.

Plasmid Construction and Bacterial Transformation.

Constructs designated pIL (4,793 bp), pIO_E.coli (4,211 bp), pIO_Sodalis (4,214 bp), and pIO_Sodalis-Luc (5,867 bp) encode the firefly luciferase gene, the E. coli OmpA gene, the Sodalis OmpA gene, and the Sodalis OmpA and luciferase genes, respectively, under transcriptional control of Sodalis' insulinase promoter (Fig. S1). Details on the engineering of these constructs are described in SI Text.

All E. coli transformations were performed via electroporation according to a standardized protocol (38). Sodalis were also genetically transformed via electroporation using a published protocol (8). After transformation, recSodalis and recE. coli cells were plated onto MM and LB agar plates, respectively (supplemented with antibiotics) (39). Single colonies of each bacterial strain were then placed into liquid culture.

Luciferase Reporter Assays.

All luciferase reporter assays were performed according to the manufacturer's protocol (Promega) by using a Sirius single-tube luminometer (Berthold Detection Systems). Overnight cultures of both bacterial species were subcultured to an OD₆₀₀ of 0.05. Subsequently, OD₆₀₀ readings were taken for recSodalis_pIL every day (Sodalis ' maximum OD₆₀₀ = 1.0) and recE. coli_pIL every 2 h (E. coli's maximum OD₆₀₀ = 1.8) until both strains had grown to approximately midlog phase. For in vitro assays, 5 × 10⁵ rec cells from each time point were subjected to luciferase quantification. This procedure was repeated twice with two clonal populations of each bacterial strain. RLU per cell was then determined by dividing RLU expression by cell number.

RLU per fly was determined by homogenizing individuals in 100 μl of PBS and then pelleting the cellular debris. Supernatant (90 μl) from each sample was then assayed for luciferase expression as described above (n = 3 individuals per time point, with each value representing the average RLU expression of the 3). Based on the assumption that luciferase expression remains constant both in vitro and in vivo, we determined the rec bacterial cell number for each species at designated time points by dividing RLU values per fly by the average number of RLU produced per cell in vitro (8.0 for recSodalis_pIL and 2.2 for recE. coli_pIL). For all in vivo experiments, background luciferase expression (from WT tsetse) was subtracted from experimental values.

Infection of Tsetse and Drosophila with rec and Mutant Bacteria.

Septic infection of tsetse and Drosophila was achieved by anesthetizing individuals with CO₂. Subsequently, tsetse were injected with 1 × 10³ CFU of live bacterial cells by using glass needles and a Narashige IM300 microinjector, whereas Drosophila were pricked with a tungsten needle dipped into a concentrated culture of the appropriate bacterial strain (SI Text, note 1). Control groups and sample sizes for all infection experiments are indicated in the corresponding figure legends. Heat-killed cells were incubated for 90 min at 80°C.

For tsetse feeding experiments, a blood meal was supplemented with 3.3 × 10⁴ CFU of live bacterial cells per milliliter of blood (which equaled 1 × 10³ cells per 30 μl, the quantity consumed by a tsetse fly per feeding, as determined by weight gain).

Outer-Membrane Protein A Alignments.

A 690-bp OmpA fragment from SOPE (S. oryzae principle endosymbiont; GenBank accession no. EU426969) and C. melbae secondary symbiont (CMS; GenBank accession no. EU684475) was amplified from insect total genomic DNA by using the following primer pair: OmpA-F, 5′-ATGAAAAAGACAGCTATC-3′; OmpA-R, 5′-GTCAGACTTCAGGGTGAA-3′. Putative OmpA protein sequences from the above-mentioned microbes, as well as Sodalis (BAE74305), H. defensa, (EU682308), P. luminescens (NP_929054), Y. pestis (NP670036), E. coli 536 (UPEC; CP000247), S. typhimurium (X02006) and S. flexneri (AF234271), were aligned by using ClustalX 1.83 (http://bips.u-strasbg.fr/fr/ClustalX/) and BioEdit (www.mbio.ncsu.edu/BioEdit/bioedit.html) software packages.

Immunological Assays.

Western blot and in situ antibody staining protocols are described in detail in SI Text. For Western blots, 1 × 10⁴ CFU of cultured WT Sodalis, WT E. coli, recSodalis_pIL, recSodalis-pIO_E.coli, recE. coli^ΔOmpA-pIO_E.coli and recE. coli^ΔOmpA-pIO_Sodalis were pelleted, reconstituted in protein loading buffer, boiled, and placed on ice. Hemolymph (20 μl) from both infected fly groups was similarly treated. Anti-E. coli OmpA and anti-T7 epitope primary antibodies were diluted 1:20,000. For in situ antibody staining, smears were prepared by using 1 × 10⁸ CFU of recSodalis-pIO_E.coli and recSodalis_pIL, and the anti-T7 epitope primary antibody was diluted 1:5,000.

Analysis of Immunity-Related Gene Expression.

Sample preparation and quantitative real-time PCR (qRT-PCR) were performed as described (8). Amplification primers are listed in Table S3. Quantitative measurements were performed on biological samples in duplicate, and results were normalized relative to tsetse's constitutively expressed β-tubulin gene (determined from each corresponding sample). Fold-change data are represented as a fraction of average normalized gene expression levels in bacteria-infected flies relative to expression levels in PBS-injected controls. Values are represented as the mean (±SEM), and statistical significance was determined by using a Student's t test and Microsoft Excel software.

Tsetse RNA Interference and Subsequent Bacterial Infection.

PGRP-LB and GFP double-stranded (ds) RNA was prepared according to the manufacturer's protocol (Ambion). Gene-specific PCR primers, each encoding a 5′ T7 RNA polymerase binding site, are listed in Table S3. Treatment and control flies were injected with 4 μg of dsPGRP-LB and dsGFP RNA, respectively. Four days after injection, both groups of flies were infected with 1 × 10³ CFU of E. coli K12 as described previously. RNAi was performed in two separate experiments for both treatment and control flies (n = 25 flies per experiment).

For quantification of pgrp-lb expression after dsRNA injection, RNA was harvested from three treatment and control flies 4 days after treatment. qRT-PCR was performed as described above. pgrp-lb expression in treatment and control flies was normalized relative to tsetse's constitutively expressed β-tubulin gene.