Chronic bacterial infections are widespread in nature and exhibit a large range of interactions with their host, from mutualism to parasitism. In insects, symbiotic intracellular bacteria (endosymbionts) are deeply integrated into host cell biology and development as they are transmitted maternally through hundreds of host generations, and early on in the insect embryogenesis they invade specialized host cells called bacteriocytes that sometimes form a specific organ, the bacteriome 123. Physiological and molecular investigations have provided evidence that endosymbionts supply their host's diet with limiting nutrients, thereby improving their invasive power and their ability to settle on nutritionally poor sources and habitats, such as blood (Glossina, Rhodnius), plant sap (aphids, psyllids, whiteflies, mealybugs), and cereal grains (Sitophilus) 24.

However, while the physiological and the evolutionary aspects of insect endosymbiosis have been thoroughly investigated over the past two decades, very little is known about the molecular mechanisms that permit the establishment, then the maintenance and the regulation of such beneficial interactions. One striking question concerns the interaction between the bacteria and the host innate immune system, an area which has been relatively well investigated in pathogenic relationships compared with mutualistic associations that have been recently approached in only a few systems 5678.

To combat infection, insects rely on multiple innate defense reactions. Insect immunity includes physical barriers, together with local and systemic immune responses involving both cellular and humoral pathways (reviewed in 9). Activation of the humoral pathway consists of microbe-associated molecular pattern recognition by pattern recognition receptors, such as peptidoglycan recognition proteins (PGRPs), and the activation of intracellular signaling pathways, such as the Toll and the Immune deficiency (Imd) pathways. These pathways activate, in particular, the production and the secretion of a panel of antimicrobial peptides (AMPs) in response to Gram-positive and Gram-negative bacteria. In addition to AMP production, the insect humoral response also involves a proteolytic cascade leading to prophenoloxidase (PPO) activation and subsequent melanin synthesis at the site of cuticular injury. This reaction, called melanization, plays a key role in wound healing, encapsulation, sequestration of microbes and production of toxic intermediates 10. While the systemic response is by far the best characterized pathway among immune reactions, the local immune response (also known as epithelial immunity) was only recently shown to significantly contribute to protection against invading microorganisms in the alimentary tract and tracheae. In the gut, for example, there is an inducible local production of AMPs. This response is triggered upon natural infection by Gram-negative bacteria and is mediated by the Imd pathway 11. The gut epithelium also expresses amidase PGRP. It has been shown that these PGRPs, which scavenge peptidoglycan released by commensal bacteria, reduce gut immune reactivity and avoid a state of permanent immune activation in this tissue 1213.

In a previous work on the mutualistic interaction between the weevil Sitophilus zeamais and a γ-proteobacterium called S. zeamais primary endosymbiont (SZPE) (see 14 for a review on the model), we discovered the overexpression of a member of the PGRP gene family in the bacteriome tissue of the host 1516. We showed that wPGRP is induced after a bacterial challenge and that wPGRP gene expression depends on bacterial growth. Moreover, we have shown that the wPGRP gene was induced concomitantly with an endosymbiont release from the bacteriocytes during nymphosis, demonstrating that the PGRP gene family is involved in host-symbiont interaction 16.

This study was dedicated to enlarging the panel of insect immune genes involved in the host-symbiont interaction. We first identified the host immune-relevant genes by suppressive subtractive hybridization (SSH), and analyzed in silico their full-length coding sequences completed with RACE-PCR. We then measured their steady-state levels in aposymbiotic larvae challenged with SZPE and in the bacteriocyte cells of naturally infected larvae. We show that experimental infection of larvae with SZPE results in a significant induction of AMPs and that only few immune gene transcripts, including homologs to an antibacterial peptide and immunomodulators, are accumulated in the bacteriome tissue. These data reveal that endosymbionts are perceived as intruders while being present in hemolymph and that there is a local immune gene expression in the bacteriome. This study provides the first indication of how insects may maintain endosymbionts within the bacteriome and prevent their invasion into insect tissues.

There is compelling evidence that intracellular bacterial infections can establish chronic non-septic relationships with host cells for many generations. Some bacteria (for example, Chlamydophila and Salmonella) invade and proliferate rapidly in their host, causing diseases, while others (for example, Mycobacterium tuberculosis) remain latent and may reactivate months or years after initial exposure to cause chronic disease 31. In the insect world, some intracellular bacteria can generate a long-term relationship within the host without causing any disease or invading any host tissues except the germ cells and the bacteriocytes. The overexpression of the wPGRP1 gene in the weevil bacteriome suggests a close interaction between the host immune system and the endosymbionts.

To determine how the insect immune system recognizes the bacteria and prevents bacterial invasion of insect tissues, we first identified, by SSH technology, the immune genes induced in insects challenged with E. coli. Bioinformatic analysis scored a relatively high proportion of ESTs with similarities to immune-related genes, including several antibacterial peptides, lysozymes, PGRPs and proteins of the PPO pathway (Table 2), which highlights the efficiency of the subtractive approach. Further in silico analysis on full-length coding sequences has confirmed the identification of an i-type and a c-type lysozyme gene (INF-152, INF-282) and at least four antibacterial peptide genes: two members of the coleoptericin family (INF-18, INF-145), a defensin (INF-217) and a homolog of luxuriosin (INF-479), an antimicrobial peptide from Acalolepta luxuriosa 20. We also validated a homolog of Tollip, a gene so far only studied in mammals that seems to be absent from the Drosophila genome, and a PGRP gene (wPGRP2, INF-441) additional to the one previously described (wPGRP1 gene, INF-9). The expression profile of all the putative immune genes was then characterized by larvae challenge experiments combined with RT-PCR transcript quantification (Table 2). Thus we have demonstrated that all the genes studied were inducible, except for homologs of Tollip (INF-359) and hemolymph proteinase 17 (INF-459), which appear to be constitutive.

To elucidate the particular bacteriome immune features, we first examined the systemic host response following aposymbiotic larvae injection with SZPE (Figure 2) and we then measured immune gene expression in the bacteriome of naturally infected larvae (Figure 3). Experimental infection of the larvae led to a systemic immune response with the up-regulation of all the genes with similarities to AMPs, providing evidence that SZPE is recognized as a microbial intruder while present in the hemolymph. It is noteworthy that in contrast to SZPE, Spiroplasma does not induce host transcripts encoding AMPs in Drosophila while present in hemolymph. However, the lack of bacterial cell wall structure of Spiroplasma is probably the best explanation for the apparent absence of host humoral response to this bacterium 32.

In the weevil symbiotic larvae, in addition to the wPGRP1 gene, one antibacterial peptide induced by bacterial challenge (that is, the coleoptericin gene) is overexpressed in the bacteriome. These data confirm the existence of an immune response in the bacteriome tissue and provide the first elements to explain how host-endosymbiont association can persist with regard to the host immune system and bacterial proliferation.

Since all AMP genes are up-regulated following a hemolymph challenge with SZPE, the unique overexpression of the coleoptericin gene in the bacteriome suggests a constitutive expression in this tissue rather than an induction by SZPE, unless coleoptericin gene regulation involves a separate pathway from those that trigger synthesis of other AMPs. We can only speculate about this question, as comparison between symbiont-full and symbiont-free bacteriocytes cannot be conducted since no symbiont-free bacteriome has been detected in Sitophilus aposymbiotic insects so far. Taking into consideration that some AMPs are expressed constitutively in cells potentially in contact with the environment 113334, and that the coleoptericin gene encodes a signal peptide, these data suggest that the coleoptericin gene is expressed constitutively and secreted outside the bacteriocyte. Coleoptericin is currently under investigation to determine whether this peptide may prevent host tissue invasion by the endosymbiont, and/or may protect the endosymbiont population against infection with environmental bacteria, as the bacteriome tissue is intimately attached to the insect foregut. Moreover, another interesting issue would be to investigate whether coleptericin expression in the bacteriome can protect the host from pathogens. This would represent an additional example of 'symbiont-mediated protection' 35.

Although the absence of a host humoral immune response to an intracellular symbiont and a constitutive expression of coleoptericin in the bacteriome are consistent with current knowledge of the insect immune response, the overexpression of gene homologs to PGRP-LB (that is, wPGRP1) and Tollip is quite intriguing. PGRP-LB gene inhibition in Drosophila was indeed shown to result in a significant induction of antibacterial peptides 13, while Tollip has been implicated as a negative regulator of the mammalian immune response 2425. Moreover, these two genes are overexpressed in the gut epithelia and have been proposed to play a role in the control of immune reactivity of the host to the presence of bacteria in the gut. Therefore, the overexpression of these two genes in the bacteriome may be consistent with an immune modulation that inhibits the production of AMPs in this organ, as it was described in the gut epithelia. Together with the overexpression of an AMP (that is, coleoptericin), these results uncover striking similarities between the bacteriome immune profile and the local immune response in gut epithelium that is in permanent contact with commensal and mutualistic bacteria 91113333637. A gene silencing study in weevil is currently under development to validate coleoptericin, wPGRP1 and Tollip gene function and their contribution to the maintenance of symbiosis.

It is noteworthy that this work was performed on a recently established symbiosis where, contrary to ancient symbiotic associations known for a drastically reduced endosymbiont genome size such as in the aphid/Buchnera association 38, SZPE exhibits similar features to those of free-living Gram-negative bacteria 3940414243. Hence, whether and how the host immune response evolves in parallel to symbiont genome reduction constitutes a key aspect in the understanding of host-symbiont interaction in the course of evolution. Recently, a transcriptomic study was performed on the aphid bacteriocyte to identify a gene of interest in intracellular symbiosis 30. Intriguingly, no homolog to any known immune genes was shown to be expressed in the aphid bacteriocyte except for i-type lysozymes, whose function in bacteriocytes remains unknown. Since both the weevil i-type and c-type lysozymes were shown to be weakly expressed in the bacteriocytes, these data suggest that the bacteriocyte immune profile may have evolved with bacterial coevolution with the host, unless the immune response of weevils (holometabolous insects) has diverged significantly from that of aphids (heterometabolous insects).

Whatever the antimicrobial effectors (that is, antibacterial peptides, lysozyme), their permanent expression in bacteriocytes could represent a common mechanism to restrict the localization of mutualistic symbionts, which is probably necessary to optimize the host-symbiont interaction. Without this confinement, intracellular symbiotic bacteria may invade the whole organism without inducing any antibacterial peptide synthesis, as has been shown in Drosophila infected by Wolbachia, a parasitic symbiont widely distributed in host tissues 44.

This work provides the first immune gene expression profile in the insect bacteriome and reveals the overexpression of at least three homologs to immune genes: a member of the PGRP gene family involved in bacteria recognition, an antibacterial peptide involved in bacterial clearance and a gene involved in immunomodulation in mammals. This immune profile uncovers some striking similarities between the bacteriome and the gut epithelium, which is in constant contact with environmental bacteria. Taking into account that the endosymbiont is recognized as an intruder in the host hemolymph, these findings also indicate that the host immune system may prevent endosymbiont invasion into insect tissues in a manner similar to the gut local immune response that helps to confine the microbiota to the gut, avoiding a permanent systemic response to the commensal bacteria.

CA designed and performed experiments, analyzed data (statistics and bioinformatics), wrote the paper and participated in bioinformatic analysis of the subtracted library; VPB participated in molecular studies and sequenced the subtracted library. AV carried out dissections and real-time RT-PCR, and CVM participated in molecular studies and sequence analysis. DC carried out bioinformatics analysis and the URL construction for the subtracted library; AL and AM sequenced the subtracted library and provided critical comments on the manuscript. AH conceived the study, coordinated the work and helped to draft and write the manuscript. All authors read and approved the final manuscript.

During the preparation of this manuscript, the experimental identification of genes that are induced in response to septic injury in the red flour beetle Tribolium castaneum using the SSH method has been published 47. By comparing EST sequence to Tribolium sequence databases, the authors have identified 75 immune-inducible genes in T. castaneum potentially involved in immune defense, signaling, and other immunity-linked cellular processes including homologs of, for example, Toll, PGRP-SC, lysozyme, and multiple isoforms of defensins and thaumatin-like peptides. They have also performed a qRT-PCR to determine transcriptional regulation of selected genes (that is, defensin, thaumatin and stress-related genes) in response to either septic versus sterile injury or heat shock.

While this study explored the Tribolium immune response considering all cellular processes, we focused our study on Sitophilus immune pathways and effectors. It should be noted that, in contrast to this study, we have identified members of different antibacterial peptide families (for example, coleoptericin, defensin). We also identified some homologs of thaumatin-like peptides (for example, INF-475, INF-332, INF-CL57Contig1), but were unable to quantify them without specific amplification.