Recently, the use of invertebrate models of infection has given exciting insights into host–pathogen interaction for a number of bacteria. In particular, this has revealed important factors of the host response with remarkable parallels in higher organisms. Here, we review the advances attained in the elucidation of virulence determinants of a major human pathogen, Staphylococcus aureus, in relation to the invertebrate models thus far applied, the silkworm (Bombyx mori), the fruit fly (Drosophila melanogaster) and the roundworm (Caenorhabditis elegans). Also, the major pathways of host defence are covered in light of the response to S. aureus and the similarities and divergences in innate immunity of vertebrates and invertebrates. Consequently, we comparatively consider pathogen recognition receptors, signal transduction pathways (including Toll, Imd and others), and the humoral and cellular antimicrobial effectors. The technically convenient and ethically acceptable invertebrates appear as a valuable first tool to discriminate molecules participating from both sides of the host–S. aureus interaction as well as a high throughput method for antimicrobial screening.
Bacterial infectious disease and its debilitating or fatal outcomes are by-products of the microbe successfully colonizing its ecological niche: the animal or plant host. To unveil the basis of host–pathogen interaction, several animal and plant models have been developed based on their experimental advantages, i.e., low cost, ease of maintenance, high throughput capabilities, similarity to natural hosts, etc. Another preferred feature for any model host is the knowledge already accumulated about its biology to be able to monitor and evaluate changes following the challenge with a pathogen. Ultimately and obviously, the key characteristic for any host to qualify as a model for bacterial pathogenesis is to respond to the infectious agent in a similar manner to the host it is modelled after. The validity of a pathogen–host pair as a model is to be provided by comparative scrutiny of the findings for a given pathogen in various hosts including its natural host.
2Staphylococcus aureus: the pathogen
The gram-positive bacterium Staphylococcus aureus is the most frequent pathogen isolated from community- and hospital-acquired infections in the bloodstream, lower respiratory tract, and skin/soft tissue worldwide . It is an opportunist pathogen, common resident of nasal membranes and skin of warm-blooded animals with a rich palette of virulence factors and great versatility to efficiently colonize a variety of systems and organs. As a consequence, it is the causative agent of a variety of diseases in humans and animals ranging from sub-acute superficial skin lesions and wound infections to more serious conditions, such as osteomylietis, endocarditis and septicemia . Additionally, S. aureus has been proficient at developing resistances to the newest generation of antibiotics, and the increase is on the rise. In the United States, Methicillin Resistant Staphylococcusaureus (MRSA) in hospital infections has nearly doubled since 1997, more than any other nosocomial pathogen including vancomycin resistant enterococci or Pseudomonas aeruginosa. The study of staphylococcal virulence, as well as the defence and pathophysiological responses in the host in the many diseases due to S. aureus has prompted the development of the corresponding animal models.
Small mammals are the predominant predecessors of invertebrates as models of bacterial human pathogenesis. Since the early 1960s rabbits, rats, and mice have been used to determine various aspects of the pathophysiological characteristics, immunological responses and efficiency of therapeutics during staphylococcal infections in various organs and systems. Although rats remain the preferred model of staphylococcal endocarditis  and rabbits for Toxic Shock Syndrome (TSS)  mouse models have been developed for a variety of the pathophysiologies generated by S. aureus, including: wound infection , arthritis and sepsis , mastitis , kidney infection , and nasal colonization . Consequently, the murine model has been a hallmark contributor towards discerning the breadth of virulence effectors and regulators identified in staphylococcal species. Despite the unquestionable advantages in the usage of mammals for bacterial pathogenesis studies there are still important problems. The use of large numbers of mammals is difficult for logistical, ethical and financial reasons, and large numbers are critical to statistically overcome the intrinsic variability of the otherwise essential ‘in vivo’ models. Alternatives have been explored to allow easier determination of host effects.
4Of invertebrates and humans: notes on comparative genomics
The convenient simplicity of invertebrates as animal models of infectious disease exposed them as appealing candidates to overcome the logistical limitations of the mouse model. The question obviously is how similar are insects, worms or amoebae to mammals, particularly humans, to use them as models? We certainly do not look alike and the evolutionary distance between us ranges from more than 900 to over 2000 million years . However, it is widely acknowledged that the percentage of human protein homologs found within the various invertebrates is in some cases remarkably high, in the vicinity of, e.g., Drosophila melanogaster (fruit fly), 60%, Caenorhabditis elegans (roundworm) 55%, Dictiostelyum discoideum, 23% (TaxPlot, 170 cutoff, National Center for Biotechnology Information; http://www.ncbi.nlm.nih.gov/sutils/taxik2.cgi). Importantly, a similar number of predicted proteins in humans (62%), flies (51%) and worms (49%) are contained within clusters of orthologous genes . The significance of these findings lies in the fact that orthologous proteins constitute a core of conserved functions and contribute to basic biological processes. Importantly, amongst the significantly conserved genes and pathways reside those involved in bacterial pathogen recognition, signal transmission and effector molecules of the first barrier of the immune defence: the innate immune response.
5Innate immune response: a vertebrate inheritance from invertebrate ancestors?
The innate immune response consists of immediate recognition of the bacterial pathogen and deployment of humoral and cellular effectors aimed to its elimination . Whether the protection systems against bacterial invaders in nematodes, insects and mammals have developed divergently from common ancestor molecules or is a case of convergent evolution from originally different molecules is not clear [14,15]. Taxa-specific differences do exist, e.g., vertebrates lack the melanization, nodulation and encapsulation systems used by insects and insects are deprived of the adaptive immune response encountered in jawed vertebrates [13,16–19]. Nevertheless, there is striking and remarkable homology between vertebrates and invertebrates on the humoral and cellular responses of their innate immune systems (Fig. 1) [15–22]. The following summary of these similarities is important to provide a suitable context whereto analyse the applications and results of invertebrate models in staphylococcal pathogenesis.
6The humoral response
The effectors of the humoral innate immune response can be divided in two groups, general factors commonly used for house-keeping functions and adapted to anti-microbial functions, and dedicated factors for recognition and destruction of pathogens (antimicrobial peptides).
6.1‘General purpose’ anti-microbials
The phagocytic vacuoles within the human neutrophils contain a variety of microbial-degrading enzymes including lysozyme, proteinases (elastase, cathepsin G, gelatinase), nucleases, saccharidases, phospholipases, and enzymes for lipopolysaccharide (LPS) degradation . Similarly, lysozyme is produced within the phagocytic haemocytes in diptera and lepidoptera  and it has also been shown to be present in C. elegans. It degrades the peptidoglycan layer, releasing sugars and exposing teichoic acids and lipomannans, which are recognised by BDL-1 carbohydrate binding proteins (lectins) in cockroaches . The subsequent stages are also similar from insects to mammals involving an oxidative burst of oxygen radical intermediates characteristic of macrophage physiology, including a role for nitric oxide . Importantly, in Drosophila, as in mammals, lysozyme, lipases and metalloproteases (like the mammalian gelatinases or collagenases) amongst other proteins are induced by and protect against gram-negative microorganisms [26,27]. Members of the transferrin family of glycoproteins, involved in iron homeostasis and known antimicrobial activity, like the human lactoferrin, are also present in diptera (D. melanogaster) and lepidoptera (Manduca sexta and Bombyx mori) .
6.2‘Custom’ antimicrobial peptides
More than 800 peptides with anti-microbial activity (AMPs) against a broad range of bacteria and fungi have been isolated from unicellular and multicellular organisms: bacteria, amoebae, plants, lepidoptera, amphibians, fish and human . Over 500 of them have activity against gram-positive bacteria and more than 200 have been reported to act on S. aureus (http://www.bbcm.univ.trieste.it/~tossi/search.htm). For instance, ABF-2 in C. elegans, moricin from the silkworm , and β-defensin-3 from humans . In vertebrates the most abundant AMPs are the defensins, β-sheet or β-hairpin-like cationic peptides with internal disulfide bridges . They are also widely spread amongst arthropods [14,19], and the C. elegans genome sequence has revealed a few AMP candidates with characteristics resembling insecta and mollusca defensins .
Although AMPs as a whole have a broad spectrum of microbicidal activities against gram-positives, gram-negatives, fungi, protozoa and viruses, selectivity does exist and it is exerted at two levels. On one hand, structural properties of the peptide itself (primary sequence, net charge, etc.) will recognize specific targets . Also, gram-negatives, gram-positives and fungi induce the expression of different AMP-encoding genes with preferential activity for the inducing pathogen through pathways notably conserved from C. elegans to humans (see below; Fig. 1). In Drosophila, gram-positive bacteria induce drosomycin and other AMPs through the Toll pathway, while gram-negatives induce diptericin, attacin, cecropin and many others through the Imd pathway . Cecropins are also induced in other invertebrates, like M. sexta. However, cross-talk seems to occur. For instance, S. aureus, L. lactis, M. luteus and Bacillus sp. are able to induce production of cecropins, diptericins and attacins, and P. aeruginosa, E. coli and Erwinia carotovora induce expression of drosomycin . Nevertheless, S. aureus is especially resilient to a wide range of AMPs (human alpha-defensins, protegrins, tachyplesins and magainin II) through the generation of gene products responsible for the modification of teichoic acids and phospholipids as well as efflux pumps . Components of both subsets of the humoral response are intimately connected, for instance, lysozyme (and cecropin) action in insects might be facilitated by attacin .
7Recognition by the host: pattern recognition receptors
The so-called pathogen-associated molecular patterns (PAMP) are molecules exclusive to microorganisms therefore absent from eukaryotic cells (e.g., LPS, lipoteichoic acids (LTA) and peptidoglycan (PGN)) and are identified by pattern recognition receptors (PRR) of the immune system of both, vertebrates and invertebrates [15,19,34]. The best studied invertebrate PRRs are the peptidoglycan recognition proteins (PGRP) and the gram-negative bacteria-binding proteins (GBNP) [18,19,35].
PGRPs are soluble or transmembrane proteins containing a domain similar to the bacterial amidases (which are involved in hydrolysis of bacterial cell wall fragments) . PGRPs are sub-grouped in various families depending on their length and the presence of signal peptides, transmembrane segments, and various other domains . The PGRP-S and PGRP-L families are common to various mammals and insects [15,19,35,36].
Different PGRP in a given organism show strong preference, but not necessarily exclusivity, towards specific PAMPs and their activities may be concerted with other PRRs (Fig. 1). In Drosophila, soluble PGRP-SA (similar to the discovered in silkworms, moths, mosquitoes and mammalian) circulates in the hemolymph recognizing the PGN variant common to S. aureus and other gram-positives (lysine in the third position of the PGN peptide chain) and in conjunction with gram-negative binding protein 1 (GNBP1), activate the Toll signal transduction pathway [34,37]. Very recently, another PRR, PGRP-SD has been shown to also mediate the response to S. aureus and other gram-positive bacteria in Drosophila. The result is the synthesis of AMPs, e.g., drosomycin [15,34,37,39]. In contrast, a PGRP-LC isoform in the membrane (PGRP-LCx) [40–42] and the soluble PGRP-LE  recognize the diaminopimelic acid-containing peptide of the peptidoglycan encountered in gram-negative and some gram-positive bacteria (Bacillus and Clostridium) activating the immune deficiency pathway (Imd) [15,39,44]. Certain promiscuity in the interactions exists; PGRP-LCx also weakly recognizes gram-positive-type PGN and PGRP-S can bind with low affinity gram-negative-type PGN as well as LPS and LTA . Intracellular staphylococci are also recognized by the mammalian Nod2 protein but not by Nod1 proteins, which are intracellular PRRs for gram-negative PGN . Nod homologs are found in other organisms, including plants.
PGRP homologs in different organisms may differ in the downstream pathways of choice. For instance, PGRP-SA from B. mori and PGRP-LE from Drosophila activate the cascade leading to the insect-specific mechanism of pro-phenoloxidase (PPO) activation (Fig. 1), but not the PGRP characterized from M. sexta, which may participate in a different immune pathway [16,18,35,36,43]. Similarly there are differences regarding the interacting partners in different species. The Drosophila PGRP-SA acts as a PAMP co-receptor in conjunction with another receptor, Toll. However, the intra-cellularly located PGRP-S in mammals does not interact with Toll-like receptors (TLRs), which can themselves interact directly with PAMPs . Importantly, PGRP-S can eliminate non-pathogens or pathogens of moderate virulence, like B. subtilis or Micrococcus luteus, in an unknown but non-amidase-mediated fashion, but it does not affect highly pathogenic microorganisms like S. aureus and E. coli, likely influenced by the decreased binding affinity to the latter . In Drosophila, PGRP-LC seems to also play a role in phagocytosis of E. coli but not of S. aureus.
Microbial carbohydrates (d-mannose, N-acetylglucosamine, LTA) are recognized by soluble PRRs i.e., mannose binding lectins (MBLs) and ficolins, which in turn activate the complement cascade (Fig. 1) [17,47]. Ficolins and C-type lectins (CTLs), like the mammalian MBL, can be found in higher and lower vertebrates as well as in horsehoe crab, ascidians, and other invertebrates. In D. melanogaster, C. elegans and M. sexta CTLs are induced following bacterial immune challenge [18,22,26,27]. Like in human, MBL or ficolins bound to their cognate microbial substrates activate the serine proteases of the complement cascade, immulectins from M. sexta bind to serine proteases in plasma activating pro-phenoloxidases (PPOs), an early stage in the process of melanization . Interestingly, PPOs have sequence similarity to the thio-ester region of the vertebrate complement proteins C3 and C4 . As in vertebrates, thio-ester-containing proteins (TEPs) are also found to play a role as opsonins in the cellular immune response of insects.
8The pathogen is here, now what do we do?
In a variety of organisms, including mammalia, crustacea, insecta and nematoda, following recognition of the non-self (PAMP/PRR), serine protease cascades similar to the PPO activation in insects, opsonization in crustacea or complement activation in humans, are implicated in transmission and amplification of the signal . Serine proteases containing the so-called CLIP domains are modulated by the presence of serine protease inhibitors (serpins) (Fig. 1) . The information is transferred to membrane receptors and subsequently to intracellular phosphorylation cascades resulting in transcriptional upregulation of AMP-encoding genes. There are two main signal transduction pathways i.e., Toll and Imd in Drosophila, which are analogous to the mammalian Toll/IL-1R and TNF-signalling in humans [19,48]. The Imd pathway appears to mainly control the expression of anti-gram-negative AMPs, i.e., diptericins and drosocins, while Toll preferentially induces the expression of the anti-gram-positive and anti-fungal AMPs like drosomycin.
8.1The Toll pathway
In Drosophila, PGRP-SA and GNBP1 by an unknown mechanism activate the CLIP (Persephone) likely to be involved in the cleavage of another CLIP, Spätzle, which binds directly to and activates the transmembrane receptor Toll (Fig. 1) [19,21,49]. Alternatively, the activated PRRs may inactivate the serpin inhibitors (like Necrotic) responsible for blocking the action of the serine protease cascade . Serpins are also involved in the regulation of melanization in D. melanogaster[21,51], M. sexta, and A. gambiae. Besides Drosophila, Toll-like receptors (TLRs) have been identified in mosquitoes, silkworm, mouse, human, etc., and shown to be regulated by PAMPs and live bacteria [35,52].
The Drosophila Toll upon activation, recruits three death domain proteins Myd88, Tube and Pelle, which are homologs of the human Myd88, Mal (functional equivalent) and IRAK, respectively . Out of the nine Drosophila Toll proteins, only Toll plays a role in immune defence inducing drosomycin expression, while roles in infection have been assigned to most mammalian TLRs . The toll pathway does not seem to be functionally conserved in C. elegans since homologs of the Drosophila Toll, dTraf, pelle and cactus (tol-1, trf-1, pik-1 and ikb-1, respectively) do not seem to play a role in immune defence .
The activation of the Pelle kinase through the Toll receptor-adapter complex triggers the hydrolysis of Cactus, a Drosophila homolog of the human NF-κβα inhibitor (IKβα) responsible for sequestering in the cytoplasm Dif and Relish, two members of the NF-κβα/Rel family of human transcription factors (Fig. 1) . Interestingly, there is no NF-κβ homolog in C. elegans. Ultimately, the release and subsequent translocation of these transcription factors entails expression of AMPs [55,56]. Northern blot analysis indicates that S. aureus and M. luteus induce cecropin expression using the Toll pathway through Dif . Alternatively, Dif activation for these bacterial species may occur independently of Toll. The induction by S. aureus was not mediated by its LTAs. In contrast, LTAs from S. aureus and S. pneumoniae bound to the human LPS-binding protein, CD14 and TLR2, but not to TLR4 or MD2, activated human mononuclear phagocytes . Abundant evidence indicates that TLR2, TLR6, CD14, Myd88 and NF-κβ are key players in the response to S. aureus infections [58–61]. Nod1 and Nod2 also induce NF-kβ activation  Further homologies in the signal cascade pathway are noticeable as shown in Fig. 1. Orthologs of Toll, Spätzle, Myd88, Tube, Pelle and Dif have also been identified in the mosquito Anopheles gambiae[16,35,62].
8.2The Imd pathway
The Immune Deficiency pathway receives its name after its first intracellular effector in Drosophila, Imd, and it is analogous to the TNFα signalling pathway in humans . However, similar to the latter various death-domain-containing proteins (Imd, dFADD and DREDD) are recruited following transduction of the PGRP-LC/PGRP-LE signal by a yet unclear mechanism (Fig. 1) [19,40,41,43]. The Drosophila Imd, dFADD and DREDD have homolog counterparts in mammals (RIP, FADD and caspase-8, respectively) as well as the downstream kinases dTAK1, ird5, Kenny (TAK-1, IKKβ, IKKγ; in mammals, respectively) . The final stage of this cascade in Drosophila is the phosphorylation of Relish, a member of the Rel family of transcriptional regulators characterized by containing in the same molecule the Iκβ/NF-κβ couple . The components of the Imd pathway are conserved in the genome sequence of invertebrates like A. gambiae but not in C. elegans. As raised by various authors, the level of complexity of the immune defences is greater than that reported. For instance, Relish and/or the Imd pathway seem to be responsive and necessary not only against gram-negative, but also against the gram-positive S. aureus and Enterococcus faecalis.
Besides the Toll and Imd signalling cascades, other phosphorylation pathways present in vertebrates and invertebrates are typically involved in developmental or stress resistance processes and also induced in response to bacterial infections. They contribute to the host defence through the induction of cytokines, immune-related agonist receptors, general antimicrobials like lysozyme or oxygen radical-detoxifying enzymes, etc. The JAK/STAT pathway branches out from the Imd pathway in fruit flies and is activated in response to gram-negative challenge regulating expression of TEPs (Fig. 1) . A similar involvement of the JAK/STAT pathway in immunity is observed in mosquitoes . Its composition is essentially identical in vertebrates (where it has been shown to also be modulated by staphylococcal enterotoxin) and conserved in lower eukaryotes [16,20,35]. Similarly, the p38 MAPK pathway is present in phylogenetically diverse organisms, from mammals to the roundworm, playing a role, amongst others, in immune defence . The roundworm, seemingly lacking Imd- and Toll-mediated defence mechanisms, has a homolog of the p38 kinase (pmk-1), which has been proven to be an upstream regulator of programmed cell death (PCD) in response to, and important in defense against, E. faecalis and S. aureus (Fig. 1) [63,64]. Likewise, mammalian neutrophils activate p38 MAPK and induce apoptosis after S. aureus phagocytosis . Interestingly, neutrophil apoptosis in mammals seems to also be involved in infection resolution . Fruit flies, the roundworm, yeast, mice and humans further complete their immune defence repertoire with the Insulin-Like (I-LR) and TGF-β receptor pathways [22,67,68]. C. elegans mutants in I-LR genes (daf-2 and age-1) are more resistant to S. aureus and S. faecalis and bacterial challenge induces expression of lipases, lysozyme and caenopores through TGF-β[26,27,69,70]. Resistance to pathogens appears to occur through the interplay between immune defence, apoptosis and general stress mechanisms.
9The cellular response
Phagocytosis is a hallmark of the cellular branch of the immune response and it exhibits remarkable commonalties in vertebrates and invertebrates. Plasmatocytes in Drosophila, granulocytes in Aedes aegypti or their mammalian counterparts macrophages/monocytes recognize invaders to subsequently engulf and dispose of them [16,71]. Although in C. elegans coelomocytes could not be shown to be involved in phagocytosis of bacteria, they have such activity in other nematodes . The Drosophila PGRP-LC and Croquemort (a human CD36 homolog) participate in the recognition and phagocytosis of gram-negative (but not gram-positive) bacteria (Fig. 1) [42,72]. The complement-like thioester-containing opsonin TEP1 promotes phagocytosis of gram-negative bacteria in a mosquito haemocyte-like cell line . Sequence homology reveals the presence of a TEP-related gene in C. elegans.
In insects and mammals opsonic ligands bind to the surface of the particle, which is followed by recognition by specific receptors and subsequent phagocytosis. Work with cockroaches has shown that following gram-negative infection, N-acetylglucosamine (GlcNa)-specific lectins (BDL-2 lectins) recognise and bind to peptidoglycans on the bacterial cell surface. These bind to plasmatocytes and facilitate phagocytosis . In M. sexta the process also seems to be mediated through lectins . Surface receptors on lepidopterous phagocytes (e.g., G. mellonella) are similar to receptors on mammalian phagocytes and are responsible for phagocytosis . The insect proteins Malvolio and dSR-C1 show a high degree of homology to mouse natural resistance associated macrophage protein-1 (NRAMP-1). And Croquemort in D. melanogaster mediates the breakdown of apoptotic cells, similarly to murine macrophage proteins [72,73].
10Invertebrate models of human bacterial pathogenesis
The group of F. Ausubel has been a pioneer in the use of plants and invertebrates as models to identify bacterial virulence genes, i.e., the cress Arabidopsis, G. mellonella and C. elegans for P. aeruginosa[74–76], and recently, C. elegans as a model host to identify virulence factors from E. faecalis, S. pneumoniae and S. aureus[64,77]. The roundworm has also been used by other authors for Burkholderia pseudomallei, Salmonella typhimurium, Serratia marcescens,  and Yersinia. The wax moth M. sexta has been utilized as an insect model for the insect pathogen Photorhabdus luminescens and the soil-living amoeba D. discoideum to study intracellular pathogens, e.g., Legionella pneumophila. Silkworm larva was introduced as a model of infections caused by P. aeruginosa, Vibrio cholerae and S. aureus. D'Argenio and colleagues  applied the fruit fly as model host for P. aeruginosa infections and De Gregorio and co-workers performed a genome-wide screening of Drosophila genes expressed in response to in E. coli and M. luteus. The gram-negative S. marcescens and the gram-positive Listeria monocytogenes and Mycobacterium marinum[85,86] were also assayed in D. melanogaster. We have recently reported the use of the fruit fly to discriminate virulence determinants in S. aureus and also demonstrated its applicability to S. agalactiae, S. pyogenes, B. megaterium, E. faecalis, S. epidermidis, B. subtilis and S. pneumoniae.
Investigations on the involvement in virulence of gene products from pathogens in a given host, particularly natural pathogens, may monitor highly specific phenotypes, like the histopathological observations of subcellular distribution of P. luminescens in M. sexta. Results from these studies may allow further progress on the molecular basis of the interaction between bacteria and plants or invertebrates establishing fundamental mechanisms not unlikely to be preserved higher up the phylogenetic scale. It may also be a tool to tackle the study of attenuated microbial pathogens. Bacterial proliferation within a host is also a rather easy parameter to assess and has been determined in various studies [74,77,83,84,87,88]. Awareness of microscopic or sub-microscopic phenotypes would provide an insight as to the underlying causes of infection, and further the parallelisms between higher and lower order animals. However, a dramatic ‘life/death’ end-point phenotype of an invertebrate host confronted by a pathogen is a preferred clear-cut and high-throughput screening used in most studies [22,79,83,87].
Since the early 1960s invertebrates like C. elegans and D. melanogaster were introduced as animal models to study genetics, development, or the nervous system because although primitive they could provide answers to fundamental questions in higher order organisms. They are multicellular with a short life cycle, easy to cultivate and small enough to be handled in large numbers. The economic impact of some other less amenable invertebrates also prompted their study and became models of invertebrate physiology. For instance, the silkworm (B. mori) or pests like the tobacco hornworm (M. sexta) and the wax moth (G. mellonella). The properties and breadth of knowledge acquired from those and other invertebrates made them appropriate as workhorses for bacterial pathogenesis studies. In addition, these models are commercially available and can be maintained at a minimal cost in the laboratory.
The reasonably sized larvae of the lepidopteran G. mellonella, B. mori, or M. sexta (ca. 2, 5 and 7 cm respectively) are attractive model systems whereto easily inject the microbial pathogen of choice and convenient for studies of drug pharmacodynamics. The knowledge accumulated on the lepidopteran although significant it is not yet comparable to that obtained from D. melanogaster or C. elegans. More importantly genetic studies are comparatively impractical because of more restricted tools and deficient genomic knowledge. Nonetheless, the sequencing project of the B. mori genome scheduled to be completed by 2004 and genetic tools and techniques for lepidopteran investigations are quickly accumulating  (http://www.ab.a.u-tokyo.ac.jp/lep-genome).
C. elegans is transparent, can generate isogenic lines and although it is much smaller (1 mm) it can feed on bacterial lawns, a convenient route of inoculation . Nonetheless, C. elegans is rather selective towards growth medium and bacterial species, which may complicate analysis of results [78,90]. For instance, although BHI medium seems to be optimal to generate E. coli highly pathogenic for C. elegans, S. aureus grown in BHI does not display its full lethal potential in the same host [64,77]. S. aureus was more pathogenic in TS agar . Moreover, C. elegans isolates vary in their responses towards a specific pathogen, as happens with B. thuringensis. Nevertheless, this is a source of variability likely to apply to other invertebrate models. Worm age plays a role in susceptibility to pathogens . Age problems with C. elegans can be overcome by generating synchronous populations . Selectivity can also be present in the physical barriers at the surfaces. For instance, certain C. elegans mutants with altered body surface antigenicity or internal physical barriers differ in their susceptibility to the Microbacterium nematophilum, P. aeruginosa and S. enterica[63,79,91]. Killing efficiency of C. elegans was also strain-specific for various microorganisms: S. aureus[64,83,87], P. aeruginosa[76,90,92], E. faecalis, S. enterica, and B. pseudomallei[78,94].
The fruit fly (1–3 mm) is bigger than the roundworm but significantly smaller than caterpillars. It is normally injected on the dorsal thorax with the bacterial pathogen suspensions . Females are of larger size and naturally more resistant to the pricking than males in our hands and those of other authors [84,87]. We have observed that the oral route does not modify the lifespan of Drosophila. Nevertheless, like in C. elegans it may be selective route for certain pathogens. S. marcescens in the food kills Drosophila.
Bacterial pathogenesis experiments with the roundworm, caterpillars and fruit flies have been performed at optimal rearing temperatures ranging between 22 and 30 °C. However, expression of certain virulence factors in bacterial pathogens is temperature-regulated. For instance, during cold-shock S. aureus overproduces CspA, which has been shown to mediate protection from human AMPs . Hence, it may influence studies of bacterial pathogenesis in invertebrate models. Interestingly, the amount of CspA at conventional growth temperature conditions (37 °C) varies significantly between different Staphylococcus strains . In this regard, models able to be kept at 37 °C, like M. sexta would have advantages .
11Antimicrobial screening in invertebrate models
Another important aspect to consider is the capability of the different models to assess the efficacy of therapeutic agents. In B. mori, oral administration or injection in the midgut of vancomycin or kanamycin was ineffective against S. aureus and Stenotrophomonas maltophilia. However, vancomycin, was effective against MSSA or MRSA only if injected in the hemolymph . Chloramphenicol, tetracycline ampicillin and oxacillin were effective when provided orally [83,98]. Vancomycin does not penetrate the midgut epithelial membrane, while it was permeable to chloramphenicol . Our own work reveals that S. aureus infections in fruit flies can be cured with methicillin or tetracycline provided orally matching the ED50 obtained in mice .
12Staphylococcal infection in invertebrate animal models
To validate invertebrates as suitable models of human staphylococcal infections, there is a need to examine whether S. aureus is pathogenic in the different models and whether previously characterized virulence determinants are required for pathogenicity in those models. Wild type S. aureus kills C. elegans[64,77,99,100], D. melanogaster and B. mori[83,98], as long as it is fed to roundworms, or injected into the body cavity of fruit flies or silkworms. Half of the roundworms died in 2 days . By contrast, the lethal phenotype in fruit flies and the silkworm was more dramatic all were dead within a day [83,87]. In all three models no significant differences could be observed between the MSSA and MRSA strains used. Although, the killing efficiencies varied between different S. aureus strains, there is no evidence as to whether it is due to toxins or another components. Strain diversity or the growth medium variation mentioned above may explain the abolished or severely diminished virulence of S. aureus in some C. elegans model experiments . Sifri and co-workers report to have found such variability with a subset of clinical isolates .
The death-determining number of bacterial CFUs in Drosophila was higher for S. aureus (2 ×107–2 ×109) than for P. aeruginosa (1 ×106–4 ×107) [84,87]. Obviously, the exact time of death is unknown, and the differences in these numbers might be a by-product of the differential abilities of both pathogens for post-mortem colonization and maximum cell density at stationary phase. Alternatively, the starting infective dose may play a role; it was lower in the P. aeruginosa experiments (400–2000 CFU) than in the S. aureus experiments (1 ×104 CFU). The number of S. aureus ingested or accumulated in C. elegans has not been determined . However, for another non-persistent strain, E. faecium, the population was stably maintained in the roundworm intestine at about 1 ×105 enterococci cells per intestine. Importantly, the relevance of the infective dose is related to the volume available in the host to be colonized. For instance, the death rates provided above for the B. mori experiments, correspond to caterpillars seeded with a higher dose of staphylococcal cells 3 ×107. However, they were also distributed in a much larger volume. Decreased S. aureus inocula in B. mori or D. melanogaster resulted in extended host survival times .
C. elegans or D. melanogaster were insensitive to heat-killed S. aureus[64,87]. Live cells of S. aureus targeted the C. elegans intestine for infection . There, they accumulated, like P. aeruginosa, but unlike E. faecalis or S. enterica did not seem to proliferate and persist [76,93]. Conversely, in B. mori or in D. melanogaster, proliferation is involved in disease [83,87]. S. aureus developed a systemic infection as shown by monitoring the distribution in tissues of staphylococci fluorescently labelled with either green fluorescent protein (GFP) or by immunostaining (Fig. 2) [83,87]. Although the killing mechanism remains a mystery in any of the models, using the nomenclature introduced by Tan and co-workers  for the P. aeruginosa, C. elegans model, S. aureus carries out ‘slow-killing’, in other words not mediated by toxins but rather by growth or accumulation.
13Do invertebrates and animals share common Staphylococcal virulence factors?
More than 50 gene products in S. aureus are known factors of pathogenicity and a similar number of genes appear to encode putative candidates [101,102]. There is certain variation as to the virulence factors encoded by different species . Three studies published thus far have addressed the study of virulence determinants of S. aureus in invertebrate models [64,87,99]. Two of them evaluate the implication of known S. aureus virulence factors in C. elegans and D. melanogaster[64,87]. Bae and co-authors (2004) have taken a genome-wide approach to identify novel virulence factors in C. elegans, generating in the process an ordered library of S. aureus mutants, a highly valuable resource for staphylococcal research. Some of the virulence factors unveiled by the latter screening of S. aureus have not been previously described for S. aureus but for other species; e.g., srr in Streptococcus[99,103]. Based on a lack of function mutation analysis, Sifri and co-workers  showed that sarA (staphylococcal accessory regulator) was attenuated in C. elegans. SarA is a regulator of multiple genes required for full staphylococcal pathogenesis in animal models . SarA also modulates transcription of agr (accesory gene regulator; RNAIII), which is itself a regulator of virulent genes in S. aureus. Interestingly, S. aureus agr mutants were more highly attenuated that sarA mutants in the roundworm . Surprisingly, a sarA agr double mutant was no more attenuated than the single sarA-defective strain. Mutants in alpha-hemolysin (Hla) and V8 protease (SspA), which are also regulated by SarA and RNAIII, were found to be attenuated. An SspA-deficient strain has been shown to be attenuated in the mouse abscess model . Very recently contradictory data has been obtained, although a sarA homolog (sarH2) as well as hla were identified as virulence factors in the C. elegans model, neither agr nor sspA seemed to have a role . Similarly, our investigations showed no pathogenic effect of RNAIII, hla or sspA in the survival of fruit flies . Nonetheless, in vivo expression studies shown increased transcription from the hla and agr/RNAIII promoters fused to a lacZ reporter gene . Some virulence determinants in mice were involved in pathogenesis in Drosophila but not in C. elegans (perR, pheP), and vice versa (saeS, saeR, acnA, ftsH) [105–108]. Similarly, Drosophila and C. elegans did not recognize as virulent various other known virulence determinants in mice, e.g., fur, katA, sodA, sodM, and sigB[87,99,109,110]. Nevertheless, such variation in sigB or sodM has also been observed within different rodent models [64,109–114].
A positive correlation between bacterial virulence genes in insects and mammals has been reported previously for P. aeruginosa. The genome-wide study by Bae and co-workers  did not pick up all the genes previously shown to be involved in virulence in mice, like clfA or srtA. However, it did reveal a significant number of them, including surface proteins like capsular polysaccharide genes . Interestingly, most of the genes seemingly involved in S. aureus pathogenesis in the roundworm belong to the metabolic genes class, including ftsH, acnA, and clpP[107,108,117].
Pathogenesis studies with organisms like S. aureus that are not natural to the invertebrate host might be battled through general stress and/or taxa-specific immune exclusion responses, like encapsulation in insects, perhaps cloaking an additional inducible response analogous to the mammalian pathway. Thus, it is reasonable to speculate that some of the virulence factors affecting mammals may go underscored in invertebrate experimental screenings. It is therefore worth considering the use invertebrate species with mutations in physical barriers and physiological housekeeping genes, like the use of coronin mutants in D. discoideum to enhance pathogenicity of intracellular pathogens like L. pneumophila. On the other hand, the use of the life/death endpoint monitoring for any invertebrate model as compared to, for instance, size and appearance of a skin lesion in a mouse may intrinsically differ in their levels of sensitivity.
Invertebrate models are not directly comparable to mammals, and within the latter responses before a given pathogen and virulent factor also vary amongst different species and tissues. The consideration of bacterial virulence factors as such is subject to the response by the animal model employed, therefore being affected by innate susceptibility, route and mechanism of inoculation, pathogenic strain utilised, growth conditions, and the reasons for toxicity or lethality. In consequence, given the intrinsic variability of the systems, the application of more than one invertebrate model for a given pathogen to discern a bacterial component as being involved in virulence before moving onto higher order animal models seems an advisable option. In summary, the pathogenesis models available are valuable, especially when used in combination, in the assessment of virulence factors. As important is elucidation of the host response, which has and is continuing to yield information on conserved mechanisms.