Entomopathogenic nematode-associated microbiota: from monoxenic paradigm to pathobiome

Background The holistic view of bacterial symbiosis, incorporating both host and microbial environment, constitutes a major conceptual shift in studies deciphering host-microbe interactions. Interactions between Steinernema entomopathogenic nematodes and their bacterial symbionts, Xenorhabdus, have long been considered monoxenic two partner associations responsible for the killing of the insects and therefore widely used in insect pest biocontrol. We investigated this “monoxenic paradigm” by profiling the microbiota of infective juveniles (IJs), the soil-dwelling form responsible for transmitting Steinernema-Xenorhabdus between insect hosts in the parasitic lifecycle. Results Multigenic metabarcoding (16S and rpoB markers) showed that the bacterial community associated with laboratory-reared IJs from Steinernema carpocapsae, S. feltiae, S. glaseri and S. weiseri species consisted of several Proteobacteria. The association with Xenorhabdus was never monoxenic. We showed that the laboratory-reared IJs of S. carpocapsae bore a bacterial community composed of the core symbiont (Xenorhabdus nematophila) together with a frequently associated microbiota (FAM) consisting of about a dozen of Proteobacteria (Pseudomonas, Stenotrophomonas, Alcaligenes, Achromobacter, Pseudochrobactrum, Ochrobactrum, Brevundimonas, Deftia, etc.). We validated this set of bacteria by metabarcoding analysis on freshly sampled IJs from natural conditions. We isolated diverse bacterial taxa, validating the profile of the Steinernema FAM. We explored the functions of the FAM members potentially involved in the parasitic lifecycle of Steinernema. Two species, Pseudomonas protegens and P. chlororaphis, displayed entomopathogenic properties suggestive of a role in Steinernema virulence and membership of the Steinernema pathobiome. Conclusions Our study validates a shift from monoxenic paradigm to pathobiome view in the case of the Steinernema ecology. The microbial communities of low complexity associated with EPNs will permit future microbiota manipulation experiments to decipher overall microbiota functioning in the infectious process triggered by EPN in insects and, more generally, in EPN ecology.


Background
Host-microbe symbioses, which range from mutualistic to pathogenic interactions, are ubiquitous in both marine and terrestrial environments. The host microbiota and its variations have recently been shown to be important for the outcome of interaction between symbiotic microbes and plants [1], vertebrates [2] or invertebrates [3]. When symbiotic microbes are transmitted by macro-organism hosts, symbiont dispersion may also be influenced by the cohort of micro-organisms carried by the host [4,5]. If we are to understand host-microbe interactions, we must therefore consider the cohort of micro-organisms associated with the host.
The pathobiome is defined as the pathogenic agent together with its overall microbial environment, which contributes to pathogenesis [6]. Within the context of mammalian gut microbiota, this concept leads to the definition of several new categories on the commensal-to-pathogen spectrum: accessory pathogens, which act synergistically, under certain conditions to enhance the virulence of the main pathogenic bacteria and pathobionts, which are generally benign within the indigenous community and become pathogenic when homeostasis is disrupted [7].
Invertebrates rely on symbionts for many of their life history traits [8,9]. Moreover, their microbial communities can be analysed with methodologies that cannot be used on vertebrates. Invertebrates are therefore the favoured models for explorations of the pathobiome concept. For instance, experimental inoculations or transverse transplantations of gut microbiota in honey and bumble bees have been shown to lead to changes in susceptibility to the protozoan Lotmaria passim [10] and the trypanosomatid parasite Crithidia bombi [11]. In the lepidopteran Spodoptera littoralis, the silencing of host immune pathways by RNAi revealed that Serratia and Clostridium species in the gut microbiota switched from asymptomatic symbionts to haemocoel pathogens during Bacillus thuringiensis infection [12].
Invertebrates may themselves be parasites. The infectious process results not only from their own virulence, but also from their cohort of associated microorganisms. Many examples are provided by parasitic helminths and nematodes in particular [13]. For instance, the obligate endosymbiont Wolbachia plays a key role in the pathogenesis of the filarial nematodes responsible for lymphatic filariasis and river blindness [14]. Other emblematic nematode-bacterium associations include those between entomopathogenic nematodes Heterorhabditis and Steinernema and the ϒ-Proteobacteria Photorhabdus and Xenorhabdus, which inhabit their gut. The bacteria are involved in both insect killing and the lifecycle of the nematodes [15].
Steinernema and its intestinal symbiotic bacterium, Xenorhabdus, have long constituted a model system for investigating pathogenesis and mutualism [16][17][18]. They establish a sustainable association in lifecycles with three main steps [16][17][18]. The infective juveniles (IJs) of Steinernema are the free soil-dwelling forms of the nematode harbouring the symbiotic bacteria in their gut (step 1). On encountering a living insect larva, the IJ enters the insect gut and perforates the insect gut wall, which allows the injection of the Xenorhabdus symbiont into the insect hemocoel (body cavity) before the passage of the IJ through the insect midgut [19]. In the hemolymph (insect blood), Xenorhabdus multiplies and subsequently kills the insect (step 2). The penetration of IJs also permits insect gut microbiota translocation and proliferation into the insect hemocoel [20]. Within the insect cadaver (step 3), Xenorhabdus bacteria contribute to nematode maturation and reproduction, by degrading the insect tissues, feeding the nematode and outcompeting soil micro-organisms for nutrient acquisition in the insect cadaver. Once the nutrient resources are depleted, new IJs are produced, which specifically reassociate with their cognate Xenorhabdus symbiont and return to the soil (step 1).
The Steinernema IJs, which are the vector of Xenorhabdus between insect hosts, have a number of unusual features. They are non-feeding third-stage larvae enclosed within the second-stage cuticle, which closes the intestinal orifices (mouth and anus) and is lost only when the nematodes reach the insect gut [19,21,22]. The Xenorhabdus symbiont is carried in a dedicated part of the anterior intestine of the IJ, known as the receptacle [23][24][25]. Observations of these anatomical and histological aspects led to the idea of natural monoxeny in the Xenorhabdus-Steinernema symbiosis [26]. However, several culture-based studies have reported associations of bacteria other than the cognate Xenorhabdus symbiont with Steinernema IJs [26][27][28][29][30][31][32]. Unidentified bacteria have been observed in the intercellular space between the secondand third-stage cuticles [26].
Community profiling has progressed considerably over the last 10 years, making it possible to circumvent some of the limitations of culture-based methods (see for example [33][34][35]). These methods have been used to describe the microbiota of the free soil-dwelling nematodes Acrobeloides maximus and Caenorhabditis elegans [36][37][38]. Next-generation sequencing has recently been used to profile bacterial community dynamics in the cadavers of insect larvae infected with the entomopathogenic nematode Heterorhabditis, but the bacterial community associated with the IJs has not been extensively studied [39], and such approaches have never been used on Steinernema.
In this study, we investigated the hypothetical monoxeny of the Xenorhabdus-Steinernema symbiotic interaction. We first used a metabarcoding approach to describe the microbial community associated with several Steinernema species reared in the laboratory. We then focused on the well-characterised S. carpocapsae nematode, for which we have access to diverse strains and multiplication batches from different laboratories. We provide the first demonstration that, in this model, bacteria other than the symbiont X. nematophila are sustainably associated with S. carpocapsae IJs. We validated the pertinence of this frequently associated microbiota (FAM) by performing metabarcoding analysis on IJs freshly isolated from their natural habitat, the soil. We isolated a large range of bacterial strains from the FAM and explored some of their possible functions in the parasitic lifecycle of Steinernema. Two of the species identified, Pseudomonas protegens and Pseudomonas chlororaphis, kill insects as efficiently as the cognate Xenorhabdus symbiont, suggesting that they may belong to the IJ Steinernema pathobiome.

Results
The IJs of Steinernema are living in soils but cannot directly be isolated from their habitat. They are first captured by ex situ Galleria trap (Fig. 1a). Once isolated, they are stored and multiplied on laboratory Galleria trap (Fig. 1b).
Optimization of a metabarcoding pipeline for molecular description of the bacterial microbiota of Steinernema IJs reared in laboratory DNA was extracted from an IJ nematode sample (S. carpocapsae SK27 batch 23_08_16; Additional file 1) from our laboratory stock, from controls (Kitome_QE, Kitome_MN, Tap water and Ringer) and from Galleria larvae. Amplification and sequencing of the V3V4 region of the 16S rRNA gene generated a total of 1,907,256 reads (mean of 45,411 reads per sample; number of samples = 42). Based on rarefaction curves, an adequate sequencing depth was reached at 15,000 reads per sample (Additional file 2). α-Diversity indicators for the bacterial communities associated with the IJs, Galleria and control samples highlighted the low diversity of the bacterial communities within the different samples (Additional file 3).
The community structure observed differed significantly between IJs, Galleria and control samples (Additional file 4). The IJ sample and control samples contained mostly Proteobacteria, including many members of the Burkholderiaceae. However, only one OTU, affiliated to the Sphingomonas genus, was common to both the control and IJ samples (Fig. 2a). As previously reported in different Lepidoptera [20,39,40], the Galleria samples consisted mainly of Firmicutes, and an OTU affiliated to the genus Enterococcus was consistently detected (Additional file 4). No OTU common to the IJ and Galleria samples was identified (Fig. 2a). We therefore (i) removed the OTUs affiliated to the genus Sphingomonas from subsequent analysis and (ii) concluded that the multiplication of IJs in Galleria larvae does not influence the Steinernema microbiota. In each replicate of the IJ samples, the OTU identified as X. nematophila, the S. carpocapsae symbiont, was one of the 30 most abundant OTUs (Top30), but other OTUs were also present ( Fig. 2b), demonstrating the association of a bacterial community much more complex than the symbiont with the IJ samples.
In metabarcoding analysis based on the V3V4 region of the 16S marker, only 57 and 12% of the OTUs could be associated with a genus or species name, respectively (Additional file 5). For refinement of the taxonomic assignment of OTUs, we used the rpoB marker, a 435-bp region of the housekeeping rpoB gene [41]. The sequencing of rpoB marker amplicons generated 96,825 reads (mean = 16,137 reads per sample; number of samples = 6). Rarefaction curves generated on quality-filtered reads indicated that sequencing depth was sufficient (Additional file 2). As previously observed [41], the rpoB marker gave a better taxonomic assignment, with 56% and 32% of the OTUs associated with a genus or species name, respectively (Additional file 5). Based on these results, we pursued the two-marker approach.
Microbial community patterns segregate according to the Steinernema species Are IJ microbiota composition and richness influenced by the host species? We tested this hypothesis by describing the microbial communities associated with IJ from different Steinernema species (Fig. 3a). As an outgroup of the genus Steinernema [42], we included Heterorhabditis bacteriophora TT01, another entomopathogenic nematode harbouring symbiotic Photorhabdus luminescens bacteria. After quality filtering, we were left with a mean of 21,635 sequences per sample (n = 45) corresponding to 317 OTUs for the V3V4 marker, and 14,935 sequences per sample (n = 34) corresponding to 241 OTUs for the rpoB marker. Rarefaction curves confirmed that sequencing coverage was adequate (Additional file 2). Alpha diversity was similar for the two markers (Additional file 3). The community structure found in the various IJs ( Fig. 3b and Additional file 6) differed significantly between the outgroup H. bacteriophora and Steinernema samples (Permanova V3V4, Df = 1, R 2 = 0.28, p value = 10 −4 ; Permanova rpoB, Df = 1, R 2 = 0.38, p value = 10 −4 ), regardless of the marker considered, but also between the Steinernema species (Permanova V3V4, Df = 3, R 2 = 0.37, p value = 10 −4 ; Permanova rpoB, Df = 3, R 2 = 0.59, p value = 10 −4 ). As previously shown for S. carpocapsae SK27_batch 23_08_ 16, Proteobacteria were prominent members of the IJ microbiota ( Fig. 3c and Additional file 6). Microbial patterns seem to be correlated to the nematode host's phylogenic tree (Fig. 3a), suggesting co-evolution of the microbiome and its host. Further analyses including a larger number of strains would be necessary to confirm this phylosymbiosis signal.
Despite these variations and regardless of the marker used, we detected recurrent OTUs in addition to the cognate symbiont X. nematophila among the 30 most abundant OTUs of the whole set of S. carpocapsae Soil was sampled at a depth of 0-20 cm and transferred to 1 L plastic containers. Five last-instar Galleria mellonella larvae were placed in each container (ex situ Galleria trap). The containers were stored in the dark at 18°C. After 7 days, dead G. mellonella were placed in a White trap consisting of an overturned 35-mm Petri dish covered with a piece of material (in blue) and placed in a larger Petri dish containing Ringer solution. After the production of a few generations of nematodes in G. mellonella cadavers, the IJs left the cadaver and migrated into the Ringer solution via the wet material (black dashes in Ringer solution). b IJ storage and laboratory Galleria trap method. IJs emerging from Steinernema and Heterorhabditis species were stored in 250-mL flasks containing 80 mL of Ringer solution supplemented with 0.1% formaldehyde at 9°C, and flasks containing 80 mL of Ringer solution supplemented with 0.01% formaldehyde at 15°C, respectively. Every 6 months, stocks were multiplied by adding 50-100 IJs to Galleria larvae placed on filter paper in a Petri dish (laboratory Galleria trap). Once the insects had died, their cadavers were placed in a White trap, as described above. Multiplication batches were labelled with the infestation date samples (Additional file 8). For each abundant OTU (read number > 0.1% of the sample read), we represented the frequency of abundance as a function of frequency (Fig. 5). For OTUs with a high occurrence (present in at least 70% of the samples), we identified two groups. In the first group, the X. nematophila OTU was always abundant (present in at least 90% of the samples), confirming its status as the core symbiont of S. carpocapsae (Fig. 5). In the second group, the OTUs were abundant in less than 90% of the samples. We named this set of OTUs the "frequentlyassociated microbiota" (FAM). After removal of the OTU found in the controls (Sphingomonas), the FAM common to the two markers encompassed the genera Alcaligenes, Stenotrophomonas, Pseudomonas and the Rhizobiaceae family (Fig. 5).
We checked for the FAM of laboratory-reared S. carpocapsae in the IJs of S. carpocapsae GRAB recently sampled from soils by the ex situ Galleria trap method (Fig. 1a) in an apple orchard in November 2017 (Additional file 1). Metabarcoding analysis with both V3V4 and the rpoB markers was performed on GRAB IJs shortly after trapping (only one round of multiplication in Galleria larvae). Bacterial diversity was higher than that in nematode lineages multiplied for long periods in the laboratory, but both markers identified the core symbiont Xenorhabdus nematophila and the FAM members Stenotrophomonas, Pseudomonas and Rhizobiaceae, but not Alcaligenes, suggesting that the identification of this last taxon might be due to a bias caused by laboratory rearing (Additional file 9).
Bacteria associated with S. carpocapsae IJs are cultivable We performed a large survey of the isolation on nutrient-rich culture media of bacteria from ground S. carpocapsae IJs or from the content of G. mellonella cadavers infested with S. carpocasae IJ. We isolated only Gram-negative taxa. We regularly sampled isolates from the following eight genera: Brevundimonas, Ochrobactrum, Pseudochrobactrum, Achromobacter, Alcaligenes, Stenotrophomonas, Xenorhabdus and Pseudomonas (Additional file 1 and Fig. 6). These taxonomic groups matched with the OTUs identified by metabarcoding in 70% of the S. carpocapsae samples with one of the two markers (Fig. 5). Bacterial isolation on rich agar medium cannot be used for the quantification of taxa. However, for some samples, Xenorhabdus colonies were less abundant on NBTA plates than the other taxa, such as Stenotrophomonas, Alcaligenes and Pseudomonas. Culture-dependent approaches thus validate the S. carpocapsae FAM identified by the molecular approach.  The ability of S. carpocapsae to kill insects is thought to be due to the presence of the symbiont X. nematophila [18]. However, other bacteria associated with the S. carpocapsae IJs might also contribute to entomopathogenicity. We assessed the pathogenicity of several of the bacteria frequently associated with the S. carpocapsae IJs. When 10 2 -10 3 CFU/mL were injected into larvae of the lepidopteran insect Spodoptera littoralis, the highly pathogenic symbiont Xenorhabdus nematophila XnSc_F1 and Pseudomonas protegens strain PpSc_PP-SC-10, both isolated from S. carpocapsae SK27, took similar time to kill 50% of the larvae (LT 50 ): 27 and 29 h, respectively (S. carpocapsae microbiota in Table 1). The larvae were not killed by other bacteria from the S. carpocapsae FAM (S. maltophilia, Alcaligenes faecalis, Pseudochrobactrum, Ochrobactrum), even at doses of 10 5 -10 6 CFU/mL.
It is also generally asserted that X. nematophila outcompetes other bacterial strains or species during its lifecycle, particularly during the infestation phase of insect cadavers [20,45]. We therefore monitored the in vitro antibiosis of the X. nematophila XnSc_F1 strain against the cultivable bacterial members of the S. carpocapsae microbiota. Only two of the six bacteria isolated (Alcaligenes faecalis Alcf_SC and Ochrobactrum sp. OchSc_ALL4) displayed slight susceptibility to the production of antimicrobials by X. nematophila XnSc_F1 (Additional file 10). By contrast, five of these six bacteria were susceptible to the antimicrobial production of P. protegens PpSc_PP-SC-10 (Additional file 10). During this study, we also recovered a panel of bacterial strains from IJs of three other entomopathogenic nematodes: S. glaseri SK39 and S. weiseri 583 that carried the symbionts X. poinarii XpSg_G6 and X. bovienii XbSw_ CS03, respectively, and H. bacteriophora TT01 that carried the symbiont the symbiont P. luminescens PholHb_ TT01. X. poinarii XpSg_G6 and X. bovienii XbSw_CS03 were previously reported to be attenuated in virulence [46][47][48], while P. luminescens PholHb_TT01 is highly virulent [49]. As in S. carpocapase, bacterial isolates from Steinernema IJs were generally more susceptible to the antimicrobials produced by P. protegens or P. chlororaphis than those produced by Xenorhabdus (Additional file 10). We assessed the pathogenicity of these bacterial isolates by injection in S. littoralis larvae (Table  1). For isolates from S. glaseri SK39 and S. weiseri 583, only P. protegens and Pseudomonas chlororaphis strains were able to kill the insect larvae (Table 1). They had LT 50 values between 19 and 29 h, in the same range as the values obtained with reference strains isolated from rhizosphere environments, P. protegens CHAO T and P. chlororaphis CFBP2132 T [50][51][52][53] (Table 1 and Fig. 7).

Discussion
This exploration of the monoxenic paradigm for the Xenorhabdus-Steinernema symbiotic interaction provided us with an opportunity to explore, for the first time, the bacterial microbiota associated with entomopathogenic nematodes of the genus Steinernema. To this end, we used two approaches to analyse a large panel of Occurrence (x axis) and frequency of high abundance (y axis) of OTUs associated with laboratory-reared S. carpocapsae IJs. Abundant OTUs with read numbers accounting for more than 0.1% of the total reads for the sample were plotted. The darker the dot, the more OTUs with these characteristics are found at the position concerned. Among the OTUs with high occurrence rates (present in more than 70% of the samples), the yellow and orange areas delimit two groups. Orange group: frequency of high abundance > 90% (core symbiont). Yellow group: frequency of high abundance < 90% (frequently associated microbiota or FAM). OTU identity is indicated to the right of the graph (bootstrap confidence ≥ 0.9). *OTU present in the control samples Steinernema infective stages. We first molecularly characterised the microbiota of four Steinernema species with two markers, the V3V4 region of the 16S rRNA gene and a 435 bp region of the rpoB gene, because multigene approaches have been shown to improve the reliability of bacterial community description in ecosystems of low complexity [41]. We also characterised the Steinernema microbiota by culture-based methods. This approach differed from that used in previous culture-dependent studies in the considerable effort made to isolate a large set of colonies (71, selected from more than 350 isolates, in this study) and to perform molecular taxonomic characterization on these colonies.
In the previous culture-dependent studies, questions were frequently raised about the source of the other bacteria than the symbiont found in Steinernema IJs. The authors often concluded that these taxa were contaminants [26], except for Ochrobactrum strains that were considered to be involved in a dixenic association with Photorhabdus in the EPN Heterorhabditis [54]. Indeed, as the isolated taxa are frequently detected in the healthy insect gut, in soils and in the haemolymph of insects infected with Steinernema [20,32,55,56], it is generally assumed that the non-symbiotic bacteria randomly "hitchhike" in IJ vectors via the cuticle or intercuticular space and are introduced in the insect haemocoel during IJ penetration [27,29].  6 Distance phylogenetic tree of 62 bacterial isolates from S. carpocapsae IJs. The phylogenetic tree of taxa isolated from S. carpocapsae IJs was constructed with the 16S rRNA gene sequences (1377 nucleotides), with the Kimura two-parameter model [43] and the neighbour-joining method [44] included in SeaView 4.7 software. Bootstrap values (percentages of 1000 replicates) of more than 90% are shown at the nodes. Twelve type strains (in bold) of the Xenorhabdus, Pseudomonas, Stenotrophomonas, Alcaligenes, Ochrobactrum, Pseudochrobactrum, Achromobacter and Brevundimonas genera were added. The bar represents 1% sequence divergence Here, we clearly described the microbiota associated with Steinernema carpocapsae reared in laboratory as composed of the core symbiont, Xenorhabdus, and a frequently associated microbiota (FAM) consisting of cultivable Proteobacteria also identified in IJs freshly isolated from soil, their natural habitat. The only exception was Alcaligenes, which was detected only in laboratory-reared IJs. A large number of reads attributed to Alcaligenes was generally correlated with a high level of abnormal morphology (Ogier and Pagès, unpublished data). Poinar also described morphological and behavioural abnormalities in the nematode Heterorhabditis heliothis when IJs developed in presence of Alcaligenes faecalis [57], suggesting that a high abundance of this genus reflects drift due to laboratory rearing in batches of sick nematodes.
The Proteobacteria of the FAM are frequently encountered in the microbiota of organisms or compartments associated with soils. For instance, we detected taxa previously identified by culture-independent studies on other parasitic or soil nematodes. Pseudomonas has been detected and isolated from soil-dwelling forms of the nematode Pristionchus, a parasite of scarab beetles [58]. Stenotrophomonas, Ochrobactrum and Pseudomonas have repeatedly been identified in the soil-dwelling Acrobeloides maximus and Caenorhabditis elegans nematodes [36][37][38]. Likewise, Proteobacteria is the most abundant phylum in bacterial communities associated with plant roots [59,60] as well as in plant-covered soils, such as the rhizosphere or litterfall [61,62].
The composition of the Steinernema microbiota seems to be shaped by the ecological niche, but the question remains as to whether the establishment of the association between the bacterial community and the nematode is due to the chance alone or to specific mechanisms of association. In this study, we provide an array of arguments in favour of specific association with a larger microbiota than the core symbiont alone. First, Proteobacteria frequently encountered in the microbiota of organisms or compartments associated with soils were identified in nematodes that had been reared in the laboratory for years, suggesting that the nematodemicrobiota association is robust enough to be sustained in non-natural (laboratory rearing in the non-natural host Galleria) and very different (France and USA) environments. Second, microbiota profiles distinguished between the genera Heterorhabditis and Steinernema, and also between the four Steinernema species studied here. Further studies are required to determine whether IJ microbiota composition actually reflects the entomopathogenic species phylogeny, as demonstrated for the gut bacteria of the wasp Nasonia [63], for coralassociated bacteria [64] and for insect viromes [65]. However, these preliminary data support the notion that some members of the Steinernema microbiota may be vertically (or pseudovertically) transmitted to the progeny, as reported for the core symbiont Xenorhabdus [66]. This would not rule out the possibility of hybrid transmission, with more "accessory" microbiota members transmitted horizontally, as reported for a nematode colonising the vertebrate gastrointestinal tract, Haemonchus contortus [13]. The description of an IJ-associated microbiota for S. carpocapsae and other Steinernema species is important because it improves our understanding of the ecological functioning of EPN. Indeed, the killing efficiency of nematodes was long thought to depend principally on the action of Xenorhabdus [16][17][18]. However, several studies have cast doubt on this view, because some entomopathogenic pairs have been found to have attenuated virulence or to be non-virulent when injected alone into insects [46-48, 67, 68]. These previous findings and the results presented here call into question the notion of the "pathobiome" in the case of the EPN ecology. Pseudomonas protegens and Pseudomonas chlororaphis, two rhizospheric and entomopathogenic species from this genus [69], are two good candidate members of the IJ pathobiome. An entomopathogenic Pseudomonas fluorescens strain has already been isolated by Lysenko and Weiser from S. carpocapsae IJs [28]. At the time P. fluorescens taxonomy was far from fully elucidated, but we assume that the authors actually identified a strain of P. protegens or P. chlororaphis.
The EPN lifecycle is not restricted to the killing of the insect. The nematode completes its sexual reproduction within the killed insect, then re-associates specifically with the symbiont, and possibly other members of its microbiota, after which it survives in the soils [16][17][18]. If P. protegens or P. chlororaphis belong to the IJassociated microbiota, how do they interact with the other members of the IJ microbiota? An assessment of in vitro antimicrobial activity on agar plates showed that the Xenorhabdus strains have weak antimicrobial activity against bacterial members of Steinernema microbiota, whereas P. protegens and P. chlororaphis strains displayed higher levels of antimicrobial activity against both Xenorhabdus strains and other bacterial members of the Steinernema microbiota. If antibiotic activity measured in vitro accurately reflects the antibiotic activity in vivo, it suggests that antimicrobial production from species other than Xenorhabdus may be involved in in vivo interspecies competition. These data also suggest that the various members of the Steinernema microbiota might be subject to spatial compartmentation.

Conclusions
Finally, increasing numbers of scientists in the community working on microbial ecology are stressing the importance of working on natural microbial communities of low genotypic, functional or environmental complexity for the purposes of experimentation and theoretical modelling [70]. We are convinced that the microbial communities associated with EPNs meet this need perfectly. Three main questions could be explored in future studies of EPN models. Where are the members of this frequently IJ-associated microbiota located? How are these bacteria transmitted between EPN generations? What impact does each member of the microbiota have on entomopathogenic nematode fitness, survival and the capacity for adaptation in changing environments (soils, insects, etc.)?

Biological material
All the entomopathogenic nematodes (EPN) strains used here are listed in Additional file 1. Bacteria isolated from EPNs or EPN-infested Galleria cadavers and used for pathology and/or antibiosis assays are listed in Additional file 11. Bacteria used for the construction of the phylogenetic distance tree are listed in Additional file 12. Galleria mellonella (Lepidoptera: Pyralidae) insect larvae were reared at 28°C in the dark, on honey and pollen. Spodoptera littoralis (Lepidoptera: Noctuidae) insect larvae were reared on an artificial diet [71] at 23 ± 1°C, with a photoperiod of 16 h of light and 8 h of darkness (L16:D8) and relative humidity (RH) of 40 ± 5%.

Isolation, multiplication and storage of EPNs
EPN were isolated by ex situ Galleria trap (Fig. 1a) as previously described [72]. We used laboratoryreared EPN stocks, except S. carpocapsae GRAB, which was isolated in November 2017 from an apple orchard near of Avignon, France (43°54′20.4″ N 4°53′09.0″ E). Briefly, soil samples were collected at a depth of 0-20 cm (temperature 5.5 to 8.2°C; RH 15.2 to 21.4%) and placed in 1-L plastic containers containing five last-instar Galleria larvae, which were then stored in the dark at 18°C for seven days. The Galleria cadavers were placed on White traps (Fig.  1a). IJs emerging from cadavers and their cognate symbiont were identified molecularly (see below for details) as S. carpocapsae (100% identity for the ITS1-2 region) and X. nematophila (100% identity for the 16S rRNA), respectively. IJs were stored in Ringer's solution (Merck). Laboratory-reared EPN stocks were multiplied every six months by infestation of last instar of Galleria as previously described [73]. Briefly, the EPNs were placed in contact with the Galleria larvae in Petri dishes. When the Galleria larvae died, the cadavers were placed on a White trap (Fig. 1b). Emerging IJs of Steinernema and Heterorhabditis species were stored in Ringer's solution at 9°C and 15°C, respectively.

Isolation, culture and storage of bacteria isolated from EPNs
The bacteria were isolated by the hanging drop technique [23] (for Xenorhabdus and Photorhabdus only), from content of G. mellonella cadavers after IJ infestation or by crushing IJs. For the crushing method, we placed 20 IJs in a 1.5-mL Eppendorf tube containing 200 μL of LB broth and three 3-mm glass beads and subjected them to three cycles of grinding (1 min, 30 Hz, followed by 1 min without agitation) in a TissueLyser II apparatus (Qiagen, France). Bacteria were isolated on nutrient agar (Difco) plates or on nutrient bromothymol blue agar (NBTA) plates [74] and incubated at 28°C for 48 h. Bacteria were also routinely grown at 28°C in Luria-Bertani (LB) broth. The bacteria were stored at − 80°C with 16% glycerol (v/v).

Pathogenicity assays
Bacterial pathogenicity was assessed by injection into S. littoralis, as previously described [77]. Briefly, bacterial cultures in LB broth (#OD = 0.8) were diluted in the culture medium and 20 μL of the resulting bacterial suspension, containing 10 3 to 10 6 colonyforming units (CFUs), was injected into the haemolymph of 20 fifth-instar larvae of S. littoralis. The number of bacterial cells injected into the larvae was determined by plating on nutrient agar and counting the CFUs. After the bacterial injection, the insect larvae were incubated at 23°C and mortality was monitored for up to 72 h. The pathogenicity of bacterial isolates was determined by measuring the time required for 50% of the insect larvae to be killed (LT50).

Antimicrobial activity
The antimicrobial activity of the Xenorhabdus, Pseudomonas protegens and Pseudomonas chlororaphis strains against other members of the EPN microbiota was assessed in vitro as previously described [74]. Briefly, 5 μL of an exponential growing culture of the producer strain was spotted onto Mueller-Hinton agar plates (Biokar). The producer strain was grown at 28°C for 48 h and killed by exposure to chloroform for 30 min, followed by air drying for 15 min. Indicator strains were diluted (1.5% overnight culture) in 10 mL of top agar (LB broth; 0.7% agar) and poured over the producer strain plates to form an overlay. Plates were incubated for 24 h at 28°C for all indicator strains and at 37°C for Micrococcus luteus, which was used as a positive control. The diameter of the zone of inhibition was measured in millimetres. The assays were performed three times.

DNA extractions from IJ
We sampled 5000 IJs from a storage batch (Fig. 1b). In preliminary protocol tests, we treated nematodes with bleach solutions of different concentrations (from 2% to 10%) as previously performed [19,32], but microscope observation showed that it resulted in the destruction of the external cuticle in many IJs. To avoid the alteration of the IJ integrity and consequently the bias in their microbiota composition, we decided to minimize contamination with micro-organisms from the body surface by rinsing the IJs thoroughly with tap water, on a filter. The washed IJs were recovered from the filter with a sterile pipette, transferred to 10 mL of sterile ultrapure water and immediately frozen at − 80°C for future use. We compared several different DNA preparation methods, because EPNs are resistant to lysis because of their double cuticle. DNA recovery rates were highest for mechanical grinding of IJs with three 3-mm glass beads (three cycles of 30 Hz for 2 min, followed by 1 min without agitation) in a TissueLyser II apparatus (Qiagen, France) followed by DNA extraction with the Quick Extract kit from Epi-centre, USA (QE method) or with the Tissue extraction kit from Macherey-Nagel, Hoerdt, France (MN method). Briefly, for the QE method, frozen samples were rapidly thawed, heated at 80°C for 20 min and centrifuged (3.000×g, 10 min) to collect the IJs as a pellet. The IJ pellet was placed in 200 μL of QE lysis buffer and ground. We then added 2 μL of Ready-Lyse Lysozyme Solution (Epi-centre, USA) and incubated the samples at room temperature until the solution cleared (48 to 72 h). Samples were again heated at 80°C for 20 min, and IJs lysis was checked under a light microscope. When lysis was complete, the sample was treated with 20 μL of RNaseA 20 mg/ml (Invitrogen PureLinkTM RNaseA, France). It was then subjected to phenolchloroform extraction followed by chloroform extraction alone. DNA was precipitated in 70% ethanol, resuspended in 50 μl ultrapure water and stored at − 20°C. The MN method was performed according to the kit manufacturer's instructions, except that the column step was replaced by phenol-chloroform and chloroform extractions, as described above. We assessed contaminant bacterial DNA levels, by generating several control DNA preparations: in ultrapure water with the QE method (Kitome_QE control), in ultrapure water with the MN method (Kitome_MN control), in the tap water used during the IJ washing step, with the QE method (Tap water control) and in the Ringer medium used for IJ storage, with the QE method (Ringer control). For each sample, three to six technical replicates were prepared.
DNA extraction from G. mellonella larvae As IJ multiply within Galleria larvae, we also extracted DNA from insect larvae. Three G. mellonella larvae were surface-sterilised with 70% (vol/vol) ethanol, were placed in a pot with a 2.5 mm tungsten bead and 1.5 mL PBS buffer and were crushed by three grinding cycles (1 min at 30 Hz followed by 1 min without agitation) in the Tis-sueLyzer II apparatus. The homogenate was centrifuged (400×g, 2 min) to remove larval debris and the supernatant was immediately frozen at − 80°C until use. Samples were rapidly thawed, heated at 80°C for 20 min and split into five fractions. The QE method (see the procedure for IJs described above) was used for the preparation of independent DNA extracts from each technical replicate.

Library preparation and sequencing
Amplicon libraries were constructed following two rounds of PCR amplification. The first amplification step was performed on 10 to 100 ng of DNA, with iProof TM DNA Polymerase (Bio-Rad), in a Bio-Rad thermocycler. The hyper variable V3V4 region of the 16S rRNA gene was targeted with the universal primers F343 (5′-CT TTCCCTACACGACGCTCTTCCGATCTTACGGRAG GCAGCAG-3′) and R784 (5′-GGAGTTCAGACGTG TGCTCTTCCGATCTTACCAGGGTATCTAATCCT-3′), and the rpoB fragment was targeted with the primers Univ_rpoB_F_deg (5′-CTTTCCCTACACGA CGCTCTTCCGATCTGGYTWYGAAGTNCGHGAC GTDCA-3′) and Univ_rpoB_R_deg (5′-GGAGTTCA GACGTGTGCTCTTCCGATCTTGACGYTGCATGT TBGMRCCCATMA-3′) [41]. The Illumina adapters are indicated in non-bold characters. We performed 30 amplification cycles with annealing temperatures of 65°C and 60°C for the V3-V4 region (~460 bases), and the rpoB region (~435 bp), respectively. Single multiplexing was performed at the Genomic and Transcriptomic Platform (INRA, Toulouse, France), with 6-bp index sequences, which were added to the reverse primers during a second 12-cycle PCR. Amplicon libraries were sequenced with Illumina MiSeq technology (MiSeq Reagent Kit v2) according to the manufacturer's instructions. FastQ files were generated at the end of the run. The program FastQC [78] was used for quality control checks on raw sequence data and the quality of the run was internally checked by adding 30% of PhiX sequences at 12.5pM. Each pair-end sequence was assigned to its sample with the help of the previously integrated index, and pair-end reads were assembled with FLASH [79]. Unassembled reads were discarded.

Sequence data processing and analyses
The sequencing reads obtained were processed according to the FROGs pipeline [80]. A pre-processing tool was used to remove sequences that did not contain both primers, to trim the primers and to remove all sequences containing an ambiguous base. Sequence clustering was performed with the Swarm algorithm [81]. Chimeric sequences were detected with the VSEARCH algorithm, by the de novo UCHIME method [82,83] and were removed. A filtering tool was used to remove clusters, which had a read number abundance of less than 0.005 % of all reads. Sequences were assigned with RDP Classifier [84] and the 16S rRNA database Silva [85] for V3V4 reads. For sequence assignment with the rpoB marker, we constructed a reference database including 45,000 sequences; this database is available from the FROGs website (http://frogs.toulouse.inra.fr/).

Phylogenetic analysis
Phylogenetic analysis of the ITS sequences and of the V3V4 and rpoB amplicons, sequence alignment and maximum likelihood analysis with the GTR model were performed as previously described [47]. The Mega7 tool [89] was used to generate a circular phylogenetic tree. Phylogenetic analysis of the 16S rRNA genes from 47 bacterial isolates from S. carpocapsae IJs, sequence alignment and neighbour-joining analysis with the Kimura two-parameter model were performed as described by Tailliez et al. [90].
or the V3V4 region of the 16S rRNA gene were inferred with SEAVIEW 4.0 [91], using a PhyML-based maximum likelihood algorithm [92] and the GTR model. The sum of read numbers for the six replicates is indicated after the OTU name. Only abundant OTUs with read numbers accounting for more than 0. 1% of the reads for the sample in at least one replicate are included in the phylogenetic tree. OTUs belonging to the same bacterial genus are shown in the same colour. Additional file 10. Antimicrobial activities of Xenorhabdus (a) and Pseudomonas (b) strains against bacterial taxa from the IJ-associated microbiota of S. carpocapsae SK27, S. glaseri SK39 and S. weiseri 583.
Additional file 11. Bacterial strains isolated in the study and used in pathology and/or antibiosis assays.
Additional file 12. List of the 62 bacterial strains, including 50 bacterial isolates from S.carpocapsae IJs, used in this study to build the 16S rRNA gene phylogenetic tree.
Additional file 13. a. List of OTUs obtained with the V3V4 marker. b. List of OTUs obtained with rpoB marker. c. Venn diagram table detailing compositions of core and non-core OTUs of Figure 2a. d. List and frequency of OTUs associated with the S. carpocapsae IJs obtained with the V3V4 marker for which abundance in the whole dataset exceeded 0.1% (plotted in Figure 5.b). e. List and frequency of OTUs associated with S. carpocapsae IJs obtained with the rpoB marker for which abundance in the whole dataset exceeded 0.1% (plotted in Figure 5.a).