Long-Term Persistence of Yersinia pseudotuberculosis in Entomopathogenic Nematodes

Entomopathogenic nematodes (EPNs) are small worms whose ecological behaviour consists to invade, kill insects and feed on their cadavers thanks to a species-specific symbiotic bacterium belonging to any of the genera Xenorhabdus or Photorhabdus hosted in the gastro-intestinal tract of EPNs. The symbiont provides a number of biological functions that are essential for its EPN host including the production of entomotoxins, of enzymes able to degrade the insect constitutive macromolecules and of antimicrobial compounds able to prevent the growth of competitors in the insect cadaver. The question addressed in this study was to investigate whether a mammalian pathogen taxonomically related to Xenorhabdus was able to substitute for or “hijack” the symbiotic relationship associating Xenorhabdus and Steinernema EPNs. To deal with this question, a laboratory experimental model was developed consisting in Galleria mellonella insect larvae, Steinernema EPNs with or without their natural Xenorhabdus symbiont and Yersinia pseudotuberculosis brought artificially either in the gut of EPNs or in the haemocoel of the insect larva prior to infection. The developed model demonstrated the capacity of EPNs to act as an efficient reservoir ensuring exponential multiplication, maintenance and dissemination of Y. pseudotuberculosis.


Introduction
Entomopathogenic nematodes (EPNs) are microscopic soil worms exclusively feeding on insect preys. They have the ability to cause death in a huge variety of insects, making them powerful candidate biopesticides in agriculture and horticulture [1,2]. EPNs owe their insecticidal properties to symbiotic bacteria belonging to two genera of Enterobacteriaceae, namely Xenorhabdus and Photorhabdus. These bacteria are hosted in the gastro-intestinal tract of the nematode-located in an intestinal vesicle in the case of Xenorhabdus [3] at the infectious free-living stage, called infective juveniles (IJs). Upon invasion of an insect prey, the symbiotic bacteria are expelled from the IJ's digestive tract. These bacteria multiply in the insect haemocoel and release insecticidal toxins as well as degradative enzymes able to digest the insect macromolecules, thereby feeding their EPN partners which mature to the adult stage through 4 larval stages named J1 to J4 and undergo several reproduction cycles [4]. Moreover, the symbiont prevents microbial competitors growth inside the insect's cadaver by releasing antibiotic and antifungal compounds [5]. After all insect macromolecules have been exhausted, a few symbiotic bacteria enter the intestinal tract of the mature IJs just before they emerge from the dead larva and seek another prey [6,7].
While a few EPNs are generally sufficient to kill an insect prey, up to half a million of IJs can emerge from a single infected host upon completion of their reproductive life cycle inside the insect cadaver [8]. Each of these freshly emerged IJs is able to infect a new insect prey. IJs can survive in the soil for several months thanks to their protective cuticle and a huge lipid supply they can store [9].
In 2008 Heermann and Fuchs have shown that Photorhabdus luminescens, the bacterial symbiont of Heterorhabditis bacteriophora, shares a number of unique genes with the taxonomically related, yet ecologically different, Yersinia enterocolitica [10]. The shared genes are for some of them clustered in so-called "High Pathogenicity Islands" described in Enterobacteriaceae including Yersinia [11]. Many of these genes are either involved in pathogenicity toward insects, like the insecticidal toxin complex (Tc) [12], or in colonisation of eukaryotic cells, like the YplA phospholipase [13]. The recently discovered type 6 secretion system (T6SS) involved in toxin secretion and in mutualism between bacteria [14] is also conserved between P. luminescens and Y. enterocolitica [10]. Unlike P. luminescens which can cause casual infection in humans [15], Y. enterocolitica as well as Y. pseudotuberculosis are mammalian pathogens causing gastrointestinal diseases in infected hosts. These two Yersiniae are regularly isolated from meatespecially pork meat-and root vegetables [16,17]. However, they have also been found in the gut lumen of adult flies and fly larvae, suggesting that they can use insects as passive vectors [18,19,20]. In addition, in vitro experiments have shown that both Y. enterocolitica and Y. pseudotuberculosis are able to colonize insect cells [21] and even to kill insect larvae like Galleria mellonella [22]. It is well known that Yersinia pestis, the third mammalian pathogenic Yersinia and etiological agent of plague, is able to colonize insects since it uses fleas as vectors. Hinnebusch et al. demonstrated the essential implication of the Yersinia murine toxin (Ymt) in flea colonisation [23]. Interestingly, it has been suggested that Y. pestis acquired ymt gene from P. luminescens or from a close relative [24]. Moreover, Photorhabdus asymbiotica which can infect either insects or humans, possesses a plasmid related to pMT-1 found in Y. pestis [25].
Besides their animal hosts, Y. enterocolitica as well as Y. pseudotuberculosis are commonly found in water, soil and vegetables [16,17,26]. Several studies have shown that Y. pestis can also be found in soil [27,28]. Moreover, several experiments have highlighted the survival of Y. enterocolitica, Y. pseudotuberculosis and Y. pestis in free living soil amoeba [29,30]. Since pathogenic Yersiniae are able to persist in soil and are phylogenetically very close to the bacterial symbionts of EPNs, we wondered whether Yersiniae would be able to intrude the symbiotic relationship associating EPNs and their natural symbiont. In order to test this hypothesis, we used an experimental model consisting of insect larvae of the species Galleria mellonella used as prey for an African species of entomopathogenic Steinernema hosting its natural Xenorhabdus symbiont as well as a Y. pseudotuberculosis field isolate naturally resistant to the anti-microbial compounds produced by Xenorhabdus. We show that Y. pseudotuberculosis can be successfully transmitted by the EPN carrier inside an insect larva in which it persists and multiplies. Moreover, EPNs emerging from the insect cadaver after 10 to 15 days where found to host large numbers of Y. pseudotuberculosis cells in their gastro-intestinal tract. These EPNs were in turn able to transmit Y. pseudotuberculosis to a new insect larva and so on for at least 7 successive infectious cycles (14 weeks). If they turn out to have an ecological significance, these findings may reveal an unexpected biotic reservoir for the long-term persistence and dissemination of pathogenic Yersiniae in the environment.

Bacterial strains, plasmids and growth conditions
Enterobacteriaceae were grown in LB liquid broth with strong agitation (150rpm) or on LB agar or McConkey agar plates. The NBTA plates (nutrient agar supplemented with 25 mg l −1 bromothymol blue and 40 mg l −1 triphenyltetrazolium chloride) [31] were also used to check the phase I of the Xenorhabdus species used in this study. The incubation temperature was 37°C except for Yersiniae and Xenorhabdus sp. which were grown at 28°C. Antibiotics were added at the following concentrations: Kanamycin (Km): 30μg ml −1 , Nalidixic acid (Nal): 25μg ml −1 , Streptomycin (Sm): 50μg ml −1 ; Ampicillin (Ap): 100μg ml −1 . Nalidixic-acid resistant (Nal R ) bacteria were obtained in three consecutive steps by plating 10 7 to 10 9 CFU per agar plate supplemented with increasing concentrations of Nalidixic acid (5μg/ml; 20μg/ml; 50μg/ml). Bacterial strains used in this study are listed in Table 1.
Some Enterobacteria, listed in table 1, were fluorescently labelled with GFP-mut2 [32] using a mini-Tn5 transposon [33]. Mini-transposon labelling was conducted by conjugating a nalidixic-acid resistant variant of the target bacterium with E. coli S17/1 λ pir hosting a transposon delivery suicide vector [34]. Transconjugants were isolated on selective agar plates and tested for GFP fluorescence. Integration of gfp-mut2 was further confirmed by PCR with primers mut2-GFP_F (GGG ATC TTT CGA AAG GGC AGA TTG TGT GG) and mut2-GFP_R (GGA GAG GGT GAA GGT GAT GCA ACA TAC GG). The size of the amplified fragment was 543 bp. For dual labelling experiments, the gfp-mut2 gene of Y. pseudotuberculosis 4N1G was substituted for mCherry, encoding a red-fluorescent protein, by allelic exchange. The replacement cassette consisted in mCherry flanked by the beginning and the end of the gfp-mut2 nucleotide sequence. The upstream and downstream flanking parts consisted of 244 and 223 base pairs of gfp-mut2, respectively, obtained by PCR amplification. A ribosome binding site was added upstream of the mCherry open reading frame to ensure optimal translation. The replacement cassette was cloned into the mobilizable suicide vector pKNG101, which confers resistance to streptomycin and carries the counter-selectable marker sacBR [35]. The recombinant suicide plasmid termed pSGCG was then transferred to Y. pseudotuberculosis 4N1G from the conjugative strain E. coli SM10λpir. Allelic exchange was conducted in two steps. Initial integration of pSGCG was first selected on specific agar plates containing streptomycin (100μg ml −1 ). After purification of the recombinant strain, allelic exchange was selected on agar plates containing sucrose (100μg ml −1 ) and recombinant colonies expressing mCherry but not GFP were controlled by both epifluorescence microscopy and PCR.

Mini-Tn5 transposon insertion mapping
The mini-Tn5 transposon used here to tag the Yersinia pseudotuberculosis derivatives 4N1G and 4N1C was mapped by TAIL-PCR using the method of Liu and Wittier [36] and by sequencing the amplified fragment. The mini-Tn5-gfp was found inserted in the chromosome at codon 60 of the fimbrial A protein gene in the same transcriptional orientation (ORF YPK_0694 as described in the annotated genome of Y. pseudotuberculosis YPIII, Accession number NC_010465.1). The transposon-specific primers used for TAIL-PCR were the following: SP1: CGC GAA AGT AGT GAC AAG TGT TGG CCA TGG; SP2: GTA TAA CAT GTC TTA TAC GCC CGT GTC AAC C; SP3: AGA TCC CCG GGT ACC GAG CTC GAA TTC GCG. The arbitrary degenerated primers used here were the same described by Liu and Whittier [36]. Final confirmation of the insertion point of the mini-Tn5 transposon was obtained by amplifying chromosomal fragments covering part of the transposon and part of the fimbrial A protein gene using the PCR primers SP1, SP2 or SP3 together with the fimbrialspecific primer CCG GTT CTA TCA TTG AAG CAC CTT GTT C. Infection of the larvae was conducted by incubating 500 to 1000 Steinernema sp. MW8B IJs suspended in 1ml physiological water (NaCl 9g L −1 ) with four to six larvae confined in a closed Petri dish. Upon emergence from the dead larvae which occurred after 10 ± 2 days later, IJs were collected and stored at room temperature in physiological water.

Galleria mellonella in vitro infection model
For the first infection cycle, 6 G. mellonella larvae were injected with 10 6 CFU of the studied bacterium (Yersinia sp., GFP-labelled or not) using sterile 1-ml syringes bearing 0.

Microscopic observations
A 10-μl suspension of infective juveniles (IJs) was observed on a microscope slide (covered with a 18×18mm coverslip) with an epifluorescence optical microscope (Olympus BX-40-FX) with objectives 10x/0.25 (for EPNs) and 40x/0.75 (for EPNs and bacteria). Samples were observed under visible and UV light (Hg) adjusted for optimal GFP or mCherry fluorescence.

Bacterial counts
To quantify the amount of bacteria contained in one IJs pool, and to avoid any contamination (either from the passage in G. mellonella or from the environment), IJs were surface sterilized following a standardized procedure. In brief, IJs were immersed in a 1.5ml eppendorf for 3 minutes with 1ml of a sterilization solution (as described previously in the M&M) with gentle agitation. After 1min centrifugation at 4000rpm in a minicentrifuge, supernatant was discarded and IJs were rinsed thrice with physiological water (0.9% NaCl). Finally, surface sterilised IJs were crushed and plated on selective agar. The number of IJs present in the pool was estimated by microscopic counting of a representative sample (50μl). Alternatively, a non-sterile method was used to assess the number of targeted bacteria associated with the IJs: G. mellonella cadavers were rinsed with physiological water to collect the freshly emerged IJs in suspension. The number of IJs per larva was estimated by microscopical count on 50-μl drops from this suspension. Serial dilutions of the supernatant were then plated on selective agar medium and bacteria were enumerated.
Theoretical count of Y. pseudotuberculosis In order stress out Y. pseudotuberculosis multiplication throughout the successive infection cycles, theoretical counts (T Y ) that would be observed starting from the same inoculum if no bacterial division would occur have been calculated. For this, theoretical volumes of 0.5ml (V Gm ) and 0.8nl (V IJ ) have been assigned per G. mellonella larva and Steinernema sp MW8B IJ, respectively, and a mean EPN emergence yield of 50,000 EPNs (N IJ ) per larva was considered.
The infection dose was fixed to 125 IJs per G. mellonella larva. The dilution factor is the ratio V Gm /(V IJ × N IJ ). T Y is calculated by dividing the number of Y. pseudotuberculosis CFUs infecting a larva by the dilution factor.

Susceptibility testing towards Xenorhabdus sp. antimicrobial compounds
To estimate the resistance of enterobacteria towards Xenorhabdus sp. TZ01 antibiotic compound production, growth of the tested bacteria was monitored every 30 minutes during 6 hours by optical density (OD) measurement in the presence of Xenorhabdus culture extracts. For this purpose, LB liquid medium was inoculated with an inoculum of the target bacterium derived from a fresh culture to reach an initial OD 600 of 0.05-0.1. Prior to inoculation, LB was supplemented with 4% or 8% (v/v) of 0.2-μm filtered supernatant of a 48h liquid culture of X. sp. TZ01 grown at 28°C with shaking at 150rpm (Cell Free Supernatant). X. sp. TZ01 liquid cultures used for supernatant preparation were stopped when OD 600 reached 11-13. For growth curve analysis, OD 600 was plotted into a log2 scale in order to obtain a linear graph for the exponential phase of the curve. The slope of this line (calculated with GraphPad Prism 6) defines the growth rate of a given bacterium in the defined culture conditions.

Results
Yersinia pseudotuberculosis resists to antimicrobial compounds produced by Xenorhabdus sp.
Susceptibility of several strains of Y. pseudotuberculosis, pathogenic E. coli, Salmonella enterica subsp. enterica serovar Enteritidis (S. Enteritidis) and Serratia marcescens towards antimicrobial compounds produced by X. sp. TZ01 was tested by growth curve analysis. Growth of Y. pseudotuberculosis (strains 4N1 and IP2777) was slightly delayed with 4% X. sp. TZ01 supernatant compared to a control growth without X. sp. TZ01 supernatant, but subsequent growth was merely unaffected when up to 8% X. sp. TZ01 supernatant was added. Likewise, S. marcescens EE016 was characterized by a delayed growth while its capacity to grow with up to 8% X. sp. TZ01 supernatant was unaffected. To the contrary, both S. Enteritidis SE01 and Vero-toxigenic E. coli VT01 strains were drastically inhibited with 4% X. sp. TZ01 supernatant (data not showed) and totally unable to grow with 8% added supernatant (Fig. 1A). Growth rates of Y. pseudotuberculosis 4N1 and IP2777 were both slightly affected when X. sp. TZ01 supernatant was added. At 4% supernatant, slopes decreased by 12% and 18% for Y. pseudotuberculosis 4N1 and IP2777, respectively. These values dropped respectively by 30% and 35% when 8% of X. sp. TZ01 supernatant was added. S. marcescens EE016 growth was unaffected with either 4% or 8% added X. sp. TZ01 supernatant (Fig. 1B).

Yersinia pseudotuberculosis colonizes the gastro-intestinal tract of EPNs and survives long-term EPN storage
To assess the ability of Y. pseudotuberculosis to colonize Steinernema sp. MW8B, 7 independent Galleria mellonella infection experiments were conducted with a Y. pseudotuberculosis GFPmut2-labelled strain (4N1G strain) ( Table 2).
In all experiments, Steinernema sp. MW8B IJs 1 exhibited GFPmut2 fluorescence along the entire length of their gut ( Fig. 2A.1). However, in 2 out of 7 experiments (29%), IJs1 emerged from Y. pseudotuberculosis 4N1G-infected G. mellonella larvae failed to invade and kill new naive G. mellonella larvae in spite of the fact that they were massively colonized by Y. pseudotuberculosis 4N1G as attested by the bright GFP fluorescence they displayed. In 2 out of 7 experiments (29%), IJs that were both fluorescent and infective emerged from dead G. mellonella cadavers after 4 consecutive infection cycles ( Fig. 2A.2). One experiment (14%) led to the emergence of fluorescent/infective IJs after the fifth infection cycle and were so even after 7 consecutive infection cycles which lasted 14 weeks ( Fig. 2A.3).
Directly after the first emergence, freshly emerged IJs 1 were stored at 4°C, 16°C and 28°C in physiological water. IJs 1 were observed in epifluorescence microscopy to monitor the presence of Y. pseudotuberculosis 4N1G. These observations were made every day during the first week post emergence (PE) then once a week during 13 weeks. At 4°C, stored IJs 1 did not survive a week and were all dead by day 8 PE (Fig. 2B.1 and 2B.2). Nevertheless, Y. pseudotuberculosis 4N1G was still alive-and did even multiply slowly inside the IJs cadavers-since GFP fluorescence was still observed 6 weeks PE (Fig. 2B.3). No differences were observed between IJs 1 stored either at 16°C or at 28°C. In these samples, microscopic observations showed that Y. pseudotuberculosis 4N1G was still present inside IJs 1 of Steinernema sp. MW8B, either in the gut or in the inter-cuticular space, after 14 weeks of storage at either 16°C or 28°C (Fig. 2C).
To check the ability of Y. pseudotuberculosis 4N1G to remain associated with Steinernema sp. MW8B after several infection cycles, IJs 4 were also kept at 28°C. At 3 weeks PE, 23 +/− 3 IJs 4 were crushed and counted on selective agar plates. An average of 5.0 × 10 3 CFUs of Y. pseudotuberculosis 4N1G per IJs 4 was measured. Compared to IJs 4 at 0 day PE (8.6 × 10 3 CFU/IJs 4 ), the number of Y. pseudotuberculosis 4N1G CFUs per IJ decreased by 41%. However, quantitative results obtained for IJs 4 at week 3 PE and IJs 1 at day 0 PE are comparable.
Similar G. mellonella infection experiments were conducted three times with a GFPmut2labelled Y. enterocolitica O:3 strain (YE03). This strain is more sensitive towards antimicrobials produced by X. sp. since Y. enterocolitica O:3's growth is totally inhibited with the presence of 8% of X. sp. supernatant (data not shown). Microscopic observations showed EPN colonization by Y. enterocolitica YE03 during at least 2 consecutive infection cycles for one out of the three experiments conducted. Confocal microscopic observations localized Y. enterocolitica YE03 in the gut lumen after 2 infection cycles (Fig. 3). However, no GFPmut2 fluorescence was observed after 3 consecutive infection cycles with Y. enterocolitica YE03. GFPmut2-labelled

EPNs support dramatic multiplication and dissemination of Yersinia pseudotuberculosis
To confirm quantitatively the maintenance of Y. pseudotuberculosis 4N1G in the experimental model, CFU counts were determined at different time points (Table 2). After the first infection cycle, an average of 5.0 × 10 3 Y. pseudotuberculosis 4N1G CFUs per Steinernema sp. MW8B IJ were found. Similar counts were determined during 7 consecutive infection cycles, with an average of 8.6 × 10 3 CFUs of Y. pseudotuberculosis 4N1G per IJ still found after the 4 th infection cycle and 5.6 × 10 3 CFUs after the 7 th infection cycle ( Table 2). Knowing the number of CFU per IJ and the number of IJs emerged from dead larvae, we calculated the total increase in Y. pseudotuberculosis 4N1G counts after the various infection cycles. Starting with 1.9 × 10 6 CFUs of Y. pseudotuberculosis 4N1G directly injected in the larva, Y. pseudotuberculosis 4N1G counts after one cycle increased by two orders of magnitude and reached 2.5 × 10 8 CFUs. These counts were similar after the fourth infection cycle (3.5 × 10 8 CFU) and started to decrease from the  . This confirms our model predictions which suggest that in the absence of active multiplication, Y. pseudotuberculosis 4N1G counts would drastically decrease and would become undetectable after two infection cycles (Fig. 4). The experimental counts hence reflect an active multiplication of Y. pseudotuberculosis 4N1G in the studied laboratory model.

Yersinia pseudotuberculosis cannot replace Xenorhabdus sp. TZ01 as EPN symbiont
After having demonstrated the colonisation and multiplication capacity of Y. pseudotuberculosis 4N1G in the gut of Steinernema sp. MW8B IJs, we wondered whether Y. pseudotuberculosis 4N1G could substitute for X. sp. TZ01 as a bacterial symbiont in this EPN species. To address this question, we obtained axenic Steinernema sp. MW8B EPNs by collecting surface sterilised eggs from gravid females. Prior to G. mellonella infection, axenic EPNs were incubated with the mCherry-labelled Y. pseudotuberculosis 4N1C strain on Wouts Agar plates in order to obtain IJs exclusively colonised by Y. pseudotuberculosis 4N1C. A pool of such "monoxenic" IJs displaying red fluorescence (Fig. 5A) was divided into 4 equal groups. Two groups were incubated separately with 6 G. mellonella larvae in empty containers. One group was deposited onto a sterile Wouts Agar plate with 4 G. mellonella larvae and the last group was deposited onto a Wouts Agar plate without any larva. At day 3 post infection (PI), only G. mellonella larvae grown on Wouts Agar were found dead. At day 8 PI, all G. mellonella were dead. Three G. mellonella larvae recovered from Wouts agar plates showed emergence of EPNs. The emerged EPNs displayed no mCherry fluorescence when observed microscopically. When crushed and plated onto selective agar plates, no Y. pseudotuberculosis 4N1C could be retrieved from these EPNs neither. At day 10 PI, emergence of EPNs could be observed in 2 larvae incubated in empty containers, but again none of the emerged EPNs exhibited red fluorescence and no Y. pseudotuberculosis 4N1C could be retrieved after EPN crushing and plating on selective agar. In contrast, EPNs grown freely on Wouts agar still exhibited red mCherry fluorescence 10 days after plating. The same experiment was conducted with the GFPmut2-labelled X. sp. TZ03. Microscopic observations showed not only that X. sp. TZ03 was able to colonise the symbiotic vesicle of Steinernema sp. MW8B axenic EPNs (Fig. 5B), but also that X. sp. TZ03 maintained in its host after 2 consecutive cycles (Fig. 5C) and probably much more (not tested).

Discussion, Conclusion, and Perspectives
Compared to other enterobacteriaceae tested so far, the capacity of Y. pseudotuberculosis 4N1 to colonize Steinernema sp. MW8B is remarkably efficient and suggests that a number of biological functions required for its successful dissemination through this host during and between infection cycles are present and functional in this bacterium. We showed that enterobacteria sensitive to the antibiotics secreted by Xenorhabdus. sp. TZ01 have no ability to colonize the EPN gut. Two strains of Y. pseudotuberculosis (4N1 and IP2777) as well as one S. marcescens strain (EE016) isolated from a Steinernema sp. MW8B-infected G. mellonella larva were found naturally resistant to X. sp. TZ01 secreted antibiotics and were tested for their ability to colonize Steinernema sp. MW8B EPNs with the model developed herewith. The Entomopathogenic Nematodes Hijacked by Yersinia pseudotuberculosis growth rate of both Y. pseudotuberculosis strains was slightly affected by the presence of X. sp. TZ01 supernatant, while the S. marcescens EE016 was not. Likewise, Ochrobactrum tritici strain EE10.1 isolated from a Steinernema sp. MW8B-infected G. mellonella larva in our laboratory displayed a similar capacity to resist to X. sp. TZ01 antibiotics (data not shown). Despite this capacity, S. marcescens EE016 was unable to sustain EPNs life cycle completion since no IJ emergence occurred from S. marcescens-injected G. mellonella larvae. This suggests that Serratia and Ochrobactrum may accidentally reach the gut of Steinernema sp. MW8B but are unlikely able to colonize and multiply within the EPN gut as Yersinia pseudotuberculosis does. It has been shown that Serratia marcescens uses the type VI secretion system to neutralize bacterial competitors [39]. Injection of 10 6 S. marcescens CFUs may therefore impair Xenorhabdus growth in G. mellonella larvae. This could explain why Steinernema sp. MW8B cannot complete its reproductive cycle within S. marcescens-infected G. mellonella. S. marcescens has been shown to be pathogenic towards the free-living nermatode Caenorhabditis elegans but beneficial to the entomopathogenic nematode C. briggsae [40,41]. Zhang   Entomopathogenic Nematodes Hijacked by Yersinia pseudotuberculosis possibly after horizontal gene transfer [42,43]. Dixenic associations have been described, such as P. luminescens and Ochrobactrum spp. found together in tropical species of Heterorhabditis [44]. Genomic comparison between S. nematodiphila and other (non-symbiotic) Serratia spp. could provide interesting insights in the discovery of genes involved in the symbiotic association with EPNs.
In this study we showed that Y. pseudotuberculosis 4N1G is able to colonize and maintain for several generations inside a Steinernema species for long-term periods (14 weeks). Quantitative data showed that EPNs support efficient multiplication of Y. pseudotuberculosis 4N1G during this period. Indeed, counts of Y. pseudotuberculosis CFUs carried away by EPNs emerged from dead larvae are roughly multiplied by a factor 10 3 at the term of each infection cycle, a number which is probably underestimated as it does not take into account Y. pseudotuberculosis bacteria left over in the dead cadaver. Y. pseudotuberculosis 4N1G colonizes mainly the gut of Steinernema sp. MW8B but can be found in the inter-cuticular space as well after 3-month storage in physiological water. The localisation of Y. pseudotuberculosis 4N1G in Steinernema sp. MW8B IJs' gut is quite different from the normal localisation of the symbiotic Xenorhabdus sp. TZ03. Indeed, the natural niche of Xenorhabdus inside its Steinernema hostbefore infecting an insect prey-is a so-called symbiotic vesicle located along and separated from the EPN gut [3]. Our observations on EPN colonization are consistent with axenic EPNs experiments, which demonstrated that Y. pseudotuberculosis 4N1G does not replace the X. sp. TZ01 symbiont during EPN infection cycle but more likely hijack the symbiotic relationship between Xenorhabdus and EPNs. Indeed, when the natural symbiont of Steinernema sp. MW8B is absent, EPNs colonized by Y. pseudotuberculosis 4N1C alone are unable to develop properly in a G. mellonella larva. In control experiments where axenic EPNs are supplemented with GFP-labelled X. sp. TZ03, EPNs recover their ability to indefinitely multiply and feed on G. mellonella larvae. Moreover, unpublished results showed that IJs colonized by both X. sp. and Y. pseudotuberculosis strains labelled with two different fluorescent protein markers do contain both bacteria.
The mini-Tn5 transposon used to tag Y. pseudotuberculosis 4N1G and 4N1C in our experiments was mapped in the fimbrial A protein gene. This gene is found in two intact copies in the Y. pseudotuberculosis genome meaning that the protein is probably still expressed in the GFP-tagged strains. Fimbrial proteins are known to act as colonization factors for Y. pseudotuberculosis and are involved in the attachment to epithelial cells [45]. The capacity of the 4N1G and 4N1C tagged strains to colonise the EPN's gut, as demonstrated throughout our study, argue in favour of a non-detrimental effect of the transposon insertion compared to the wild-type strain.
Interestingly, it has been shown that Y. pseudotuberculosis and Y. pestisthe etiological agent of plaguecan infect or form a biofilm mainly around the head of Caenorhabditis elegans, a well-studied nematode laboratory model [46,47,48]. Given the fact that Y. pestis evolved quite recently from Y. pseudotuberculosis [49], it would be interesting to know whether Y. pestis can also resist to antimicrobial substances produced by Xenorhabdus/Photorhabdus spp. and colonize EPNs. If these findings turn to have an environmental significance, it would provide new insights in the understanding of long-term persistence of Y. pestis in plague endemic areas worldwide [28,50].
Future work should focus on the identification of Yersinia genetic determinants required to colonize and maintain inside EPNs. Several genes shared by Yersinia and the EPN's natural symbionts are good candidates to play this role such as the phospholipase A encoded by yplA, structural genes of the type 6 secretion system (T6SS) and possibly others [10]. Likewise, genome comparisons between Serratia, Ochrobactrum, Yersinia and Xenorhabdus should help deciphering the critical genetic determinants required for EPN colonization.