Spatiotemporal expression of the putative MdtABC efflux pump of Phtotorhabdus luminescens occurs in a protease-dependent manner during insect infection

Photorhabdus luminescens is an enterobacterium establishing a mutualistic symbiosis with nematodes, that also kills insects after septicaemia and connective tissue colonization. The role of the bacterial mdtABC genes encoding a putative multidrug efflux system from the resistance/nodulation/cell division family was investigated. We showed that a mdtA mutant and the wild type had similar levels of resistance to antibiotics, antimicrobial peptides, metals, detergents and bile salts. The mdtA mutant was also as pathogenic as the wild-type following intrahaemocoel injection in Locusta migratoria, but had a slightly attenuated phenotype in Spodoptera littoralis. A transcriptional fusion of the mdtA promoter (PmdtA) and the green fluorescent protein (gfp) encoding gene was induced by copper in bacteria cultured in vitro. The PmdtA-gfp fusion was strongly induced within bacterial aggregates in the haematopoietic organ during late stages of infection in L. migratoria, whereas it was only weakly expressed in insect plasma throughout infection. A medium supplemented with haematopoietic organ extracts induced the PmdtA-gfp fusion ex vivo, suggesting that site-specific mdtABC expression resulted from insect signals from the haematopoietic organ. Finally, we showed that protease inhibitors abolished ex vivo activity of the PmdtA-gfp fusion in the presence of haematopoietic organ extracts, suggesting that proteolysis by-products play a key role in upregulating the putative MdtABC efflux pump during insect infection with P. luminescens.


Introduction
Photorhabdus luminescens is a member of the family Enterobacteriaceae. This bacterium is highly pathogenic to many insects and maintains a mutualistic relationship in the gut of freeliving entomopathogenic nematodes [1]. On  also substrates for this pump along with bile salts derivatives, SDS and novobiocin [23,28,33,34]. The overlapping substrate specificities of Mdt transporters among enterobacteria for novobiocin and bile salts suggest that these compounds may be similar to the natural substrates of these pumps such as sodium tungstate and flavonoids [35]. Pletzer and Weingart demonstrated that the expression of the mdtABC operon in the phytophatogen E. amylovora was induced by the plant polyphenol tannin in a BaeSR-dependent manner. They further showed that MdtABC was upregulated in vivo during bacterial growth in planta suggesting a role of this efflux pump in resistance towards antimicrobial plant compounds, such as flavonoids [36]. Similarly in P. aeruginosa, expression of the mexAB-oprM genes is increased in the upper layer of a biofilm upon colistin exposure so that a spatially distinct subpopulation of metabolically active cells develop tolerance to this AMP [37]. For instance, there is no data available about RND efflux pumps of entomopathogenic nematode-associated bacteria, and here we propose to monitor their expression and to assess their roles in the life cycle of Photorhabdus luminescens within its insect hosts.
In the present study, we identified genes from P. luminescens encoding a putative MdtABC efflux pump, which shows sequence similarity to other proven efflux pumps of enterobacteria. The mdtA mutant showed wild-type levels of MDR and virulence in insects. We then showed that the mdtABC operon was specifically induced during late stages of insect infection, within bacterial aggregate formations in the connective tissue of L. migratoria HO and that of S. littoralis midgut. Finally, we described a protease-dependent induction of the mdtABC promoter ex vivo in a HO extract-containing medium.

Construction of the P. luminescens ΔmdtA mutant
We created a stable chromosomal mutation in the mdtA gene of P. luminescens TT01 (mdtA Pl ) by allelic exchange, by constructing a derivative of pJQ200KS carrying the upstream part of the target mdtA gene (690 bp extending from positions-532 to + 158 with respect to the mdtA translation initiation site), an ΩCm antibiotic resistance cassette [42], and the downstream part of the target gene (603 bp extending from positions-453 to + 130 with respect to the mdtA stop codon). TT01 genomic DNA fragments were amplified by PCR with the oligonucleotide primer pairs 5' -GCGCTCTAGAATCATCGGAAGCCGTATCTG-3' / 5' -CGCGGATCCG GGGGGTAAAGGGGAATGTCT-3' (upstream region) and 5'-CGCGGATCCGGTTGTTAGT GCCTGGGATCG-3' / 5' -GCGCGAGCTCCCACTTCCGGTAATGCAGAC-3' (downstream region). The PCR products were restricted with XbaI/BamHI and BamHI/SacI enzymes, the recognition sites of which are underlined. The ΩCm antibiotic resistance cassette was isolated as a 3.5 kb BamHI fragment from the pHP45-ΩCm plasmid. This 3.5 kb DNA fragment was then ligated into the BamHI site, between the upstream and downstream mdtA regions, the resulting fragment being inserted into the pUC19/XbaI-SacI vector to generate pUC-mdtA-ΩCm. The full-length 4.8 kb mdtA-ΩCm DNA fragment was gel-purified, digested with XbaI and SacI and inserted into the corresponding restriction sites of the pJQ200KS vector, yielding pJQ-mdtA-ΩCm. The pJQ200KS plasmid is a derivative of pACYC184 carrying the sacB gene and the mob site from RP4. The pJQ-mdtA-ΩCm plasmid was used to transform E. coli strain S17.1 and was introduced into P. luminescens TT01 by mating, as previously described [43]. Cm R and Sac R exconjugants were selected on 3% sucrose and LB agar supplemented with chloramphenicol. The 612 bp mdtA deletion and the omega insertion were checked by PCR analysis and DNA sequencing (MACROGEN, Seoul, Korea). The polar effect of mdtA mutation was checked by reverse transcription PCR (RT-PCR). The resulting recombinant clone was named ΔmdtA.

Construction of vectors for complementation assays
The entire mdtABC operon of P. luminescens TT01 (plu2774-plu2776) (https://www.genoscope. cns.fr/agc/mage) was amplified by PCR from TT01 genomic DNA with the high-fidelity Herculase DNA polymerase (Stratagene) and the oligonucleotide primers 5'-GCGCCTGCAGGC CTTGCGTTTGAGAGAATG-3' and 5'-GCGCTCTAGACCACTGATGCCGAGTTTCAT-3', which include recognition sites for PstI and XbaI, respectively. The resulting 7.5 kb fragment was digested with PstI and XbaI and ligated with the pBBR1MCS-5 vectors digested with the same enzymes. E. coli XL1-Blue MRF' was transformed with the ligation product, according to standard protocols [44]. The resulting plasmid, pBB-mdtABC, was checked by restriction and sequencing (MACROGEN, Seoul-Korea) and transferred into ΔmdtA by mating [43]. We then checked, by RT-qPCR, that the mdtABC operon was expressed in the complemented mutant.

Determination of the minimum inhibitory concentrations of toxic compounds
In vitro susceptibility tests to determine minimum inhibitory concentrations (MICs) for wildtype P. luminescens and its derivatives were performed for a range of antimicrobial peptides

Construction of transcriptional fusions with the green fluorescent protein gene gfp[AAV]
The P mdtA -gfp [AAV] and P lac -gfp[AAV] transcriptional fusions were constructed by inserting the promoter regions of the mdtA gene (P mdtA ) and of the lactose operon (P lac ) upstream from a promoter-less gfp[AAV] gene [46] located on a low-copy plasmid, pPROBE'GFP-AAV [46]. Briefly, the 332 bp P mdtA corresponding to the sequence extending from positions-338 to-6 relative to the translation start site of the mdtA gene was amplified by PCR from P. luminescens TT01 genomic DNA with sense 5'-GCGCGAATTCCGCAGGAATATGTTCTATACACTATGG-3' and antisense 5'-CGCGGATCCGACATTCTCTCAAACGCAAGG-3' primers, in which the EcoRI and BamHI recognition sites, respectively, are underlined. Similarly, the P lac promoter (120 bp) was amplified by PCR from the bluntII-TOPO vector (Invitrogen), with sense 5'-G CTCTAGAGCGCAACGCAATTAATGTG-3' and antisense 5'-GGGGTACCAGCTGTTTCCTGT GTGAAATTG-3' primers, in which the XbaI and KpnI recognition sites, respectively, are underlined. The PCR products were purified and inserted upstream from the gfp [

In vivo infection assays
The Noctuidae species studied (S. littoralis and S. frugiperda) and locusts, L. migratoria, were reared on an artificial diet at 23˚C and on grass at 30˚C, respectively, with a photoperiod of 12 h. Larvae were selected and surface-sterilised with 70% (v/v) ethanol before intrahaemocoel injection. All the bacterial strains used were grown to late exponential growth phase in LB broth, washed and diluted in LB medium supplemented with 20 µg.ml -1 km, when necessary. We injected 20 µl each of bacterial suspension, containing 10 3 bacteria (for S. littoralis and S. frugiperda) or 10 4 bacteria (for L. migratoria), into groups of 20 larvae. LB medium was injected as a control into groups of 10 larvae. Treated larvae were incubated for up to 96 h, and the time of death of the insects was recorded. Bacterial concentrations, in CFU, were determined by plating dilutions on nutrient agar. Statistical analyses were performed, as previously described [43], with SPSS (SPSS, Chicago, IL), for the comparison of survival rates.

Insect dissection and histology
We assessed the in vivo expression of the gfp[AAV] transcriptional fusions in the body cavity of the insect by dissecting, at regular time points, the larval cuticle dorsally, removing pieces of the organs of interest (the haematopoietic organ in L. migratoria and mid-gut connective tissues of S. littoralis and S. frugiperda), squashing them between the slide and cover slip and viewing them under fluorescence conditions. 10 µl of insect haemolymph were examinated at regular intervals, by fluorescence microscopy (Leica).

In vitro reporter gene assays
We analysed the kinetics of gene expression in vitro by inoculating fresh LB containing the appropriate concentrations of kanamycin and a potential inducer to be tested, in black-sided clear bottomed 96-well plates (Greiner), with a 1/500 dilution of an overnight culture of wildtype TT01 strain carrying one of the various GFP fusions. The use of such a high dilution (1/ 500) ensured that GFP proteins present in the inocula were diluted out by the first cell divisions and do not interfere with the monitoring of gene expression. Plates were incubated for 48 h at 28˚C, with orbital shaking, in an Infinite M200 microplate reader (Tecan). A 600nm and GFP fluorescence intensity, with excitation at 485 ± 4.5 nm and emission at 520 ± 10 nm, were measured every 30 minutes. The effects of various chemical components on P mdtA activity were determined at maximum fluorescence intensity. Each of the follwing was added to an appropriate final concentration, as follows: 25 µg.ml -1 S. frugiperda cecropin B, 25 µg.ml -1 cecropin A (Sigma), 1 mM MgCl 2 , 0,5 mM FeSO 4 , 2 mM CuSO 4 . CuSO 4 was added at 2 mM to the bacterial culture at mid-exponential growth phase (O.D 600 = 0.4). The pH of the culture medium supplemented with these compounds was adjusted to that of LB medium, and we checked that cell growth was unaffected. We also investigated the effects of insect-derived molecules on P mdtA activity by removing HOs cleanly from 20 L. migratoria larvae and grinding them in 500 µl of cold PBS, with a microtissue grinder (Ultra-turrax, IKA). The mixture was centrifuged at 13000 g for 5 minutes and the supernatant was collected, filtered (0.2 µm pores) and stored at-80˚C until further use. For induction tests, the HO-derived stock solution was diluted (10-fold) in fresh LB medium. For the inhibition of proteases in bacterial and insect HO cells, Complete Mini and Complete Mini EDTA-free tablets (Roche) were added to the LB medium to yield a 0.4x solution. Plasma was collected from locust larvae by centrifuging haemolymph at 13000 g for 5 minutes, for the elimination of circulating cells. Plasma was then filtered (0.2 µm pores) and stored at-80˚C or used directly as a medium for bacterial inoculation. Growth data were recorded in triplicate for each set of conditions and A 600nm and fluorescence recorded over 48 hours of incubation were averaged for the plotting of curves and histograms. Fluorescence intensities were then normalized with respect to bacterial cell density, by dividing the fluorescence unit value (FU) recorded at a given time and in a specific condition by the corresponding A 600nm , to obtain a specific fluorescence value. For data analysis, Relative Fluorescence Units (RFUs) were determined by the ratio of the P mdtA -gfp[AAV] specific fluorescence value to that of the P lac -gfp[AAV] in a given condition and at maximum fluorescence intensity.

RNA preparation and quantitative RT-PCR analysis
Total RNA was extracted from P. luminescens TT01 wild-type strain cultured in LB broth (at an optical density of 0.4, corresponding to the middle of the exponential growth phase) in TRIzol reagent, according to the manufacturer's instructions (Invitrogen). Total RNA was then purified with the High Pure RNA Isolation kit (Roche), including incubation with DNase I. RNA concentration was determined by measuring absorbance at 260 nm. For each RNA preparation, we assessed DNA contamination by carrying out a control PCR before RT-qPCR analysis.
RT-qPCR was performed in two steps. First, cDNAs were synthesised from 1 µg of total RNA with the Super Script II Reverse Transcriptase from Invitrogen and random hexamers (100 ng.µl -1 ) from Roche Diagnostics. Quantitative PCR was then performed, in triplicate, with the Light Cycler 480 SYBR Green I Master kit from Roche Diagnostics, with 1 µl of cDNA synthesis mixture and 1 µM of specific gene primers for mdtA (5'-GTTTTACAAG CTCAGAAAACGAATC-3' and 5'-ACTGATTTGGGAATAAGATGCTTTC-3') or 16S rRNA, used as the control gene (5'-AATGGCATCTAAGACTGGTTGACTG -3' and 5'-TACCAGGGTATCTAATCCTGTTTGC -3'). The enzyme was activated by heating for 10 minutes at 95˚C. The reaction mixture was then subjected to 45 cycles of 95˚C for 5 s, 60˚C for 5 s and 72˚C for 10 s, with the Light Cycler 480 system (Roche) used for monitoring. Melting curves were analysed for each reaction, and each curve contained a single peak. We determined the amounts of PCR products generated from standard curves obtained for PCR with serially diluted genomic DNA from P. luminescens TT01. Data were analyzed as a ratio, with 16S rRNA gene used as the control gene (95% confidence limits).

Identification and genomic environment of the mdtABC operon in Photorhabdus
Analysis of the P. luminescens TT01 genome sequence [47] revealed the presence of a mdt locus containing three open reading frames, mdtA, mdtB and mdtC (plu2774 to plu2776) ( Fig  1). The mdtA gene encodes a 401-amino acid (aa) putative membrane fusion protein (MFP) 64.29% (E-value of 1e -166 ) identical to the MdtA (YegM) protein of E. colithat serves as periplasmic "adaptor" connecting the inner membrane transporter (MdtB-MdtC) to an outer membrane channel such as TolC. MdtA of P. luminescens is also 81% identical to the TT01 acrA (acriflavin resistance protein A precursor)-encoded proteins AcrA. Both mdtB and mdtC genes encode homomultimer RND-type transporter proteins MdtB and MdtC that share 48.04% of identity and are 77.81% (E-value of 0,0) and 76.13% (E-value of 0,0) identical to MdtB (YegN) and MdtC (YegO) of E. coli, respectively. MdtB and MdtC are each partially identical to TT01 AcrB (acriflavin resistance protein B) (28.42% and 28.9%, respectively) and to the plu0758 gene product (28.26% and 27.74%, respectively) which is predicted to be similar to a RND-type multidrug transporter AcrB/AcrD/AcrF with a highly conserved AcrB domain. The genome sequences of E. coli and Salmonella also contain a gene encoding an additional putative MFS-type drug exporter, MdtD. mdtD is located between the mdtC and baeS genes. However, there is no mdtD gene in the genome of Photorhabdus (Fig 1). In E. coli and Salmonella, the function of the MdtABC transporter requires the outer membrane channel TolC that is also found in the Photorhabdus genome (plu3954). According to Joshi and Xu (2007) in their measure of functional similarity based on Gene Ontology annotation, if two proteins have sequence identity more than 70%, they have about 90% probability or more to share the same biological process [48]. In line with this and in light of the high amino acid sequence similarity of Mdt proteins from P. luminescens and E. coli, as well as the remarkable similarities in genomic organization of the mdtABC operon among enterobacteria, the MdtABC complex of P. luminescens is likely to be an export system whose function also requires TolC. Moreover, the 121 nucleotides downstream from the TT01 mdtC stop codon constitute the baeS/baeR genes, which are thought to encode putative sensor kinase BaeS and response regulator BaeR, which are 61.09% and 70.74% identical to BaeS and BaeR of E. coli, respectively (Fig 1). The presence of short intergenic sequences, an absence of putative terminators between the genes and the presence of a putative terminator downstream from baeR strongly suggest that the baeSR operon may be transcribed with mdtABC as a monocistronic RNA, as in E. coli and Salmonella [28,34].

Phenotypic characterisation of the isogenic P. luminescens ΔmdtA mutant
We investigated the role of the mdtABC operon in P. luminescens TT01 by inactivating the mdtA gene by allelic exchange. The phenotypic traits of the resulting ΔmdtA mutant and the wild-type strain were then compared. The ΔmdtA and wild-type TT01 grew similarly in LB medium at 28˚C, indicating that the mdtA mutation did not affect bacterial growth. We then assessed the role of the mdtABC operon in the resistance of P. luminescens to various antimicrobial compounds (bile salts, antibiotics, AMPs, detergents and metals), by determining MICs for growth for the wild-type and ΔmdtA strains ( Table 2). The ΔmdtA and wild-type strains had similar susceptibilities to all the antimicrobial compounds tested. RT-qPCR experiments showed that mdtABC expression was ten times stronger in the complemented strain (ΔmdtA/pBB-mdtABC) but no increase in MIC was observed. Thus, mdtABC inactivation is not sufficient to reduce the metal or antimicrobial resistance of P. luminescens.
We also investigated the effect of efflux pump inhibitors (EPI) on both wild-type TT01 and ΔmdtA strain resistance to some antimicrobial agents. Phenyl-arginine-β-naphthylamide (PAβN) acts as broad-spectrum EPI in various Gram-negative bacteria [49], whereas arylpiperidines, such as 1-(1-naphthylmethyl)-piperazine (NMP) moderately decrease MDR in E. coli strains overexpressing RND-type efflux pumps [50,51]. Both PAβN and the NMP EPIs block RND-type efflux pumps, such as E. coli AcrAB, specifically. We established the EPI concentrations to be used in assays with antimicrobial compounds at concentrations of 50 µg.ml -1 for NMP and 25 µg.ml -1 for PAβN, values one quarter and one eighth the MICs of these inhibitors for wild-type TT01 and ΔmdtA strains, respectively. Only PAβN significantly decreased, to various degrees, the MICs in wild-type TT01 and ΔmdtA of SDS, rifampicin, novobiocin, nalidixic acid and copper, with respect to the control values obtained in the absence of EPIs (Table 3). This suggests that these compounds are substrates of RND-type efflux transporters, but not solely of the MdtABC efflux pump.
We also investigated the role of the mdtABC operon in the pathogenicity of P. luminescens in insect larvae by injecting similar doses of the wild-type TT01 strain or its isogenic mdtA mutant directly into the haemocoel of each Spodoptera littoralis, S. frugiperda and Locusta migratoria larva, and by monitoring the insect mortality over time after the injection. Surprisingly, only bacterial injections into S. littoralis larvae showed the mdtA mutant to be slightly attenuated for virulence (Fig 2). These larvae infected with mdtA mutant cells died about 9 h later than larvae infected with the P. luminescens wild-type strain. The mdtA mutation had no effect on insect survival in another lepidopteran, S. frugiperda, or the locust.

Copper increases P. luminescens mdtABC promoter activity in vitro
For characterisation of the expression of the P. luminescens mdtABC operon, we fused P. luminescens P mdtA to the reporter gene gfp[AAV] (which encodes a destabilized GFP with a half-life of~60 min in E. coli) of pPROBE'-gfp[AAV] [46], a highly stable plasmid maintained without selection pressure in a sister taxon of P. luminescens, Xenorhabdus nematophila [52]. The expression of mdtABC in Salmonella is responsive to metals [34]. We first assessed the role of metals, such as copper, magnesium and iron, in regulating mdtABC gene expression in P. luminescens, by adding these metals to bacterial exponential growth-phase cultures, as described by Nishino et al. (2007). Similarly, we investigated the effects of AMPs from insects, such as cecropin A and Sf cecropin B. Only the addition of copper (2 mM) to a mid-exponential growth phase culture of P. luminescens-P mdtA -gfp[AAV] significantly enhanced fluorescence emission without affecting bacterial growth. At the maximum rate, recorded six hours after the addition of copper, specific fluorescence was 2.3 times higher than that in the control LB culture to which no copper was added (Fig 3).

The mdtABC operon is expressed within specific insect tissues and induced ex vivo by insect haematopoietic organ extracts
For the characterisation of mdtA expression in vivo, we injected the TT01 strain carrying the P mdtA -gfp[AAV] fusion or the P lac -gfp[AAV] fusion (positive control) into the haemocoel of L. migratoria last-instar larvae (10 4 bacteria/larva) and S. littoralis fifth-instar larvae (10 3 bacteria/larva). We then monitored the presence of bacteria for up to 30 hours after infection in haemolymph, the locust haematopoietic organ (HO), and S. littoralis mid-gut connective Table 3. tissues. We first localized P. luminescens-P lac -gfp[AAV] fluorescent bacteria to the sites of preferential colonization: the haemolymph (Fig 4A, 4B, 4E and 4F), the HO of L. migratoria ( Fig  5A), and the mid-gut connective tissues of S. littoralis (Fig 5K) from 20 hours to 30 hours after infection. After injection of the P. luminescens-P mdtA -gfp[AAV] strain, only 5% of fluorescent bacteria were detected after recovery from haemolymph at 20 to 24 hours (S. littoralis) and 30 hours post-injection (L. migratoria) (Fig 4D and 4H). However, All P. luminescens-P lac -gfp[AAV] recovered from the haemolymph at the same time points were fluorescent (Fig 4B and 4F). Thus, P mdtA activity is almost entirely absent from the haemolymph in both Locusta and Spodoptera. In Locusta, numerous nodules were observed in the HO after the injection of P. luminescens-P lac -gfp[AAV] (Fig 5A and 5B). Constitutively fluorescent P. luminescens bacteria colonised the connective tissues of the locust HO and the lepidopteran mid-gut as bacterial aggregates, as observed under a fluorescence microscope (Fig 5A, 5C, 5E, 5F, 5H and 5J). The localisation of bacterial aggregates was not necessarily associated with the presence of nodules (Fig 5B-5E). Moreover, fluorescence microscopy observations of HO and mid-gut connective tissues showed that bacteria harbouring the P mdtA -gfp[AAV] fusion were fluorescent within aggregates (Fig 5F, 5H and 5J). Not all the fluorescent bacteria observed were in contact with nodules and the trachea in Locusta HO and Spodoptera, respectively, suggesting that the P mdtA activity was not related to these structures (Fig 5I, 5J, 5S and 5T). For further confirmation of MdtABC induction ex vivo, we monitored the fluorescence of P. luminescens-P mdtA -gfp[AAV] grown in the plasma (cell-free haemolymph) of locust larvae or in HO homogenate-containing medium. We found that P mdtA activity was almost entirely absent when bacteria were grown in plasma, but very strong in locust HO extracts (Fig 6). Indeed, the specific fluorescence of P. luminescens-P mdtA -gfp [AAV] in the presence of locust HO extracts measured in vitro was at least 12 times stronger than that recorded in the plasma of the same insect (Fig 6). These experiments clearly support the in vivo observations and indicate that the P mdtA activity is tissue-specific. Moreover, we observed under fluorescence microscopy bacteria from culture-based induction assays at different time intervals and during peak expression. We first showed that both planktonic and aggregated bacterial cells from HO-containing cultures strongly expressed GFP. However, bacteria grown in either plasma or LB showed slight expression similar to what we have observed in the plasma from infected insects. These results suggestthat mdtABC expression in locust HO was unrelated to aggregate formation. Such a transient (late stages of infection) and spatially (connective tissues) specific expression of mdtABC operon most probably improve bacterial dissemination and persistence into the insect cadaver. Moreover, giving the strong activity of the P mdtA reported both ex vivo and in vivo in the locust HO, we wanted to test the phenotype of Photorhabdus under these conditions regarding MICs changes in vitro. P. luminescens mdtA mutant and wild-type strains exhibited the same MICs for SDS, Kanamycin, copper, and novobiocin, in the presence or absence of HO extracts (S1 Table).

HO-mediated mdtABC expression is reduced by protease inhibitors
For preliminary characterisation of the HO-derived putative signals for mdtABC expression, the TT01 strain carrying the P mdtA -gfp[AAV] fusion in LB supplemented was grown with HO extracts untreated or previously heated at 60˚C. The use of heated HO extracts resulted in significantly lower levels of P mdtA -gfp[AAV] activity (S1 Fig), indicating that the stimuli upregulating mdtABC expression are non-heat stable and could be proteinaceous in nature (proteins or peptides).
Photorhabdus produces high numbers of proteases during the colonization of insect host tissues [5]. We therefore focused on the protease-dependent modulation of mdtABC gene  We monitored the expression of both P mdtA -gfp [AAV] and P lac -gfp[AAV] fusions in TT01 strain, by recording GFP fluorescence in LB supplemented with HO extracts in the presence and absence of complete mini EDTA-free protease inhibitor, which fully inhibits a broad spectrum of serine and cysteine proteases (Fig 7A). We first showed that a 0.4x solution of EDTA-free protease inhibitor had no effect on bacterial growth (data not shown). We then showed that expression of the P mdtA -gfp [AAV] fusion was inhibited by the addition of EDTAfree protease inhibitor to the culture medium containing the inductive HO extract, no such the vicinity of the midgut in S. littoralis (K and P), each injected with Photorhabdus. Pieces of HO in which nodule structures were visible (B and G) or not visible (D and I) were observed under the epifluorescence microscope (C and E, H and J, respectively). Brightly fluorescent green bacteria were detected in both cases, demonstrating the HO site-specific expression of P mdtA -gfp[AAV] within closed bacterial aggregates, independently of nodule structures. The circle (G) indicates the limit of a bacterial aggregate and arrowheads (H) indicate isolated bacteria. All these observations were made 20 to 28 hours postinjection for S. littoralis and at 30 hours post-injection for L. migratoria, and correspond to results of at least three independent experiments. ECM, extracellular matrix. Red scale bar represents 10 µm. White scale bar represents 50 µm. https://doi.org/10.1371/journal.pone.0212077.g005

Fig 6. Level of expression of Photorhabdus luminescens mdtABC operon in Locusta migratoria plasma and medium containing HO extract in vitro.
Overnight cultures of P. luminescens carrying the P mdtA -gfp[AAV] fusion were diluted (1/500) in black-sided clear-bottomed 96-well plates containing one of the following media: fresh LB medium, plasma or fresh LB medium supplemented with L. migratoria HO extract. A 600nm and GFP fluorescence were monitored, in real time, in triplicate for each set of conditions, at 28˚C, with orbital shaking, in an Infinite M200 microplate reader (TECAN). Specific fluorescence was obtained by dividing GFP fluorescence units (FU) by the A 600nm value at 20 hours of growth. Relative Fluorescence Units (RFUs) were determined by the ratio of the P mdtA -gfp [AAV] specific fluorescence value to that of the P lac -gfp [AAV]. Asterisks indicate statistically significant differences ( � , P <0.05) in the paired Student's t test. The expression rate is the ratio between specific activity obtained after incubation with HO extracts and that obtained after incubation with insect plasma. inhibition being observed with non-supplemented control medium (Fig 7A). As a control, expression of the P lac -gfp[AAV] fusion remained very strong in the presence of HO extracts and inhibitors (Fig 7B). Thus, expression of the mdtABC operon in Photorhabdus is dependent on proteolysis by-products.

Discussion
We examined the role of the putative MdtABC efflux pump in P. luminescens MDR and virulence. P. luminescens ΔmdtA was as resistant as the wild-type strain to common substrates of multidrug transporters, such as DOC, SDS and novobiocin. We therefore suggest that the MdtABC multidrug transporter alone is not required for MDR in P. luminescens, at least in vitro. However, this is not surprising because the deletion of single RND drug efflux pump genes, including the mdtABC operon of S. enterica and E. coli, has no significant effect on drug resistance, other than in acrAB mutants [25,34]. As mentioned in results section, other putative RND pumps-encoding genes are present in the genome of P. luminescens TT01 such as acrAB and acrB/acrD/acrF homologues, that probably contribute to MDR resistance in Photorhabdus as described in other enterobacteria [22,25,53,54]. This hypothesis is strongly supported by our findings that RND inhibitors significantly enhanced bacterial susceptibility to SDS, rifampicin, novobiocin and nalidixic acid, all of which are common substrates of RND efflux pumps. Although some substrates of the MdtABC efflux pump were characterized in some bacteria, most of its known substrates, such as antibiotics and detergents, are not encountered in the natural environment of these bacteria. In this study, we showed that the genetic control of the MdtABC-mediated efflux is characterized by a well regulated expression of the mdtABC genes in specific spatial (connective tissues) and temporal (late infection stages) patterns. Therefore, it seems likely that the MdtABC efflux system may have evolved for specific purposes during the interaction of Photorhabdus with its invertebrate hosts from symbiosis with nematodes to pathogenicity for insects. In this regard, it is noteworthy that the natural environment of Photorhabdus, especially the insect cadaver, is enriched in antimicrobial compounds produced by Photorhabdus itself, such as potentially toxic metabolites [55][56][57], bacteriocins [58], and stilbene [59][60][61], as well as other antibiotics that are released by the competitive microbiota of both the nematode and the insect [62,63]. Bacterial multidrug efflux pumps of the RND family have been shown to be important for virulence in model animals, following infection via several routes, including ingestion, inhalation and the intraperitoneal injection of bacteria [64,65]. We show here that P. luminescens ΔmdtA is as virulent as the wild-type strain after injection into haemocoel of the locust and the lepidopterans Spodoptera. These findings were expected given that Photorhabdus has a plenty of different and redundant virulence factors ranging from potent toxins and enzymes to PhoPregulated LPS modification systems mediating resistance to AMPs produced in insect haemolymph upon infection [9]. Therefore, it is often hard to identify an effect on insect pathogenicity when just a single gene or operon that might contribute to virulence is inactivated. The only genes currently described as essential for P. luminescens resistance to AMPs and virulence in insects are phoP and pbgPE [66,67], that were recently shown to govern the main virulence strategy of an AMP resistant subpopulation of P. luminescens causing septicemia in insects [68]. Previous studies of the colonization of insect larvae by P. luminescens have shown that the extracellular matrix of the connective tissues and haemolymph are initially more extensively colonized than other tissues [4,5]. We showed, by transcriptional fusion with the gfp gene, that the P mdtA activity of Photorhabdus mdtABC operon was clearly increased in the connective tissues, but not in the plasma, during the infection of three insect species. We mimicked the 'host-like' conditions triggering the MdtABC efflux system encountered during bacterial colonization of the extracellular matrix, by dissecting the HO from Locusta. Only a moderately developed connective tissue enclosing the belts of muscle around the mid-gut is found in lepidopterans, such as Spodoptera. By contrast, a well-developed connective tissue with specialist cells and extracellular material, often organised into structured fibres (collagen), has been described in the HO along the dorsal vessel in Locusta migratoria [69]. Our ex vivo findings provide strong evidence for a specific signal within the HO but apparently absent from insect plasma. Collagen is a normal constituent of the animal extracellular matrix, including that of the insect L. migratoria, which has been shown to possess two types of collagen: a predominant fibrous collagen similar to collagen I from mammals and a minor collagenous component very similar to basement membrane collagen IV [69]. Preliminary assays showed that mammalian collagen IV and collagen I had no effect on mdtABC expression. Our studies have shown that expression of the lopT gene, encoding an antiphagocytic type three secretion system (TTSS) effector, is induced only at sites of cellular defence reactions, such as nodules in the HO of L. migratoria [4]. We show here that mdtABC expression is also site-specific but independent of these defence reactions, because fluorescent bacteria were not only associated with nodule formation in the HO.
Our data revealed that the inhibition of cysteine and serine proteases in HO-supplemented medium was sufficient to inhibit the ex vivo expression of P mdtA -gfp [AAV]. Indeed, Photorhabdus produces a wide range of proteases, including metalloproteases and non-metalloproteases, in culture media or in contact with insect tissues during infection [70][71][72]. PrtA, a metalloprotease inhibited by disulfide bridge-reducing agents and SH group reagents, is a good candidate that is produced within the connective tissues close to the midgut epithelium during late infection stages in insects when septicaemia is well established [73,74]. PrtA degrades basal lamina, providing the bacteria with access to the underlying tissues for bioconversion [5]. Moreover, the genome of the TT01 strain contains other sequences potentially corresponding to cysteine and serine protease-or peptidase-encoding genes [47]. The peptides and/or amino acids generated by the proteolytic action of these bacterial proteases on the HO tissue probably trigger the production of the MdtABC pump in vivo.
Sensing of and responding to environmental fluctuations via two-component systems are of vital importance for bacteria [75,76], especially in the specific context of MDR [75,77,78]. Indeed, the BaeSR two-component system of E. coli and Salmonella detects specific envelopedamaging compounds (indole, zinc, copper and certain flavonoids) and induces the MdtABC efflux pump, which can expel these inducing compounds, as well as some antibiotics and bile salt derivatives, thereby maintaining bacterial envelope homeostasis [23,34,35,79,80]. The baeS and baeR genes are found immediately downstream from mdtABC in P. luminescens TT01 genome, suggesting that BaeSR may also regulate the expression of the mdt operon. This supports the hypothesis that the mdtA-inducing compounds within the insect HO are sensed by the Photorhabdus BaeSR pathway. Our data indicate that mdtABC expression is modulated by copper in vitro, but no change in P. luminescens copper resistance was observed with the mdtA mutant strain. This finding further suggests that the MdtABC pump may be one of multiple redundant systems contributing to metal homeostasis in Photorhabdus as demonstrated for Salmonella, and that other transporters may function independently or in synergy with MdtABC to increase copper resistance. This hypothesis is supported by our finding that mdtA gene was weakly expressed in the widely used laboratory LB medium, and that such expression was slighlty increased after copper supplementation. Copper in insects may derive from phenoloxidase that is a metalloenzyme involved in nodule formation [81]. As we already showed that the P mdtA activity was slightly increased (2.3 folds) by copper in vitro, we investigated whether copper induces MdtABC site-specific expression within the HO ex vivo. This seems consistent with the fact that proteases such as PrtA produced at later stages of infection could cause copper release from phenoloxidase in the HO, resulting in the increase in steady state levels of mdtABC expression. Nevertheless, we demonstrated that addition of a copper chelator to the HO extracts-containing medium did not affect GFP levels emitted by P. luminescens-P mdtA -gfp[AAV] comparing to a copper chelator-free medium, thereby suggesting that copper is not responsible of activating mdtABC expression within the HO.
The aim of this study was to elucidate the inducing cues and function of the putative MdtABC efflux pump in P. luminescens. Although no clear phenotype for the mdtA mutant has been identified, it is tempting to speculate that the Bae stress response pathway involving MdtABC may be involved in fostering Photorhabdus survival to the complex stresses of its hosts [82]. Microbes that can grow within the constraints of the insect cadaver may, themselves, produce stress through competitive or antagonistic interactions. Therefore, microbial activity in host tissues, or interactions with specific microbial taxa within the community, may induce multiple stresses that symbiotic Photorhabdus should overcome in order to increase their fitness and specific recruitment by the infective juvenile nematodes. These hypotheses are highly consistent with our findings concerning the site-specific expression of MdtABC during late infection stages prior to insect death. Further studies are required to determine the precise nature of signals involved in MdtABC induction as well as its natural substrates in P. luminescens. For instance, we suggest that these inducers could belong to by-products resulting from the action of proteases on HO matrix compounds. This would constitute a novel type of cross-talk, highlighting the versatility of efflux pumps in responses to the various compounds generated or encountered by bacteria in their surrounding environment.
Supporting information S1 Table. Minimum growth-inhibitory concentrations (MICs) a of SDS, Kanamycin, copper, and novobiocin, for wild-type P. luminescens TT01 (TT01 WT), the ΔmdtA mutant and its derivatives harbouring pBBR1-MCS5 or pBB-mdtABC, in the presence or absence of HO extracts. (PDF) S1 Fig. Heating of HO