The Global Transcription Factor Lrp Controls Virulence Modulation in Xenorhabdus nematophila

ABSTRACT

The bacterium Xenorhabdus nematophila engages in phenotypic variation with respect to pathogenicity against insect larvae, yielding both virulent and attenuated subpopulations of cells from an isogenic culture. The global regulatory protein Lrp is necessary for X. nematophila virulence and immunosuppression in insects, as well as colonization of the mutualistic host nematode Steinernema carpocapsae, and mediates expression of numerous genes implicated in each of these phenotypes. Given the central role of Lrp in X. nematophila host associations, as well as its involvement in regulating phenotypic variation pathways in other bacteria, we assessed its function in virulence modulation. We discovered that expression of lrp varies within an isogenic population, in a manner that correlates with modulation of virulence. Unexpectedly, although Lrp is necessary for optimal virulence and immunosuppression, cells expressing high levels of lrp were attenuated in these processes relative to those with low to intermediate lrp expression. Furthermore, fixed expression of lrp at high and low levels resulted in attenuated and normal virulence and immunosuppression, respectively, and eliminated population variability of these phenotypes. These data suggest that fluctuating lrp expression levels are sufficient to drive phenotypic variation in X. nematophila.

IMPORTANCE Many bacteria use cell-to-cell phenotypic variation, characterized by distinct phenotypic subpopulations within an isogenic population, to cope with environmental change. Pathogenic bacteria utilize this strategy to vary antigen or virulence factor expression. Our work establishes that the global transcription factor Lrp regulates phenotypic variation in the insect pathogen Xenorhabdus nematophila, leading to attenuation of virulence and immunosuppression in insect hosts. Unexpectedly, we found an inverse correlation between Lrp expression levels and virulence: high levels of expression of Lrp-dependent putative virulence genes are detrimental for virulence but may have an adaptive advantage in other aspects of the life cycle. Investigation of X. nematophila phenotypic variation facilitates dissection of this phenomenon in the context of a naturally occurring symbiosis.

INTRODUCTION

Bacteria commonly utilize phenotypic variation (also known as clonal variation) to adapt at the community level to rapid changes in environment, such as those that may be encountered by pathogens during host colonization (1–3). Broadly defined, phenotypic variation occurs when an isogenic population exhibits two or more distinct phenotypes. This nonconformity is heritable and can arise through genetic or epigenetic mechanisms (4–6). For many pathogens, phenotypic variation allows diversification of surface-exposed antigens and consequent evasion of host immunity (7) as in Neisseria gonorrhoeae, which displays variation in immunogenic pilin gene expression through a recA-dependent gene conversion at the pilE gene locus (8). In some bacterial species, stochastic variation in regulatory cascades leads to heterogeneous populations with cell-to-cell variation in gene expression profiles and phenotypes. This is the case in Pseudomonas aeruginosa, wherein the gene bexR, encoding a LysR-type transcription factor, is positively autoregulated. This autoregulation is thought to lead to bistable expression of BexR, with some cells expressing elevated levels of BexR and BexR-dependent gene products (9). Presumably, phenotypic variation systems allow pathogens to balance the need to respond rapidly to a changing environment while minimizing energetic costs of and negative host responses to virulence factor expression. The prevalence of phenotypic variation systems controlling virulence factor gene expression supports this idea.

The bacterium Xenorhabdus nematophila has a triphasic life cycle that includes both pathogenic and mutualistic host interactions, and like other pathogens, it exhibits phenotypic variation. X. nematophila infects insects by virtue of its association with the nematode Steinernema carpocapsae. The nematode infects and releases X. nematophila into the insect's blood-containing body cavity, wherein both partners contribute to suppressing insect immunity and causing insect death (10–13). Both partners multiply by consuming nutrients within the cadaver, putatively liberated by bacterially produced exoenzymes, such as proteases and lipases (14). The bacteria also produce compounds that inhibit the growth of microbial competitors and deter scavengers (15–17). Upon nutrient depletion, progeny nematodes and bacteria reassociate and exit the insect cadaver to hunt for a new host.

Prolonged incubation in the laboratory leads to X. nematophila phenotypic variation between two cell types termed primary and secondary (18). The mechanism driving this phenomenon is as yet unknown. Although the consequences of phenotypic variation can vary by strain, for X. nematophila ATCC 19061 (the strain used in this study), primary-form cells produce antibiotics and exoenzymes, such as proteases and hemolysins, bind bromothymol blue dye, and are motile. Secondary-form cells generally display reduced exoenzyme production and dye binding and are nonmotile (19, 20). Despite reduced production of putative virulence factors, such as hemolysin, secondary-form cells do not exhibit reduced virulence against insects. Primary- and secondary-form cells of this strain are also similarly capable of colonizing nematodes, and the selective advantage of the primary to secondary conversion is as yet unclear.

X. nematophila also exhibits phenotypic variation with respect to virulence. X. nematophila is a pathogen in its own right: injection of bacteria directly into the insect's blood-containing body cavity does not induce common insect immune responses, such as hemocyte aggregation and nodule formation, and results in the suppression of antimicrobial peptide gene expression followed by insect death (21–23). In a phenomenon termed virulence modulation (VMO), wild-type populations can harbor both virulent and attenuated subpopulations, both with primary-form characteristics (21). The virulent subpopulation is capable of host immune suppression, preventing both cellular (hemocyte aggregation/nodulation) and humoral (antimicrobial peptide production) responses (21).

The genes involved in VMO and the processes driving virulence heterogeneity are as yet unknown, but a likely candidate is the global regulator leucine-responsive regulatory protein (Lrp), which also has been implicated in the primary/secondary-form pathway (24). lrp mutants exhibit several but not all characteristics typical of secondary-form cells, suggesting that Lrp may be involved in the switch between the primary and secondary forms (24, 25). Furthermore, distinct from secondary-form cells, the lrp mutant exhibits defects in virulence and immunosuppression (24), raising the possibility that X. nematophila Lrp also could function in VMO. Lrp controls phenotypic variation pathways in other bacteria, such as in Escherichia coli (26), where it affects bias of type I fimbrial gene expression by binding in tandem with integration host factor (IHF) protein at the invertible promoter region (fimS), “trapping” the promoter in the on position (27). E. coli Lrp also regulates expression of genes encoding the pyelonephritis-associated pili (pap) by competing with Dam methylase for binding of two regions in the pap promoter. Depending on which region is bound by Lrp, transcription is either activated or repressed, resulting in stochastic expression and phenotypic variation (28). Based on the precedence for Lrp functioning as global regulator of metabolism and symbiosis in X. nematophila, as well as its role in governing variation of virulence traits in E. coli, we investigated its role in the VMO phenotypic variation pathway.

MATERIALS AND METHODS

Bacterial strains and growth conditions.

The bacterial strains and plasmids used in this study are listed in Table 1. Bacteria were grown in lysogeny broth (LB) culture medium (29) at either 30°C (for X. nematophila) or 37°C (for E. coli). The media used to culture X. nematophila were either stored in the dark or supplemented with 0.1% sodium pyruvate (30). Mini-Tn7 and lrp::kan constructs were delivered on plasmids into X. nematophila via conjugative transfer with E. coli, as described previously (25, 31). Antibiotics were used, as necessary, at the following concentrations: ampicillin (Ap), 150 μg ml⁻¹; chloramphenicol (Cm), 30 μg ml⁻¹ (E. coli) or 15 μg ml⁻¹ (X. nematophila); erythromycin (Erm), 200 μg ml⁻¹; and kanamycin (Km), 50 μg ml⁻¹.

TABLE 1.

Strains and plasmids used in this study

Strain or plasmid	Relevant characteristic(s)	Source or reference
Strains
S17-1(λpir)	E. coli strain for cloning and conjugations	58
HGB009	Bacillus subtilits AD263	A. Driks
HGB1087	X. nematophila wild-type ATCC 19061	21
HGB1941 to -1950	Independent isolates of HGB1087	This study
HGB1962	HGB1087 lrp::kan	This study
HGB1969	X. nematophila virulent (vir) isolate of HGB1087 transformed with pKV69 vector	This study
HGB1970	X. nematophila attenuated (att) isolate of HGB1087 transformed with pKV69 vector	This study
HGB1951 to -1960	Independent isolates of HGB1087 attTn7::Plrp-lacZ	This study
HGB1962	HGB1961 attTn7::Plrp-lacZ	This study
Plasmids
pKR100-lrp::kan	Suicide vector with lrp::kan disruption construct; Cmr	24
pKV69	Multicopy mobilizable vector; Cmr Tetr	59
pEH54	pKV69 with lrp in opposite orientation relative to Plac	This study
pEH56	pKV69 with lrp under control of Plac	This study
pTMO82	PCR2.1-TOPO with promoterless lacZ; Apr Kanr	60
pEVS107	oriR6K mini-Tn7 delivery vector; Ermr Kanr	61
pEH11	pTMO82 with lrp promoter (776 bp) inserted upstream of lacZ	This study
pEH13	pEVS107 with Plrp-lacZ cloned into mini-Tn7 from pEH11	This study

Construction of lrp reporter strains.

The entire intergenic promoter region upstream of the lrp gene was amplified from the genome of X. nematophila using the following primers: plrp large F (5′-AAAGGGCCCGATGAGTTTACTGTGTTTGGCTGTG-3′) and plrp R (5′-AAAACTAGTGGTGACAGACCTACTCGCTTGG-3′) (Integrated DNA Technologies, Coralville, IA) and standard PCR techniques with the PrimeSTAR high-fidelity polymerase (Clontech Laboratories, Inc., Mountain View, CA). The promoter region was introduced upstream of lacZ in the plasmid pTMO82 using restriction enzymes according to the manufacturer's directions (Promega, Madison, WI). The entire P_lrp-lacZ construct was removed from pTMO82 and introduced into the mini-Tn7 vector pEVS107 via standard restriction digestion and ligation. The resulting plasmids were sequenced at the University of Wisconsin—Madison Biotechnology Center using BigDye version 3.1 (Applied Biosystems, Foster City, CA). X. nematophila strains containing the lrp reporter were analyzed for β-galactosidase activity using standard methods (29).

Insect virulence assays.

Manduca sexta eggs (Carolina Biological Supply Company, Burlington, NC) were surface sterilized in a 0.6% bleach solution, washed with sterile water, and then reared on an artificial Gypsy moth wheat germ diet (MP Biomedicals, Aurora, OH) as previously described (32). For each of three experimental replicates, three biological replicates per bacterial strain were grown from single colonies overnight in LB (with added Cm in the case of strains carrying pKV69 or its derivatives). Strains were washed in phosphate-buffered saline (PBS), diluted in 10-fold serial dilutions, and plated on LB agar before and after injection to enumerate CFU. Approximately 10³ CFU of each replicate bacterial strain was injected into each of 10 fourth-instar M. sexta larvae using a 30-gauge syringe (Hamilton, Reno, NV). Larvae were monitored for mortality for at least 72 h subsequent to injection.

Western blotting to detect Lrp.

To determine Lrp protein levels in X. nematophila, strains were grown from single colonies to the stationary phase in LB (with added Cm in the case of strains carrying pKV69 or its derivatives). Cell extracts were generated by suspending cell pellets in lysis buffer (12.5 mM Tris-HCl [pH 6.8], 4% SDS, 0.2 mM phenylmethylsulfonyl fluoride [PMSF]), followed by 3 freeze-thaw cycles and centrifugation to remove cell membranes and membrane-associated material. Total cellular protein was quantified using standard Lowry assays (33) and mixed with 0.25 volumes of 4× sample buffer (200 mM Tris-HCl [pH 6.8], 8% SDS, 40% glycerol, 0.04% bromophenol blue, 5% β-mercaptoethanol). After boiling for 6 min, the indicated amounts of total cellular protein were separated by SDS-PAGE on 10% polyacrylamide gels. In the case of semiquantitative Western blotting experiments, cell extracts were diluted such that equal volumes were loaded for each sample. Following separation, proteins were transferred to Immun-Blot polyvinylidene difluoride (PVDF) membrane (Bio-Rad, Hercules, CA) in a Mini Trans-blot cell (Bio-Rad, Hercules, CA) for 1 h at 100 V.

Subsequent to transfer, blots were blocked in Tris-buffered saline solution containing 0.05% Tween 20 (TBST) with 5% nonfat dry milk (Lab Scientific, Livingston, NJ) overnight. Blots were then incubated for 1 h at room temperature in 20 ml TBST containing a polyclonal rabbit antiserum specific to E. coli Lrp (courtesy of R. M. Blumenthal, University of Toledo) in a 1:5,000 dilution. Following three washes with TBST, blots were incubated 0.5 h at room temperature with horseradish peroxidase (HRP)-conjugated goat anti-rabbit IgG antibody from Pierce (Thermo Fisher Scientific, Inc., Rockford, IL) in a 1:5,000 dilution in TBST and then washed three times with TBST again. Visualization was achieved using Pierce ECL (enhanced chemiluminescence) Western blotting substrate (Thermo Fisher Scientific, Inc., Rockford, IL) on Kodak BioMax XAR film (Carestream Health, Inc., Rochester, NY) per the manufacturer's instructions. Band intensity was quantified using ImageJ software (National Institutes of Health, Bethesda, MD), and then the intensity per nanogram of total protein was calculated and normalized against the lowest-intensity band within the compared group. All samples compared directly were analyzed on the same gel, and images were not manipulated in any way other than to remove samples not discussed in the text (negative controls and additional replicates).

Creation of fixed-lrp-expression vectors.

The lrp gene and its ribosomal binding site (but not promoter) were amplified from the X. nematophila genome with the primers lrp F apaI (5′-AAAGGGCCCGGGAAAATGTTATGGGTGTAGG-3′) and lrp R kpnI (5′-AAAGGTACCGCATCTGTACATTTCAGACAAAAG-3′) (Integrated DNA Technologies, Coralville, IA), using standard PCR techniques and PrimeSTAR high-fidelity polymerase (Clontech Laboratories, Inc., Mountain View, CA). Instead of using the engineered restriction sites, the lrp gene was cloned into pKV69 via blunt-ended ligation into the SphI-digested restriction site after filling in with DNA polymerase I Large (Klenow) fragment per the manufacturer's instruction (all enzymes from Promega, Madison, WI). This placed lrp downstream of the P_lac promoter native to pKV69, and the orientation of lrp was determined via restriction analysis. Two plasmids were selected for further study: one carrying lrp in the opposite orientation relative to P_lac (pEH54) and one carrying lrp in the same orientation as P_lac (pEH56). Plasmids were sequenced as described above. Plasmids pKV69 (vector control), pEH54 (low Lrp), and pEH56 (high Lrp) were introduced into X. nematophila by transformation as previously described (34), using LB medium containing Cm to select for transformants.

Analysis of Lrp-dependent phenotypes.

All assays were conducted by plating of stationary-phase cultures derived from three independent single colonies per strain. Protease activity (35), hemolysis of horse red blood cells (36), dye uptake on NBTA (nutrient agar supplemented with triphenyltetrazolium chloride and bromothymol blue) (37), lipase activity (38), and antibiotic activity (39) against Bacillus subtilis AD623 (courtesy of Adam Driks, Loyola University, Chicago, IL) were assayed as previously described. Bacterial motility was assessed by measuring the diameter of the migratory ring formed by each strain on 0.25% (wt/vol) LB agar approximately 16 h after plating. Congo red dye binding was determined by plating cultures on LB medium containing 20 μg/ml Congo red and 10 μg/ml Coomassie brilliant blue.

Detection of cecropin-6 transcripts in M. sexta.

Fourth-instar M. sexta larvae were reared and injected as described above. Approximately 7 h after injection, whole insects were snap-frozen and homogenized in TRIzol reagent (Invitrogen, Grand Island, NY). Five micrograms of total RNA was treated with RQ1 RNase-free DNase I (Promega, Madison, WI) and used as the template for reverse transcription using the Mg primer (21, 40), 5′-CGGGCAGTGAGCGCAACGTTTTTTTTTTTT-3′ (Integrated DNA Technologies, Coralville, IA), and avian myeloblastosis virus (AMV) reverse transcriptase (Promega, Madison, WI). Reverse transcriptase quantitative PCR (RT-qPCR) was performed with SYBR green supermix (Bio-Rad, Hercules, CA) on a Bio-Rad iCycler using a 1/10 dilution of cDNAs as a template. cecropin-6 transcript levels were measured and normalized against rpS3 using the following primers: cecropin-6 F (5′-GGTCAAAGGATTCGTGACGC-3′) and cecropin-6 R (5′-TTTGATTGTCCTTTGAAAATGGCG-3′) and rpS3 F (5′-ACTTCTCAGGCAAGGAGTGC-3′) and rpS3 R (5′-GTCACCAGGATGTGGTCTGG-3′). Data were normalized using the formula 2^Cq(rpS3)/2^{Cq(cecropin-6)} and are presented as a ratio of infected versus PBS-injected larvae, where C_q represents the quantification cycle.

Assay for M. sexta plasma antimicrobial activity.

Fourth-instar M. sexta larvae were injected with 10⁴ CFU of bacteria of the groups lrp::kan plus low Lrp (pEH54), lrp::kan plus high Lrp (pEH56), lrp::kan plus vector (pKV69), or PBS (control). Plasma from 10 larvae per group was harvested as described previously (41). To assess antimicrobial activity, cultures of E. coli DH5α at an optical density at 600 nm (OD₆₀₀) of 1.0 were diluted 1:100 in 99 μl of each plasma sample. Cultures were kept at 30°C with shaking, and the OD₆₀₀ was measured in a BioTek PowerWave 340 microplate reader (Biotek, Winooski, VT) every 30 min for 48 h. Growth inhibition (relative to PBS) was determined by dividing the OD₆₀₀ reading of E. coli grown in hemolymph from naive insects (injected with PBS) by the OD₆₀₀ reading of E. coli grown in hemolymph from infected insects. A value close to or below 1 indicates no inhibition of E. coli growth, and values above 1 indicate inhibition of E. coli growth. Data from three biological replicates were used for statistical analysis.

Characterization of growth in insect plasma.

Plasma from fourth-instar M. sexta larvae was obtained as described above. Three biological replicates per bacterial strain were grown from single colonies overnight in LB with Cm and subsequently diluted to an OD of approximately 0.02 in extracted plasma with Cm in a 96-well microplate. Cultures were grown at 30°C with shaking, and the OD₆₀₀ was measured in a BioTek PowerWave 340 microplate reader (Biotek, Winooski, VT) every 30 min for 48 h.

Microarray design and analysis.

A tiled microarray was designed by Roche NimbleGen (Madison, WI) based on the genomic sequences of X. nematophila strain ATCC 19061. Consecutive probes of 50 bp were used to represent the entire bacterial genome. For both ends of each probe, there were 25-bp sequences overlapping the adjacent probes. After removal of repetitive probe sequences from the array, a total of 192,509 probes (176,235 probes for chromosome sequences, 6,166 probes for plasmid sequences, 10,000 random sequence probes, and 108 probes for external control genes) were synthesized in duplicates on 385K chips. To prepare hybridization samples, overnight cultures were diluted 1:100 into 30 ml of LB supplemented with 0.1% pyruvate and 50 μg/ml ampicillin in 125-ml glass flasks and grown for 12 h to the early stationary phase (OD of 2 to 2.1) at 30°C at 150 rpm on a shaker. One milliliter of each culture was used to extract total RNA using the Qiagen RNeasy minikit, and on-column DNA digestion was performed using the Qiagen RNase-free DNase set according to the manufacturer's protocol (Qiagen, Valencia, CA). RNA quality and integrity were evaluated, and the samples were then submitted to Roche NimbleGen for microarray analysis. For each strain, the average signal strength of all random probes was used as the baseline signal level, and genes with an average signal strength higher than 5 times this value were considered expressed. The baseline signal strength value was subtracted from the average signal strength of each gene, and the resulting values were normalized using the values for the recA gene across the strains. The normalized values were then used for comparison between different strains, and a 2-fold change in average signal strength was used as the cutoff level for determining the significance of changes in gene transcript levels.

fliZ expression in X. nematophila.

Total RNA from stationary-phase cultures of X. nematophila was isolated using TRIzol reagent (Invitrogen, Grand Island, NY). Extracted RNA was treated with RQ1 RNase-free DNase I (Promega, Madison, WI) and used as the template for cDNA synthesis using random hexamers (Integrated DNA Technologies, Coralville, IA) and AMV reverse transcriptase (Promega, Madison, WI). RT-qPCR was performed with iTaq Universal SYBR green supermix (Bio-Rad, Hercules, CA) on a Bio-Rad CFX96 real-time PCR detection system using cDNAs as the template. fliZ transcript levels were measured and normalized against recA using the following primers: fliZ F (5′-TGTCTGTTACAACACAAAAGAAACG-3′), fliZ R (5′-AAGTGTTTTATGGCAGTGCAAA-3′), recA F (5′-TGTCCGTTTGGATATCCGCC-3′), and recA R (5′-CCCAGAGTATTAATACCTTCCCCAT-3′). Data were normalized using the formula 2^Cq(recA)/2^Cq(fliZ) and are presented as a ratio of the test sample versus the control (lrp::kan plus vector).

Statistical analyses.

Insect survival curves were plotted using the Kaplan-Meier method and analyzed by the log rank test using the Prism 5 software from GraphPad (GraphPad Software, Inc., La Jolla, CA). Significant differences in β-galactosidase activity and motility were distinguished by one-way analysis of variance (ANOVA) with a Tukey's post hoc test for multiple comparisons, also in Prism 5. For RT-qPCR analyses, data were transformed using the natural logarithm of the normalized C_q values and statistically analyzed using mixed-effect ANOVA with a Dunnett's or Tukey's post hoc test for multiple comparisons with Statistical Analysis System (SAS) version 9.3 (SAS Institute, Cary, NC). To determine if growth inhibition of E. coli was significantly different between the lrp::kan plus vector control and the other treatments, a Kruskal-Wallis one-way ANOVA with a Dunn's post hoc test was performed in Prism 5.

RESULTS

X. nematophila virulence is inversely correlated with Lrp protein levels and lrp promoter activity.

The established roles of the global transcription factor Lrp in X. nematophila virulence and as a modulator of phenotypic variation in other bacteria prompted us to examine its involvement in VMO. We first determined if Lrp protein levels vary between virulent and attenuated wild-type cells. Six individual colonies from a single wild-type population (HGB1087) were isolated and streaked onto LB plates to isolate three colonies from each of the original six isolates. These 18 colonies were grown in liquid LB to stationary phase, and the cultures were split for use in both a virulence assay using larval M. sexta and Lrp-specific Western blotting. Three of the six wild-type isolates killed all of the injected insects within 21 h postinjection, whereas the remaining three isolates exhibited attenuated virulence: two required 73 h to kill approximately 90% of insects, while the other killed only half of the insects by the end of the experiment (Fig. 1A). The replicates of each isolate exhibited consistent virulence phenotypes, suggesting that the parental (historical) phenotype is predominant among progeny cells within the VMO pathway. These results are consistent with previous reports from our laboratory (21).

FIG 1.

Lrp protein levels and virulence modulation. (A) Six wild-type (WT) X. nematophila isolates were grown in triplicate to the stationary phase, and approximately 2 × 10³ cells were injected into M. sexta (n = 10 for each replicate). Animals were subsequently analyzed for survival for 73 h. Different letters denote statistically significant differences via log rank analysis (P < 0.03). (B) Representative replicates of the same six wild-type isolates were analyzed by Western blotting for Lrp protein levels using an antibody specific to Escherichia coli Lrp (see Materials and Methods). For each sample, 10 ng of total cellular protein was loaded. The experiments depicted in panels A and B were independently repeated at least 3 times each.

Western blotting to detect steady-state Lrp levels in the 18 wild-type cultures revealed variability in Lrp levels among the 6 isolates (Fig. 1B) (data not shown). Surprisingly given the importance of Lrp in X. nematophila virulence (24), the isolates with the lowest steady-state levels of Lrp (isolates 1 to 3) were the most virulent, while the isolates with high levels of Lrp (isolates 4 to 6) were attenuated in virulence. More than 10 additional isolates were tested and yielded similar results (data not shown). Thus, our data indicate that X. nematophila populations exhibit cell-to-cell variation of both Lrp levels and virulence, suggesting Lrp is linked to VMO.

To determine if variability in Lrp protein levels is due to variability at the level of transcriptional regulation, we constructed lrp promoter-lacZ reporters integrated in single copy at the attTn7 site (25). Multiple (>10) exconjugants were selected for analysis subsequent to conjugation and transposition of the Tn7 reporter to ensure representation of all variants. The selected strains were then grown and the cultures split for injection into M. sexta to determine the virulence phenotype and for β-galactosidase activity assays, to assess lrp promoter activity. Of the four representative exconjugants depicted in Fig. 2A, two (E2 and E4) were fully virulent, killing ≥90% of M. sexta larvae within 27 h postinjection. The remaining two were attenuated, with >50% of insects surviving infection after 120 h. β-Galactosidase activity assays confirmed that the virulent subpopulation exhibits approximately 2.5-fold-reduced lrp promoter activity relative to the attenuated subpopulation (Fig. 2B). These results show that variability in Lrp protein levels is evident at the level of transcription, although additional posttranscriptional regulation events may also occur.

FIG 2.

lrp promoter activity and virulence modulation. (A) Four X. nematophila exconjugants carrying the lrp promoter-lacZ reporter were grown in triplicate to the stationary phase, and approximately 10³ cells were injected into M. sexta (n = 10 for each replicate). Animals were subsequently analyzed for survival for 110 h. Different letters denote statistically significant differences via log rank analysis (P < 0.0001). (B) The same four exconjugants were analyzed for lrp promoter activity via β-galactosidase assay (in triplicate). Different letters denote statistically significant differences via one-way ANOVA and a Tukey's multiple-comparison test (P < 0.05). One representative of at least 3 independent experiments is shown for both panels A and B.

In E. coli and Salmonella, lrp expression exhibits negative autoregulation (42–44). Regulatory feedback can result in multistable gene expression, suggesting a potential mechanism for Lrp-dependent X. nematophila VMO switching (6). To assess the autoregulatory capacity of Lrp, we monitored lrp promoter-lacZ reporter expression in an X. nematophila lrp mutant strain. The lrp mutant and the attenuated wild-type isolate displayed similar levels of lrp promoter reporter expression, both at higher levels than those observed in a virulent wild-type isolate (Fig. 3). A similar pattern among the strains was detected regardless of growth phase (data not shown). These results show that, as in E. coli, X. nematophila lrp is negatively autoregulated, which may contribute to the establishment of variable Lrp levels within the population.

FIG 3.

Autoregulation of lrp promoter activity. Attenuated (WT exconjugant 1) and virulent (WT exconjugant 2) wild-type and lrp mutant strains carrying the lrp promoter-lacZ reporter were analyzed for promoter activity via β-galactosidase assay (in triplicate). One representative from at least 3 independent experiments is shown, and different letters denote statistically significant differences via one-way ANOVA and a Tukey's multiple-comparison test (P < 0.05).

Fixed lrp expression eliminates virulence modulation.

To determine if varying the amount of cellular Lrp protein is sufficient to cause cell-to-cell differences in virulence, we constructed strains expressing lrp from plasmids at fixed high (pEH56) and low (pEH54) levels by cloning the lrp gene under the control of, or inversely oriented to, the lac promoter, respectively. The X. nematophila lrp mutant carrying each of these plasmids was then assessed for Lrp levels and virulence.

To confirm the effects of each multicopy plasmid on Lrp protein levels, we conducted semiquantitative Western blotting using an Lrp-specific antibody (Fig. 4) The virulent wild-type isolate has approximately 1.4-fold-lower steady-state levels of Lrp protein than the attenuated wild-type isolate (Fig. 4A), suggesting that VMO can occur within a narrow range of cellular Lrp levels. The lrp mutant carrying the low-lrp plasmid contains approximately 7.6-fold-lower steady-state levels of Lrp than even the virulent wild-type strain (Fig. 4B), whereas the high-lrp strain yields about 1.6-fold-greater Lrp than the attenuated wild-type strain (Fig. 4C). Generally, these relative expression trends remained consistent across multiple (>3) independent transformants (data not shown), suggesting that potential variability in plasmid copy number would not hinder the use of these strains as fixed low- and high-lrp populations.

FIG 4.

Lrp protein levels in fixed-expression strains. Representative replicates of the virulent and attenuated wild-type and fixed lrp expression strains were analyzed for Lrp protein levels via semiquantitative Western blotting using an antibody specific to Escherichia coli Lrp (see Materials and Methods). For each strain, the lightest band in the dilution series was quantified (see Materials and Methods), and band intensity was normalized to the lower-intensity band of the two compared strains. In panel A, total protein isolated from wild-type strains was compared using serial 2-fold dilutions. In panel B, total proteins isolated from wild-type virulent and low-lrp strains were compared using serial 4-fold dilutions. In panel C, total proteins isolated from wild-type attenuated and high-lrp strains were compared using serial 4-fold dilutions.

We next examined the effects of fixed lrp expression on X. nematophila virulence against M. sexta. Both virulent and attenuated wild-type strains carrying the vector control exhibit virulence phenotypes consistent with previous experiments (Fig. 5A), indicating that the multicopy vector has no effect on virulence modulation. Consistent with previous findings (24), the lrp mutant strain carrying the vector control plasmid was attenuated in virulence, taking 32 h to kill approximately 90% of injected insects compared to the 23 h required by the virulent wild-type strain.

FIG 5.

Effect of fixed lrp expression on virulence. (A) Wild-type or lrp mutant strains of X. nematophila carrying the vector control or low- or high-lrp plasmids were grown in triplicate to stationary phase, and approximately 10³ cells were injected into M. sexta (n = 10 for each replicate). Animals were subsequently analyzed for survival for 96 h. All strains exhibit significantly distinct virulence phenotypes via log rank analysis (P < 0.001). (B) Wild-type and lrp mutant strains were transformed with the vector, low-lrp, or high-lrp plasmids, and multiple independent transformants were selected and grown to stationary phase. Approximately 10³ cells were injected into M. sexta (n = 10 for each independent transformant), and animals were monitored for survival. Calculations of the time required to kill 20% of injected insects (LT₂₀) were performed using linear regression and plotted such that independent transformants are represented by individual data points. The dotted line represents the time at which the experiment was ended (i.e., the limit of detection of death). Data points above the dotted line represent experiments in which >80% of insects remained alive at the time the experiment concluded (66 h).

The low- and high-lrp plasmids have distinct effects on the virulence of X. nematophila. When introduced into the lrp mutant strain, the low-lrp plasmid restores virulence to the bacterium: more than 90% of insects succumb to infection within 25 h postinjection (Fig. 5A). In contrast, the strain carrying the high-lrp plasmid exhibits an even greater virulence defect than the lrp mutant parental strain, requiring 68 h to kill 95% of insects (Fig. 5A), demonstrating that the high levels of Lrp are sufficient to attenuate virulence in X. nematophila. However, the attenuated wild-type strain is still significantly less virulent than the high-lrp strain, despite the fact that the former expresses 1.6-fold-lower steady-state Lrp levels than the latter (Fig. 4B). This lack of direct proportionality between Lrp levels and virulence may be due to complexities in the control of downstream genes by this global regulator. Multiple independent transformants of the lrp mutant parental strain carrying either the high or low-lrp plasmids displayed consistent virulence phenotypes, measured as the time required to kill 20% of the injected insects (LT₂₀) (Fig. 5B). Transformants of the wild-type strain with the vector control plasmid exhibit characteristic VMO, dividing into two populations with a greater than 3-fold difference in LT₂₀ between them, resulting in a >27.5-h standard deviation from the mean (Fig. 5B). In contrast, the lrp mutant strain carrying the vector, as well as the low- and high-lrp strains, each exhibit standard deviations of ≤3 h among their LT₂₀ values. These results indicate that fixed expression of lrp is sufficient to eliminate the variability characteristic of VMO and further establish a role for Lrp in controlling phenotypic variation of virulence in X. nematophila.

High lrp expression levels eliminate host immune suppression.

In addition to virulence, the ability of X. nematophila to suppress insect immunity, including antimicrobial gene expression, is subject to variability (21). To further establish the role of lrp expression levels in VMO switching, we analyzed the ability of the fixed-lrp-expression strains to suppress production of the antimicrobial peptide cecropin-6 in M. sexta hosts. Consistent with the virulence phenotype, insects injected with the attenuated wild-type, lrp mutant, and the high-lrp strains expressed higher levels of cecropin-6 (∼150-, 680-, and 380-fold induction over PBS controls, respectively) than those injected with the virulent wild-type and low-lrp strains (50- and 40-fold induction, respectively) (Fig. 6). In addition, we monitored overall antimicrobial activity, measured as inhibition of E. coli growth, in hemolymph extracted from insects injected with high- and low-lrp strains. Plasma derived from hemolymph extracted from insects injected with the low-lrp strain had significantly lower antimicrobial activity than the lrp mutant carrying the vector control. In contrast, the high-lrp strain was not significantly different from the vector control strain, consistent with the idea that the lrp mutant and the high-lrp strain are defective in suppressing induction of antimicrobial activities. Taken together, these results suggest that while immune suppression is an Lrp-dependent phenomenon, high levels of lrp expression are sufficient to either prevent or counter this immunomodulation.

FIG 6.

Antimicrobial activity in insects injected with X. nematophila expressing different levels of Lrp. (A) cecropin-6 levels in fourth-instar M. sexta larvae injected with PBS (control), an X. nematophila lrp::kan mutant carrying the vector or low- or high-lrp plasmids, or virulent and attenuated strains of X. nematophila carrying a vector control. Total RNA was isolated 7 h postinjection. M. sexta cecropin-6 transcription is presented as relative abundance of transcript normalized to M. sexta rpS3. The data represent the means from five independent experiments. Asterisks denote significant difference (P < 0.05) in mean values of treatments compared to the PBS control via a mixed-effect ANOVA with a Dunnett's post hoc test. (B) Growth inhibition of E. coli cells subcultured into plasma harvested from fourth-instar M. sexta larvae injected with X. nematophila expressing different levels of Lrp. Insect plasma was harvested from injected insects 20 h postinjection and used as the growth medium for E. coli DH5α. Growth inhibition is directly correlated with antimicrobial activities in plasma from infected insects. Plasma from lrp::kan-plus-vector-injected insects served as a positive control for antimicrobial activity, while plasma from naive insects (PBS injected) served as a negative control. The data represent the means from three biological replicates. An asterisk denotes significant difference (P < 0.05) in mean values of treatments compared to the lrp::kan plus vector control via a Kruskal-Wallis one-way ANOVA with a Dunn's post hoc test.

Fixed expression of lrp influences additional characteristic phenotypes of X. nematophila.

Lrp positively regulates many genes involved in various characteristic X. nematophila phenotypes believed to play a role in host association. These include motility, antibiotic, hemolysin, and lipase production, an additional phenotypic variation pathway involving a switch from primary- to secondary-form cells, and host immune suppression (21, 24) (Table 2). We analyzed the fixed high- and low-lrp strains with respect to these phenotypes (Table 2) to help clarify the components of the Lrp regulon that contribute to VMO switching. A characteristic of primary-form cells is their ability to bind bromothymol blue dye, rendering their colonies blue on NBTA agar plates compared to the red color of secondary-form colonies (such as the lrp mutant strain) on this medium (19, 20). While both the wild-type and the high-lrp strains of X. nematophila appear blue on NBTA, the low-lrp strain appears purple, representing an intermediate phenotype.

TABLE 2.

Selected phenotypes of wild-type and fixed-lrp-expression X. nematophila strains

Strain type	Plasmid type	Dye bindinga		Enzyme activityb			Antibiotic productiond	Motility (mm)c
		BTB	Congo red	Lipase	Protease	Hemolysin
lrp::kan	Vector	−	++	+/−	−	−	−	5 ± 0 A
Virulent	Vector	+	+	+	+	+	+	47.7 ± 4.1 B
Attenuated	Vector	+	−	+	+	+	++	54.0 ± 4.3 C
lrp::kan	Low lrp	+/−	+	−	−	−	−	29.0 ± 3.5 D
lrp::kan	High lrp	+	−	+	+	+	++	59.0 ± 4.5 C

^aQualitative evaluation of colony color on plates containing bromothymol blue (BTB) or Congo red 8 days after inoculation. −, no detectable dye binding; +/−, intermediate binding; +, strong dye binding; ++, stronger dye binding.

^bQualitative determination of halo formed around colony 3 days after inoculation on plates containing Tween 20, milk, or horse blood.

^cEach value indicates the mean diameter of the colony and motility ring ± standard error (SE) 22 h postinoculation on 0.25% agar plates. The letters indicate distinct statistical groups as determined by 1-way ANOVA analysis and Tukey's multiple-comparison test (P < 0.05).

^dQualitative evaluation of the zone of clearance formed by lack of B. subtilis growth surrounding the X. nematophila colony 2 days postinoculation.

The low-lrp strain exhibits reduced lipase activity and motility compared to the attenuated wild-type and high-lrp strains (Table 2). This strain also lacks detectable protease and hemolytic activities, as well as antibiotic production. Similarly, the virulent wild-type strain of X. nematophila exhibits reduced lipase activity and antibiotic production relative to the attenuated wild-type and high-lrp strains. Taken together, these results suggest that relatively greater bromothymol blue binding, motility, antibiotic production, lipase, protease, or hemolytic activity does not contribute to virulence against M. sexta. If anything, enhanced expression of the genes involved in these phenotypes may be detrimental to pathogenesis.

We noticed enhanced pellicle formation by the virulent strains of X. nematophila in culture tubes during laboratory growth (data not shown). It is well established that extracellular polysaccharide production and pellicle (or more generally, biofilm) formation can contribute to virulence (45). We analyzed the ability of our fixed-lrp-expression strains to bind Congo red, which can be an indicator of polysaccharide production (46). The lrp mutant strain appeared to have the greatest ability to absorb the Congo red dye, since its colonies had a red appearance on agar plates containing the dye, while those of the virulent wild-type and low-lrp strains were pink and the attenuated wild-type and high-lrp strains were white. Thus, polysaccharide production may play a role in virulence; however, its abundance as indicated by Congo red binding does not directly correlate with virulence, since the lrp mutant exhibits enhanced Congo red binding but attenuated virulence.

Lrp positively regulates fliZ expression.

Our data support a model in which X. nematophila (strain ATCC 19061) cell-to-cell heterogeneity of Lrp levels results in phenotypic variation of virulence and other phenotypes, including motility. A recent report demonstrated that the X. nematophila (strain F1) regulator FliZ controls heterogeneous expression of genes within the flagellar regulon and several other loci, including two, xhlBA and xaxAB, that are known to be Lrp dependent (24, 47, 48). The X. nematophila F1 FliZ regulon revealed by transcriptome sequencing (RNA-seq) analysis includes 278 coding sequences, also including several (e.g., nilR) members of the established Lrp regulon. These findings raise the possibility that FliZ and Lrp are part of an overlapping regulatory system that coordinately controls large sets of genes with known or predicted roles in host interactions. The FliZ regulon reported by Jubelin et al. does not include lrp. In contrast, microarray analyses comparing the wild type to an lrp mutant indicate that Lrp positively regulates fliZ, as well as 148 of the other 226 FliZ-regulated genes common to both strains (see Table S1 in the supplemental material). RT-qPCR assays revealed that the high-lrp strain had significantly higher fliZ transcript levels than the low-lrp strain (Fig. 7), suggesting that Lrp positively regulates fliZ gene expression.

FIG 7.

fliZ expression is modulated by Lrp. fliZ expression in stationary-phase cultures of the X. nematophila lrp::kan mutant (HGB1962) carrying the vector or low- or high-lrp plasmids. fliZ transcript levels are presented as relative abundance of transcript normalized to recA. The data represent the means from six biological replicates. Different letters denote significant difference (P < 0.05) in mean values of test samples compared to the control (lrp::kan plus vector) via a mixed-effect ANOVA with a Tukey's post hoc test.

DISCUSSION

X. nematophila lrp expression exhibits population heterogeneity and correlates with VMO.

In this work, we establish that variable expression of the gene encoding the global regulator Lrp is a key component of the VMO pathway. Approximately 2-fold fluctuations in steady-state levels of lrp gene expression occur at the level of transcription (Fig. 2B) and are reflected in Lrp protein levels (Fig. 1B and 5A). These results correlate with VMO phenotypes, where subpopulations of X. nematophila with low steady-state levels of Lrp exhibited the highest observed level of virulence, compared to the subpopulations with higher Lrp levels, which exhibited attenuated virulence against M. sexta larvae (Fig. 1A and 2A). Given that X. nematophila pathogenicity is likely multifactorial (15, 49), it follows that a global regulatory protein (and established virulence factor) such as Lrp plays a key role in modulating virulence.

Fluctuations in lrp expression may arise via negative autoregulation (Fig. 3). Previous work conducted by Becskei and Serrano using engineered gene circuits showed that autoregulatory negative-feedback loops can result in unstable gene expression when the repressor binds with low affinity (50), and indeed Wang et al. demonstrated that E. coli Lrp binds relatively weakly upstream of its own promoter (42). Furthermore, E. coli Lrp protein typically binds multiple adjacent sites at each regulated promoter and exhibits reduced sequence discrimination relative to most other transcriptional regulators (51). Lrp occurs as a mixture of octamers (favored in the presence of leucine) and hexadecamers (52), and these combinations have different DNA-binding capabilities, depending on the promoter (53). Taken together, promiscuous, quaternary structure-dependent binding by Lrp allows for multiple means by which autoregulation may introduce cell-to-cell heterogeneity in lrp gene expression. Additionally, other regulators or epigenetic modifications (such as methylation) may influence the ability of Lrp to bind its own promoter, contributing to variability.

lrp expression variation is sufficient to modulate the virulence phenotype.

We found that fixed expression of lrp at low or high levels using nonnative promoters on a multicopy plasmid removed fluctuations in virulence phenotypes (Fig. 5B). Combined with the result that the high-lrp strain exhibits attenuated virulence and the low-lrp strain exhibits wild-type virulence (Fig. 5A), these results suggest that in X. nematophila ATCC 19061, variable lrp expression is sufficient to modulate changes in virulence. The inverse relationship between Lrp levels and virulence is unexpected but intriguing and may reflect that overstimulation of the Lrp regulon is detrimental during initial stages of infection, as discussed further below. Minor differences in virulence exist between the severely attenuated wild-type strain and the somewhat less-attenuated high-lrp strain, even though the latter has higher steady-state levels of Lrp (Fig. 4C and 5A). In the wild-type strain, Lrp expression potentially is subject to multiple layers of regulation that are removed in the fixed-expression strain. Furthermore, we monitored steady-state Lrp levels under in vitro laboratory growth conditions, which do not reflect potentially dynamic changes in Lrp expression that may be occurring in the wild-type attenuated strain during insect infection. It is possible that in the insect environment, this strain responds to environmental conditions to modulate Lrp expression such that in vivo, populations of this strain express higher levels of Lrp than those of the high-lrp strain. Future work will focus on the development of a single-cell reporter of VMO switch status to analyze the timing, frequency, and effects of its occurrence in vivo.

Our data show that the X. nematophila ATCC 19061 Lrp regulon includes a subset of those genes identified as FliZ dependent in X. nematophila F1 (see Table S1 in the supplemental material) (24, 47, 48). Furthermore, Lrp positively regulates FliZ (Fig. 7), demonstrating that FliZ is an Lrp-dependent gene. X. nematophila F1 displays bimodal expression of FliZ-dependent genes, with levels of FliZ corresponding to the observed “on” and “off” states of gene expression (24, 47, 48). Although strain differences in regulatory hierarchies may exist, taken together, these data suggest that Lrp levels may influence the heterogeneous expression of some genes through regulation of fliZ. However, the specific details of this regulatory relationship await a more detailed analysis of direct, indirect, and reciprocal effects of Lrp and FliZ on each other's expression, as well as characterization of both regulons within a single strain background.

Lrp and the VMO phenotype.

Our data show that low levels of Lrp protein are optimal for X. nematophila virulence, while high levels are detrimental. The physiological basis of this difference awaits elucidation. The high-lrp strain exhibits no defect in growth relative to the low-lrp strain in cell-free hemolymph isolated from insect larvae (see Fig. S1 in the supplemental material), suggesting the virulence attenuation is not due to failure to utilize insect nutrients for growth. In addition, our phenotypic analyses do not support the idea that the high-lrp strain is expressing reduced virulence or immunosuppressive factor production, since each tested Lrp-dependent bacterial activity was expressed at greater levels in the high-lrp strain than in the low-lrp strain (Table 2). However, we have not ruled out the possibility that high levels of Lrp inhibit the expression of specific key virulence factors. Indeed, in E. coli, expression of the glutamate synthase operon (gltBDF) is activated by low levels of Lrp and inhibited at high cellular Lrp levels (43).

The inability of the high-lrp strain to suppress the induction of antimicrobial activities in the insect (Fig. 6) suggests an alternative hypothesis that attenuation is caused by this strain's susceptibility to immune clearance. The high-lrp strain may prematurely trigger a successful immune response—for example, via elevated expression of immune-stimulating factors or degradative factors that cause tissue damage, before the immune response can be circumvented by Lrp-dependent immunosuppressive activities. In this model, the negative effects (e.g., premature immune stimulation) of high levels of Lrp protein would negate the effects of increased virulence factor production. This underlying cause of attenuated virulence may be distinct from that for the lrp mutant strain, which presumably is the loss of virulence factor expression, and these distinct causes of attenuation may explain their distinct attenuated virulence phenotypes (Fig. 5A). Future experiments will help define if VMO-attenuated virulence and immunosuppression are due to reduced production of specific factors or overwhelming immune stimulation through direct or indirect pathways.

Although X. nematophila Lrp-dependent proteases, lipases, antibiotics, hemolysins, and motility have each historically been considered likely virulence determinants, several studies have supported the alternate view that these activities are expressed and functional after insect death, potentially functioning to promote nutritional acquisition by both the bacterium and nematode within the cadaver (16, 17, 54). The virulent wild-type isolate and the low-lrp strain both express lower levels of these activities than do the attenuated strains (wild type or high lrp). This raises the possibility that the high-lrp variants are better adapted than their low-lrp counterparts to postcadaver environmental conditions. Additionally, these data further support the idea that the virulent phenotype of the low-lrp variants is due either to reduced levels of these potentially immunogenic activities or to increased expression of as-yet-unidentified bona fide virulence determinants. Future analysis of the expression profiles of variants should help distinguish between these possibilities and yield further insights into molecular factors used by X. nematophila to engage in either mutualistic or pathogenic host interactions and how these two phenomena are coordinately regulated.

The relationship of the VMO pathway with phenotypic variation pathways in X. nematophila and related species.

Previous studies of X. nematophila and the related bacterium Photorhabdus luminescens reported a phenotypic variation pathway wherein a distinct subpopulation of cells, termed “secondary form,” arise upon prolonged stationary-phase culturing. These secondary-form cells differ from the other, “primary-form” population in many ways, including reduced motility, decreased ability to bind dyes, such as bromothymol blue, and altered production of proteases, lipases, antibiotics, and protoplasmic paracrystalline inclusions containing crystal proteins (20, 35, 55, 56). Despite these differences, primary- and secondary-form variants of strain ATCC 19061 (the strain used in this study) were equally virulent against M. sexta (20). Lrp appears to regulate the primary/secondary-form phenotypic variation pathway in X. nematophila, based on the fact that an lrp mutant strain exhibits most of the known characteristics of secondary-form populations (24). However, although both phenotypic variation pathways are Lrp regulated, they appear to be distinct from each other, since (i) virulence is affected by VMO but not the primary/secondary-form switch, and (ii) both virulent and attenuated variants exhibit primary-form characteristics (Table 2).

An additional virulence-related phenotypic variation pathway occurs in P. luminescens, an insect-pathogenic bacterium closely related to X. nematophila that also engages in mutualism with nematodes (57). In P. luminescens, an invertible promoter controls expression of the genes encoding maternal adhesion (Mad) fimbriae, necessary for association with nematode hosts. Switching between promoter orientations (on or off) results in two phenotypically distinct populations. The M-form cells, characterized by the expression of Mad fimbriae, are attenuated in virulence but can colonize nematodes and produce low levels of exoenzymes, such as antibiotics and hemolysins. The P-form cells do not express Mad fimbriae and cannot colonize nematodes but are virulent and produce increased amounts of exoenzymes relative to the M form. The role of Lrp in this pathway, if any, is unknown. However, this “Mad switch” variation pathway is different from X. nematophila VMO, in which the virulence-attenuated population exhibits higher levels of exoenzyme production relative to the virulent population (Table 2). This parallel does raise the interesting question of whether or not the X. nematophila VMO pathway functions to control the transition from mutualistic to pathogenic lifestyles. Our laboratory and others are investigating the contribution of individual bacterial factors to either mutualism or pathogenesis.

Conclusions.

This work establishes a novel role for the global transcriptional regulator Lrp in X. nematophila virulence against the insect host, such that expression of Lrp, regulated at the transcriptional level, is variable and inversely correlated with virulence. The VMO phenomenon may have implications in X. nematophila mutualism with its nematode host, as the subpopulation expressing high levels of Lrp protein exhibits increased production of phenotypes associated with promoting nematode fecundity. This idea presents a model in which the VMO pathway facilitates the transition of X. nematophila populations between two distinct hosts (nematode and insect) and host interaction strategies (mutualism versus pathogenesis).