Characterization of a lipoprotein, NilC, required by Xenorhabdus nematophila for mutualism with its nematode host

Authors


Summary

Xenorhabdus nematophila is a γ-proteobacterial mutualist of an insect-pathogenic nematode, Steinernema carpocapsae. X. nematophila requires nilC, a gene predicted to encode an outer membrane lipoprotein of unknown function, for colonization of its nematode host. Characterization of NilC, described here, demonstrated it is a 28 kDa lipoprotein directed to the periplasm by an N-terminal signal sequence. Lipidation and processing of NilC occurs by a mechanism that is conserved in proteobacteria. This work also showed NilC is membrane associated and oriented towards the periplasm of X. nematophila and is produced as an outer membrane-associated protein when expressed in Escherichia coli. Expression analyses revealed that nilC transcription is directly or indirectly repressed by Lrp, and this regulatory link may explain the nematode mutualism defect of a previously identified lrp::Tn5 mutant. An lrp::Tn5 mutant produces an additional nilC transcript, not observed in wild-type cells growing in vitro, and produces ≈ 75-fold more nilC than wild-type cells in late stationary phase. These fundamental characterizations of nilC expression and nilC localization and processing events have provided firm bases for understanding the role of this colonization factor in the X. nematophila/S. carpocapsae microbe–host interaction.

Introduction

Xenorhabdus nematophila, a member of the Enterobacteriaceae, has co-evolved a mutualistic relationship with an insect parasitic nematode, Steinernema carpocapsae. The free-living infective juvenile (IJ) stage of S. carpocapsae nematodes is non-feeding and has an intestinal region, termed the vesicle, colonized by a monoculture of ≈ 20–200 X. nematophila bacteria (Bird and Akhurst, 1983; Heungens et al., 2002; Martens et al., 2003). Nematode IJs infect an insect prey by breaching the cuticle and entering the blood system, where X. nematophila is then released from the vesicle by defecation. After killing the insect, X. nematophila and S. carpocapsae reproduce until the insect cadaver is spent, at which point they re-associate into the colonized IJ form and migrate into the soil to infect a new insect host. This re-association event occurs in a species-specific manner, the mechanism of which is not understood, even though S. carpocapsae has access to other microbes within the insect cadaver (Maxwell et al., 1994; Forst and Clarke, 2002). The process of colonization occurs through repeated bursts of growth by a small population of founding cells, resulting in IJs that are colonized by mono- or biclonal populations of X. nematophila bacteria (Martens et al., 2003). Once the bacterial population has fully grown, it is maintained in a viable state in the vesicle for periods of up to several months (Kung et al., 1990).

We have been utilizing X. nematophila as a model to better understand general processes underlying the development of long-term, mutually beneficial relationships between animals and microbes. As part of this effort, we have identified X. nematophila genes required for normal colonization of the vesicle. Among these are several genes (rpoS, rpoE, lrp and nilD) encoding putative regulators of gene expression (Vivas and Goodrich-Blair, 2001; Heungens et al., 2002), the downstream targets of which are, as yet, undiscovered. In addition to these regulators, we previously identified three genes, nilA, nilB and nilC (nematode intestine localization), which are involved in nematode colonization. nilA, nilB and nilC are grouped in a 3.5 kb gene cluster termed SR1 (Symbiosis Region 1) and putatively encode an integral inner membrane protein, an outer membrane β-barrel-type protein and an outer membrane lipoprotein respectively (Heungens et al., 2002). A nilA mutant is attenuated in colonization of S. carpocapsae nematodes (to approximately one-twentieth that of wild-type levels), and nilB and nilC mutants are unable to colonize nematodes to detectable levels in our assays. The function(s) of the nilA, nilB and nilC gene products in nematode colonization are unknown.

NilC, the focus of this study, is predicted to be an outer membrane-associated lipoprotein based on the fact that its N-terminus is probably a hydrophobic α-helix that is followed by cysteine and serine residues (Heungens et al., 2002). In various proteobacteria, lipoproteins are involved in microbe–environment interactions, serving in such functions as nutrient acquisition (Janulczyk et al., 1999; Cornelissen, 2003), signal transduction (Otto and Silhavy, 2002) and adhesion (Jin et al., 2001). Lipoproteins are defined by the covalent attachment of three acyl tails to their N-termini that allow them to associate with the periplasmic leaflet of the inner membrane or either leaflet of the outer membrane (Wu, 1996). Another general feature of lipoproteins is an N-terminal signal sequence that directs the premature protein from the cytoplasm to the periplasm where, after attachment of a diacyl glycerol moiety to a cysteine residue adjacent to the signal sequence, the signal sequence is removed by signal peptidase II (Paetzel et al., 2002). After signal sequence removal, the amino group of the newly exposed N-terminus is acylated (Wu, 1996). The Lol sorting system directs mature lipoproteins to their final destination in the cell based on the identity of the amino acid residue after the lipoprotein's acylated N-terminal cysteine (Yakushi et al., 1998; 2000; Narita et al., 2002; Hara et al., 2003). Typically, proteins with a serine in this position are sorted to the outer membrane, whereas proteins with an aspartic acid in this position remain in the inner membrane (Yamaguchi et al., 1988). However, this residue is not always the protein's sole sorting determinant (Terada et al., 2001).

Because sequence information has yet to provide clues to NilC function (Heungens et al., 2002), we chose to conduct a fundamental characterization of this colonization factor. Here we describe expression and biochemical analyses of NilC regulation, processing and localization.

Results and discussion

NilC is a 28 kDa protein that is necessary for nematode colonization

To facilitate genetic manipulation of nilC, we deleted the entire SR1 locus from the chromosome of wild-type X. nematophila. As expected, this mutant, designated XnΔSR1, was unable to colonize S. carpocapsae nematodes to a detectable level (Fig. 1A) and did not produce a protein of the approximate size predicted for NilC (28 kDa), as detected by immunoblot with anti-NilC antibodies (Fig. 1B). To stably introduce a single copy of the SR1 locus into the XnΔSR1 chromosome, we utilized a mini-Tn7 transposition method (Bao et al., 1991; Stabb and Ruby, 2002; Martens et al., 2003). The SR1 locus was cloned into a mini-Tn7 transposon and transposed into the attTn7 site of the XnΔSR1 chromosome. The resulting strain, XnΔSR1(Tn7/SR1), colonized S. carpocapsae nematodes at wild-type levels (Fig. 1A) and produced roughly wild-type levels of NilC, detected by immunoblot (Fig. 1B).

Figure 1.

nilC is necessary for colonization and production of a 28 kDa protein. (A) Colonization of S. carpocapsae nematodes by and (B) immunoblot detection of NilC in wild-type (white, lane 1), XnΔSR1 (black, lane 2), XnΔSR1(Tn7/SR1) (grey, lane 3) and XnΔSR1(Tn7/SR1/CMZ) (checkered, lane 4) cells. ‘L.D.’ denotes the limit of detection for the colonization assay. Error bars indicate standard deviations of five replicate plates. Numbers to the left of the blot shown in (B) indicate the migration points of protein size standards.

The nilC open reading frame encodes nine methionine codons that could serve as potential start codons. To determine whether, as predicted, the 5′-most ATG codon serves as the start codon for nilC, it was mutated to a TAG stop codon in the pTn7/SR1 construct. This mutant construct (pTn7/SR1/CMZ) was integrated into the attTn7 site of the XnΔSR1 chromosome. The resulting strain was unable to colonize S. carpocapsae nematodes to a detectable level (Fig. 1A) and did not produce the 28 kDa protein that was detected by immunoblot in wild-type cells (Fig. 1B). These results indicate that the 5′-most ATG of the nilC open reading frame is probably the nilC start codon and verify the specific requirement for NilC in nematode colonization.

NilC levels are maximal during stationary phase and are greatly increased in an lrp mutant

Our previous work indicated that nilC is expressed in the colonization-defective rpoS::Tn5 and lrp::Tn5 mutants (Heungens et al., 2002). However, in these studies, the timing and levels of nilC expression were not quantified, nor was an anti-NilC antibody available to allow measurement of NilC protein levels. In addition, nilC transcript levels were not examined in the nilD::Tn5 mutant, a nematode colonization mutant with a disruption in a gene encoding a small RNA (NilD RNA) with possible regulatory functions (A.W. Andersen, K. Heungens, K.M. Wassarman and H. Goodrich-Blair, unpubl. data). To monitor the expression profile of NilC and to test whether the putative regulators of colonization have an effect on this expression, NilC levels were quantified by immunoblot analysis and compared in parallel cultures of wild-type, lrp::Tn5, nilD::Tn5 and rpoS::Tn5 strains. Wild-type, nilD::Tn5 and rpoS::Tn5 cells each exhibited ≈ 20- to 40-fold higher NilC levels in late stationary phase versus log phase (Table 1). Late-stationary-phase NilC levels in the nilD::Tn5 and rpoS::Tn5 mutants, quantified in eight parallel cultures, were not grossly different from wild type (Table 1). Finally, all strains examined showed lower NilC levels in late log phase, relative to all other points during growth (Table 1). We believe that the drop in NilC levels from early log to late log phase reflects dilution and/or degradation of the higher levels of NilC that accumulate in cells during late stationary phase to achieve a lower steady-state level during log phase. This general trend is routinely observed when monitoring expression of genes whose expression is increased during stationary phase relative to log phase (Lange and Hengge-Aronis, 1991).

Table 1. NilC levelsa during LB growth.b
StrainEarly logLate logEarly stationaryLate stationary
  • a. 

    Micrograms of NilC/OD600 ± standard deviation from the mean of two replicate samples, except where noted by (c).

  • b. 

    Early log: OD600∼0.2; late log: OD600∼1.0; early stationary: OD600∼3.5; late stationary: 16 h after early stationary time point.

  • c. 

    Data represent measurements from eight replicate samples.

Wild type0.12 ± 0.020.09 ± 0.010.21 ± 0.064.20 ± 0.07c
nilD::Tn50.20 ± 0.040.14 ± 0.030.33 ± 0.024.04 ± 0.28c
rpoS::Tn50.24 ± 0.050.13 ± 0.030.26 ± 0.06 5.0 ± 0.16c
lrp::Tn52.21 ± 0.400.46 ± 0.023.49 ± 0.05 318 ± 38

The lrp::Tn5 mutant also showed an accumulation of NilC during late stationary phase, relative to levels during log phase. However, NilC levels in this mutant were much higher than in wild-type cells at every point during growth. In late-stationary-phase lrp::Tn5 cultures, NilC was present at ≈75-fold higher levels, relative to late-stationary-phase wild-type cultures (Table 1). These results indicate that NilD RNA and RpoS are probably not transcriptional regulators of nilC expression. Lrp, on the other hand, appears to play a major role, either directly or indirectly, in regulating NilC levels.

Lrp represses nilC transcription

We sought to determine whether differences in NilC levels between strains could be explained by differences in nilC transcription. To this end, we constructed a nilC::lacZ transcriptional fusion and integrated it into the wild-type, nilD::Tn5, rpoS::Tn5 and lrp::Tn5 chromosomes. To our knowledge, this is the first reported use of lacZ as a reporter for gene expression in X. nematophila. We therefore included a wild-type control lacking the nilC::lacZ fusion to demonstrate a lack of endogenous β-galactosidase activity. Wild-type, nilD::Tn5 and rpoS::Tn5 cells containing the nilC::lacZ fusion all showed similar β-galactosidase activity to each other and these levels increased during late stationary phase (Table 2). The lrp::Tn5 mutant, on the other hand, showed elevated β-galactosidase activity at all points during growth, relative to that seen in wild-type cells (Table 2). β-Galactosidase activity in the lrp::Tn5 mutant was maximal in late stationary phase and was ≈20-fold higher than the same point in wild-type cells (Table 2).

Table 2. β-Galactosidase activitya during growth.
StrainGrowth phaseb
Early logLate logEarly stationaryLate stationary
  • a. 

    Average Miller units ± the deviation from the mean of duplicate cultures.

  • b. 

    Early log: OD600≈ 0.2; late log: OD600≈ 1.0; early stationary: OD600≈ 3.0; late stationary: 24 h after the cells reached early stationary.

HGB081 (wild type) 0.22 ± 0.010.00 ± 0.01 0.02 ± 0.01  0.11 ± 0.00
HGB081 (Tn7/nilC::lacZ) 3.98 ± 0.033.78 ± 0.15 4.15 ± 0.02 10.10 ± 0.07
nilD::Tn5 (Tn7/nilC::lacZ) 3.87 ± 0.083.58 ± 0.05 4.21 ± 0.03  8.85 ± 0.65
rpoS::Tn5 (Tn7/nilC::lacZ) 3.90 ± 0.033.73 ± 0.14 3.81 ± 0.12  8.04 ± 0.28
lrp::Tn5 (Tn7/nilC::lacZ)10.50 ± 0.0610.8 ± 0.3131.90 ± 1.01197.00 ± 15.40

NilC protein levels in wild-type cells, measured by immunoblot, are 35-fold higher in stationary phase versus early log phase (Table 1). In comparison, nilC::lacZ expression, measured by reporter enzyme activity, in stationary-phase cells is only 2.5-fold higher than in early log cells (Table 2). While it is injudicious to quantitatively compare data from such different techniques, the relatively larger-stationary-phase increase in NilC protein levels compared with nilC transcription may indicate that nilC transcript and/or NilC protein become more stable, or that nilC translation increases in stationary phase. Regardless, our data clearly show that NilC is growth phase regulated and that this regulation is independent of nilD RNA and rpoS. Furthermore, our data show that Lrp functions as a repressor (direct or indirect) of nilC transcription.

lrp::Tn5 mutants colonize nematodes at reduced levels (15- to 892-fold) relative to wild-type (Heungens et al., 2002). The data presented above indicate that the product of the lrp locus negatively regulates (directly or indirectly) another gene required for colonization, nilC and comprise the first description of a regulatory pathway controlling the expression of a specific X. nematophila colonization factor. The observation that nilC expression in the lrp::Tn5 mutant is elevated well above wild-type levels during in vitro growth suggests that induction of nilC could be a normal, regulated event that occurs during a specific stage of X. nematophila colonization of nematodes. For example, nilC expression may be induced upon reception of a specific stimulus, perhaps one that is only found in the nematode intestinal vesicle. If true, this scenario could provide a possible explanation for the colonization phenotype of the lrp::Tn5 mutant: if the putative inducing stimulus affects nilC induction by down-modulating Lrp activity then the increase in nilC levels in the lrp::Tn5 mutant represents an inappropriate bypass of normal control by the stimulus. This inappropriate bypass may result in a colonization defect by increasing nilC expression at an inappropriate and deleterious time during colonization.

In Escherichia coli, Lrp is a global regulator affecting the expression of more than 10% of all genes (Tani et al., 2002). Most of these genes are expressed specifically on entrance into stationary phase, and include those involved in amino acid metabolism and pili biosynthesis (Tani et al., 2002; Brinkman et al., 2003). However, E. coli Lrp is the only homologue to date with demonstrated global regulatory functions (Brinkman et al., 2003); other characterized members, such as that of Haemophilus influenzae, appear to have more specific regulatory functions (Friedberg et al., 2001). Therefore, caution must be exercised in extrapolating putative NilC functions from the fact that it falls within the regulon of an Lrp homologue. However, all of the Lrp homologues studied to date regulate some aspect of amino acid metabolism (Brinkman et al., 2003), suggesting NilC may play a role in a basic metabolic process required by X. nematophila for nematode colonization. Additionally, because Lrp directly or indirectly regulates numerous genes in X. nematophila as observed by two-dimensional gel electrophoresis (H. Goodrich-Blair, unpubl. data), it cannot be determined from these experiments whether the colonization defect of the lrp::Tn5 mutant is specifically because of an inappropriate overexpression of nilC.

An lrp::Tn5 mutant initiates nilC transcription from an additional start site

Primer extension analysis was used to map the nilC transcript start site in wild-type and lrp::Tn5 cells. The nilC transcript was found to initiate from a ‘G’ residue (P1) 41 nucleotides upstream of the predicted nilC open reading frame start codon in wild-type cells (Fig. 2A, lanes 1–4). An additional site of transcript initiation (P2) was observed in the lrp::Tn5 mutant in early and late stationary phase: an ‘A’ residue 46 nucleotides upstream of the predicted nilC start codon, five nucleotides upstream of the transcription start site observed in wild-type cells (Fig. 2A, lanes 5–8). Therefore, Lrp appears to regulate nilC transcription by specifically repressing, either directly or indirectly, the P2 start site. Total nilC transcript levels (P1 and P2 combined) observed by primer extension in the lrp::Tn5 mutant did not appear to be strikingly higher than that in wild-type cells to the extent observed in nilC::lacZ transcriptional fusion (20-fold) and quantitative NilC immunoblot (75-fold) experiments. However, these primer extension analyses were not quantitative and are subject to bias from template extension efficiencies. As a result, these data may underestimate each transcript's concentration in the cell.

Figure 2.

Transcription start site and promoter analysis of nilC.
A. Primer extension analysis of nilC RNA from wild-type (lanes 1–4) and lrp::Tn5 (lanes 5–8) cells taken at OD600≈ 0.2 (lanes 1 and 5), OD600≈1.0 (lanes 2 and 6), OD600≈ 3.5 (lanes 3 and 7) and 16 h after cells reached OD600≈ 3.5 (lanes 4 and 8). ‘P1’ and ‘P2’ denote transcript start sites. Lanes labelled C, T, A and G are sequencing reaction products generated using pTn7/SR1 as template and the terminating dideoxynucleotide indicated above each lane.
B. Sequence motifs and elements in the immediate nilC upstream region. Indicated are the direct repeat (DR), first inverted repeat (IR1), second inverted repeat (IR2), P1 promoter (−10, extended −10, and +1 nucleotide below the DNA sequence), P2 promoter (−10, 16 bp spacer, −35, and +1 nucleotide above the DNA sequence), ribosome binding site (RBS) and nilC start codon (ATG). The putative Lrp binding site is boxed in grey.

The nilC upstream region encodes two potential promoters and multiple potential regulatory elements

The data presented above indicate that nilC transcription is controlled by at least two promoters, one of which is repressed (directly or indirectly) by Lrp. To identify promoter elements we searched the nilC upstream region for consensus E. coli promoter sequences. This analysis revealed two putative overlapping promoter sequences that could direct transcription from the P1 and P2 start sites. The P1 (downstream) putative promoter sequence has a moderate −10 element (CTTAAT versus TATAAT) (Harley and Reynolds, 1987), and a weak extended −10 region (TTAG versus TGTG) (Fig. 2B) (Mitchell et al., 2003). Extended −10 elements (termed TG promoters) occur in ≈ 20% of E. coli promoters (Burr et al., 2000), and tend to have longer spacer regions between the −10 and −35 elements, and fewer matches to consensus in each of these elements (Mitchell et al., 2003). The putative P1 promoter has no discernible −35 element in the area predicted by a typical 17-bp spacer region. However, a strong −35 element (TTGAAA versus TTGACA) (Harley and Reynolds, 1987) occurs 20-bp upstream of the −10 element, consistent with the > 18 bp spacer found in 39% of E. coli TG promoters (Mitchell et al., 2003). This  same −35  element  is  positioned  16 bp  upstream  of  a  putative −10 element (TAGACT versus TATAAT) that could direct transcription from the P2 (upstream) start site (Fig. 2B) (Harley and Reynolds, 1987). To our knowledge, TG promoters have not, as a class, been reported to exhibit differential expression or regulation relative to standard consensus promoters. However, it is possible that the core elements of P1, a putative TG promoter, dictate a constant low-level expression observed in wild-type and that the potentially stronger P2 promoter is repressed under these conditions except when Lrp is lacking.

There are several sequence elements present in the nilC promoter region that could be binding sites for regulatory proteins: two separate 10 bp inverted repeat sequences and an 11 bp direct repeat sequence, each with a single mismatch. The upstream inverted repeat (IR1) overlaps both proposed −10 regions with its two arms spaced 4 bp apart (ATTTAACTTAGACTTAATTTAAAT) (Fig. 2B). The downstream inverted repeat (IR2) (TTAGTTTTTTAAAAAACAAA) is located between the start site of the P1 transcript and a purine rich sequence (GGAA) (Gren, 1984) spaced appropriately to potentially serve as a ribosome-binding site (Fig. 2B) (Calogero et al., 1988). The 11 bp direct repeat (DR) sequence (AGATTTAACTT and AAATTTAACTT) occurs with one arm located between the putative −35 and −10 sequences of the P2 promoter (Fig. 2B). Many transcriptional repressors prevent RNA polymerase from accessing their regulated promoters by binding to the −35 and/or −10 regions (Gralla and Collado-Vides, 1996). Both IR1 and the direct repeat overlap the putative −10 elements of both P1 and P2 and are therefore likely sites to be bound by a repressor. The position of IR2 with in the 5′ untranslated region of the nilC transcript suggests it could either be a binding site for a repressor protein, or a stem-loop structure in the mRNA, or both. Regardless, if this element is involved in regulation it is predicted to occlude access to the proposed ribosome-binding site in a post-transcriptional regulatory mechanism (Draper, 1996).

Because Lrp negatively regulates nilC transcription, we searched the nilC promoter region (from the nilC start codon to 200 bp upstream of it) for potential Lrp binding sites. Lrp multimers tend to cooperatively bind promoters at multiple sites that typically lack perfect inverted repeat elements (Brinkman et al., 2003). Using the consensus Lrp binding site from E. coli, YAGHAWATTWTDCTR (Brinkman et al., 2003), we identified only one putative Lrp binding site with 12 of 15 nucleotides matching the consensus: TAAAAAATTTAACTT (Fig. 2B). This sequence is located in the spacer region for the putative P2 promoter and overlaps 1 bp of the putative P2 promoter −10 region. A regulatory protein binding in this position would be expected to disrupt transcription from both the P1 and P2 promoters; however, Lrp appears to specifically repress (directly or indirectly) transcription from the P2 promoter (Fig. 2A). Lrp does not have a strict requirement for its consensus binding site (Brinkman et al., 2003), and cooperative binding at suboptimal binding sites could account for high affinity binding. Furthermore, the E. coli consensus site may not match the optimal binding site of X. nematophila Lrp. Hence, the lack of additional strong matches to the E. coli consensus site neither excludes nor supports the idea that X. nematophila Lrp is a direct regulator of nilC.

The presence of myriad putative promoter elements suggests that the transcriptional regulation of nilC is tightly controlled and complex. Our future work will be aimed at understanding which, if any, of the identified putative promoter elements contribute to controlling nilC transcription and which transcriptional regulators bind these sites. An essential aspect of these analyses will be to determine which promoter elements and regulators are necessary for nematode colonization, and to what extent. This, in turn, will help elucidate the stimuli that signal changes in nilC transcription levels and will provide valuable insights into the function that NilC provides in nematode colonization.

NilC is a lipoprotein with an N-terminal signal sequence that is processed by a conserved mechanism

To determine whether, as predicted, NilC is a lipoprotein, we grew X. nematophila cells in the presence of a radioactive lipid precursor, 14C-acetate and excess amino acids, conditions under which the 14C label should only be incorporated into proteins that have been acylated. NilC was immunoprecipitated from wild-type cells and showed incorporation of the 14C label (Fig. 3A). As a control for non-specific precipitation of another acylated protein of similar size, an immunoprecipitation was performed with a parallel culture of the XnΔSR1 mutant. This immunoprecipitation showed no radiolabelled material in the 28 kDa region, indicating that the radiolabelled band observed in the wild-type cells was, indeed, NilC (Fig. 3A). These results suggest that NilC is produced as a lipoprotein in X. nematophila.

Figure 3.

NilC is a lipoprotein with a signal sequence that directs export to the periplasm.
A. Radiogram detection of anti-NilC immunoprecipitated proteins from wild-type (lane 1) and XnΔSR1 (lane 2) cells grown in the presence of 14C-acetate.
B. Alkaline phosphatase units in cells harbouring a ′phoA vector (white), a nilC::′phoA plasmid (black) or a nilCΔaa1-22::′phoA fusion plasmid (grey). Error bars indicate standard deviations from the average units of triplicate cultures per strain.
C. Immunoblot detection of NilC from cells growing in the absence (lane ‘–’) and presence (lane ‘+’) of globomycin.

The first 19 amino acids of NilC are predicted to form a hydrophobic, α-helical signal sequence that directs the remainder of the protein to be exported to the periplasm (Heungens et al., 2002). To test this hypothesis, we created a pair of nilC fusions to a signal-sequenceless E. coli phoA (′phoA) gene. phoA encodes alkaline phosphatase, an enzyme that requires export to the periplasm (by its own N-terminal signal sequence) to be active. E. coli cells harbouring a plasmid construct in which we fused codons 3–66 of nilC in frame with ′phoA showed high alkaline phosphatase activity, relative to those harbouring either the ′phoA vector alone or a nilC::′phoA construct containing codons 22–66 of nilC (Fig. 3B). These results indicate that the N-terminus of NilC is sufficient to direct protein export to the periplasm of E. coli and that residues 3–22 are necessary for this function. Therefore, the N-terminus of NilC behaves as a standard signal sequence.

The first 19 amino acids of NilC are predicted to comprise a ≈ 2 kDa leader peptide that is cleaved by signal peptidase II following the first lipidation event (Heungens et al., 2002). In proteobacteria, the signal peptides of lipoproteins are removed by a prolipoprotein-specific signal peptidase, signal peptidase II that can be specifically and potently inhibited by globomycin (Inukai et al., 1978; Dev et al., 1985). To determine whether signal peptidase II cleaves a signal peptide from NilC as it is being produced, globomycin was added to a growing culture of wild-type X. nematophila cells. Immunoblot analysis detected an additional NilC species that was slightly larger than the NilC that is detected in cells grown in the absence of globomycin (Fig. 3C). These results demonstrate that NilC is initially produced as a ≈ 30 kDa protein that is processed by signal peptidase II to yield the mature ≈ 28 kDa protein observed in X. nematophila cells. Moreover, these results demonstrate that lipoprotein signal peptide processing is similar to that in other proteobacteria.

To our knowledge, these experiments are the first-described biochemical characterizations of a lipoprotein from X. nematophila. The results suggest that lipoprotein lipidation, export and processing in X. nematophila occur by mechanisms similar to those observed in other proteobacteria. These studies will also substantiate the prediction of lipoproteins within the Xenorhabdus genomes currently being sequenced (Goldman et al., 2004).

NilC is membrane associated in X. nematophila and is expressed as an outer membrane-associated protein in E. coli

NilC is predicted to be an outer membrane-localized lipoprotein (Heungens et al., 2002). As a first test of this prediction, we determined whether or not NilC was associated with the membrane fraction of X. nematophila cells. A cell-free X. nematophila sonicate was separated into soluble and insoluble fractions. NilC was found solely in the insoluble fraction, suggesting that it is membrane associated (Fig. 4A). Although it is possible that NilC is simply produced as an insoluble protein in X. nematophila, it is more likely that NilC is membrane associated via its acyl moiety. Indeed, NilC recombinantly overproduced in E. coli as a protein lacking a signal sequence and lipid attachment site is produced as a soluble protein in the complete absence of any detergents (see NilC purification and anti-NilC antibody production in Experimental procedures).

Figure 4.

NilC is membrane associated and not surface exposed. Shown are immunoblots detecting NilC in (A) soluble (lane 1) and insoluble (lane 2) fractions of X. nematophila sonicates, (B) sarkosyl-soluble (lane 1) and sarkosyl-insoluble (lane 2) fractions of the insoluble cell-free extract fraction of X. nematophila and (C) whole X. nematophila cells (top series) and X. nematophila cell lysates (bottom series) exposed to proteinase K for the indicated amounts of time.

In E. coli, lipoproteins are sorted to either the outer or inner membrane according to the amino acid that follows the cysteine residue to which their acyl tails are attached; lipoproteins with a serine residue in this position are typically sorted to the outer membrane, whereas lipoproteins with an aspartic acid residue in this position are left in the inner membrane (Yamaguchi et al., 1988; Terada et al., 2001). NilC is predicted to have a serine in this position and is therefore predicted to be an outer membrane-associated lipoprotein. To test this hypothesis, we attempted to perform a sucrose density gradient separation (Nikaido, 1994) of X. nematophila membrane fractions. However, we were unable to sufficiently separate peak fractions of known inner and outer membrane markers (data not shown). We therefore heterologously expressed NilC in E. coli, an organism for which sucrose density gradient membrane separations have been more extensively perfected (Nikaido, 1994). A sucrose density gradient separation of E. coli inner and outer membranes from cells expressing NilC yielded an inner membrane peak fraction (fraction 14) in which the outer membrane marker (KDO) was virtually undetectable, and an outer membrane fraction (fraction 5) that had low contaminating levels of the inner membrane marker (NADH oxidase activity) (Fig. 5A). NilC was only detectable in the outer membrane peak fraction, thereby demonstrating that, in E. coli, it is associated specifically with this membrane (Fig. 5B). Because NilC lipidation, signal sequence removal and association with the insoluble fraction occur in X. nematophila with the same conventions as do E. coli lipoproteins, it is reasonable to expect that NilC is associated with the outer membrane in X. nematophila.

Figure 5.

NilC is associated with the outer membrane when expressed in E. coli.
A. NADH oxidase activities (closed circles) and KDO concentrations (open circles) of sucrose-density-gradient fractions of E. coli cells expressing NilC.
B. Immunoblot detection of NilC from the outer membrane peak fraction (lane ‘5’) and inner membrane peak fraction (lane ‘14’) of sucrose-density-gradient separated E. coli inner and outer membranes from cells expressing NilC.

Sarkosyl detergent extractions of the insoluble fraction of cell-free extracts provide an alternate means of determining the membrane association of proteins in a proteobacterium. This anionic detergent is believed to specifically solubilize the inner membrane and to be repelled from outer membrane vesicles by virtue of the greater negative charge of outer membranes (Filip et al., 1973). We performed a sarkosyl extraction of X. nematophila membrane fractions. Surprisingly, NilC was detected by immunoblot in both the sarkosyl-soluble (inner membrane) and sarkosyl-insoluble (outer membrane) fractions (Fig. 4B). This apparent dual membrane association was still observed even when several critical parameters such as pH, total protein concentration, detergent concentration and salt concentrations were widely varied (data not shown). In all cases, Coomassie-stained SDS-PAGE gels showed proper overall separation of proteins (data not shown) with patterns that were consistent with previously published findings (Leisman et al., 1995). It is plausible that NilC is inherently prone to misextraction by sarkosyl, because of the fact that, as a lipoprotein, it is only membrane associated by virtue of the three acyl tails covalently attached to its N-terminus. Indeed, misextraction of membrane proteins by this method has been previously documented (Chopra and Shales, 1980).

NilC is not a surface-exposed protein

The data presented above demonstrate that NilC is likely to be an outer membrane-associated lipoprotein in X. nematophila. At this location, NilC could be oriented either towards the extracellular environment or the periplasm. The orientation of NilC towards the periplasm or extracellular environment cannot be deduced from its primary amino acid sequence. Therefore, to address this question, whole late-log-phase X. nematophila cells and a late-log-phase SDS cell lysate were incubated in the presence of proteinase K. NilC was rapidly degraded by proteinase K in the SDS lysate, but not in whole cells, suggesting that it is located in the periplasm and protected from proteinase K degradation by the outer membrane in intact cells (data not shown). However, because some surface-exposed proteins in their native state can be resistant to proteinase K degradation (Bunikis and Barbour, 1999; El-Hage et al., 2001) but are potentially susceptible upon denaturation, we repeated the experiment with a lysate made with a non-denaturing detergent, Triton X-100. Again, we observed NilC to be degraded in the presence of proteinase K (and not in its absence, data not shown) only upon cellular lysis, indicating that NilC is oriented towards the periplasm (Fig. 4C). Understanding that NilC is located in the periplasm is important to predict its biochemical function in mutualism with the nematode host; for example, in this orientation NilC is not likely to function in direct adherence to the nematode intestinal vesicle. This orientation will also have to be taken into account when modelling protein–protein interactions with other proteins with which NilC may interact.

Concluding remarks

Xenorhabdus nematophila is a model to understand how beneficial microbes initiate and maintain long-term relationships with animal hosts and has the potential to provide insights into host range specificity, host cell binding and mutually beneficial nutrient exchange. The molecular mechanisms by which X. nematophila mutualistically colonizes its nematode host are only beginning to be elucidated. The present work has provided details on the expression and localization of an essential colonization factor, NilC. We show that X. nematophila nilC expression occurs throughout in vitro growth, and that transcription from a second promoter occurs in an lrp::Tn5 background. Furthermore, we demonstrate that NilC protein accumulates in stationary phase and is a periplasmically oriented lipoprotein anchored in the outer membrane. NilC regulation by Lrp coupled with its membrane localization may suggest a link with amino acid metabolism. Further analysis of the other colonization factors, NilA and NilB, encoded by the SR1 locus may reveal similar regulatory patterns, as well as interactions among these factors. Such studies will in turn allow the biochemical function and precise role in nematode colonization of these factors to be determined.

Experimental procedures

Strains, plasmids, media and growth conditions

The strains and plasmids used in this study are listed in Table 3. Luria–Bertani (LB) broth (Miller, 1972) was used exclusively as the growth medium at 30°C, except for nematode cocultivation assays, which used lipid agar (Vivas and Goodrich-Blair, 2001) at room temperature. Media used with X. nematophila strains was either stored in the dark or supplemented with 0.1% sodium pyruvate (Xu and Hurlbert, 1990). Table 3 also lists the plasmids used in this study. Plasmids were conjugated from E. coli S17-1(λpir) into X. nematophila strains as described (Vivas and Goodrich-Blair, 2001) and were maintained with the following antibiotic concentrations: ampicillin (Ap), 100 µg ml−1; gentamicin (Gm), 30 µg ml−1; chloramphenicol (Cm), 30 µg ml−1; kanamycin (Km), 50 µg ml−1, erythromycin (Erm) 200 µg ml−1.

Table 3. Bacterial strains and plasmids used in this study.
Strain/plasmidDescription, relevant characteristicsReference or source
E. coli strains
 S17-1(λpir) thi pro hdsR hdsM+ recA, chromosomal insertion of RP4-2(Tc::Mu Km::Tn7 ) Simon et al. (1983)
 BL21(λDE3)FompT hsdSB(rB mB) gal dcm (DE3): T7 RNA polymerase under control of lacUV5 promoter Studier and Moffatt (1986)
 CC118 araD139Δ(ara, leu)7697 ΔlacX74 phoAΔ20 galE galK thi rpsE rpoB argEamrecA1 Manoil and Beckwith (1985)
X. nematophila strains  
 HGB081Wild-type AN6/1S. Forst
 HGB315HGB081, nilD::Tn5 Heungens et al. (2002)
 HGB321HGB081, lrp::Tn5 Heungens et al. (2002)
 HGB323HGB081, rpoS::Tn5 Heungens et al. (2002)
 HGB007Wild-type, ATCC19061ATCC
 HGB777XnΔSR1: HGB007 with a kanamycin cassette replacing the SR1 locusThis study
 HGB778XnΔSR1(Tn7/SR1): HGB777 with a mini-Tn7/SR1 chromosomal insertion at the att Tn7 siteThis study
 HGB779XnΔSR1(Tn7/SR1/CMZ): identical to HGB778, but with the 5′-most ATG of nilC changed to a TAG stop codonThis study
Plasmids
 pJQ200Cloning vector; GmR Quandt and Hynes (1993)
 pKR100Mobilizable, oriR6K plasmid for allelic exchange; CmRKaren Visick
 pKV124Mobilizable, oriR6K mini-Tn10::lacZ delivery vector; CmR, KmR Visick and Skoufos (2001)
 pΔSR1pKR100 with a kanamycin resistance cassette flanked by ≈ 1 kb of DNA from the regions flanking the SR1 locus; KmR, CmRThis study·
 pTYB1Intein expression plasmid for protein purification; ApRNew England Biolabs
 pNilCexpTYB1 with nilC cloned to express aa 22–282; ApRThis study
 pNilCexcorrpNilCex with the last nilC codon changed from AAT to CAG (Asn to Gln); ApRThis study
 pEVS107Mobilizable, oriR6K mini-Tn7 delivery vector; ErmR, KmR Stabb and Ruby (2002); E. Stabb
 pTn7/SR1pEVS107 with the SR1 locus cloned into the KpnI site of the mini-Tn7 transposon; ErmR, KmRThis study
 pTn7/SR1/CMZIdentical to pTn7/SR1, but with the 5′-most ATG of nilC changed to a TAG stop codon; ErmR, KmRThis study
 pTn7/nilC::lacZpEVS107 with a nilC::lacZ fusion cloned into the KpnI site of the mini-Tn7 transposon; ErmR, KmRThis study
 pUI1156Signal sequenceless phoA (′phoA) fusion plasmid; ApR Ouchane and Kaplan (1999)
 pUI1156shiftpUI1156 with ′phoA shifted in frame with the start codon by XhoI digestion, end-filling and religation; ApRThis study
 pUI1156shift/nilCpUI1156shift with nilC bp 7–200 cloned in frame with ′phoA; ApRThis study
 pUI1156shift/nilCΔsspUI1156shift with nilC bp 67–200 cloned in frame with ′phoA; ApRThis study
 pUX-BF13Helper plasmid carrying Tn7 transposase functions; ApR Bao et al. (1991)

Molecular biological methods

Standard molecular biological methods were used for this study (Sambrook et al., 1989). All DNA constructs were fully sequenced at the University of Wisconsin Biotechnology Center to ensure correct sequence. For site-directed mutagenesis of DNA constructs, polymerase chain reactions (PCRs) were performed using Platinum Pfx (Invitrogen) and plasmid template, which was subsequently digested with DpnI before transformation of the PCR product. The primers used in this study (Integrated DNA Technologies) are listed in Table 4.

Table 4. Primers used in this study.
NameSequence (5′ to 3′ orientation)
NilCIMPACTforCATATGAGAGGAGGGGGTTCTCAC
NilCIMPACTrevGCTCTTCCGCAATTGCTGGGGTTAGGAGAC
pNilCexcorrforCTCCTAACCCCAGCCAGTGCTTTGCCAAGG GTACC
pNilCexcorrrevCCCTTGGCAAAGCACTGGCTGGGGTTAGG AGACG
EcoRIcompTGTCTAGAATTCATTAGGTCATGCG
RevKOGGATCCTTATATGGGCACTGTTGC
ForKOGCAACAGTGCCCATATAAGGATCCTTATGGT GTTACCTGACC
PstIcompCGGCTGCAGTACTCCATCTATTGAC
LSBTn7 fornewTTAGGTACCCTCCTGTGTAGTTAAACC
LSBTn7revnewTTAGGTACCTTATTAATTGCTGGGG
AttTn7EXTTGTTGGTTCACATCC
ErmAnch1TACTTATGAGCAAGTATTGTC
NilCMtoZforCAAATTGGAATCATTATTAGAATACAAAATCT AAAATC
NilCMtoZrevTTTAGATTTTGTATTCTAATAATGATTCCAATT TGTTT
NilCstart1GTATTCATATAATGATTCC
phoAfornewGCCAAGCTTTACAAAATCTAAAATCTATTTAG CTCTCGC
NilCdeltassphoAforTCAAAGCTTAGGAGGGGGTTCTCACCAC
NilCfusionrev♯1CATGGATCCTTGGAATATCCCTTAAGG
NilCssTruncRevTTGTATTCATATGATGATTCCAATTTG

Nematode colonization assays

Nematode co-cultivations were performed as described previously (Heungens et al., 2002). Briefly, sterilized nematode eggs were added to lipid agar plates containing a lawn of a bacterial strain to be assayed. Progeny nematodes were collected from the plates, surface sterilized and sonicated for 1 min. The sonicate was dilution plated to determine the number of colony-forming units of bacteria per nematode.

Deletion of the SR1 locus

The SR1 locus was deleted from the X. nematophila HGB007 chromosome and replaced with a Km cassette by allelic exchange (Tabatabai and Forst, 1995) with plasmid pΔSR1 to create XnΔSR1 (HGB777). To confirm the correct construction of this strain, we used PCR amplification across the deleted SR1 region using primers EcoRIcomp and PstIcomp. We also conducted a Southern blot analysis (ECF random prime labelling kit, Amersham) of MluI-digested wild-type and ΔSR1 genomic DNAs. Probes used were complementary to the 5′ and 3′ flanking regions (amplified with 5′ flanking side primers EcoRIcomp and RevKO and 3′ flanking side primers ForKO and PstIcomp), the deleted region (amplified with primers KHP042 and KHP047; Heungens et al., 2002), or the Km cassette (by directly labelling gel-purified kanamycin cassette) (data not shown).

Tn7 complementation

The SR1 locus was amplified from X. nematophila HGB007 genomic DNA with primers LSBTn7fornew and LSBTn7revnew and cloned into the KpnI site of pEVS107 to create pTn7/SR1. Plasmid pTn7/SR1/CMZ was created by site-directed mutagenesis of pTn7/SR1 template with primers NilCMtoZfor and NilCMtoZrev. Tri-parental conjugations were performed with XnΔSR1 and S17-1(λpir) E. coli donor strains carrying pTn7 constructs and pUX-BF13. Correct insertion of Tn7 constructs into the attTn7 site of the X. nematophila chromosome was confirmed using PCR with AttTn7EXT and ErmAnch1 primers.

NilC purification and anti-NilC antibody production

NilC was purified using the IMPACT-CN system from New England Biolabs. nilC was amplified from X. nematophila HGB007 genomic DNA using primers NilCIMPACTfor and NilCIMPACTrev and cloned into the NdeI/XbaI sites of pTYB1. To avoid potential toxicity frequently associated with overexpression of membrane proteins, the predicted signal sequence and lipid attachment site (aa 1–21) were not included in the expression clone. To allow the intein chemistry of the purification system to function, the C-terminal amino acid (Asn282) was changed to a glutamine using primers pNilCexcorrFor and pNilCexcorrRev. Approximately 1 mg of purified NilC was obtained from 1 l of BL21λDE3(pNilCexcorr) cells after 24 h of growth. Protein was purified according to the manufacturer's instructions except that 0.1% Lauryl dimethylamine oxide was substituted for 0.1% Triton X-100 and was dialysed away after column elution. A New Zealand White rabbit was injected with 0.5 mg of purified NilC and then boosted at 4 and 8 weeks with 0.25 mg of purified protein at the University of Wisconsin Laboratory Animal Resources Polyclonal Antibody Service. Serum was collected before NilC injection (pre-immune), at 2 weeks (first-bleed), 6 weeks (second-bleed) and 10 weeks (final-bleed). The first-bleed anti-serum was solely used in this study. The preimmune anti-serum showed no detection of any proteins from X. nematophila (data not shown).

Immunoblot detection and quantification of NilC during growth

Anti-NilC anti-serum was used at a 1:5000 dilution in immunoblot analyses using the ECF Western Blotting Kit (Amersham) according to the manufacturer's instructions. Fluorescence was detected on a Storm860 Phosphorimager (Molecular Dynamics). Quantitative immunoblots were performed with the ECL Plus Western Blotting Kit (Amersham) and a goat, anti-rabbit IgG HRP-conjugated secondary antibody (Pierce, cat ♯31460) using 1.0, 5.0 and 10.0 ng standards of purified NilC. Samples were taken from 0.5 l of cultures of HGB081, lrp::Tn5, nilD::Tn5 and rpoS::Tn5 at the indicated times. Cells were harvested and washed with phosphate-buffered saline (PBS) (Sambrook et al., 1989).

β-Galactosidase activity assays

A mini-Tn7-based nilC::lacZ fusion was created by first PCR amplifying the SR1 locus ending at the 5′-most ATG of nilC with primers LSBTn7fornew and NilCssTruncRev. The PCR product was cloned into pCR4Topo (Invitrogen) with nilB and nilA oriented opposite the plac promoter. This plasmid and plasmid pKV124 were digested with NotI and ligated together, such that the nilC promoter region was facing the lacZ gene of pKV124. The SR1/lacZ KpnI fragment was subcloned into the KpnI site of pEVS107 to yield pTn7/nilC::lacZ. This mini-Tn7 construct was integrated into the attTn7 site of HGB081, lrp::Tn5, nilD::Tn5 and rpoS::Tn5 cells as described above for similar mini-Tn7 integrations (see Tn7 complementation). Cells were inoculated into duplicate 125 ml of shaking cultures to an optical density (OD600) of 0.01. Samples of cells were collected from these cultures at indicated times and β-galactosidase activity assays were performed according to Miller (Miller, 1972) except that chloroform was eliminated from and 0.1% SDS was added to the assay buffer to improve cell lysis. Each cell sample was assayed in triplicate to yield an average activity per culture.

Primer extension mapping of the 5′ end of the nilC transcript

RNA was isolated from growing cultures of HGB081 and lrp::Tn5 cells at the indicated times using Trizol (Invitrogen), according to the manufacturer's instructions. A PAGE-purified primer (NilCstart1) was end-labelled using T4 Polynucleotide Kinase (Promega) and [γ-32P]-ATP (Perkin Elmer), according to the kinase manufacturer's instructions. The labelling reaction was cleaned using the Qiagen Nucleotide Removal Kit. Primer and RNA template were hybridized by placing them in a tube together with a proportional amount of AMV-RT reaction buffer at 80°C for 10 min and then allowing the heating block to slowly cool to 30°C over the course of 1 h. Reverse transcription reactions using AMV-RT (Promega) were performed according to the manufacturer's instructions at 42°C for 15 min using 20 µg of RNA in a 60 µl reaction. The reactions were ethanol precipitated and resuspended in loading buffer before separation through a 12% urea/PAGE sequencing gel. A sequencing ladder was generated using the same primer and pTn7/SR1 as template using the fmol sequencing kit according to the manufacturer's instructions (Promega).

14 C-acetate incorporation and NilC immunoprecipitation

Lipid-containing proteins from X. nematophila HGB007 were labelled using 14C-acetate as a lipid precursor. Overnight cultures of HGB007 and HGB777 were subcultured 1:100 in LB and grown to OD600 = 1.0. The cells were then subcultured 1:10 into 1 ml of LB containing 10 µCi of 14C-acetate (47.5 mCi mmol−1) (Moravek Chemicals), and grown for an additional 4 h, after which time they were washed twice with PBS. The washed cells were resuspended in 100 µl PBS and lysed by the addition of 900 µl of lysis buffer (1% Triton X-100 and 0.1% SDS in PBS) with slow inversion of the lysates at 4°C for 15 min. The lysates were spun at top speed in a refrigerated microfuge for 30 min and the supernatants were removed to clean tubes. Fifty microlitres of rabbit anti-NilC anti-serum were added to each tube and the tubes were inverted slowly for 2 h, followed by the addition of 50 µl of pre-swelled protein A-sepharose beads (Sigma), and continued inversion for 2 h. The sepharose beads were washed three times in lysis buffer. Fifty microlitres of SDS-PAGE loading buffer (Sambrook et al., 1989) were added to each tube, followed by boiling for 10 min. The samples were separated by SDS-PAGE, dried onto Whatman 3 mm paper, and radioactivity was detected with Kodak BioMaxMS film exposed for 40 days.

Alkaline phosphatase assays

Plasmid pUI1156shift was created by digesting pUI1156 with XhoI, filling in the overhanging ends with Klenow and ligating the new ends to shift the entire ′phoA gene into the same reading frame. Plasmid pUI1156shift/nilC was created by cloning bp 7–200 of the nilC open reading frame (amplified using primers phoAfornew and NilCfusionrev♯1) into the HindIII and BamHI sites of pUI1156shift. Plasmid pUI1156shift/nilCΔss was created by cloning bp 67–200 of the nilC open reading frame (amplified using primers phoAfornew and NilCdeltassphoAfor) into the HindIII and BamHI sites of pUI1156shift. Alkaline phosphatase assays of triplicate cultures of CC118 cells harbouring pUI1156shift, pUI1156shift/nilC or pUI1156shift/nilCΔss were performed (each culture in triplicate) according to previously described methods (Manoil, 1991), and alkaline phosphatase activities were calculated using the standard equation presented in Manoil (1991).

Globomycin inhibition of lipoprotein signal peptide processing

Globomycin was added to a log-phase culture (OD600 = 1.0) of X. nematophila HGB007 cells to a final concentration of 200 µg ml−1 and incubated for 1 h. Cells were harvested and analysed by Western blot as above.

Xenorhabdus nematophila membrane purification and sarkosyl extractions

To purify X. nematophila HGB007 membrane vesicles, 5 ml of an overnight culture was inoculated into 500 ml of media and the cells were grown with vigorous shaking to an OD600 of ≈1.0. The cells were pelleted, washed once with PBS and resuspended in a minimal volume of PBS. Phosphate buffer (20 mM, pH 7.5) was added to a total volume of 15 ml, followed by the addition of DNaseI (100 U, Boehringer) and RNase (1 mg from bovine pancreas, Sigma). The cells were lysed by a single passage through a French Pressure Cell at 16 000 psi. Unbroken cells were removed by centrifugation and the cleared lysate was spun at 28 000 r.p.m. for 2 h in a Beckman SW40Ti rotor to pellet membrane vesicles. The supernatant was removed (a portion was saved as the soluble fraction) and the remaining membrane pellet was resuspended in 2 ml of phosphate buffer with rapid stirring. For Sarkosyl extraction of the inner membrane fraction, 200 µl of the resuspended membrane vesicles were added to 1.8 ml of 0.5% Sarkosyl (in 20 mM phosphate buffer, pH 7.5) and spun in a Beckman TLA100.2 rotor for 30 min at 100 000 r.p.m. The supernatant was removed and saved separately from the pellet, which was resuspended in SDS-PAGE loading buffer at a dilution that would equal the treatment of the supernatant.

Sucrose-density-gradient separation of E. coli inner and outer membrane fractions

To separate the inner and outer membranes of E. coli cells harbouring a plasmid carrying the nilC gene, DH5α(λpir)(pTn7/SR1) cells were grown in 0.5 l of LB + Km with vigorous shaking to an OD600 of 0.5. Cell washing and lysis, lysate clearing, and membrane vesicle pelleting and resuspension were all carried out as above. Two hundred and fifty microlitres of the resuspended membrane fraction were layered onto a discontinuous sucrose gradient constructed with 60% (3 ml), 45% (5 ml) and 15% (2 ml) sucrose in phosphate buffer (Sambrook et al., 1989). The gradient was topped with phosphate buffer and spun at 4°C for 16 h in a Beckman SW40Ti rotor at 28 000 r.p.m. The tube was then pierced with a needle and 0.5 ml of fractions were collected. NADH oxidase assays were performed by adding 5 µl of each fraction to 95 µl of NADHred (0.125 mg ml−1 in 10 mM Tris pH 7.5, Sigma), allowing the sample to react for 30 min at 25°C, and determining the change in absorbance at 340 nm (Osborn et al., 1972). Fractions with NADH oxidase activities (nmol of NADHred oxidized per min per 5 µl of each fraction) that consumed more than 40% of the available substrate were diluted and re-reacted in triplicate. The sucrose from a 200 µl sample of each fraction was removed by diluting 1:10 in phosphate buffer and pelleting the membrane vesicles in a TLA100.2 rotor as above. The pellets were washed in 0.5 ml of phosphate buffer and re-pelleted in a microfuge at top speed for 20 min at 4°C. The final membrane pellet was resuspended in 50 µl of phosphate buffer and its KDO content was assayed as described (Keleti and Lederer, 1974). The fraction with the highest NADH oxidase activity was saved as the inner membrane fraction for further analysis and the fraction with the highest KDO content was saved as the outer membrane fraction.

Proteinase K assays

An overnight culture of X. nematophila HGB007 was washed with PBS and resuspended to an OD600 of 10.0. The culture was split into two equal samples, one of which was lysed by the addition of SDS to 1% or Triton X-100 to 0.1%. Cells lysed immediately in SDS, but required incubation at 30°C for 1 h to lyse in Triton X-100. Proteinase K was added to a final concentration of 50 µg ml−1 and the samples were incubated at 37°C. Samples were removed at the indicated time points and the protease reaction was stopped by the addition of PMSF to 5 mM, followed by the addition of SDS-PAGE sample buffer and heating to 95°C for 10 min. Samples were separated using SDS-PAGE electrophoresis and detected using immunoblots, as above.

Acknowledgements

We would like to thank M. Inouye for graciously providing us with the globomycin used in this study. T. Gaal was very helpful in discussing promoter elements. We thank S. Kaplan for providing us with the ′phoA fusion plasmid, pUI1156, and S. Orchard and K. Cowles for their review of the manuscript and helpful discussions. C.E. Cowles is supported by an NRSA Traineeship from the National Institutes of Health Grant T32 AI007414. This research was supported by a  National  Institutes  of  Health  Grant  GM59776  to H. Goodrich-Blair.

Ancillary