Identification of Xenorhabdus nematophila genes required for mutualistic colonization of Steinernema carpocapsae nematodes



One stage in the symbiotic interaction between the bacterium Xenorhabdus nematophila and its nematode host, Steinernema carpocapsae, involves the species-specific colonization of the nematode intestinal vesicle by the bacterium. To characterize the bacterial molecular determinants that are essential for vesicle colonization, we adapted and applied a signature-tagged mutagenesis (STM) screen to this system. We identified 15 out of 3000 transposon mutants of X. nematophila with at least a 15-fold reduction in average vesicle colonization. These 15 mutants harbour disruptions in nine separate loci. Three of these loci have predicted open reading frames (ORFs) with similarity to genes (rpoS, rpoE, lrp) encoding regulatory proteins; two have predicted ORFs with similarity to genes (aroA, serC) encoding amino acid biosynthetic enzymes; one, designated nilB (nematode intestine localization), has an ORF with similarity to a gene encoding a putative outer membrane protein (OmpU) in Neisseria; and three, nilA, nilC and nilD, have no apparent homologues in the public database. nilA, nilB and nilC are linked on a single 4 kb locus. nilB and nilC are > 104-fold reduced in their ability to colonize the nematode vesicle and are predicted to encode membrane-localized proteins. The nilD locus contains an extensive repeat region and several small putative ORFs. Other than reduced colonization, the nilB, nilC and nilD mutants did not display alterations in any other phenotype tested, suggesting a specific role for these genes in allowing X. nematophila to associate with the nematode host.


Stable relationships between microbes and hosts are ubiquitous, often specific and typically essential for one or all partners (McFall-Ngai, 1998). One approach to understanding the molecular components that mediate stable relationships is to study experimentally tractable systems in which a single microbe associates specifically with a plant or animal host (McFall-Ngai, 1998), such as the mutually beneficial symbiosis between the Gram-negative bacterium Xenorhabdus nematophila (formerly X. nematophilus; Euzéby and Boemare, 2000) and the entomopathogenic nematode Steinernema carpocapsae (Vivas and Goodrich-Blair, 2001). Together, these two organisms are able to infect, kill and reproduce within the larval stage of many types of insects (Poinar, 1979). The mutualistic relationship is based on the fact that X. nematophila is necessary for both insect host killing and nematode development within the host (Poinar and Thomas, 1967; Boemare et al., 1983), whereas the nematode is required as a vector to bring X. nematophila into an insect (Poinar and Thomas, 1967) and possibly for protection outside the insect (Poinar, 1979; Morgan et al., 1997).

The soil-dwelling, infective form of the nematode, known as the infective juvenile (IJ), is non-feeding and carries an extracellular monoculture of X. nematophila bacteria in an intestinal region termed the vesicle (Bird and Akhurst, 1983). X. nematophila bacteria appear to have the maximal ability to occupy this niche; Xenorhabdus species other than X. nematophila are not able to colonize S. carpocapsae intestines efficiently (Akhurst, 1983a; C. E. Cowles and H. Goodrich-Blair, unpublished results). It has been reported that Xenorhabdus bacteria are not required for the development of the vesicle in their nematode hosts (Bird and Akhurst, 1983). However, possible subtle changes in host vesicular cells, such as in surface structure profiles, microvillar density or cell size, have not yet been examined. Upon infecting insect larvae, S. carpocapsae IJ nematodes travel to the haemocoele and release X. nematophila, which kill the insect. Within the insect cadaver, the nematode carries out several cycles of vegetative development and reproduction until several tens of thousands IJ progeny emerge, carrying X. nematophila, to forage for new hosts (Woodring and Kaya, 1988). In soil, the non-feeding IJs are capable of surviving and maintaining their pathogenicity (i.e. viable bacteria) for as long as 16 weeks (Kung et al., 1990a), with the rate of decline depending on soil type, pH, temperature, oxygen level, humidity and the abundance of natural predators (Ishibashi and Kondo, 1986; Kung et al., 1990a,b; 1991).

The physiological state of X. nematophila as it resides in the intestine of IJs has not been examined, and our work is aimed towards a better understanding of this, as well as characterizing the molecular foundation of X. nematophila–nematode interactions. Little has been published to date on the molecular features promoting X. nematophila colonization of, and survival within, S. carpocapsae nematodes. The expression profile of abundant X. nematophila outer membrane proteins (OMPs) has been characterized (Forst et al., 1995; Tabatabai and Forst, 1995; Forst and Leisman, 1997), but roles for specific OMPs have not been reported in nematode colonization. In addition, evidence exists that X. nematophila produces a diffusible quorum-sensing signal, possibly hydroxybutanoyl homoserine lactone, required for virulence towards insects (Dunphy et al., 1997). However, the genes required for production and response to the putative quorum-sensing signal have not been identified, and experiments testing possible roles for such a signal in the mutualistic association with the nematode have not been published. Recently, we reported that an insertion in the rpoS gene, encoding a ‘stress’ sigma factor, σs, prevents X. nematophila from colonizing nematode intestines, but does not severely affect the stress resistance of this bacterium or its virulence towards insects (Vivas and Goodrich-Blair, 2001). These results suggest that a σs-dependent gene(s) may be required for colonization, and characterization of the σs regulon of X. nematophila is under way to identify such genes.

To gain further insight into the molecular mechanisms of X. nematophila–nematode interactions, and to reveal possible mechanisms of communication between them, we screened bacterial mutants for loss of the ability to associate with the vesicle of the nematode. In nature, S. carpocapsae nematodes depend on X. nematophila for reproduction within dead insect larvae. Non-feeding IJ progeny, carrying X. nematophila symbionts, emerge from spent carcasses to forage for new hosts. This process can be mimicked in the laboratory in the absence of insects by cultivating nematodes on lawns of X. nematophila bacteria (Wouts, 1984). Such co-cultures can be initiated by surface-sterilized nematode eggs that are completely free from bacteria, resulting in IJ progeny that are colonized only by the X. nematophila bacteria present on the plate, if those bacteria are colonization competent (Vivas and Goodrich-Blair, 2001). The technique of X. nematophila–nematode co-culturing was combined with a signature-tagged mutagenesis (STM) technique to create a powerful screen for X. nematophila mutants unable to colonize IJ nematodes mutualistically. The STM process allows multiple mutants to be pooled together for the screening process because each individual mutant contains a transposon ‘tag’ (variable sequence) that distinguishes it from all other mutants in the pool (Hensel et al., 1995). The tagged region can be polymerase chain reaction (PCR) amplified and prepared as a probe for an arrayed blot containing each tag. The loss of a hybridization signal for a particular tag after growth under selective conditions indicates that the mutant marked with that tag was unable to pass the selection criterion (in this case, the ability to colonize nematodes). STM was originally developed to enhance identification of virulence genes in bacterial pathogens of mammals (Hensel et al., 1995). This technique has since been applied to approximately 20 different bacterial and fungal pathogens and has been the subject of several reviews (Holden and Hensel, 1998; Lehoux and Levesque, 2000; Shea et al., 2000) but, until now, has not been applied to the analysis of a mutualistic symbiosis.

This paper describes the identification, through the use of the STM technique, of several X. nematophila genes involved in the mutualistic colonization of S. carpocapsae nematodes. Some of these genes are genetically linked and are essential for colonization, indicating that they may play a major role in this host–microbe interaction.


Bacterial mixture complexity and adaptation of the STM technique to study mutualistic X. nematophila colonization

Although X. nematophila populations in individual nematodes are near clonal (K. Heungens and H. Goodrich-Blair, unpublished results), the large number of IJ progeny that can be gathered from a single co-culture (≈ 105 from each plate) allows for the retrieval from nematodes of all colonization-competent X. nematophila in a complex mixture of bacteria. We developed a technique to recover the bacteria from the vesicles with an efficiency of ≈ 75% (see Experimental procedures) and estimated that the average number of bacteria per nematode is ≈ 60 when grown on plate co-cultures (IJs harvested from insects contain over 100 cfu per nematode). In our STM assay, we screened pools of 48 mutants (i.e. 48 tags) per co-culture and harvested ≈ 45 000 cfu from 10 000 nematodes to isolate the template DNA for amplifying tag probes from each pool. This means that, theoretically, a tag would no longer be represented in the isolated DNA when the bacteria containing that tag were present at ≈ 1000-fold (45 000/48) reduced level compared with the other bacteria in the pool. The actual detection level of a mutant (i.e. loss of a specific tag's hybridization signal) was expected to occur even when a mutant was present in a pool at higher levels than 1:1000, as not all tags will amplify with equal efficiency. This expectation was confirmed experimentally in a preliminary experiment in which a tagged mutant was present as a 10-fold or 100-fold minority within a total population. These experiments revealed that signal from the tagged mutant was detected when it was present as a 10-fold minority, but not when it was present as a 100-fold minority (data not shown). Taken together, our initial experiments suggested that the adaptation of the STM technique to this system should give reproducible results at an appropriate detection level (i.e. loss of signal when a mutant had 100-fold reduced levels of colonization).

Identification of X. nematophila mutants defective in nematode colonization

Initially, sets of 95 X. nematophila signature-tagged mutants, derived from a preselected set of 95 Escherichia coli transposon donors, were pooled, co-cultured with S. carpocapsae nematodes and screened for loss of the ability to colonize IJ progeny. After screening ≈ 450 mutants, from which no colonization-defective mutants were retrieved, it became clear that many of the tags in the set gave insufficient hybridization signal and/or were able to hybridize to multiple tags present in the pool. Therefore, a subset of 48 E. coli STM donors was selected based on strength of signal and lack of cross-hybridization of their tags. Using this new donor set, 3000 X. nematophila mutants were created and screened. Of those, 182 (6.1%) candidates were selected for retesting based on the absence of their tag in the bacteria recovered from nematodes. Putative colonization mutants were cultured individually (in the absence of any other bacteria) with nematodes, and progeny IJs were analysed for the presence of bacteria (see Experimental procedures). Fifteen mutants (0.5%) were reproducibly defective in their ability to colonize nematode intestines and were characterized further. Southern hybridization revealed that each of the 15 mutants had a single transposon insertion, and the genomic regions flanking the transposon insertions were sequenced. Sequence comparisons showed that several mutants had transposons in the same locus, although none of the insertions was in identical positions within a locus. Based on this information, we were able to divide the mutants into nine sequence classes, of which three had multiple mutant representatives (Table 1).

Table 1. Colonization-defective mutants, disrupted alleles and quantification of colonization defects.
StrainaAlleleSequence classVesicle colonizationb
  • a.  

    Mutants with colonization defect.

  • b.  

    Vesicle colonization expressed as average number of bacteria per nematode assessed by the grinding method (see Experimental procedures ). Ranges reflect variations in absolute numbers of bacteria recovered over multiple experiments and/or from different strains.

HGB081Wild typeN.A.44.6 ± 11.7
HGB310 nilA1 ::Tn 5 nilA 1.2–3.2
HGB311 nilB2 ::Tn 5 nilB < 0.005
HGB312 nilB3 ::Tn 5   
HGB313 nilB4 ::Tn 5   
HGB314 nilC5 ::Tn 5 nilC < 0.005
HGB315 nilD6:: Tn 5 nilD 0.72–2.82
HGB316 serC1 ::Tn 5 serC 0.005–0.014
HGB317 aroA1 ::Tn 5 aroA 0.07–0.63
HGB318 aroA2 ::Tn 5   
HGB319 aroA3 ::Tn 5   
HGB320 aroA4 ::Tn 5   
HGB321 lrp-1 ::Tn 5 lrp 0.05–2.95
HGB322 rpoE1 ::Tn 5 rpoE 0.61–1.05
HGB323 rpoS1 ::Tn 5 rpoS < 0.005
HGB324 rpoS2 ::Tn 5   

Database homologies of the loci involved in a colonization defect

Sequence analysis of the disrupted genomic loci revealed that, in each of the 15 mutants, the transposon had inserted in a putative open reading frame (ORF). The predicted size of the proteins that would be encoded by these ORFs and the results of database searches are shown in Table 2. Three loci are homologous to genes encoding known regulatory proteins (rpoS, rpoE, lrp), two loci appear to encode amino acid biosynthetic enzymes (aroA and serC) and the product of one locus, designated nilB (nematode intestinal localization), showed similarity to a putative outer membrane protein from Neisseria meningitidis and to predicted proteins from Pasteurella multocida, Haemophilus influenza and Moraxella catarrhalis. The final three X. nematophila loci did not show homology to known genes and are designated nilA, nilC and nilD (Table 2). serC and aroA are tandemly oriented and separated by 158 nucleotides, suggesting that they may be co-transcribed as they are in E. coli (Duncan and Coggins, 1986).

Table 2. Homologies of disrupted loci in X. nematophila colonization mutants.
OrganismSize of protein
% identity
  • a.  

    Size (in amino acids) of predicted protein encoded by disrupted ORF.

  • b.   Homology based on blastp ( Altschul et al., 1997 ). In the absence of a clear protein designation, homologues are listed by accession numbers.

  • c.   The expression of at least three ORFs could be affected by the nilD6 ::Tn 5 insert. See Fig. 2 and Results and Discussion sections on SR2.

nilA 90
nilB 466AAK03599 Pasteurella multocida 50922(44)1E-23
P44087 Haemophilus influenza 36823(43)2E-20
AAD11779 (OmpU) Neisseria meningitidis 48820(40)3E-14
AAC34281 Moraxella catarrhalis 51220(39)2E-14
nilC 282
nilD 33/36/18c
serC 362SerC Yersinia pestis 36173(84)1E-156
Yersinia enterocolitica 36173(83)1E-156
aroA 428AroA Yersinia enterocolitica 42779(87)0.0
Escherichia coli 42777(85)0.0
lrp 164Lrp Proteus mirabilis 16498(98)3E-87
Escherichia coli 16497(98)9E-87
rpoE 191RpoE Yersinia pestis 19187(95)5E-93
Escherichia coli 20286(94)4E-91
rpoS 331RpoS Yersinia pestis 33284(92)1E-138
Escherichia coli 36082(91)1E-138

Phenotypic assays

The colonization defect of each mutant was quantified by surface sterilization of IJs grown on individual mutant lawns, followed by grinding (to release bacteria) and plating for cfu. Disruptions in three of the loci, nilB, nilC and rpoS, resulted in the complete absence of detectable bacteria (< 0.005 bacteria per nematode) within progeny IJs. Of the remaining six loci, the minimum colonization defect was ≈ 15-fold for nilA, nilD and lrp, 50-fold for aroA and rpoE and 3000-fold for serC (Table 1).

To determine whether the mutants displayed other phenotypic alterations in addition to the colonization deficiency, they (with the exception of rpoS, which had been previously characterized; Vivas and Goodrich-Blair, 2001) were subjected to a variety of assays to monitor growth, starvation survival (during 8 days of culturing on LA plates), behaviour and enzyme production (Table 3). HGB310 (nilA), HGB312 (nilB), HGB314 (nilC) and HGB315 (nilD) each displayed wild-type phenotypes in every assay except for a partial decrease in haemolysin activity for HGB310 (nilA). In contrast, multiple perturbed phenotypes were observed for HGB317 (aroA), HGB316 (serC), HGB321 (lrp) and HGB322 (rpoE) (Table 3). Owing to the pleiotropic defects of these mutants, we focused the remainder of our studies on the nil mutants.

Table 3. Selected phenotypes of X. nematophila colonization mutants.
Dye binding
(and growth
on LBTA)
  • a.   See Table 1 for information on designation and for colonization phenotype. For data on the rpoS mutant, see Vivas and Goodrich-Blair (2001 ).

  • b.   Slope of log 10 -transformed OD values from exponential part of growth curve in LB.

  • c.   log 10 of cfu per disk, 7 days after inoculation (see Experimental procedures ). Population size on disks, 1 day after inoculation, was 8.86 ± 0.17 (mean ± SD in log cfu/disk), averaged over all bacterial strains, and was not significantly different between strains ( P  = 0.234).

  • d.  

    Size of colony (mm), 24 h after inoculation on swim plates (0.25% agar).

  • e.  

    Size of halo (mm) surrounding the bacterial colony, 3 days after inoculation on Tween 20-containing plates.

  • f.  

    Subjective evaluation of halo surrounding the bacterial colony, 1–4 days after inoculation on plates.

  • g.  

    Size of halo (mm) surrounding the bacterial colony, 1 day after inoculation of the tester bacterium ( Bacillus subtilis ).

  • h.  

    The rate of increase in halo size was less than wild type.

  • i.  

    Appearance of halo was delayed compared with wild type.

  • Values marked with *are significantly different from the wild-type control according to Dunnett's simultaneous test (P < 0.05).

HGB081Wild typeWild typeYes0.30 ± 0.028.6 ± 0.365.3 ± 3.97.8 ± 0.2Wild typeWild type12.1 ± 0.5
HGB310 nilA Wild typeYes0.31 ± 0.038.6 ± 0.168.0 ± 0.77.8 ± 0.5Wild typeReduced10.3 ± 1.8
HGB312 nilB Wild typeYes0.34 ± 0.028.7 ± 0.065.4 ± 4.87.1 ± 0.7Wild typeWild type11.5 ± 3.2
HGB314 nilC Wild typeYes0.31 ± 0.008.7 ± 0.268.1 ± 2.88.3 ± 0.7Wild typeWild type13.6 ± 2.3
HGB315 nilD Wild typeYes0.34 ± 0.048.4 ± 0.266.4 ± 2.77.4 ± 1.2Wild typeWild type12.4 ± 4.1
HGB316 serC Smaller
Yes (little
0.21 ± 0.02*9.1 ± 0.2*37.5 ± 2.8*2.9 ± 1.3*ReducedNone 4.9 ± 1.9*
HGB317 aroA Wild typeYes (less
0.32 ± 0.067.8 ± 0.1*63.9 ± 3.45.1 ± 2.0Wild typeReducedh 8.6 ± 4.1
HGB321 lrp Smaller
 size and
No (rough
0.31 ± 0.068.6 ± 0.247.6 ± 8.3*2.4 ± 0.5*NoneReducedi 8.5 ± 0.4
HGB322 rpoE Rough
Yes (little
0.25 ± 0.057.2 ± 0.3*60.9 ± 4.45.8 ± 0.1ReducedNone 6.5 ± 2.1*

In addition to its ability to colonize the intestines of nematodes mutualistically, X. nematophila is capable of killing larval-stage insects when delivered directly into the blood system (Forst and Clarke, 2002). Mutations that prevent X. nematophila from colonizing nematode intestines might also affect its virulence towards insects. Therefore, we tested the ability of the nil mutants to kill larval-stage Manduca sexta (tobacco hornworm) larvae after injection. The percentage mortality of insects 24 or 72 h after injection with any of the nil mutants did not differ significantly from that of insects injected with the wild-type parent, indicating that nil mutants are not defective in virulence towards M. sexta insect larvae (Table 4).

Table 4. Virulence of Xenorhabdus nil mutant strains.
X. nematophila Number of injected bacteria% mortalitya
StrainDisrupted locus24 h72 h
  • a.  

    Percentage mortality of Manduca sexta insect larvae after injection of the respective strains. Data are mean ± SD of two experiments (see Experimental procedures ). No significant differences were observed between the strains ( P  = 0.24 and 0.80 for 24 and 72 h respectively) or between injection levels ( P  = 0.20 and 0.80 for 24 and 72 h respectively).

HGB081Wild type  1180 ± 38025 ± 585 ± 5
HGB310 nilA   1890 ± 19755 ± 2590 ± 0
HGB312 nilB   1030 ± 27040 ± 1085 ± 5
HGB314 nilC   802 ± 20245 ± 590 ± 10
HGB315 nilD   1013 ± 39030 ± 1095 ± 5
HGB081Wild type12 667 ± 466785 ± 595 ± 5
HGB310 nilA 19 815 ± 288570 ± 1095 ± 5
HGB312 nilB 10 517 ± 291780 ± 095 ± 5
HGB314 nilC   7317 ± 131790 ± 095 ± 5
HGB315 nilD 11 015 ± 478565 ± 595 ± 5

The nilA, nilB and nilC loci are linked

Sequencing of the DNA flanking the inserts in nilA, nilB and nilC revealed that these loci are connected in a 4 kb region that we have designated SR1 (symbiosis region 1) (Fig. 1). This genetic area contains several transposase-like sequences and is flanked 5′ by a truncated homologue of E. coli argD, encoding acetylornithine aminotransferase, and 3′ by an ORF (designated orfX) with sequence similarity to Clostridium thermocellum xynX, encoding xylanase. SR1 was subjected to a sliding-window (50, 100 and 200 bp windows were used) GC content analysis, which revealed GC contents of 39%, 29% and 45% of the nilA, nilB and nilC ORFs respectively (data not shown). A survey of ≈ 50 kb of X. nematophila sequence revealed an average GC content of 44% (K. Heungens and H. Goodrich-Blair, unpublished results), indicating that nilB and, to a minor extent, nilA are AT-rich and, in that respect, differ from the rest of the genome. nilA and nilB are tandemly oriented and separated by 109 nucleotides, suggesting that they could be co-transcribed.

Figure 1.

Schematic diagram of SR1. Map of SR1 genome area including predicted ORFs (solid arrows) and designations, location of Tn 5 inserts (see Table 1 ) and fragments used in mutant complementation assays (thick solid lines). argD′ indicates a homologue of an E. coli gene encoding acetylornithine aminotransferase, truncated at its 3′ end. orfX has sequence similarity to Clostridium thermocellum xynX , encoding xylanase. ORFs with sequence homology to transposases have been labelled as tn1 (34% identity, 52% positives, 2E-11 to Vibrio anguillarum putative transposase) and tn2 (35% identity, 49% positives, 7E-15 to Wolbachia sp. transposase; 31% identity, 53% positives, 4E-14 to Vibrio cholerae IS 1004 transposase).

SR1 complementation and expression studies

To verify that the transposon-mediated gene disruptions are responsible for the loss of colonization, SR1 mutants were transformed with plasmids carrying specific X. nematophila wild-type genomic fragments (for schematic locations of genomic subclones, see Fig. 1). In no case was a plasmid able to restore wild-type levels of colonization to a mutant. Possible causes of incomplete complementation could be inappropriate levels of nil gene expression from multicopy plasmids, variable maintenance of multicopy plasmids during the colonization assay, creation of dominant-negative alleles by transposon insertion or the presence of contributory secondary mutation(s) in each mutant. To address the multicopy plasmid concern, fragments were also subcloned into a low-copy (two or three copies per cell) vector (Keen et al., 1988) and tested for their ability to complement colonization defects. Again, in no case was a plasmid able to restore wild-type levels of colonization to a mutant (data not shown).

Although not to wild-type levels, HGB312 (nilB) colonization was clearly rescued by pBCSR1-B and pBCSR1-AB, which each contain a wild-type copy of the nilB gene, but not by any other plasmid tested (Table 5), whereas HGB314 (nilC) colonization was rescued only by the plasmid pBCSR1-C, carrying the wild-type copy of the nilC ORF (Table 5). Colonization levels of HGB310 (nilA) with no plasmid or harbouring the vector control ranged from sixfold to 38-fold lower than the wild-type strain, and HGB310 transformed with SR1 or SR2 (Fig. 2) plasmids were always within this range of colonization. The subtle colonization defect of HGB310 (six- to 38-fold reduced) coupled with the inability of plasmid constructs to complement any mutant strain fully (noted above) prohibited conclusions regarding whether the nilA1::Tn5 mutation causes the colonization defect of HGB310 and whether this mutation is polar on nilB expression. Analysis of the colonization phenotype of a strain harbouring a non-polar deletion within nilA should answer these questions conclusively. 

Table 5. Colonization complementation analysis of X. nematophila SR1 mutants.
X. nematophila Average cfu per nematodea grown on SR1 mutants containingb:
StrainDisrupted locuspBCSK+pBCSR1-ApBCSR1-BpBCSR1-ABpBCSR1-CpBCSR2
  • a.  

    Average number of bacteria per nematode; data represent mean ± SD with two replicate co-cultures per experiment.

  • b.   Clones of genomic DNA from HGB007 (see Fig. 1 ). In some experiments, co-culture on media containing Cm can lead to a general reduction in the average number of bacteria per nematode (see Experimental procedures ).

HGB081Wild type57.7 ± 2.458.0 ± 32.483.7 ± 43.463.9 ± 34.262.7 ± 0.558.0 ± 38.2
HGB310 nilA  1.5 ± 0.1 6.3 ± 4.9 9.8 ± 2.7 6.4 ± 2.9 8.1 ± 10.2 3.0 ± 2.0
HGB312 nilB < 0.005< 0.00523.5 ± 29.4 2.5 ± 2.4< 0.005< 0.005
HGB314 nilC < 0.005< 0.005< 0.005< 0.00514.7 ± 5.1< 0.005
Figure 2.

Schematic diagram of SR2. Map of SR2 genome area including predicted ORFs (solid arrows) and designations, location of Tn 5 insert ( nilD6 ::Tn 5 ) and fragments used in mutant complementation assays (SR2 and derivatives, thick solid lines). The SR2-2 region is enlarged in the lower portion of the figure and shows the location of the putative small ORFs. ‘dnaQ indicates a partially sequenced ORF with similarity to E. coli dnaQ encoding DNA polymerase III (ɛ subunit). Dashes represent 29 bp direct repeats, the sequence of which is shown below one of the dashes. Within this sequence is an 8 bp internal inverted repeat (shaded) that is also present in orf1 (light hatches). orf2 and the overlapping orf3 contain a 7 bp inverted repeat of similar, but non-identical, sequence (dark hatches, see text). SR2-9 is identical to SR2-1 except for a single basepair mutation in the overlapping putative start codons of orf1 and orf2 (see Experimental procedures ).

Preliminary analysis of transcription from nilA, nilB and nilC was performed by non-quantitative reverse transcription (RT)-PCR. PCR products corresponding to RNA transcripts of nilA, nilB and nilC were detected in late log phase cells of wild-type X. nematophila (Fig. 3A, B and C), indicating that expression of these genes can occur during laboratory culture and is not limited to a host environment. Furthermore, products specifically amplified from mRNA of each gene were also present in rpoS, lrp and rpoE mutants (Fig. 3A, B and C). Although non-quantitative RT-PCR will not reveal subtle transcriptional regulation, these initial results indicate a lack of major transcriptional control of RpoS, Lrp or RpoE over nilA, nilB or nilC. A transcript containing the sequence between nilA and nilB could not be detected by amplification, supporting the possibility that these two genes are transcribed separately (Fig. 3D).

Figure 3.

RT-PCR of transcripts expressed from the SR1 locus. Reverse transcription (RT) and PCR amplifications using primers specific for nilA (A , expected product size = 194 bp), nilB (B , expected product size = 527 bp) or nilC (C , expected product size = 255 bp) on RNA template isolated from wild type (lanes 1 and 2), rpoS1:: Tn 5 (lanes 3 and 4), rpoE1:: Tn 5 (lanes 5 and 6) or lrp-1:: Tn 5 (lanes 7 and 8). Control reactions with each primer set were performed on wild-type chromosomal DNA (lane 9). In (A–C), reverse transcriptase was added only to the reactions shown in lanes 1, 3, 5 and 7. Reactions shown in (D) were performed with wild-type RNA (lanes 1, 2, 4, 5, 7 and 8) or chromosomal DNA (lanes 3, 6 and 9) using primers specific for nilA (lanes 1–3, expected product size = 194 bp), nilB (lanes 4–6, expected product size = 527 bp) or a transcript spanning both nilA and nilB (lanes 7–9, expected product size = 853 bp). Reverse transcriptase was added to reactions shown in lanes 1, 4 and 7 only. pGEM (Promega) DNA size standards are included (M) in each blot, with the 676 bp band indicated.

Analysis of the deduced amino acid products of SR1

Predicted sizes and molecular weights of the putative Nil proteins encoded within SR1 are: 90 amino acids and 9.99 kDa for NilA, 466 amino acids and 55.02 kDa for NilB and 282 amino acids and 30.26 kDa for NilC. NilA may contain an uncleavable signal peptide sequence (signalp/psort) and two putative transmembrane (TM) domains (for prediction programs, see Experimental procedures): the first predicted to be located between amino acids 1 and 24 and the second predicted to be between amino acids 50 and 83 (exact cut-offs depended on the prediction program). Those programs that predicted two TM domains also indicated that both the N-terminus and the C-terminus of NilA are located in the periplasm, a topology that is consistent with the ‘positive inside rule’ (von Heijne, 1992), in that five of the seven positively charged residues in NilA occur between amino acids 20 and 56. However, the probability value for these TM domains does not reach the cut-off value when using tmhmm or alom, the prediction programs that result in the fewest false-positive predictions (Möller et al., 2001). NilA has no overall homology to sequences in the database (blastp), but amino acids 5–43 of its sequence show similarity to the C-terminal portion of CytB homologues from many organisms (e.g. 46% identity, 61% similarity to residues 232–267 of Thomomys monticola mitochondrial cytochrome B protein). These residues of CytB are located beyond the domain that binds to the haem group (Cramer and Knaff, 1991), precluding any inference of NilA function based on this similarity.

NilB contains a clear signal peptide sequence and is predicted to be cleaved after amino acid 25 (signalp/psort). The majority of the TM prediction programs do not assign an inner membrane-spanning domain to the mature NilB protein. psort predicts that NilB is relocated to the outer membrane, and several predicted β-sheets in the C-terminal half of the protein (psipred/phdsec) may be involved in anchoring at this location. Consistent with these predictions, the putative NilB peptide terminates with a phenylalanine, a feature typical of proteins localized to the outer membrane (Struyve et al., 1991). NilB is similar to a putative outer membrane protein, OmpU, from Neisseria meningitidis as well as to putative proteins from Pasteurella multocida, Haemophilus influenza and Moraxella catarrhalis (blastp). Alignment matches are present over most of the length of the protein, but no specific regions show a high degree of similarity. The gene encoding the M. catarrhalis NilB homologue is located between two genes encoding transferrin-binding proteins, but a role for this ORF product in transferrin binding has not been demonstrated (Myers et al., 1998). Expression of N. meningitidis ompU in E. coli in the presence of haem complemented a haem biosynthesis defect (Stojiljkovic and So, 1999).

Amino acids 1–19 of NilC are predicted to comprise a lipoprotein consensus signal sequence, with the putative lipid attachment site being Cys-20 (psort, Prosite motif). NilC is predicted to be localized to one of the membranes (psort), with the largest probability of localization in the outer membrane based on the criteria of Yamaguchi et al. (1988). NilC has no apparent homologues in the database (blastp) but does have regions of similarity to Bacillus subtilis oligopeptide-binding protein OppA precursor (NilC amino acids 175–263; 34% identities; 43% positives; e = 0.16), E. coli flagellin (NilC amino acids 67–225; 22% identities; 38% positives; e = 0.35) and Salmonella enterica major pilin protein FimA (NilC amino acids 133–227; 25% identities; 43% positives; e = 1.7).

Organization of the nilD locus

Sequencing the DNA flanking the transposon in the nilD mutant revealed that the insertion had occurred in an area, termed SR2 (symbiosis region 2), with several small putative ORFs, designated orfs1, 2, 3, 4 and 5 (Fig. 2). orf1 and orf2 are divergently oriented and have over­lapping start codons. The transposon insertion is predicted to affect the expression of orf1, orf2 and orf3; the insertion is located 4 nucleotides 5′ of the start codon of orf2 and is within the predicted coding sequences of orf1 and orf3. Ending 400 nucleotides 5′ of the Tn5 insertion is an area with 20 direct repeats of 29 bp each, within which are two 8 bp inverted repeats, and between which are 32 bp variable sequence spacers. This region is similar to the E. coli repeat sequences downstream of iap (Nakata et al., 1989). A homologue of an E. coli gene encoding DNA polymerase III (ɛ subunit) (blast e = 1E-38) is present 5′ of the repeat region. Encoded 3′ of the transposon insert is an ORF with similarity to Burkholderia pseudomallei transposase A (e = 3E-3), followed by a homologue to gloA (glyoxalase) of Clostridium acetobutylicum (e = 2E-22) (Fig. 2). The GC content of the genome area spanning the five small ORFs is 33%, considerably lower than the flanking gloA (41%) and 29 bp repeat sequence area (48%). Intriguingly, orf1 and orf2 (as well as the overlapping orf3) each contain an inverted repeat sequence. The sequence of the 8 bp orf1 inverted repeat corresponds exactly to the inverted repeat sequence found within each of the 20 direct repeats. This inverted repeat was not found elsewhere in the area. The sequence of the 7 bp orf2 inverted repeat is similar but not identical to the 8 bp inverted repeat in orf1 (5′-TTCCCGT-3′ versus 5′-TCCCCGTA-3′). No database motifs or similarities were observed for the orf1–orf5 area or the predicted Orf1–Orf5 oligopeptides. Predicted sizes and molecular weights of the putative oligopeptides in SR2 are: 33 amino acids and 3.8 kDa for Orf1; 36 amino acids and 4.3 kDa for Orf2; 18 amino acids and 2.1 kDa for Orf3; 34 amino acids and 4.1 kDa for Orf4; 32 amino acids and 3.9 kDa for Orf5.

SR2 complementation and expression studies

To determine whether the HGB315 (nilD) colonization defect is caused by the nilD6::Tn5 insertion in SR2, the mutant strain was transformed with plasmids containing various fragments of the wild-type SR2 (Fig. 2). Plasmids carrying these SR2s clearly restored nilD colonization, albeit not to the level of the wild-type strain (Table 6). HGB315 colonization was not complemented by a plasmid harbouring the SR2 fragment with a point mutation in the overlapping putative start codons of orf1 and orf2 (= SR2–9 with ATG to ACG in orf1 and ATG to GTG in orf2), confirming the involvement of the SR2 in colonization. The point mutation does not alter the predicted amino acid sequence of an orf3 translation product. A possible ribosome binding site (GGGAA) is centred 9 bp upstream of the start codon of orf1, but no obvious ribosome binding sites are apparent upstream of orf2 or orf3. Non-quantitative RT-PCR experiments were designed to determine whether the SR2 region is transcribed and, if so, which strand. Reverse transcriptase reactions were carried out using primers complementary to either the top or the bottom strands of SR2 (see Table 7; Experimental procedures) and RNA isolated from late log stage wild-type, rpoS, lrp or rpoE cells. The resulting cDNA templates were then amplified by PCR. Products of the appropriate size were detected consistent with both the top and the bottom strand being transcribed (Fig. 4) in all backgrounds tested.

Table 6. Colonization complementation analysis of HGB315 ( nilD6 ::Tn 5 ).
X. nematophila Average cfu per nematodea grown on SR1 mutants containingb:
StrainDisrupted locuspBCSKpBCSR2pBCSR2-1pBCSR2-2pBCSR2-9
  • a.  

    Average number of bacteria per nematode; data represent mean ± SD with two replicate co-cultures per experiment.

  • b.   Clones of genomic DNA from HGB007 (see Fig. 2 ). HGB315 colonization was not complemented by any SR1 clones (data not shown). In some experiments, co-culture on media containing Cm can lead to a general reduction in the average number of bacteria per nematode (see Experimental procedures ).

HGB081Wild type3.7 ± 0.88.2 ± 1.014.6 ± 0.312.0 ± 1.120.9 ± 3.4
HGB315 nilD < 0.0071.0 ± 0.1 1.7 ± 1.2 0.5 ± 0.1< 0.007
Table 7. Primers used in PCR amplifications.
PrimerSequence (5′ to 3′)TargetaPurposeb
  • a.   See Figs 1 and 2 for location of SR fragments and nil genes. For SR1 primers, forward and reverse designations indicate direction of primer relative to the ORF specified. For SR2 primers, all forward and reverse designations are relative to orf1 .

  • b.  

    Purpose: 1 = complementation; 2 = RT-PCR; 3 = construction of mutation.

  • c.  

    Underlined nucleotides represent added Mlu I restriction site.

  • d.  

    Underlined nucleotide represents mutation.

Figure 4.

RT-PCR of transcripts expressed from the SR2 locus. Reverse transcription and PCR amplification of SR2 transcripts. Reactions specifically detecting nilD top strand (A, expected product size = 180 bp) or nilD bottom strand (B, expected product size = 215 bp) were performed on RNA template from wild type (lanes 1 and 2), rpoS1 ::Tn 5 (lanes 3 and 4), rpoE1:: Tn 5 (lanes 5 and 6) or lrp-1:: Tn 5 (lanes 7 and 8). Control reactions were performed on wild-type chromosomal DNA (lane 9). pGEM (Promega) DNA size standards are included (M) in each blot, with the 222 bp band indicated. The images have been inverted to black on white to visualize the bands better. The bands of ≈ 530 bp present in lanes 1 and 5 are of unknown origin and are of variable occurrence.


This paper is the first report of the use of the STM method (Hensel et al., 1995) in a mutualistic symbiosis. The screen resulted in the identification of a set of loci required by X. nematophila to colonize the intestinal vesicle of its nematode host. The proportion of mutualistic colonization mutants we observed (0.5%) was lower than the proportion of mutants found by STM to have attenuated virulence in other systems, which is usually more than 2% and sometimes as high as 15% (Hamer et al., 2001). Technical details, such as our individual retesting of putative mutants in the absence of competition from wild-type strains and our stringent cut-off level, may explain some of this discrepancy. However, the relatively low percentage of mutations giving rise to a colonization defect might also imply that few genes are required by X. nematophila for its mutualistic association. This supposition is further supported by the fact that multiple independent insertions, each in a different location, were obtained within three of the nine loci identified in this screen. Although this could in part reflect a propensity of Tn5 to insert within these loci, it also argues that, in total, X. nematophila colonization genes comprise a small target for transposition. Additional support for this assertion lies in the finding that, although the serC aroA region of X. nematophila was targeted by five independent insertion events, our screen did not reveal any other amino acid biosynthesis mutants as being defective for colonization. This is in contrast to most STM screens involving mammalian pathogens, in which a significant proportion of the identified mutants have lesions in a wide range of amino acid biosynthetic genes (Shea et al., 2000). The probability that X. nematophila uses a relatively small set of genes during colonization highlights the authenticity of those genes that were identified by the STM method.

It is perhaps not surprising that X. nematophila would require a small set of genes for colonization; differences in the biological nature of mutualistic versus pathogenic interactions might necessitate a different number and set of genes. For example, X. nematophila resides extra­cellularly in the digestive tract of the nematode and does not appear to invade nematode host tissues (Bird and Akhurst, 1983). Therefore, genes encoding functions required for cell or tissue invasion are not likely to be necessary for X. nematophila–host interactions. However, X. nematophila might be expected to require functions for attachment to vesicular cells, nutrient acquisition or synthesis, molecular communication and resistance to host-imposed stresses, as well as regulatory mechanisms to control the expression of these functions. The colonization genes described in this report can now be analysed further to provide insight into X. nematophila physiology within the non-feeding nematode and the nature of the host environment.

One class of genes identified, designated nil, have either no homology to genes in the public sequence database or homology to genes of unknown function. The nil mutants are of particular interest to us, as they do not have obvious growth, survival, behavioural or exoenzyme activity defects. The only phenotypic defect detected in nilB, nilC and nilD mutants is loss of the ability to colonize nematode intestines efficiently, suggesting that these genes play a direct role in nematode colonization. Several mutations were clustered in a single 4 kb genomic region, SR1, that contains four putative genes, nilA, nilB, nilC and tn2. The requirement for nilB and nilC in colonization was confirmed by complementation analyses. Complete analysis of SR1 promoters should reveal operon structure and possible mechanisms of regulation, which may in turn provide insight into the functions of the gene products encoded in this region. Considering the demonstrated role of nilB and nilC in colonization and the predicted outer membrane location of their products, we speculate that NilB and NilC may be involved in a direct interaction with the nematode vesicular environment. Based on the genetic location and putative functions of its homologues in other organisms (Myers et al., 1998; Stojiljkovic and So, 1999), NilB could be involved in iron scavenging, an important survival function for microbes within hosts (Ratledge and Dover, 2000). Characterization of iron-related phenotypes of SR1 mutants, such as survival under iron-limiting conditions, may reveal whether any of the genes in this region play a role in iron homeostasis and whether this function is important for colonization. The function of NilC cannot be inferred from comparative sequence analyses because of lack of a homologue of known or suspected function. The regions of similarity to flagellin and pilin proteins could indicate that NilC forms a higher order structure and/or is involved in a physical interaction with the host vesicle. Elucidation of the role of NilC in colonization will require fundamental analyses of gene expression regulation, protein localization, mutagenesis, protein–protein interactions and NilC–vesicular cell interactions.

A second locus found to be involved in colonization was designated SR2. At this point, we are unable to identify conclusively whether an ORF in SR2 is necessary for colonization and, if so, which one. The SR2-9 fragment, which harbours a mutation that alters the overlapping start codons of orf1 and orf2 but does not affect the predicted amino acid sequence of orf3, does not complement the colonization defect of HGB315 (nilD). This result suggests that either orf1 or orf2 could be involved in colonization, but that orf3 is not likely to be required. An alternative hypothesis is that the colonization factor is messenger RNA(s) transcribed from the nilD region rather than a peptide product. There are precedents for both possibilities. Untranslated RNAs are now recognized as important regulatory elements in E. coli (Wassarman et al., 1999) and are also thought to be involved in the expression of virulence determinants in plant and animal pathogens (Altier et al., 2000; Kreikemeyer et al., 2001; Ma et al., 2001). Small, ribosomally synthesized peptides are known to be involved in regulatory cascades including sporulation in B. subtilis (Lazazzera et al., 1997) and development of nitrogen-fixing heterocysts in cyanobacteria (Yoon and Golden, 1998). Our data show that transcripts are generated from both strands of the nilD locus (Fig. 4), suggesting that both ‘sense’ and ‘antisense’ products (be they RNA or protein) could be involved in colonization. Experiments are under way to distinguish between the possible requirement for nilD-encoded small ORFs, untranslated RNA or both in X. nematophila colonization of nematodes. The potential role of the SR2 repeat region (located 400 nucleotides away from nilD6::Tn5) in colonization has not been tested. However, its proximity to the nilD locus and the presence of the inverted repeat sequences within the nilD locus raise the provocative, testable possibility that the repetitive sequences also play a part in colonization.

Xenorhabdus species have specific nematode host ranges, and the basis of this specificity has not been defined. It is possible that SR1 and/or SR2 are required for colonization because they encode specificity determinants that define the host range of X. nematophila. If so, one might predict the presence or sequence of these regions to vary among closely related species of Xenorhabdus. Indeed, preliminary evidence suggests that the SR2 region, but not the SR1 region, is present in the genomic DNA of X. beddingii ATCC49542 and X. poinarii ATCC49121 (data not shown). These bacteria colonize Steinernema nematodes ( Akhurst, 1983b, 1986 ) but do not colonize S. carpocapsae efficiently ( Akhurst, 1983a ; C. E. Cowles and H. Goodrich-Blair, unpublished results). Studies are under way to examine whether SR2 sequence variations occur among strains, and whether SR1, SR2 or both might be sufficient to expand the host range of non-cognate Xenorhabdus spp. to include S. carpocapsae.

The initiation of an intimate association between a microbe and its host requires the co-ordinated expression of multiple colonization factors. Therefore, an expected class of mutations leading to colonization defects is in genes encoding regulatory proteins. The SR1 nil gene products are not expected to be transcription or translation regulatory proteins, based on their predicted membrane localization and the apparent absence of nucleotide-binding motifs in their primary sequences. However, mutations in homologues of three genes, lrp, rpoE and rpoS, encoding known transcription regulatory proteins, were obtained from the STM screen. Although none of these three genes is essential for nil gene expression (Figs 3 and 4), their products may be involved in regulating other colonization functions. In many Gram-negative bacteria, rpoS encodes a sigma factor involved in virulence and/or stationary phase and stress survival (Loewen and Hengge-Aronis, 1994; Chen et al., 1996; Yildiz and Schoolnik, 1998). Identification of rpoS in this study is an independent confirmation of a previous report from our laboratory of the role of rpoS in nematode colonization (Vivas and Goodrich-Blair, 2001).

lrp encodes the leucine-responsive regulatory protein involved in the regulation of amino acid biosynthesis, catabolism and transport, motility, chromosome organization and osmolarity regulation in E. coli ( Newman and Rongtuan, 1995 ). Several, but not all, of the functions of lrp are dependent on leucine, which is believed to function as a general sensor of the nutritional status of the cell. It is conceivable that any of these functions could be involved in X. nematophila colonization processes. Interestingly, Lrp positively regulates the transcription of the serC aroA operon in E. coli ( Man et al., 1997 ). This raises the possibility that Lrp co-ordinately regulates serC and aroA gene expression in X. nematophila and that the colonization defect of the lrp mutant results from an inability to induce the expression of these genes adequately. It should be noted that aroA has long been implicated in the virulence of a number of pathogenic microbes, including Salmonella typhimurium ( Hoiseth and Stocker, 1981 ). Thus, aroA may encode functions important for the ability of a microbe to colonize any host, regardless of whether the eventual outcome for the host is beneficial or detrimental.

The third putative regulatory protein implicated in X. nematophila colonization is encoded by a homologue of E. coli rpoE. In E. coli, rpoE encodes the extracytoplasmic stress sigma factor, σE, which is negatively regulated by binding of an antisigma factor. An unknown indicator of extracytoplasmic stress activates σE by releasing it from antisigma factor control, ultimately resulting in the expression of stress-related proteins (Raivio and Silhavy, 2001). An ORF predicted to encode an antisigma factor is located 3′ of the rpoE homologue in X. nematophila (data not shown), identical to the genetic arrangement of these genes in E. coli. Well-conserved homologues of rpoE in several organisms have been implicated in conferring protection from oxidative stress and in mediating host interactions (Humphreys et al., 1999; Mathee et al., 1999). As has been suggested for rpoS (Vivas and Goodrich-Blair, 2001), the function of rpoE in X. nematophila colonization could be related to survival inside the vesicle, which may exert multiple stresses on microflora. Host-imposed stress might serve to discourage non-symbiont bacteria from colonizing or to inhibit uncontrolled growth of the symbiont beyond the colonization organ (McFall-Ngai, 1998).

The STM screen reported here was highly successful in revealing previously unidentified loci required for the mutualistic interactions between X. nematophila and its nematode host. Characterization focused on mutants that had colonization defects even in the absence of competitive pressure. A significant number of mutants identified in the initial screen as having colonization defects were able to colonize upon retesting. These mutants may have been falsely identified in the initial screen because of stringent washing conditions. Alternatively, the mutants may have had a colonization defect that was too small to be detected by the non-quantitative microscopic analysis used for retesting. However, some portion of the mutants are likely to have a competitive defect when tested in a background of other strains, while being fully colonization competent when individually tested. Indeed, analysis of one mutant confirmed that it was not competitive in colonization when co-cultured with wild-type X. nematophila, but colonized well under monoxenic conditions (data not shown). Preliminary evidence indicates that each nematode is colonized by one or a few bacterial clones (K. Heungens and H. Goodrich-Blair, unpublished results), suggesting that competition is not occurring routinely after colonization of the nematode intestine. However, the study of competitive mutants may yield insights into ex vivo growth and survival requirements and possibly into early events in the initiation of colonization.

The size of our mutant bank (3000 mutants) was too small to saturate the Xenorhabdus genome with transposition events. Thus, additional nil mutants may be identified when increasing the size of the mutant bank or when using different transposon systems. Even had the STM screen been saturating, there are inherent restrictions to the types of mutants that can be identified with this method, such as mutants whose defects are rescued by the presence of other strains during co-cultivations with nematodes, mutations in essential genes or mutations in redundant genes. Experimental methods such as in vivo expression technology (reviewed by Merrell and Camilli, 2000) or directed mutagenesis of specific loci may reveal additional genes required by X. nematophila for colonization that cannot be detected by the STM method. However, the application of STM to the study of X. nematophila has resulted in the identification of a concise set of genes involved in colonization. Further characterization of these genes will undoubtedly help to elucidate the molecular networks that allow a successful mutualistic interaction between X. nematophila bacteria and S. carpocapsae nematodes.

Experimental procedures

Organisms and growth conditions

Xenorhabdus nematophila strains used were HGB007 (= ATCC 19061) and HGB081 (a spontaneous rifampicin-resistant mutant of ATCC 19061 provided by S. Forst, University of Wisconsin-Milwaukee). Escherichia coli strains used were DH5α (Bethesda Research Laboratories) for plasmid maintenance, S17-1λ pir ( Simon et al., 1983 ) for conjugations and TOP10 (Invitrogen) for maintenance of PCR-amplified fragments that were cloned into pCR2.1 ® -TOPO (Invitrogen). Permanent stocks of cultures were stored at −80°C in LB broth ( Miller, 1972 ) supplemented with 10% DMSO, and cultures were grown at 30°C in LB that was kept in the dark ( Xu and Hurlbert, 1990 ). S. carpocapsae nematodes were grown at room temperature (22–26°C) on lipid agar (LA) plates that were spread with lawns of appropriate X. nematophila strains ( Vivas and Goodrich-Blair, 2001 ). M. sexta (tobacco hornworm) insect eggs were obtained from W. Goodman (University of Wisconsin-Madison) and were reared as described by Vivas and Goodrich-Blair (2001 ). Media were supplemented with Na-pyruvate (0.1%) ( Xu and Hurlbert, 1990 ), kanamycin (Km, 50 µg ml −1 ), rifampicin (Rif, 100 µg ml −1 ) or chloramphenicol (Cm, 15 µg ml −1 ), where appropriate.

DNA manipulations

Chromosomal DNA isolation, DNA digestion, electrophoresis, ligation, E. coli electrotransformation and ethanol precipitation of DNA were performed according to standard procedures (Sambrook et al., 1989). Southern blotting was carried out using the PosiBlot 30–30 Pressure Blotter (Stratagene) and the ECF random prime labelling and amplification system (Amersham Pharmacia). DNA-modifying enzymes were from Promega or Fermentas. Plasmids were isolated using standard alkaline lysis (Sambrook et al., 1989) or plasmid miniprep kits (Bio-Rad). Gel purification was performed using the QIAquick gel extraction kit (Qiagen).

PCR amplification, cloning and sequencing

PCR amplification of DNA fragments from X. nematophila ATCC 19061 was performed using ExTaq or LATaq polymerase and buffers (Takara Shuzo), with 0.2 µM each primer, 0.4 mM dNTPs, 2.5 U of polymerase and 500 ng of genomic DNA template. Cycles (30×) consisted of 20 s of denaturation at 95°C, 30 s of annealing at 55°C and 1 min kb−1 elongation at 72°C. These cycles were preceded by a 2 min denaturation at 95°C and followed by a 7 min final elongation at 72°C. For sequence verification, various fragments were PCR amplified from ATCC 19061 genomic DNA, cloned into pCR2.1®-TOPO and sequenced with the ABI BigDye terminator sequencing system (Applied Biosystems), followed by cleaning with AutoSeq columns (Amersham Pharmacia). Electrophoresis and detection was performed at the University of Wisconsin-Madison Biotechnology Sequencing Center.

For complementation analyses, genomic fragments covering the putative disrupted loci from SR1 (see Fig. 1) were PCR amplified using the following primers: P48 and P45 for SR1-A; P42 and P43 for SR1-B; P42 and P45 for SR1-AB; and P46 and P47 for SR1-C. All primer sequences are shown in Table 7. Primers were constructed with added MluI sites at the 3′ end (Table 7, underlined), although this site was not ultimately utilized. Genomic fragments covering the putative disrupted locus from SR2 (see Fig. 2) were PCR amplified using the following primers: P20 and P21 for SR2; P65 and P38 for SR2-1; and P62 and P63 for SR2-2. SR2-9 was constructed using the same primers as SR2-1 for its outer borders. However, a single basepair mutation in the putative start codon of orf1 and orf2 was introduced in the SR2-9 fragment according to the method of Link et al. (1997), using primers P85 and its reverse complement P86 (see Table 7 with introduced basepair mutation underlined). All fragments were gel purified and cloned into pCR2.1®-TOPO, creating pTopoSR1-A, pTopoSR1-B, etc. SacI- and ApaI-digested insert fragments were subcloned in the corresponding sites of pBCSK+ (Stratagene), creating pBCSR1-A, pBCSR1-B, etc. EcoRI-digested insert fragments were cloned into the EcoRI site of pRK415 (Keen et al., 1988) to create pRKSR1-A, pRKSR1-B, etc. Purified plasmids were used in complementation assays.

Construction of X. nematophila STM mutants

Plasmid pUT miniTn5-Km2 containing variable tags (Hensel et al., 1995) was obtained as a gift from D. Holden. However, 95 preselected E. coli S17-1λpir transformants, each carrying pUT miniTn5-Km2 with a variable tag inside the KpnI site (provided by R. A. Welch), were used as the transposon donors as described previously (Hensel et al., 1995). In preliminary experiments, a set of 48 donors was selected on the basis of the final signal strength of their tags and the lack of cross-hybridization with other tags. Separate conjugations were performed between the donor strains containing these 48 tags and X. nematophila HGB081. Exconjugants were plated on LB+Rif+Km and incubated for 2 days. Individual colonies were transferred into LB+Rif+Km broth in 96-well culture plates in such a way that the same tag was present only once in a block of 6 × 8 wells, and each specific co-ordinate in each plate contained a mutant labelled with the same tag. After incubation on a rotary shaker, DMSO was added to each well to a final concentration of 10%, and the cultures were frozen at −80°C until further use. The suitability of Tn5 insertion for use in X. nematophila was assessed by Southern analysis of EcoRV- or PvuII-digested genomic DNA of 20 randomly selected STM mutants using the transposon as a probe (EcoRI fragment from the donor plasmid). Each of the 20 mutants had only a single hybridizing band, and no two bands were of the same size, indicating that transposition events were occurring only once per strain and in diverse locations. The lack of plasmid integration was verified in 15 strains by arbitrary PCR sequencing (Caetano-Annoles, 1993) of the DNA flanking the transposon inserts.

Preparation of filters

Tags were separately PCR amplified as described by Polissi et al. (1998) but using 50 ng of plasmid DNA from each donor strain as template, ExTaq buffer and ExTaq polymerase. The size and quantity of the PCR products were verified by electrophoresis on a 3% agarose gel. PCR products were denatured with heat (95°C for 5 min) and with 2 N NaOH (10 min with one-third volume), neutralized (one-half volume of 1 M NaCl, 0.217 M sodium citrate, 0.5 M Tris-HCl, pH 8.0, 1 N HCl), and 1 µl of each tag was spotted on Immobilon Ny+ membranes (Millipore) in the same pattern as the stored mutants. A set of tags that had varying levels of cross-hybridizing characteristics but were not used in mutants was spotted on the side and served as indicators for hybridization washes. DNA was cross-linked to the membranes using a XL-1000 UV cross-linker (Spectronics), and blots were stored at room temperature until hybridization with DNA from output pools.

Construction of input and output pools

Sets of 48 STM mutants were grown in LB+Rif+Km starting from −80°C stocks. Cultures within a set were pooled and mixed. For each set, 5 ml of mixed cultures was used as input pools. Cells were spun down, and the pellet was stored frozen at −20°C. For the output pool, 500 µl of the pooled cultures was inoculated on each of three LA plates and incubated for 16 h. Each plate was inoculated with 1000 axenic S. carpocapsae juveniles (J1 stage) (Vivas and Goodrich-Blair, 2001) and incubated at room temperature for 11–14 days. At this time, co-cultures were put in harvesting set-ups (Vivas and Goodrich-Blair, 2001) and, after 3–5 days, IJs from the three co-cultures of each set were pooled, surface sterilized for 3 min in 2% NaOCl (Spectrum Chemical) and rinsed four times with sterile water. For each set, 104 surface-sterilized IJs were ground in a sterile Tenbroeck tissue grinder in a total of 5 ml of LB, and 100 µl of the macerate was plated onto each of five plates containing LB+Rif+Km. After incubation, bacteria were scraped from the plates with 7 ml of LB, spun down, and the pellets were frozen at −20°C until genomic DNA isolation. Once it had been determined that each tag gave the expected hybridization signal, preparation of input pools was discontinued.

Creation of STM probes

Tags were PCR amplified from the genomic DNA from each input and output pool according to the method described by Polissi et al. (1998) with the following modifications. For the first round of PCR, we ran duplicate reactions with 2.5 µg of genomic DNA as template, ExTaq buffer and ExTaq polymerase. The PCR products were pooled, and the DNA was ethanol precipitated and resuspended in 10 µl of H2O before separation by electrophoresis through 3% low-melting-point agarose (Fisher Scientific). For the second round of PCR, we used 0.4 µM of [32P]-dCTP (3000 Ci mmol−1; Amersham Pharmacia), ExTaq buffer and 1 U of ExTaq polymerase. The PCR products were digested with HindIII in a total volume of 40 µl in order to release the invariable arms. The entire digested products were separated by electrophoresis through a 4% low-melting-point agarose gel. The gel fragments containing the 40 bp bands were excised and combined in a microcentrifuge tube with 50 µl of water for use as probes.

Hybridization, washing and signal detection

Blots were prehybridized for 1 h at 55°C in 5× SSC, 2% liquid block (Amersham Pharmacia) and 0.02% SDS. Probes were denatured for 5 min at 97°C and added to the prehybridizing solution. After 16 h of hybridization at 55°C, blots were first rinsed for 20 min at 55°C and 40 min at 65°C in 5× SSC + 0.02% SDS followed by 20 min at 65°C and 10 min at 70°C in 1× SSC + 0.1% SDS. Blots were exposed to a phosphor screen for 3 days and analysed with a Storm 860 phosphorimager (Molecular Dynamics). Image files were processed with paintshoppro software (Jasc Software). Signal strength from tags that cross-hybridized to various degrees allowed adjustment of brightness and contrast settings in order to reduce the inclusion of false positives and negatives.

Retesting of putative colonization mutants

Mutants lacking signal in the output pool were individually co-cultured with nematodes, and the presence/absence of bacteria in the IJs was first determined with the guillotine assay (Vivas and Goodrich-Blair, 2001). Putative colonization mutants that were not detected in 20 IJs were selected for further quantitative analysis. Quantitative analysis consisted of surface sterilization, grinding and dilution plating of 104 nematodes as described for the construction of the output pool. Putative colonization mutants with a minimum 10-fold decrease in colonization (average number of bacteria per nematode) were subjected to at least one more round of quantitative analysis involving at least two different co-cultures to confirm their colonization defect.

Cloning of transposon insert fragments and sequence analysis of flanking regions

Southern analysis of EcoRV- and PvuII-digested genomic DNA from the colonization mutants identified single transposon inserts in all putative colonization mutants (data not shown). EcoRV and/or PvuII, EcoRI or HincII fragments containing the miniTn5 transposon were cloned in the corresponding sites of pBluescript KS+ (Stratagene) and selected based on their ability to confer KmR. The sequence of the DNA flanking the inserts was determined by outwards primer walking starting with the P6 and P7 primers (Hensel et al., 1995) and ending when the sequence of a flanking gene with predictable function was detected based on blastx sequence similarity (Altschul et al., 1997). Sequences were analysed and assembled with editview (Perkin-Elmer) and lasergene (DnaStar) software. Sequence from these clones was verified after PCR amplification and sequencing of the loci from wild-type ATCC 19061 as described above.

DNA and protein sequence analysis

Previously undescribed sequences were analysed for ORFs, GC content, repeat sequences and predicted protein size and weight using dnastrider (Christian Marck) and lasergene (DnaStar) software. Protein sequence similarity was analysed using blastp (Altschul et al., 1997). Protein sequences were analysed for predicted secondary structure using psipred ( (Jones, 1999) and phdsec (Rost and Sander, 1993) as part of the PredictProtein server (http:www.embl-heidelberg.depredictproteinpredictprotein.html). The presence of putative protein motifs was checked with profilescan (http:hits.isb-sib.chcgi-binPFSCAN) using the Prosite database (Hofmann et al., 1999). The presence of putative lipoprotein binding sites was checked with lipop as part of psortii ( (Nakai and Kanehisa, 1991). The presence of predicted signal peptides and cleavage sites was verified with signalp ( (Nielsen et al., 1997) and with psortii. The presence and topology of transmembrane domains was analysed using das ( (Cserzo et al., 1997), alom2 (part of psortii), phdhtm (Rost et al., 1996) (as part of predictprotein), hmmtop (http:www.enzim.huhmmtop) (Tusnády and Simon, 1998; 2001), tmhmm ( (Krogh et al., 2001), toppred2 (http:bioweb.pasteur.frseqanalinterfacestoppred.html) (von Heijne, 1992; Claros and von Heijne, 1994), tmap (http: (Persson and Argos, 1997), memsat (Jones et al., 1994) through the PSIPRED server and sosui ( (Hirokawa et al., 1998), each time using default settings and paying attention to the evaluation of these tools (Möller et al., 2001).

Phenotypic assays for colonization mutants

Growth rate was based on the OD600 of cultures grown in LB broth and was calculated by linear curve fitting after log transformation of the exponential part of the growth curve. Dye binding was evaluated on LBTA (Gerritsen et al., 1992). Survival was assayed on LA as described by Vivas and Goodrich-Blair (2001) but using 1.27 cm filter paper disks (Schleicher and Schuell), individually inoculated with 40 µl of overnight culture. Mutants were analysed for motility as described by Vivas and Goodrich-Blair (2001). Plate assays were performed to observe protease activity (Boemare et al., 1997), lipase activity (Sierra, 1956) and antibiotic activity (Maxwell et al., 1994; Vivas and Goodrich-Blair, 2001). Haemolytic activity was assayed using blood agar plates (Rowe and Welch, 1994) with 5% defibrinated sheep blood (Colorado Serum). Where possible, activity was quantified by measurement of the radius of the cleared area surrounding the cultures. All experiments were conducted twice with a minimum of two replicates per experiment. Data were analysed by analysis of variance (anova) using experiments as blocks. Means were compared with the wild-type control using Dunnett's simultaneous tests (Dunnett, 1955). All analyses were performed with minitab 12.1 (Minitab). Virulence of the mutants towards insects was determined as described by Vivas and Goodrich-Blair (2001) with the following modifications. Overnight cultures of each nil mutant and of the wild-type strain were diluted in phosphate buffer and 10 µl of dilutions, containing an estimated 1000 or 10 000 cfu, were injected into each of 10 insects in a double-blind experiment. The actual number of cfu injected was determined by dilution plating. Treatments were blocked for insect weight and injection time, although these factors were subsequently shown not to be statistically significant. The experiment was conducted twice, and arcsin square root-transformed proportion data were analysed by anova with minitab 12.1 using strain and injection level as factors and experiments as blocks.

Complementation assays

Plasmids carrying various fragments of X. nematophila genomic DNA (see Figs 1 and 2) were transformed into the parent strain HGB081 and into nil mutants (Xu et al., 1989). Transformants were co-cultured with nematodes on LA plates containing Rif and Cm. IJs were harvested from these co-cultures and used to determine the level of bacterial colonization. Nematodes (104) were disrupted either by grinding as described above or by 80 s of sonication in a total of 2 ml of LB in a Branson 1510 sonicating waterbath (Branson Ultrasonics) and dilution plated on LB+Rif to determine average numbers of cfu per nematode. There was a minimum of two replicate co-cultures for each treatment, and experiments were repeated at least once. In some experiments, a strain-independent decrease in overall colonization was observed in treatments in which the bacteria contained pBCSK+ or its derivatives. This seemed to be correlated with the presence of Cm in the medium, the corresponding decrease in thickness of the bacterial lawns on the lipid agar plates and the vigour of the nematode inoculum. Within an experiment, this was not an issue, as the effect was uniform and treatments could be compared with the pBCSK+-containing control. However, this observation should be taken into account when comparing experiments (e.g. data represented in Table 5 versus Table 6) and when evaluating standard deviations when the data presented were the average from more than one experiment.

RNA isolation and RT-PCR analyses

RNA was isolated from LB-grown X. nematophila cells at an OD600 of 1.0 (late log phase) with Trizol reagent (Invitrogen) according to the manufacturer's protocols except that 1.75 ml of Trizol reagent was used for 10 OD600 units of cells. Residual DNA was removed by treatment for 1 h at 37°C with 30 units DNase I (Boehringer Mannheim) per 50 µg of RNA in the presence of 5 mM MgSO4, 0.1 M NaOAc and 20 units of rRNasin (Promega). RNA was ethanol precipitated, resuspended in H2O and stored at −20°C after spectrophotometric quantification. RT-PCR was performed using 100 ng of RNA and the following primer sets: P51 and P52 for nilA; P08 and P32 for nilB; P08 and P52 for the region spanning nilA and nilB; P28 and P29 for nilC; P63 and P55 for nilD top-strand transcript; and P53N and P56 for nilD bottom-strand transcript (primer sequences provided in Table 7). We used the Access RT-PCR system (Promega) according to the manufacturer's recommendations. Template-free and reverse transcriptase-free negative controls did not result in amplification. For detection of top- or bottom-strand synthesis of SR2, cDNA synthesis was separated from amplification. The reverse transcriptase step was carried out with only P53N to detect bottom-strand (orf2 or orf3) transcripts or with only P56 to detect top-strand (orf1) transcripts. After heat inactivation of the reverse transcriptase (incubation for 2 min at 94°C), a second primer was added (P56 to bottom-strand reactions or P55 to top-strand reactions) for PCR amplification. Products were separated on 1.5% agarose gels.

Nucleotide sequence accession number

The SR1 and SR2 sequence data have been submitted to the GenBank database under accession numbers AY077465 and AY077466 respectively. The X. nematophila aroA–serC, lrp and rpoE sequence data have been submitted under accession numbers AY077462, AY077463 and AY077464 respectively.


The authors acknowledge David Holden for supplying pUT miniTn5-Km2 containing STM tags, Rod Welch and Peter Redford for providing a preselected set of E. coli donors and for helpful comments, Jo Handelsman, Gary Roberts, Diana Downs and members of the Goodrich-Blair laboratory for stimulating discussions and critical reading of this manuscript, Walter Goodman for supplying insect eggs, Karen Wassarman for advice regarding RNA analyses, Sumarin Soonsanga for technical help with the STM screen, and Joseph Gawronski-Salerno for technical help with sequence verification. This research was funded by a National Institutes of Health grant GM59776 awarded to H.G.B.