Application of Signature-Tagged Mutagenesis for Identification of Escherichia coli K1 Genes That Contribute to Invasion of Human Brain Microvascular Endothelial Cells

Abstract

Escherichia coli K1 is the leading cause of gram-negative bacterial meningitis in neonates. It is principally due to our limited understanding of the pathogenesis of this disease that the morbidity and mortality rates remain unacceptably high. To identify genes required for E. coli K1 penetration of the blood-brain barrier (BBB), we used the negative selection strategy of signature-tagged transposon mutagenesis (STM) to screen mutants for loss or decreased invasion of human brain microvascular endothelial cells (HBMEC) which comprise the BBB. A total of 3,360 insertion mutants of E. coli K1 were screened, and potential HBMEC invasion mutants were subjected to a secondary invasion screen. Those mutants that failed to pass the serial invasion screens were then tested individually. Seven prototrophic mutants were found to exhibit significantly decreased invasive ability in HBMEC. We identified traJ and five previously uncharacterized loci whose gene products are necessary for HBMEC invasion by E. coli K1. In addition, cnf1, a gene previously shown to play a role in bacterial invasion, was identified. More importantly, a traJ mutant was attenuated in penetration of the BBB in the neonatal rat model of experimental hematogenous meningitis. This is the first in vivo demonstration that traJ is involved in the pathogenesis of E. coli K1 meningitis.

In order for meningitic bacteria to cause disease, the pathogen must invade the normally nonpenetrable blood-brain barrier (BBB). The two principal causes of bacterial meningitis in the neonatal period are Escherichia coli K1 and group B streptococci (10). Earlier work has shown that E. coli K1 and group B streptococci invade brain microvascular endothelial cells (BMEC) and cross the BBB in the newborn rat model of experimental hematogenous meningitis (1, 15, 16, 20). Our previous studies have identified few E. coli K1 gene products that contribute to bacterial invasion from the circulating blood to the central nervous system (CNS). ibeA, ibeB, and outer membrane protein A (OmpA) have been demonstrated to be involved in E. coli K1 BMEC invasion in vitro and causing meningitis in the newborn rat (15, 16, 22, 23). In addition, the K1 capsular polysaccharide has been shown to play a role in bacterial survival during penetration of the BBB (14, 17). Although these factors have been identified to be necessary for efficient E. coli K1 penetration of BMEC in vitro and in vivo, they have not been shown to be sufficient. This suggests that there are yet to be identified genes that are involved in the E. coli K1 penetration of the BBB.

To facilitate the identification of E. coli K1 genes contributing to human BMEC (HBMEC) invasion, the recently developed technique of signature-tagged transposon mutagenesis (STM) was used (13). In STM, a unique DNA tag is incorporated into a transposon that enables each transposon mutant to be distinguished from others. Detection of these tags by hybridization allows for a large number of insertion mutants of E. coli K1 to be simultaneously subjected to a selective process and screened for loss of invasion using the in vitro model system of BBB (i.e., HBMEC).

We report here the isolation of seven STM mutants that exhibit decreased HBMEC invasion. Targeted mutagenesis and complementation analysis indicate that the disrupted loci contribute to E. coli K1 invasion of HBMEC. Furthermore, one mutant, JLB9 (traJ), showed a decrease in ability to penetrate the BBB in the newborn rat.

MATERIALS AND METHODS

Bacterial strains and growth conditions.

E. coli K1 strain E44 is a spontaneous rifampin-resistant mutant derived from cerebrospinal fluid (CSF) isolate RS218 (O18:K1:H7) (29). E91 is an ompA disruption mutant derived from E. coli K1 strain E44 (29). Bacteria were grown at 37°C in brain heart infusion broth (BHI; Difco). The antibiotics rifampin (100 μg/ml), kanamycin (50 μg/ml), ampicillin (100 μg/ml), chloramphenicol (25 μg/ml), and tetracycline (7.5 μg/ml) were added as appropriate.

Generation of the transposon mutant library.

A pool of random signature-tagged mTn5Km2 transposons in the pUT delivery plasmid (13) was transformed into S17-1λ pir (24). Approximately 40,000 colonies were pooled and subcultured to 1.0 unit of optical density at 600 nm, and plasmids were mobilized into E44 via conjugation as follows. In 3 ml of 10 mM MgSO₄, 0.25 ml of mid-logarithmic-phase E44 (recipient strain) was mixed with 0.25 ml of the donor strain. The mixture of bacteria was immobilized on 0.45-μm-pore-size filters. To minimize the number of sibling transconjugants, a total of 35 pools were generated from 11 independent mating experiments, and filters were incubated on M63 agar with no carbon source for approximately 16 h. Transconjugants were plated onto either Luria-Bertani or minimal M63 agar plates (containing Casamino Acids [2 mg/ml], thiamine [5 μg/ml], 1 mM MgSO₄, 0.2% glucose, and nicotinamide [5 μg/ml]) supplemented with rifampin and kanamycin. For ompA background matings, bacteria were plated as above except for the addition of tetracycline (to maintain the antibiotic marker disrupting the ompA gene). The 3,360 individual transposon insertion mutants were assembled into 35 different 96-well microtiter dishes containing BHI supplemented with appropriate antibiotics. Occasionally, for a control in the E44 background pools, less invasive ompA mutant E91 with a random signature-tagged transposon insertion was included in a designated well of a 96-well microtiter dish. All potential invasion mutants were analyzed for exclusive mTn5Km2 insertion events and not integration of the delivery plasmid by confirmation of ampicillin sensitivity (data not shown).

Tissue culture invasion screens and quantitative assays.

HBMEC were prepared as previously described (25). Qualitative invasion screening was performed as described below and in Fig. 1. Briefly, bacteria from each well of a 96-well microtiter dish were pooled and grown to early stationary phase. In duplicate or triplicate, approximately 10⁸ bacteria were added to a single well of a six-well tissue culture tray containing confluent HBMEC with a multiplicity of infection equal to 100. For pools 28 to 41, two serial invasion assays were performed before continuing on to the secondary screening. These assays were performed as previous pools except that the lysed HBMEC in the first round of screening were used as an inoculum for the subsequent invasion assay.

FIG. 1.

Outline of experimental design utilizing STM selection. Primary HBMEC tissue culture invasion (TCI) screens were performed on input pools of 96 random mutants (total of 35; 20 in wild-type [w.t.] and 15 in ompA mutant background). Bacteria (10⁸) were added to wells containing HBMEC; invasive bacteria were recovered and combined (output pool). Chromosomal DNA was prepared from input and output pools. PCR-generated probes of the unique transposon tags were prepared and used to hybridize colony blots of the original 96 transposon mutant pools. Mutants with a significant reduction or loss of hybridization signal in the output pool compared to the input pool were reassembled into new 96-well microtiter dishes. Mutants that hybridized to both the input and output pools were included as positive controls. These new pools were subjected to a secondary screen, as described above. Those mutants that repeatedly showed a significant reduction or loss of hybridization signal when probed with the output pool compared to the input pool were further tested in quantitative HBMEC invasion assays.

Quantitative invasion assays, using approximately 10⁷ early-stationary-phase bacteria, were performed as previously described (1). Percent invasion was calculated as [(number of bacteria recovered)/(number of bacteria inoculated)] × 100. Results were presented either as percent wild type or as relative invasiveness, defined as fold effect on percent invasion compared to the background bacterial strain, which was assigned a value of 1. In Table 1, invasion frequency of wild-type E44 and ompA mutant E91 were each assigned a value of 1 and routinely showed 0.1 to 1.0% and 0.02 to 0.10% HBMEC invasion, respectively. Noninvasive E. coli K-12 strain HB101 was used as a negative control and consistently showed ≤0.002% invasion (data not shown).

TABLE 1.

Characterization of STM mutants^a

E91 (ompA) background		Disrupted gene, % identityc	E44 (wt) background
Mutant	Fold decrease in invasionb		Mutant	Fold decrease in invasion
27.G10	26.0 ± 5.6	o412 (AE000144), 42–100	JLB4	4.3 ± 1.1
10.A8	4.7 ± 0.6	traJ (M19710), 96	JLB9	2.1 ± 0.1
18.D2	5.8 ± 0.4	? (cigA)	JLB13	2.9 ± 0.5
10.C7	5.8 ± 0.3	pmgI (AE000439), 60–74	JLB16	4.6 ± 0.1
		o347 (D90837), 49–100	8.B10	2.8 ± 0.5
		yaiU (AE000144), 44–100	4.A4	3.0 ± 1.4
		cnf1 (U42629), 100	43.F12	2.0 ± 0.1

^aMutants 27.G10, 10.A8, 18.D2, 10.C7, 8.B10, 4.A4, and 43.F12 were derived from independently constructed transposon libraries as described in Materials and Methods. Mutants JLB4, JLB9, JLB13, and JLB16 were constructed by targeted disruption mutagenesis in the wild-type (wt) E44 background as described in Materials and Methods. 

^bCompared to the background strain, either wild-type E44 or ompA mutant E91. Invasiveness of background strains was arbitrarily set at 1. Results are presented as the mean ± standard deviation of at least three individual experiments performed in duplicate. 

^cThe designations assigned to genes disrupted by the transposon insertion were based on homology to genes or putative ORFs entered in the GenBank database. GenBank accession numbers (in parentheses) and percent predicted amino acid identity (over the region of DNA sequenced) are given. For mutant 18.D2, whose transposon inserted into DNA with no homologous sequences entered in the GenBank database (indicated by “?”), the locus was designated cigA, for CNS invasion gene. 

DNA manipulations and hybridizations.

PCR probes were generated and hybridized essentially as previously described (13) except that the gel-purified PCR product was subjected to fluorescein-12-dCTP PCR labeling in a mixture containing 1× PE Amplitaq buffer (Perkin-Elmer), 4 mM MgCl₂, 50 μM each dATP, dTTP, and dGTP, 10 μM dCTP, 50 μM fluorescein-12-dCTP (New England Nuclear), 150 ng each of primers P2 and P4 (13), 0.01% Tween 20, ∼10 ng of DNA (melted gel slice), and 0.5 U of Amplitaq (Perkin-Elmer). Hybridized probes were detected with antifluorescein antibody conjugated to alkaline phosphatase (New England Nuclear). Hybridized blots were developed with the substrate dioxetane (New England Nuclear) and exposed to X-ray film (Kodak).

Invasion gene identification and sequencing.

To clone the DNA flanking the transposon insertion site, chromosomal DNA from each mutant was digested with a restriction enzyme that does not cut within the transposon. The resulting DNA fragments were self-ligated and subjected to inverse PCR (I-PCR) using primers P7 (5′ GCACTTGTGTATAAGAGTCAG 3′) and P9 (5′ CGCAGGGCTTTATTGATTC 3′). The resulting I-PCR products were gel purified and sequenced using primers P7 and/or P9. Then 400 to 600 bp of the I-PCR products were sequenced and analyzed using the BLAST program (National Center for Biotechnology Information at the National Library of Medicine). Sequencing was performed using an Applied Biosystems automated sequencer.

Regeneration of mutations.

Mutants JLB4 (o412), JLB9 (traJ), JLB13 (cigA), and JLB16 (pmgI) were constructed by targeted gene disruption as follows. Cloning and sequence analysis identified the gene disrupted in the ompA mutant background. Based on sequences obtained, the internal open reading frame (ORF) fragments for the gene of interest were generated by PCR. The following primers were synthesized and used to amplify internal ORFs using the original I-PCR fragments as DNA templates. All PCRs used primers P9-SacI (P9 with an additional 5′ SacI restriction site incorporated), JLB4 (5′ CATCGGGCATGATTGG 3′), JLB9 (5′ CGGACGACAAATGCAGAGCC 3′), JLB13 (5′ CATCGAGAAAGGAACCGTAG 3′), and JLB16 (5′ GGTACTGGTGATACTGG 3′). These PCR primers annealing to the 5′ or 3′ ends of the desired internal ORF also have 5′ KpnI restriction enzyme sites incorporated. Internal ORF fragments (average size, ∼300 bp) were digested with SacI and KpnI and ligated into suicide vector pEP185.2 (18). The ligation mixtures were transformed into S17-1λ pir, and the resulting plasmids were mobilized into wild-type E44 by conjugation. Transconjugants were selected on rifampin and chloramphenicol and subsequently assayed for HBMEC invasion phenotype. PCR or Southern analysis confirmed disruptions of intended loci (data not shown).

Cloning of wild-type loci.

Wild-type sequences were isolated as follows. o412, pmgI, and o347 were cloned using E44 chromosomal DNA as the template and primers designed from known E. coli K-12 sequences: o412d (5′ CGTTAACATGAGCAAGC 3′), o412c (5′ CCGCAGCTATTTGTGAATC 3′), pmgl (5′ GCCTTTCCCCTCATGG 3′), pmglb (5′ GGAGTATACCTGCGCGG 3′), o347a (5′ GGAGATGAAACGTTCGTG 3′), and o347b 5′ GATAATTAGAGATTTGCGACG 3′). To clone the finP traJ locus, primers finPa (5′ CTCTCTCCGGATAAGGG 3′) and traJb (5′ GATACATGACACTCTG 3′), based on sequences 5′ and 3′ of finP traJ loci of the F-like plasmid R1-19, respectively, were used to PCR amplify finP traJ from E. coli K1 E44 chromosomal DNA. PCR products, each containing only the designated ORF, were cloned into pCR2.1 (Invitrogen Co.), yielding po412, pfintraJ, ppmgl, and po347. To clone the cigA locus and adjacent DNA, the I-PCR product of mutant 18.D2 was used as a probe to identify a hybridizing cosmid clone p9.B4. p9.B4 contains an approximately 25-kb fragment of E. coli K1 E44 chromosomal DNA cloned into pSupercos vector (Stratagene).

Newborn rat model of hematogenous E. coli K1 meningitis.

E. coli K1 bacteremia and meningitis were induced in 5-day-old rats as previously described (1, 17). Briefly, all members of each litter were randomly divided into study groups to receive intracardiac injection of bacteria grown in BHI supplemented with newborn bovine serum (10⁷ CFU in 50 μl). One half hour later, blood and CSF specimens were obtained for quantitative cultures as previously described (17).

RESULTS

Generation and screening of the transposon mutant bank.

To identify genes that potentially play a role in E. coli K1 invasion of HBMEC, we used two approaches utilizing STM. A library of STM mutants was constructed in strain E44, a derivative of CSF isolate E. coli K1 RS218 (29). In addition, to increase the likelihood of identifying potential invasion genes that act independently of the invasion gene ompA, STM was applied to an ompA mutant of E. coli K1 strain E44, termed E91 (29). A total of 3,360 individual transposon insertions were assembled into 35 different pools (20 created in wild-type E44 and 15 created in the ompA E91 background) and subjected to serial HBMEC invasion screens as outlined in Fig. 1. In the first screen, those mutants that showed loss of or significantly reduced hybridization signals to their unique DNA tags in the output pool compared to the input pool (representing potential mutants that did not invade HBMEC) were reassembled into a new pool. For controls, random positive hybridizing input and output mutants were included in the new pools. These new pools (nine in total) were then rescreened in HBMEC invasion assays. Those mutants that showed a reproducible loss or significant decrease in hybridization signal of output pools were isolated and then subjected to quantitative invasion assays (see below). Southern hybridization of 10 random mutants using an internal fragment of mTn5Km2 as a probe revealed that all had unique and single transposon insertions (data not shown). Potential invasion mutants were assayed for auxotrophy by plating bacteria on minimal medium M63 supplemented with glucose as the sole carbon source. Three potential invasion mutants were found to be auxotrophic and thus were not further characterized.

Invasion phenotype of transposon mutants.

The transposon mutants that reproducibly failed to pass the serial invasion assay screens were then quantitatively assayed for invasion phenotypes. Invasion assays using HBMEC were performed on each of the putative invasion mutants and the parent strain (wild-type E44 or ompA mutant E91). This screening of 26 putative invasion mutants led to the identification of seven attenuated invasion mutants. As shown in Table 1, the mutants decreases in HBMEC invasion ranged from 2.0 ± 0.1- to 26.0 ± 5.6-fold compared to the parent strain. These mutants, three in the wild-type background and four in the ompA background, were further characterized.

Identification of invasion genes.

To identify the disrupted gene in each mutant that was responsible for the decreased invasion, the chromosomal DNA flanking the transposon insertion site was cloned. The nucleotide sequences were determined and subsequently analyzed by searching databases for homologous genes (Table 1). Cloning of the transposon junction of mutant 10.A8 revealed that a transposon had inserted into traJ. This locus is 96% homologous to TraJ of the F-like R1-19 plasmid transfer system and has been recently identified to play a role in E. coli K1 invasion of HBMEC in vitro (2). The mutant phenotype of strain 43.F12 was resultant of a transposon insertion in cnf1. cnf1 encodes cytotoxic necrotizing factor 1 (CNF1), a monomeric toxin that has been shown to induce bacterial phagocytosis in epithelial cells (9). One mutant, 18.D2, appeared to have DNA disrupted that was not homologous to sequences entered in the databases (Table 1). We have termed this newly identified gene cigA, for CNS invasion gene. The remaining four mutants (27.G10, 10.C7, 8.B10, and 4.A4) had transposon insertions in DNA homologous to nonpathogenic E. coli K-12 genes (o412, pmgI, o347, and yaiU, respectively). In E. coli K-12, o412, pmgI, o347, and yaiU are putative ORFs with no identified protein expressed (3). The putative ORF of E. coli K-12 o347 shows no homology to other known proteins. However, for the three other E. coli K-12 genes, based on significant homology to other proteins, their functions can be predicted. For example, E. coli K-12 pmgI encodes a hypothetical protein similar (61% predicted homology) to phosphoglyceromutase. The predicted product of E. coli K-12 ORF yaiU shows similarity to flagellin structural protein. For E. coli K-12 o412, the locus is predicted to encode a protein 51% homologous to E. coli K-12 dihydropyrimidine dehydrogenase.

Reconstruction of invasion mutations.

We sought to confirm that the attenuated invasion phenotype of the mutants in the E91 ompA background was linked to the transposon insertions and determine if these mutations (traJ, cigA, o412, and pmgI) had an additive and/or synergistic effect with the ompA mutation. For this purpose, these mutations (traJ, cigA, o412, and pmgI) were individually reconstructed in wild-type E44 by targeted gene disruption. The resulting mutants, JLB9 (traJ), JLB13 (cigA), JLB4 (o412), and JLB16 (pmgI), exhibited significantly (2.1 ± 0.1- to 4.6 ± 0.1-fold) decreased invasion compared to the wild type (Table 1), while E91 demonstrated a 3.5 ± 0.7-fold decrease in HBMEC invasion compared to the wild type. As shown in Table 1, the STM mutants in the ompA background demonstrated 5- to 26-fold decreases in HBMEC invasion compared to E91. These results suggest that (i) combining an ompA mutation with a mutation in traJ, cigA, o412, or pmgI results in both an additive (for traJ, cigA, and pmgI) and a synergistic (for o412) effect on E. coli K1 HBMEC invasion, and (ii) the decreased invasion phenotypes for these mutants are linked to the loci disrupted by the original transposon insertion.

Complementation of invasion mutant phenotype.

The observation that the mutant phenotypes originally made in an ompA background could be re-created in the wild-type background suggests that the invasion phenotypes are linked to the disrupted loci. However, there is a potential for polar effects on downstream genes when either a plasmid or a transposon has inserted in upstream DNA. In addition, we did not have evidence that the transposon mutants' phenotypes in the wild-type background (i.e., mutants 4.A4 and 8.B10) were not due to an unlinked spontaneous mutation. To address this, we attempted to obtain complementing clones for the transposon insertion mutants 8.B10 (o347) and 4.A4 (yaiU) and for regenerated mutants JLB9 (traJ), JLB13 (cigA), JLB4 (o412), and JLB16 (pmgI). For those identified loci that were homologous to known genes in the database (i.e., traJ, o412, pmgI, o347, and yaiU), sequences were PCR amplified using wild-type E44 chromosomal DNA as a template. PCR products were cloned into pCR2.1 and introduced into their mutant strains. Using this approach, we were able to clone traJ, o412, pmgI, and o347; however, multiple attempts in the cloning of yaiU were not successful. For cigA, the identified locus that was found to have no homologue in the database, an internal fragment of cigA was used as a probe to obtain hybridizing wild-type DNA cloned in an E. coli K1 E44 cosmid library. This clone, p9.B4, is the cosmid cloning vector pSupercos with an approximately 25-kb piece of E44 DNA insert. Complementation of the mutant phenotype compared to wild-type E. coli K1 harboring the relevant cloning vector was assessed in HBMEC invasion assays. As shown in Fig. 2, when wild-type DNA was supplied in trans, the invasion phenotype was fully restored for mutants JLB9 (traJ), JLB4 (o412), JLB16 (pmgI), 8B.10 (o347), and JLB13 (cigA). The complementation data for JLB9 (traJ) are consistent with our previous results (2). These data support that these particular loci (i.e., traJ, cigA, o412, o347, and pmgI) play a role in E. coli K1 invasion of HBMEC.

FIG. 2.

Wild-type DNA supplied in trans complements mutant invasion phenotype. HBMEC invasion assays were performed with strains indicated. Results are presented as relative invasion, where the value for wild-type E44 harboring each cloning vector was arbitrarily set at 100%. The data shown are from a single experiment performed in duplicate and are representative of several experiments with similar results.

Characterization of the ability of JLB9 (traJ) to penetrate the BBB in vivo.

Concurrent with these studies and utilizing a different experimental approach, the traJ locus was independently identified as a contributor to E. coli K1 HBMEC invasion in vitro (2). We therefore examined the ability of the traJ mutant (JLB9) compared to wild-type E44 to penetrate the BBB in the newborn rat model of experimental hematogenous meningitis. The traJ mutant (JLB9) and wild-type E44 were administered via intracardiac injection to 5-day-old rats; approximately 1 h after inoculation, blood and CSF specimens were obtained. As shown in Table 2, the magnitude of bacteremia was similar between the two groups; however, the occurrence of meningitis as shown by positive CSF cultures was significantly lower in animals receiving JLB9 (traJ) than those receiving wild-type E44. These findings indicate that the traJ locus influences the ability of E. coli K1 to invade HBMEC in vitro and to cross the BBB in vivo.

TABLE 2.

Bacterial counts in blood and number of animals with positive CSF cultures in groups receiving wild-type E. coli K1 strain E44 or mutant JLB9 (traJ)

Strain	n	Mean bacteremia (log CFU/ml of blood) ± SD	No. (%) of animals with positive CSF culture
E44	51	7.22 ± 0.59	34 (67)a
JLB9	50	7.10 ± 0.44	23 (46)

^aP < 0.05 by chi-square test. 

DISCUSSION

The STM technique has been used successfully to identify attenuated or avirulent transposon mutants of Salmonella spp., Vibrio cholerae, Staphylococcus aureus, Mycobacterium tuberculosis, and Legionella pneumophila in animal model systems (5–8, 13, 19, 26). Although we and others have used a well-established biologically relevant newborn rat model of experimental hematogenous meningitis for assaying the penetration of wild-type and mutant bacteria through the BBB (15, 16, 20), we found there were some limitations for applying STM to the in vivo model. In STM, the input pool must have an inoculating dose such that each mutant is well represented and the mutant pool is able to establish an infection. Likewise, the output (postselection) pool requires an adequate amount of each insertion mutant to be recovered and be represented such that a potential avirulent or attenuated mutant can be accurately detected (rather than being a false negative). In the newborn rat model of experimental hematogenous meningitis, we have previously shown that the inoculating dose is within the parameter needed for STM applications (e.g., 10⁶ to 10⁷ CFU). However, approximately 100 to 1,000 CFU are recovered from the CSF (i.e., bacteria that penetrate the BBB), and only 50 to 70% of the animals infected with invasive E. coli K1 are CSF culture positive under the experimental conditions we have used (1, 15–17). Thus, the yield of postselected bacteria would not be sufficient for adequate representation and detection of output pools. In contrast, using the in vitro model of the BBB, 10⁷ bacteria are added to a monolayer of HBMEC and the resulting output pool (those that invade) is approximately 10⁴ to 10⁵ (1, 15, 16). Thus, every transposon mutant should be adequately represented and sufficiently detected. In addition, we have previously shown a strong correlation between in vitro BMEC invasion phenotype and in vivo traversal of the BBB (1, 15, 16), and thus the application of STM to the in vitro model of BBB would be relevant in the identification of CNS invasion genes for E. coli K1.

In this study we have identified seven HBMEC invasion mutants of E. coli K1. The classification of mutants identified was diverse. One locus, traJ, was identified as a contributor to E. coli K1 HBMEC invasion, as shown by a twofold decrease of a traJ mutant (JLB9) in HBMEC invasion in vitro. More important, JLB9 (traJ) demonstrated a decrease in the ability to penetrate the BBB in the newborn rat model of experimental hematogenous meningitis compared to the wild type. These results are consistent with our previous observations that a twofold difference in E. coli K1 HBMEC invasion in vitro is biologically significant (1, 28). To our knowledge, this is the first demonstration of an in vivo phenotype for a traJ mutant. traJ belongs to a cluster of genes within the F-like plasmid R1-19 transfer region called the tra operon. In the F plasmid conjugation system, TraJ positively regulates the expression of several structural and regulatory genes necessary for DNA conjugation. FinP is an antisense RNA encoded divergently within the 5′ region traJ and negatively controls the expression of traJ (11, 27). The mutant invasion phenotype in the ompA background (mutant 10.A8) was a result of the transposon inserting into the 3′ end of traJ and is upstream of the predicted promoter region of antisense transcript finP. For the regenerated mutant in the wild-type background (mutant JLB9), the 5′ portion of the traJ gene was disrupted via plasmid insertional mutagenesis. Thus, by making a disruption mutation within the 5′ region of traJ, finP was also altered. Nonetheless, given that in E. coli K1 a mutation in traJ leads to a decrease in invasion in vitro and in vivo, one could hypothesize that the E. coli K1 TraJ homologue is necessary for the expression of a gene(s) required for efficient penetration of the BBB. Our preliminary data indicate that E. coli K1 possesses homologues of finP, traJ, and traM (encodes a DNA binding protein necessary for conjugation) (data not shown); thus, it is tempting to speculate there is an operon of E. coli K1 genes homologous to an F-like plasmid tra operon required for DNA mobilization. Of interest, a new class of macromolecule secretion components called the family of type IV transporters has been recently characterized (4). Most members of the family function primarily to mobilize DNA, but additional members have been found to have alternate functions including putative transport of multi-subunit proteins across membrane barriers (4). Further characterization of E. coli K1 finP traJ is necessary in order to delineate the relationship between this locus and E. coli K1 invasion of the CNS. In addition, whether E. coli K1 possesses a type IV secretion system and this transport system participates in the pathogenesis of meningitis remains to be determined.

Mutant 43.F12, which showed a twofold decrease in HBMEC invasion in vitro, was the resultant of the transposon inserting into cnf1. CNF1 is a protein toxin produced by several pathogenic E. coli strains and has been shown to induce bacterial phagocytosis in epithelial cells (9). In cultured epithelial cells, CNF1 activates Rho GTPase, leading to the constitutive activation of Rho and induction of actin polymerization (12). Recent work has demonstrated that invasive E. coli K1 induces actin polymerization of BMEC, and prevention of actin polymerization by cytochalasin D inhibits E. coli K1 invasion of BMEC (21). It remains to be determined if the mechanisms for CNF1 inducing phagocytosis in epithelial cells and E. coli K1 invasion of HBMEC are similar.

Several of the mutations isolated had transposons inserted into loci which were homologous to nonpathogenic E. coli K-12 genes. These loci, o412, pmgI, o347, and yaiU, were identified in E. coli K-12 as putative ORFs with no documented function or products (3). Two mutations (pmgI and o412) appeared to be results of a transposon insertion into genes that, by sequence homology, may participate in biosynthetic or metabolic pathways. In addition, one mutation appeared to have a transposon inserted into a gene homologous to E. coli K-12 yaiU. In E. coli K-12, the predicted protein product of yaiU shows a low level of homology to flagellin structural protein (3). Wild-type E. coli K1 strain E44 is motile; however, in vitro motility assays revealed that mutant 4.A4 (yaiU) has no altered motility phenotype compared to the wild type (data not shown). Lastly, mutant 8.B10 (o347) was the resultant of a transposon inserting into DNA homologous with E. coli K-12 that has no known product or predicted function. Until these loci are further characterized, their precise roles in E. coli K1 HBMEC invasion can only be speculated.

One mutant isolated, 18.D2, was the result of a transposon insertion into DNA sequences that are unique to E. coli K1. Although this locus showed no significant homology to other sequences in the databases, the predicted amino acid sequences did show a low percentage of homology to a putative tyrosine kinase-like protein of Caenorhabditis elegans. We termed this locus cigA, for CNS invasion gene. Of interest, preliminary sequence data for this locus have revealed that DNA within several kilobases of cigA has a significantly lower G+C content (38%) than the remainder of the E. coli K1 genome (49%), suggesting this invasion locus may have been acquired via horizontal gene transfer. Because thus far only complementation with the large cosmid library clone containing cigA has been tested, additional studies are needed to determine whether cigA or an operon of genes including cigA is required for efficient E. coli K1 invasion of HBMEC.

In summary, we have identified several loci of E. coli K1 that contribute to invasion of HMBEC, a prerequisite step in meningitis. In addition, we have shown that a traJ mutant is attenuated in the ability to cause meningitis in the neonatal rat. The precise mechanism(s) of TraJ in the pathogenesis of E. coli K1 meningitis, possibly involving a type IV secretion system, is currently being investigated. A better understanding for the molecular basis of E. coli K1 penetration of the BBB could potentially lead to the development of novel therapeutic and preventative strategies for E. coli K1 meningitis.