INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Haemophilus influenzae is a natural inhabitant of the human upper respiratory tract. Encapsulated strains, in particular those with a serotype b polysaccharide, are important pathogens causing systemic disease such as meningitis, epiglottitis, cellulitis, arthritis, sepsis, and pneumonia. Nonencapsulated (nontypeable) H. influenzae is a frequent cause of respiratory tract infections, including otitis media and sinusitis, and persists in the lower respiratory tracts of patients with chronic obstructive pulmonary disease (COPD) and cystic fibrosis (CF) (13, 23, 26, 35, 36). Vaccination of large groups of children with an H. influenzae type b conjugate vaccine eradicated systemic disease and also reduced oropharyngeal carriage of H. influenzae type b (11, 22, 28). However, antibodies against H. influenzae type b are not effective against nontypeable H. influenzae since these bacteria lack the capsule antigen.

Passage of H. influenzae type b through the nasopharyngeal epithelium is assumed to occur prior to systemic disease as a step toward invading the bloodstream (25). Also, nontypeable H. influenzae passes through the respiratory epithelium and has been observed in the subepithelial layers of the respiratory tract during carriage and during chronic lower respiratory tract infections in CF and COPD patients (10, 17, 24). Infections in CF and COPD patients are common despite the presence of specific antibodies, suggesting that H. influenzae can circumvent antibody-mediated defense mechanisms. Detection of H. influenzae between the epithelial cells in lung tissue of these patients and findings obtained with tissue culture model studies indicate that H. influenzae penetrates between epithelial cells and resides beneath and between these cells (30, 33, 38). The effect of bactericidal antibiotics and bactericidal activity mediated by specific antibodies in the presence of complement on H. influenzae concealed in epithelial cell layers in vitro was significantly less than that on H. influenzae in the apical fluid (37). Therefore, penetration between the epithelial cells may play an important role in the persistence of H. influenzae in the lower respiratory tract.

In this study we constructed a library of chromosomal DNA of nontypeable H. influenzae strain A960053 in Escherichia coli to identify genes of H. influenzae involved in paracytosis. Clones penetrating between epithelial cells were subcloned, and two genes likely involved in paracytosis in nontypeable H. influenzae strain A960053 were disrupted. We show that disruption of open reading frame (ORF) HI0638 in this strain reduced paracytosis through the epithelial cells, and therefore named the associated protein paracytin.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Bacterial strains and plasmids. Nonencapsulated H. influenzae strains A960053 and CF47A were isolated from sputum samples of CF patients, and H. influenzae strain A950006 was obtained from the sputum of a COPD patient (14, 23). H. influenzae strain Rdrec1, a recombination-deficient mutant of strain Rd, has been described before (32). For cloning purposes, E. coli strains DH5 and NM522 and plasmids pGJB103 (34), pUC21 (39), pUC19, and pBR322 were used. The TA vector pCR2.1 was supplied with the TA cloning kit (Invitrogen, Leek, The Netherlands). pUC19HS was constructed by digesting pUC19 with HindII and SmaI followed by self-ligation.

Growth conditions. H. influenzae strains were cultured overnight on chocolate agar plates at 37°C in humid air enriched with 5% CO₂. E. coli strains were grown on Luria-Bertani (LB) agar plates. Strains were stored at 70°C in peptone containing 15% glycerol. Standard concentrations of antibiotics were 100 µg/ml for ampicillin (pUC vectors), 12.5 µg/ml for tetracycline (pGJB103), and 50 µg/ml for kanamycin (pBR322-kana). The kanamycin-resistant knockout mutants of H. influenzae A960053 were grown in the presence of 20 µg of kanamycin/ml.

Cell culture. The epithelial cell line NCI-H292 (ATCC CRL 1848) (2), originating from a human lung mucoepidermoid carcinoma, was maintained in HEPES-buffered RPMI 1640 medium (Gibco, Life Technologies, Breda, The Netherlands) supplemented with 10% fetal calf serum (Boehringer, Mannheim, Germany) without antibiotics (RPMI) and passaged as described previously (38). Cell layers for paracytosis assays were prepared on a 10-mm-diameter tissue culture insert (Nunc, Roskilde, Denmark) with 0.2-µm pores as described previously (37).

Paracytosis assay. The number of bacteria present between the epithelial cells was determined as the number of bacteria surviving gentamicin treatment of the cell layer after infection of the cell layer as described previously (37). Briefly, H. influenzae or E. coli cells grown overnight on chocolate or LB agar plates, respectively, were suspended in phosphate-buffered saline (PBS) to an optical density at 600 nm of 1 and 40 µl of the bacterial suspension was added to 360 µl of apical fluid. The apical fluid now contained 0.5 × 10⁸ to 2 × 10⁸ CFU/ml for H. influenzae strains and 0.2 × 10⁸ to 0.5 × 10⁸ CFU/ml for E. coli DH5. The infected cell layers were incubated at 37°C in a CO₂ incubator for 4 h or overnight. Then, the infected cell layers were washed three times with 200 µl of RPMI and incubated with 450 µl of RPMI containing gentamicin (Centrafarm). H. influenzae strains were exposed to 200 µg/ml for 2 h, and E. coli DH5 was exposed to 50 µg/ml for 1 h. The concentrations used were at least 10 times the MICs for the strains. The apical fluid was removed, and the number of bacteria was determined. The cell layers were again washed and lysed with 1% saponin solution in PBS. The numbers of bacteria in the various fractions were determined by counting the numbers of CFU per milliliter of appropriate dilutions of samples of the fractions on chocolate agar plates. Alternatively, cross sections of the cell layers were examined by light microscopy or transmission electron microscopy. These sections were prepared as described before (38).

Construction of a chromosomal library of H. influenzae in E. coli DH5. H. influenzae chromosomal DNA was isolated by the phenol-chloroform method (19). Sau3A partial-restriction digests of chromosomal DNA were separated by agarose gel electrophoresis, and DNA fragments in the 8- to 12-kb range were used to construct a library by ligation into the unique BglII site of pGJB103 (34). The ligation mixture was transformed to E. coli DH5 by electroporation.

Subcloning techniques. Plasmids were isolated using the Wizard Plus SV miniprep kit (Promega). Restriction endonuclease digestions and gel electrophoresis were performed according to standard techniques (31). Fragments were isolated from the agarose gels using the QIAquick gel extraction kit (Qiagen). DNA ligations were done according to standard techniques (31), with addition of 1 mM ATP. Ligated plasmids were introduced into E. coli NM522 by electroporation and later transformed to E. coli DH5. Transformants containing pUC19 plus an insert were selected by blue/white screening on LB plates containing IPTG (isopropyl--D-thiogalactopyranoside)-X-Gal (5-bromo-4-chloro-3-indolyl--D-galactopyranoside) and 100 µg of ampicillin/ml. The plasmids were isolated and transformed into E. coli DH5. PCRs were performed using the Ready To Go beads (Pharmacia Biotech).

Construction of HI0636 and HI0638 knockout mutants of strain A960053. HI0636 and HI0638 knockout mutants of H. influenzae strain A960053 were constructed by insertion of the kanamycin resistance gene from pBR322-kana into the genes by homologous recombination. To clone ORFs HI0636 and HI0638 with the flanking regions, several oligonucleotide primers were generated as indicated in Table 1 and Fig. 1. To construct the kanamycin resistance gene in ORF HI0636, the 2-kb product obtained by PCR with primer pr2-0636BAM on chromosomal DNA of H. influenzae strain A960053 was cloned into the TA vector and transformed into E. coli, resulting in clone pHIP12. Primers pr1-0636BAM and pr1-0638 were used in a PCR with clone 9.1 to amplify the sequence downstream of HI0636. The resulting product was purified from gel and ligated into the TA vector, which was transformed into E. coli resulting in clone pHIP13. The cloned fragment was obtained from pHIP13 by digestion with XbaI and BamHI and ligated into XbaI- and BamHI-digested pUC21 together with the kanamycin box from BamHI-digested pBR322-kana, resulting in clone pHIP15b. The upstream region of HI0636 was ligated into the BamHI site upstream of the kanamycin box in clone pHIP15b, resulting in clone pHIP15. To construct the kanamycin resistance gene in ORF HI0638, a 3-kb PstI fragment from clone 9.1 containing HI0638 and approximately 1 kb of the flanking region at both sides was ligated in pUC19 and transformed into E. coli, resulting in clone 0638FL. To obtain a BamHI restriction site in HI0638, the upstream and downstream sequences of HI0638 were amplified separately by performing PCRs with the primer sets M13forward and pr1-0638BAM and M13reversed and pr2-0638BAM on clone 0638F1. Both PCR products were purified from gel and cloned into the TA vector, resulting in pHIP10 and pHIP11, respectively. The cloned fragment of pHIP11 was obtained by digestion with PstI and BamHI and ligated into PstI- and BamHI-digested pUC19, resulting in clone pHIP14a. The cloned fragment of pHIP10 was obtained by digestion with SacI and BamHI and ligated into SacI- and BamHI-digested pHIP14a, resulting in pHIP14b. The kanamycin box was obtained from pBR322-kana by digestion with BamHI and ligated into the BamHI site in pHIP14b, resulting in pHIP14. Plasmids pHIP15 and pHIP14 were linearized with NdeI and transformed into H. influenzae strain A960053 using the method described by Herriott et al. (16), resulting in kanamycin-resistant colonies. The presence of the kanamycin box in HI0636 or HI0638 was checked by PCR with the primer combinations prHI0636D and prMS0637 and prHI0638UP and prHI0638D, respectively.

View larger version (15K): [in this window] [in a new window]   FIG. 1.   Relative positions of the primers on the chromosomal fragment of clone 9.1.

DNA sequence analysis. DNA sequence analysis was performed using the big dye terminator cycle sequencing ready reaction kit (Perkin-Elmer, Beaconsfield, Great Britain) according to the manufacturer description. Primers pGJB-P1 and pGJB-P2 hybridizing to regions near the BglII site of pGJB103 were designed to sequence the chromosomal fragment in the pGJB103 clone (Table 1). The sequences of several subclones in pUC19 were determined using the M13forward and M13reversed primers. DNA sequence analysis was performed with an automated fluorescent DNA sequencer, model 310 (Perkin-Elmer). Data were analyzed with Auto Assembler, version 2.0 (ABI Prism), and aligned with known sequences using Clone Manager, version 5.0. The sequences of H. influenzae strain A960053 were in most cases compared to sequences of H. influenzae strain Rd-Kw20, the complete genome of which has been sequenced (9) and is available in GenBank (accession no. L42023). Prediction of protein localization sites of the amino acid sequences was performed using the PSORT program available online (http://psort.nibb.acc.jp).

Statistics. Data on the number of bacteria in the paracytosis assays were evaluated using the Student t test. The data on the E. coli clones were evaluated pairwise because of a large day-to-day difference in the number of bacteria of the negative control isolated from the cell layer. For the paired t test, each pair consisted of a test clone and the negative control in the same experiment. P values of <0.05 were considered statistically significant.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Survival of bacteria in the cell layer as measured by gentamicin assay. To identify the bacterial component responsible for paracytosis, we chose to clone a library of H. influenzae chromosomal DNA in a paracytosis-negative bacterium. Two clinical H. influenzae isolates, strains A960053 and A950006, were used as donor strains, since H. influenzae strain A960053 did not adhere to the cell line and H. influenzae strain A950006 adhered well. H. influenzae strain Rdrec1 containing pGJB103 and E. coli strain DH5 containing pGJB103 were tested to see if they could be used as recipients for cloning. Selection of clones penetrating into the epithelial cell layers, determined by survival in the presence of gentamicin, has been described before (37). In order to test the chromosomal library in E. coli DH5, we adjusted the incubation time from 24 to 4 h to prevent damage of the cell layer by E. coli. After 4 h of incubation the number of bacteria in the apical medium was 0.5 × 10⁸ to 2 × 10⁸ CFU/ml for all strains, similar to that described earlier for strain A960053 (37). After 4 h of incubation, approximately 10⁵ CFU of the cell-associated bacteria of the three H. influenzae strains A960053, A950006, and Rdrec1/ml were found (Fig. 2), whereas no bacteria in the apical medium survived the gentamicin treatment, indicating that these H. influenzae strains had penetrated into the cell layer. The number of bacteria of H. influenzae strain A950006 was slightly higher than the number of bacteria of strain A960053, probably due to the adherence of strain A950006 to the cells. Although the number of bacteria of strain Rdrec1 from the cell layer was significantly lower than those of strains A960053 (P < 0.02) and A950006 (P < 0.01), we considered the difference in numbers of bacteria between Rd and the clinical isolates too small to use Rdrec1 as a recipient for the chromosomal library. The number of bacteria recovered from the cell layer after infection with E. coli DH5 was 10- to 100-fold lower than that of either of the H. influenzae strains (P < 0.01) (Fig. 2), and E. coli DH5 was therefore used for further cloning.

View larger version (16K): [in this window] [in a new window]   FIG. 2.   The numbers of bacteria cultured from of the lysed epithelial cell layers after paracytosis, 4 h postinoculation with H. influenzae strains A960053 (CF), A960005 (COPD), and Rd and E. coli DH5 containing plasmid pGJB103.

Selection of E. coli DH5 clones with elevated levels of paracytosis. Since the adherent H. influenzae strain A950006 showed slightly higher numbers of bacteria penetrating into the cell layers and since adherent strains had been shown earlier to pass through the cell layers in larger amounts (37, 38; M. van Schilfgaarde, submitted for publication), we cloned a chromosomal library of A950006 DNA in E. coli DH5. The adhesin of A950006 was identified earlier as an HMW protein (van Schilfgaarde, submitted). Selection of this library by three subsequent paracytosis enrichment cycles led to identification of two clones containing plasmids with different restriction endonuclease patterns and showing a 10-fold-higher penetration rate when tested separately than E. coli DH5 (data not shown). However, these two clones and five clones from another selection round were positive for hmwA when tested in a PCR (van Schilfgaarde, submitted), indicating that this adhesin facilitates penetration of E. coli.

View larger version (16K): [in this window] [in a new window]   FIG. 3.   The numbers of bacteria cultured from of the lysed epithelial cell layers after paracytosis, 4 h postinoculation with E. coli DH5 expressing pGJB103 and E. coli clones 9.8 and 9.1. P values were determined with the paired t test.

View larger version (89K): [in this window] [in a new window]   FIG. 4.   Light micrographs of cross sections of NCI-H292 cell layers on filter inserts after incubation with E. coli clones obtained after selection for paracytosis. (A) Cell layer 4 h postinoculation with an hmw-positive adherent clone from the chromosomal library of H. influenzae strain A950006. (B) Cell layer 4 h postinoculation with E. coli clone 9.1, obtained from the chromosomal library of H. influenzae strain A960053.

Restriction endonuclease mapping of clones 9.1 and 9.8. The 5' and 3' ends of the fragments in clones 9.1 and 9.8 were sequenced, and the sequences were compared to the GenBank database using the Blast algorithm (1). The sequences were found to be homologous to sequences of H. influenzae Rd, whose whole genome has been sequenced (9) (Fig. 5). The DNA sequences obtained with primer PGJB-P1 on clone 9.1 and the sequence obtained with primer PGJB-P2 on clone 9.8 were identical and matched sequences of ORF HI0642 of the Rd genome. The sequence obtained with primer PGJB-P2 on clone 9.1 was homologous to 100 bp upstream of HI0636 and contained a CCAA repeat similar to that found in the Rd genome. The sequence obtained with primer PGJB-P1 on clone 9.8 was homologous to the sequence of HI0637. The lengths of the cloned fragments corresponded to the sizes predicted by the Rd genome sequence, suggesting a comparable genetic organization. This implied that clone 9.1 contained six complete ORFs (Table 2), HI0636 to HI0641, and that clone 9.8 contained four complete ORFs (HI0638 to HI0641) in the reverse orientation compared to that of clone 9.1 ORFs (Fig. 5).

View larger version (20K): [in this window] [in a new window]   FIG. 5.   Schematic representation of clones 9.1 and 9.8 and the subclones obtained from clone 9.1.

Cloning and sequencing the upstream sequence of ORF HI0636. To identify the sequences upstream of ORF HI0636, a PCR using primer prMS0637 together with prRDHI0635P1 or prRDHI0635P2 was designed. The primers were based on the sequence of HI0635 of H. influenzae strain Rd. PCRs using these primers with chromosomal DNA of strain A960053 gave no product. In contrast, when the PCR was performed with chromosomal DNA of H. influenzae strain Rd and two other nontypeable H. influenzae strains, products of the expected lengths were amplified (data not shown). This indicated that the genetic organization of strain A960053 upstream of ORF HI0636 differed from those of the other H. influenzae strains. To obtain upstream sequences, we performed an inside-out PCR with primers pr1-0636 and pr2-0636 and completely digested and religated chromosomal DNA of strain A960053. Unexpectedly, this PCR gave a 2-kb product irrespective of the enzyme we used to cut the DNA. A control experiment showed that primer pr2-0636BAM alone also produced this product. The 2-kb product was purified, ligated into the TA vector, and transformed into E. coli DH5. Sequencing showed that the upstream region of clone 9.1 was highly homologous to hhuA, encoding the H. influenzae hemoglobin-haptoglobin binding protein, which is not present in Rd (9).

Characteristics of E. coli subclones containing ORFs HI0636, HI0637, and HI0638 from H. influenzae strain A960053. Since ORFs HI0637, HI0639, HI0640, and HI0641 had assigned functions which identified them as "housekeeping genes" (Table 2), they were likely not to be involved in paracytosis. To determine whether ORFs HI0636 and HI0638, which encoded proteins of unknown functions, were involved in penetration into the cell layer, several subclones were constructed (Fig. 5). Since ORFs HI0636 and HI0637 have the same orientation and since the distance between these ORFs was only 22 bp, they possibly form one operon and were therefore cloned together in one subclone. Clone 9.1B contained a 3.5-kb HincII-XbaI fragment with the complete ORFs HI0637 and HI0636 ligated in pUC19. Clone 9.1H contained a 1.7-kb PstI-EcoRI fragment with the complete ORF HI0638. Clone 9.1A contained a 1.5-kb PstI-EcoRI fragment with the truncated ORF HI0638. The MIC of gentamicin for the subclones was unchanged compared to that for DH5 containing pUC19 or pGJB103. Alignment of the sequences from the 5' and 3' ends of the cloned fragments of the subclones with the Rd genome sequence confirmed the presence of the expected ORFs. In the paracytosis assay, clones 9.1B and 9.1H passed into the cell layer in higher numbers than the negative control containing only pUC19 (P < 0.01 and P = 0.02, respectively) (Fig. 6), while the paracytosis of clone 9.1A was similar to that of the negative control containing pUC19. These results indicated that not only ORFs HI0638 and HI0636 but also ORF HI0637 may be involved in penetration of the epithelial cell layers.

View larger version (17K): [in this window] [in a new window]   FIG. 6.   The numbers of bacteria cultured from of the lysed epithelial cell layers after paracytosis, 4 h postinoculation with E. coli DH5 expressing pUC19 and E. coli clones 9.1A, 9.1B, and 9.1H. P values were determined with the paired t test.

Paracytosis of H. influenzae strain A960053 HI0636 and HI0638 knockout mutants. To determine the contribution of ORFs HI0636 and HI0638 to paracytosis in a H. influenzae background, HI0636 and HI0638 knockout mutants of H. influenzae strain A960053 were constructed. The numbers of bacteria of H. influenzae strain A960053 and its HI0636 knockout mutant penetrating into the cell layer were similar (Fig. 7). In contrast, the HI0638 knockout mutant did not penetrate into the cell layer since a 100-fold-lower number of bacteria were cultured from the cell layer compared to the number for the parent strain (P < 0.01) (Fig. 7). The susceptibility to gentamicin of the HI0638 knockout mutant, as determined by E-test, was similar to that of the parent strain. However, the HI0638 knockout mutant showed a slightly reduced growth rate compared to that of wild-type strain A960053. Therefore, the survival of the HI0638 knockout mutant was compared with the survival of H. influenzae strain CF47A, a slowly growing CF isolate. H. influenzae strain CF47A and the A960053 HI0638 knockout mutant did not grow in the apical medium during the assay, and similar numbers of bacteria (0.2 × 10⁸ to 10⁸ CFU/ml) were cultured from the apical medium after overnight incubation with these strains. However, H. influenzae strain CF47A showed significantly more paracytosis than the A960053 HI0638 knockout mutant (P < 0.01) and penetrated the cell layer in numbers similar to those for H. influenzae A960053 (Fig. 7). This indicates that a reduced growth rate does not necessarily lead to a reduced penetration of H. influenzae into the cell layer. Therefore, the reduced paracytosis rate of the HI0638 knockout mutant may be related to a specific function of the HI0638 gene product in paracytosis.

View larger version (20K): [in this window] [in a new window]   FIG. 7.   The numbers of bacteria cultured from of the apical medium (shaded circles) or from the lysed epithelial cell layers (open circles) after paracytosis, 4 h postinoculation with H. influenzae A960053, its HI0636 and HI0638 knockout mutants, and H. influenzae strain CF47A. P values were determined with the Student t test. NS, no significant difference.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

In this study we describe the cloning of bacterial genes involved in paracytosis of H. influenzae through lung epithelial cell layers. We constructed chromosomal libraries in adherent and nonadherent nontypeable H. influenzae strains in E. coli DH5. Our first attempt, using the library of the adherent H. influenzae strain, resulted in selection for adherent E. coli clones. Earlier, we showed that adherence to the epithelial cells in vitro leads to elevated penetration rates but that 11 H. influenzae isolates tested all penetrated into the cell layers irrespective of their adherence characteristic (37, 38). The fact that we isolated an adhering clone indicates that adherence enhances paracytosis, although it is not a prerequisite. Since adherence to the epithelial cell surface concentrates the bacteria at the cell surface, this is not unexpected.

Screening a chromosomal library of nonadherent H. influenzae strain A960053 in E. coli DH5 resulted in the isolation of two clones with enhanced paracytosis, referred to as clones 9.1 and 9.8. Alignment of DNA sequences obtained from these clones showed that the fragments cloned in clones 9.1 and 9.8 were overlapping and homologous to a region present in H. influenzae Rd (9). Based on the genome sequence of Rd, the lengths of different PCR products, and restriction endonuclease patterns, it was concluded that clone 9.1 contained six complete and two truncated ORFs. In earlier experiments using incubation times of 24 h it was shown that the level of paracytosis of H. influenzae strains is strain dependent within a certain range (37). Although H. influenzae strain Rd showed a significantly lower level of paracytosis than the two clinical strains, the level was comparable to that of H. influenzae strain d1 (data not shown). Transmission electron microscopy analysis of the paracytosis of strain d1 also showed clusters between the cells (38), and we expect similar results for Rd. Therefore, the presence of these genes in the Rd genome was not unexpected.

Of the six complete ORFs two (HI0636 and HI0638) encode small proteins of unknown function. The other ORFs encode ribosomal proteins (HI0640 and HI0641), tryptophanyl-tRNA synthetase (HI0637; trpS), and adenylosuccinate lyase (HI0639; purB). A mutation in purB of Salmonella spp. was shown to be associated with reduced virulence in mice (21). However, this is probably due to an effect of this mutation on the survival of the bacterium and not to a specific function of the gene product in virulence such as interaction with the epithelial cells during infections. Therefore, we focused on ORFs HI0636 and HI0638 for their role in the paracytosis model. Constructing subclones containing HI0636 or HI0638 and sequencing these fragments showed that clone 9.1 indeed contained these ORFs, similar to strain Rd. The predicted amino acid sequences for ORFs HI0636, HI0637, and HI0638 of H. influenzae strain A960053 showed high homology (96, 99, and 98%, respectively) to those for the corresponding ORFs of the sequenced genome of H. influenzae strain Rd. The genome organization of strain A960053 upstream of ORF HI0636 differed from that of strain Rd. The sequencing of the region upstream of HI0636 of A960053 and a database search showed that HI0635 of strain A960053 is homologous to the hhuA gene, coding for hemoglobin-haptoglobin binding protein (20). This gene is absent in strain Rd. Although ORF HI0635 of Rd also seems to encode a hemoglobin binding protein, there is only poor homology (43%) to hhuA.

Since ORF HI0638 was present in clone 9.1 and clone 9.8 and since these clones both showed an increased survival in the cell layers, this gene was the most likely candidate to encode a paracytin. However, higher numbers of clone 9.1 than of clone 9.8 bacteria from within the cell layer were cultured, indicating that additional sequences in clone 9.1 may influence the paracytosis efficiency. Results of paracytosis assays with two E. coli subclones containing either ORF HI0638 or ORFs HI0636 and HI0637 confirmed that these ORFs were indeed involved in increased penetration of the E. coli clones. The subclone containing ORFs HI0636 and HI0637 showed higher numbers of bacteria in the cell layer than the clone containing HI0638. The combined results of these experiments indicated a role in penetration for HI0636 as well as HI0638.

Knockout mutation of HI0636 did not result in lower numbers of bacteria in the cell layers, suggesting that in the E. coli clones HI0637 was important for the interaction with the cell layer. However we consider this very unlikely due to its functional assignment as trpS. H. influenzae Rd contains another ORF (HI0400) with an associated deduced amino acid sequence showing 61% homology to that associated with ORF HI0636. It is possible that HI0400 has the same function as HI0636 and has taken over its function in the HI0636 knockout mutant. Knockout mutation of HI0638 significantly reduced the penetration of the cell layer since a 100-fold-lower number of bacteria were cultured from the cell layer. Because HI0638 and HI0639 are located in one operon, we cannot exclude an indirect effect of the HI0638 knockout mutation on expression of HI0639. However, since HI0639 was not required for paracytosis of E. coli clones, it can be excluded as the paracytin, and this leaves HI0638 as the candidate for a paracytin.

The passage between the epithelial cells, a characteristic for H. influenzae, seems to involve selective disclosure of intercellular junctions (38). Other bacteria that employ paracytosis or interjunctional passage are, e.g., the spirochetes Borrelia burgdorferi and Treponema pallidum (15). The mechanisms underlying this passage are unknown. Various intestinal pathogens perturb the paracellular barrier by production of toxins, such as the zonula occludens toxin (ZOT) and the hemagglutinin/protease of Vibrio cholerae (3, 6, 40), toxin A and toxin B of Clostridium difficile (7), and VacA of Helicobacter pylori (29). The increase in permeability of the epithelial cell layer induced by these toxins mostly coincides with a decrease in transepithelial resistance and with a rearrangement of (F-)actin. However, each of the toxins exerts a different effect on the paracellular barrier. VacA increases paracellular epithelial permeability only to low-molecular-mass (<350- to 440-Da) molecules, without an effect on the junctional proteins that maintain the structure of the intercellular junctions (29). ZOT has been shown to alter the structure of intercellular junctions, apparently by triggering an intracellular cascade which involves protein kinase C (6). As a consequence of ZOT-induced modification of the transepithelial permeability, the intestinal mucosa becomes permeable to water and electrolytes, resulting in diarrhea but not in passage of bacteria through the tight junctions. Toxins A and B of C. difficile enhance paracellular transmigration of different bacterial species through cell layers of an intestinal epithelial cell line, indicating a major change in permeability (7). However these toxins exert a pathologic effect on epithelial cells, which was never observed during passage of H. influenzae between the epithelial cells.

Alternatively, the mechanisms by which H. influenzae bacteria influence the intercellular permeability of epithelial cells in a subtle, reversible way which permits whole bacteria to protrude may resemble the mechanism by which leukocytes penetrate epithelial cell layers. Leukocytes penetrate the cell layers after upregulation of the intercellular adhesion molecule ICAM1 by tumor necrosis factor alpha (TNF-). In addition, TNF- affects the tight-junction region between epithelial cells (27). The increased paracellular permeability of brain endothelial cells induced by TNF- increased penetration of human immunodeficiency virus by the paracellular route (8). After infection with Mycobacterium tuberculosis an increased permeability of the epithelial cell layer was associated with the production of TNF- by the epithelial cells (41). Interaction of H. influenzae (4) or H. influenzae lipopolysaccharide (18) with different lung epithelial cell lines also induced the production of TNF- as well as other inflammatory mediators, such as interleukin 6 (IL-6) and IL-8, by the epithelial cells and the expression of ICAM1. However, the transepithelial permeability was either unaltered, as in our model, or decreased, as in the study using human bronchial epithelial cells (18), and it is unknown whether H. influenzae can interact with the ICAM1 molecules.

Analysis of the predicted amino acid sequences of ORF HI0636 did not show any signal sequences or membrane-spanning region, and the product was predicted to be a cytoplasmic protein. The predicted amino acid sequence for ORF HI0638 contains the N-terminal sequence ALAGVCQS (positions 10 to 17), which resembles the lipoprotein consensus sequence L(AV)-L-A(S)-G(A)-C-X-(S,D) of the outer membrane auxiliary protein family (5). The serine residue at the second amino acid position following the cysteine residue indicates that ORF HI0638 encodes a lipoprotein that is targeted to the outer membrane (5). As a surface-exposed protein, the product of HI0638 may have a direct effect on a signaling pathway involved in intercellular junction regulation or may modulate a cytokine response of the epithelial cells and affect the intercellular permeability indirectly. The predicted amino acid sequence for ORF HI0638 is homologous to YcfC of E. coli, a 22.9-kDa membrane-associated protein of unknown function (12). When the ORF HI0636 product has an auxiliary function related to that of the HI0638 product in H. influenzae, it may also act on YcfC in E. coli, explaining the effect on paracytosis in the subclone containing only HI0636. Expression of HI0638 by E. coli also led to an increase in paracytosis, probably because of overexpression of HI0638 from the plasmid compared to the normal level of expression of YcfC. Additional research may elucidate the exact functions of these proteins in the paracytosis process.

In conclusion, we identified the paracytin gene of H. influenzae involved in paracytosis through lung epithelial cells layers. This allows us to characterize the paracytin and how it regulates the intercellular permeability. Knowledge of the modification of the function of the intercellular junctions by the H. influenzae paracytin may enlarge our understanding of the normal physiological regulation of intercellular junctions and how normal intercellular permeability is regulated.