INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

Haemophilus influenzae infections, including otitis media, sinusitis, pneumonia, and persistent infections in chronic bronchitis patients, are preceded by airway colonization, which is a process facilitated by fimbria (8, 24, 32, 39, 54). Four families of fimbriae have been distinguished morphologically (5). Only the so-called LKP fimbriae are associated with the adherence of H. influenzae to human cells (5, 28, 42, 47), and LKP-positive bacteria agglutinate human erythrocytes that express the AnWj antigen (29, 30, 40). The binding of fimbriate H. influenzae to oropharyngeal epithelial cells is dependent on sialic acid-containing lactosylceramide epithelial cell receptors with monosialoganglioside-3 (GM3) as the minimal structure (41). LKP fimbriae also mediate binding to a variety of cells encountered by H. influenzae during carriage and disease. These include pseudostratified columnar and ciliated columnar epithelial cells of the (naso)pharynx, adenoid, and bronchi (36). Finally, fimbriae facilitate binding to extracellular matrix proteins (52). Mucus interferes with fimbria-mediated adherence (7), although fimbriate bacteria do not bind more readily to mucus than do nonfimbriate bacteria (7, 35, 55). Therefore, fimbriae seem to be multifunctional in the pathogenesis of H. influenzae infections. It is not known, however, whether fimbriae are involved in the pathogenesis of persistent infections and, if so, whether its function is at the onset or during persistence of infection.

The fimbria gene cluster of H. influenzae type b containing the genes hifA to hifE has previously been analyzed and sequenced completely (49). hifA encodes the fimbria subunit that mediates binding to the ganglioside receptor (50). Comparing the hifA genes of various H. influenzae type b and nontypeable H. influenzae strains revealed the presence of conserved and variable regions (6, 10, 11, 22, 23, 49, 50). hifB encodes a protein of the chaperone family (17, 49), hifC encodes a putative assembly platform protein (49, 53), and hifD and hifE encode two minor subunits that are highly homologous to hifA (49, 50). The presence of hifD and hifE affects fimbria expression, but their exact role(s) is not known (49). The transcription of all fimbria genes is determined by reversible changes in the number of TA repeats in the bidirectional promoter region localized between hifA and hifB in the fimbria gene cluster (48). The gene cluster also contains multiple inverted and direct repeats (22, 48). These structures may cause hairpins that are involved in the stabilization of mRNA, and the inverted repeats at the 5' and 3' ends of the gene cluster may make the cluster a transposon-like element (49).

Here, we report the compositions of various fimbria gene clusters of nonencapsulated H. influenzae strains isolated from chronic bronchitis patients, acute otitis media patients, and healthy carriers and compare them with the composition of H. influenzae type b. We show that all of the fimbria gene clusters of nonencapsulated H. influenzae strains contain hifA to hifE. The gene clusters are located at the same chromosomal position between purE and pepN as that for the LKP-4 fimbria gene cluster of H. influenzae type b. The promoter region of the gene cluster was found to contain various numbers of TA repeats when fimbria-positive and -negative strains were compared. In contrast to H. influenzae type b, nonencapsulated strains have sequences that show variation inserted between purE and the fimbria genes but only a small number of repetitive elements between hifB and the promoter. The composition of these elements did not change during persistent infections in chronic bronchitis patients. The significance of the flanking regions for the excision and insertion of the gene cluster in the chromosome is discussed.

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

Bacterial strains and plasmids. Nonencapsulated H. influenzae strains (n = 58) were isolated as described before from sputum samples of chronic bronchitis patients who had been infected for up to 2 years (12). Persistent strains were isolated at least twice from a patient over a period of longer than 6 months, although the periods in which sputum cultures were positive for that strain were often intermingled with periods in which sputum cultures were negative. Acute (nonpersistent) strains were isolated only once from sequentially taken sputum samples of a patient collected over a period of more than 6 months and at least not during the 3 months before or after that isolation date. Persistent H. influenzae strains were identical in their random amplified polymorphic DNA (RAPD) patterns obtained by PCR with ERIC primers but included outer membrane protein P2 and P5 variants (45). Nonencapsulated H. influenzae strains were also isolated from middle ear fluid samples of patients with otitis media (n = 13) and from throat swabs of healthy adult carriers (n = 12). Fimbria gene cluster-positive H. influenzae type b strains 770235 (identical to AM30), A920006, Eagan, A920019, A840049, A860268 and the fimbria gene-negative H. influenzae type b strain 760705 from patients with invasive diseases and strain Rd were included for references (42, 44, 48). All strains were identified as H. influenzae by their growth dependence on NAD and hemin. Strains were kept as frozen stocks at 70°C in broth containing 15% glycerol, usually after less than five passages from the primary culture plate.

RAPD analysis. To determine the genetic relationship between fimbria gene cluster-positive H. influenzae strains, bacterial strains were analyzed by random PCRs with either primer ERIC2 (45, 51) or HPO1 or HPO2 (1) as previously described (45). Chromosomal DNA was isolated by the protocol of Boom et al. (4). The PCR mixture (100 µl) consisted of 1× PCR buffer from a Perkin-Elmer kit, 50 pmol of primer, 0.2 mM (each) deoxyribonucleotide triphosphates, 50 pmol of chromosomal DNA, and 0.5 U of Taq DNA polymerase (AmpliTaq; Perkin-Elmer). PCR consisted of the following steps: denaturation (5 min, 94°C) and 35 cycles of denaturation (1 min, 94°C), annealing (1 min, 25°C), and extension (4 min, 74°C). Finally, an additional DNA extension step (10 min at 74°C) was included. PCR was performed with a Biometra II apparatus (Perkin-Elmer). After agarose gel electrophoresis of PCR products, the banding patterns on gels obtained for each primer were analyzed with Gel Compar (Applied Maths, Kortrijk, Belgium) as previously described (46). CLUSTAL analysis was performed with the Dice coefficient and UPGMA (unweighted pair group method using arithmetical averages) clustering.

Enrichment for bacteria expressing fimbriae. None of the H. influenzae strains from patients agglutinated human erythrocytes, as determined before, indicating that they did not express fimbriae (28-30). Enrichment for bacteria adhering through fimbriae was performed by a hemagglutination enrichment procedure previously described with AnWj-positive human erythrocytes (29, 30, 40, 41). The specificity of hemagglutination for fimbriae was assessed by showing the absence of hemagglutination of cord blood erythrocytes (since they are AnWj negative) and inhibition of hemagglutination by monosialoganglioside-1 (GM1) (40, 41). The functional activity of fimbriae was determined semiquantitatively by the hemagglutination titer. A bacterial suspension of 10¹⁰ CFU per ml was diluted stepwise twofold in phosphate-buffered saline and subsequently incubated for 5 to 15 min at room temperature on World Health Organization plates as previously described (48). Hemagglutination was read over a light, and the maximal dilution of the suspension that caused hemagglutination was taken as the titer (50). Fimbria expression was also analyzed by electron microscopy after negative staining, as described before (42).

General DNA techniques. Unless stated otherwise, DNA manipulations were carried out by standard procedures (34). All enzymes (Boehringer Mannheim and Pharmacia) were used according to the instructions provided by the manufacturers. Oligonucleotides (Table 1) were available from previous studies or were ordered from Applied Biosystems and used without further purification. PCR was performed with DNA isolated by the phenol-chloroform method (21) or with bacterial lysates obtained by heating an H. influenzae colony suspended in 1 ml of H₂O for 5 min at 100°C and then cooling on ice. PCR products were purified with Gene Clean II (Bio 101, La Jolla, Calif.) or Qiaex (Qiagen Gmbh, Hilden, Germany), followed by phenol-chloroform extraction, before they were used for sequence analysis or DNA hybridization.

Identification of fimbria gene clusters. A thick suspension (2 µl) of H. influenzae cells grown on chocolate agar plates was spotted onto nitrocellulose filters and baked for colony hybridization, essentially as previously described (34). The fimbria gene cluster excised from plasmid pMH140 (47) was used as a probe after being labelled with digoxigenin (Boehringer, Mannheim, Germany). The presence of fimbria gene clusters was also assessed by PCR with primers derived from the conserved flanking regions of the purE and pepN genes (Table 1) or two PCRs with a primer set consisting of purE and hifC primers and a set consisting of hifC and pepN primers. PCR was performed with thermostable Taq DNA polymerase and Pwo DNA polymerase (Boehringer) according to the instructions of the manufacturer. The lengths of reaction products were determined by agarose gel electrophoresis. PCR products derived from fimbriate H. influenzae type b strain 770235 and fimbria-negative strain 760705 were included as controls.

Identification of the individual genes from hifA to hifE. The presence of hifA, hifB, hifD, and hifE was determined by Southern blotting with purified DNAs containing the single genes as probes. Chromosomal DNA was digested mostly by restriction enzymes EcoRI, XhoI, and BglII, and digested DNA was separated by agarose gel electrophoresis. The hifA, hifB, hifD, and hifE gene probes were prepared by excising the respective genes from plasmids pMH042, pMH212, pMH046, and pMH156 (49, 50). The probes used for hybridizations were labelled with digoxigenin. The hifC gene could not be identified by hybridization with the hifC gene or a fragment of the hifC gene purified from pMH140 after EcoRI and EcoRV treatment, since high background levels were obtained. Therefore, the presence of the hifC gene was determined by PCR with a primer set complementary to sequences in hifC. Strains that were negative in this PCR were further analyzed by PCR with one primer set consisting of a primer specific for the 5' end of the hifC gene and a primer specific for purE and another set consisting of a primer complementary to the same sequence inside the hifC gene and a primer specific for pepN (Table 1). H. influenzae type b strains 770235 and 760705 were used as positive and negative controls, respectively. The conditions used for hybridizations and PCRs are summarized in Table 1.

Analysis of regions of the fimbria gene cluster containing repeat sequences. The presence of long stretches of intergenic DNA similar to those in the fimbria gene cluster of H. influenzae 770235 was determined by PCR with primers flanking the repeat regions of H. influenzae type b strain 770235 (Fig. 1; Table 1). PCR products were characterized by agarose gel electrophoresis and partial DNA sequence analysis.

View larger version (10K): [in this window] [in a new window]   FIG. 1.   Schematic representation of the fimbria gene cluster of H. influenzae, in which the primers for PCRs have been indicated in the 5'-to-3' direction. Relevant restriction enzyme sites are indicated above the schematic.

Sequence analysis of hifA. Fragments of the fimbria gene cluster were obtained by PCR with appropriate primers. These fragments were used directly as templates for automated double-stranded DNA sequencing (model 370A; Applied Biosystems) with a Taq dye-terminator cycle sequencing kit (Applied Biosystems), fluorescent-dye-labelled dideoxynucleotides, and the same primers. Fragments were sequenced in both directions. Sequences were analyzed by using computer programs included in the program package PC/GENE (IntelliGenetics, Inc.). DNA sequences were aligned with the program CLUSTAL by the methods of Higgins and Sharp (16).

Nucleotide sequence accession numbers. The hifA nucleotide sequences of strains 602, 4564, and 6173 have been deposited in the EMBL, GenBank, and DDBJ nucleotide sequence databases under accession no. U91944, U91945, and U91946, respectively.

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

Occurrence of fimbria gene clusters in nonencapsulated H. influenzae strains. Persistent strains (n = 27), acute strains (n = 8), and strains not clearly identifiable as acute or persistent (n = 23) were isolated from infected chronic bronchitis patients. Furthermore, strains from patients with otitis media (n = 13) and healthy carriers (n = 12) were obtained. All these strains were screened for the presence of fimbria genes by colony blot hybridization with the gene cluster of H. influenzae type b strain 770235 as the probe. The results of this screening are summarized in Table 2. The proportion of strains positive by hybridization varied considerably for persistent and acute strains from chronic bronchitis and otitis media patients and healthy carriers, although the differences were not significant (x² [Yates' correction], 4 df, P = 0.15). The percentage of positive strains from acute chronic bronchitis patients was the highest. The lack of significance may be due to the small number of fimbria-positive strains in the total sample of bacterial isolates.

View larger version (51K): [in this window] [in a new window]   FIG. 2.   Hybridization of DNAs of 12 nonencapsulated H. influenzae strains from chronic bronchitis patients (lanes 1 through 12) with the complete gene cluster of strain 770235 from plasmid pMH140 as the probe. The controls were DNA of fimbriate H. influenzae type b strain 770235 (origin of the probe) (lane 14) and DNA of fimbria-negative strain 760705 (lane 13). Chromosomal DNAs of bacteria were cut with restriction enzymes EcoRI, XhoI, BglII, and AccI before electrophoresis and Southern hybridization. ME, meningitis isolate; CB, chronic bronchitis isolate; P, persistent isolate; A, acute isolate; ?, acute or persistent isolate.

hifA to hifE in nonencapsulated H. influenzae strains. The presence of each gene from hifA to hifE was determined for the 15 strains positive for the fimbria gene cluster. DNA hybridization with the hifA probe revealed positive reactions with DNAs of all 15 strains and the reference H. influenzae type b strain 770235 and no reaction with the negative control strains 760705 and Rd. The PCR products obtained with hifA-specific primers were about 800 bp long, which is similar to those of reference strain 770235 (Fig. 3). These products were longer than the hifA gene (650 bp), since primer fgHifA1 annealed outside the hifA gene (Fig. 1; Table 1) (49). Minor variations in the lengths of the products obtained for other strains were observed, possibly as a consequence of DNA sequence variation (see below). hifB, hifD, and hifE were also identified by DNA hybridization, showing genetic diversity similar to that observed with the total gene cluster as the probe. The number of bands per sample that hybridized with the hifB probe was lower than the number of bands seen after hybridization with the total gene cluster, as can be expected when fimbria genes other than hifB are located on separate fragments. The reactions were specific for the tested hif genes, since the hifA, hifB, hifD, and hifE probes did not cross-react with other hif probes by hybridization. The results obtained for hifB (Fig. 4) are an example. The hifC probe, however, showed background reactivities with other fimbria genes. The problem of cross-hybridization was circumvented by using a PCR with two primers that involve hifC sequences (fgHifC1 and fgHifC2) (Table 1). For 8 of 15 strains, a product of the expected length was obtained. For the other seven strains, a product was obtained by PCR with one primer derived from a hifC sequence (primer fgHifC1 or complementary primer fgHifC1Comp1) and another derived from sequences upstream or downstream of this gene (purE or pepN [Fig. 1; Table 1]). Fimbria-negative strain 760705 was negative in these reactions. Nonencapsulated H. influenzae strains were either positive for all the fimbria genes, hifA through hifE, or negative for all of these genes.

View larger version (79K): [in this window] [in a new window]   FIG. 3.   PCR for hifA of H. influenzae type b strains from meningitis patients (ME) (lanes 1 through 4), two nonencapsulated H. influenzae strains from otitis media patients (OM) (lanes 5 and 6), strains from chronic bronchitis patients (CB) (lanes 7 through 14), and positive control strain 770235 from a meningitis patient (lane 1). Lane C, negative control strain 760705; lane M, 100-bp ladder. The hifA primers were fgHifA1 and fgHifA2. P, persistent isolate; A, acute isolate; ?, acute or persistent isolate.

View larger version (53K): [in this window] [in a new window]   FIG. 4.   Hybridization of 10 representatives of fimbria gene cluster-positive nonencapsulated H. influenzae strains (lanes 1 through 10) with the hifB gene of strain 770235 from plasmid pMH212 as the probe. Chromosomal DNAs of bacteria were cut with restriction enzymes EcoRI, XhoI, and BglII before electrophoresis and Southern hybridization. The controls were fimbria-negative strain 760705 (lane 11) and H. influenzae type b strain 770235 (origin of the probe) (lane 12). Abbreviations are identified in the legend to Fig. 3.

Sequence analysis of hifA. Since HifA of H. influenzae type b had previously been shown to be an adhesin for the ganglioside GM3 receptor, we analyzed the hifA gene in detail after sequencing the PCR products obtained with chromosomal DNA as the template. The DNA and derived protein sequences of three strains were compared. Open reading frames coding for proteins of 208, 214, and 216 amino acids were obtained for the three strains. Such lengths are similar to those of the hifA genes of H. influenzae type b strains (10, 22, 49) and the published sequence of hifA from a nonencapsulated strain (6). The lengths of hifA genes varied by at most 24 bases (4%). An alignment of the derived protein sequences and the sequences published before (Fig. 5) revealed the presence of conserved regions, interchanged with variable regions, in agreement with earlier reports (6, 10, 49). According to the analysis of Hopp and Woods (18), the protein regions encoded by variable regions appeared to be hydrophilic and to contain potentially antigenic sites.

View larger version (55K): [in this window] [in a new window]   FIG. 5.   Alignment of the amino acid (AA) sequences derived from DNA sequences of hifA genes from strains 602, 4564, and 6173 (this study) and those from strains 770235, MinnA, and (M)37 (6, 10, 22, 47). Amino acids are indicated by one-letter codes. Amino acids that are identical in all of the proteins listed are indicated by asterisks. Similar amino acids are indicated by dots. Dashes indicate gaps. Three hydrophilic domains are also shown.

Expression of fimbriae. None of the 15 clinical strains that hybridized with the fimbria gene cluster agglutinated human erythrocytes significantly. Of these strains, five were randomly chosen for hemagglutination enrichment in order to obtain fimbria-expressing bacteria. The nonhemagglutinating variant of H. influenzae 770235 was included as a positive control for enrichment. The results are summarized in Table 3. After three enrichment cycles, the positive control strain agglutinated AnWj-positive erythrocytes. After five cycles, strain A930105 from a healthy carrier and strain 6 from a chronic bronchitis patient were positive for hemagglutination. After 9 to 12 enrichment cycles, three strains from chronic bronchitis patients were still nonhemagglutinating, as was control strain 760705, which lacks fimbria genes. Since the hemagglutinating strains did not agglutinate AnWj-negative erythrocytes and hemagglutination was inhibited by GM1, we concluded that LKP-4 fimbriae were responsible for hemagglutination (41). Subsequently, the levels of fimbria expression of nonencapsulated H. influenzae strains 6 and A930105 and H. influenzae type b strain 770235 were semiquantified by determining hemagglutination titers. A 1:1,024 diluted bacterial suspension of 10¹⁰ CFU of strain 6 per ml was the highest dilution that caused hemagglutination. For strains A930105 and 770235, 1:64 and 1:512 dilutions, respectively, were the lowest concentrations to cause hemagglutination. Electron microscopy showed that fimbriae were abundantly present on the cell surfaces of strain 770235 after enrichment (positive control) but that <1 and 10 fimbriae per 20 cells were present on the cell surfaces of strain A930105 and strain 6 (Fig. 6), respectively, indicating that these nonencapsulated bacterial strains express fewer fimbriae than does H. influenzae type b.

View larger version (75K): [in this window] [in a new window]   FIG. 6.   Electron microscopy of fimbriae of nonencapsulated H. influenzae strain 6 from a chronic bronchitis patient (A), strain A930105 from the throat of a healthy individual (B), and H. influenzae type b strain 770235 (C). Bar (all panels), 0.5 µm.

TA repeats in the promoter and fimbria expression. By using primers HifARep2 and fgBrep1, PCR products were obtained for five patient strains not expressing fimbriae and two hemagglutinating variants of these strains. Sequencing of the products starting from the primer fgBrep1 sequence revealed that the promoter region, containing two overlapping 10 and 35 sequences, was similar to that of H. influenzae type b strain 770235 (48), and contained TA repeats. The number of TA repeats has previousy been shown to be important for transcription (48). The numbers of TA repeats in the strains examined are summarized in Table 3. Four fimbria-negative strains had nine TA repeats, and one (strain 6173) had surprisingly only four repeats. In H. influenzae type b, nine TA repeats reduces promoter strength and thereby fimbria expression (48). Hemagglutinating variants of strains 6 and A930105 had 10 TA repeats, the optimal number for fimbria expression (48).

Long direct repeats between hifA and hifB are not required for fimbria expression. In H. influenzae type b, the region between the promoter and hifB contains 10 long direct repeats (49). By using primers HifARep2 and MH006, products that varied considerably in length were obtained (Fig. 7). Of the fimbria gene cluster-positive strains from chronic bronchitis patients, six gave PCR products of 550 bp, two gave PCR products of 650 bp, two gave PCR products of 700 bp, and one gave a PCR product of 750 bp. The lengths of these products were much shorter than those for products from H. influenzae type b strains 770235 (49) and A920006 (1,011 bp) but were similar to those for products from H. influenzae type b strains Eagan (600 bp), A920019 (650 bp), and A840049 (680 bp). The PCR products of two strains from otitis media patients were 850 and 1,250 bp in length. Short PCR products were obtained from strains A930105 (700 bp) (data not shown) and 6 (650 bp). Starting from the promoter in the direction of hifB, the DNA sequence was conserved for the four nonencapsulated strains analyzed and for H. influenzae type b strain 770235 until the first 6 bp of the long repeats (70 bp). Downstream from that point, the DNA sequences of the four nonencapsulated strains and H. influenzae type b strain 770235 showed only limited homologies, except for the 30 bp in front of hifB. The composition of the region between the promoter and hifB is schematically represented in Fig. 8. Since we found diversity in the promoter-hifB region, we analyzed whether this region was unstable during persistent infections in chronic bronchitis patients. By PCR with primers HifARep2 and MH006, the length of the region between the promoter and hifB was determined for 21 H. influenzae strains, including 4 antigenic variants isolated sequentially from four chronic bronchitis patients during infections that persisted after 4, 11, 13, and 30 months. No differences in the lengths of PCR products were observed for any of the sequential strains.

View larger version (93K): [in this window] [in a new window]   FIG. 7.   Electrophoresis of PCR products of the intergenic regions between hifB and the promoter of the gene cluster obtained with primers MH006 and HifARep2. Lane M, 100-bp ladder; lanes 1 and 2, nonencapsulated H. influenzae from otitis media patients; lanes 3 through 8, H. influenzae type b strains from meningitis patients; lanes 9 through 20, nonencapsulated H. influenzae strains from chronic bronchitis patients. Abbreviations are identified in the legend to Fig. 3. Dash, disease not known.

View larger version (45K): [in this window] [in a new window]   FIG. 8.   Schematic representation of the fimbria gene clusters of nonencapsulated H. influenzae strains. H. influenzae type b strain 770235 is included for comparison. Intergenic sequences are drawn to scale, except for hifB-hifE regions, which are indicated as inserts, since they are identical in size. Homologous DNAs are marked identically. For the sizes of genes, see Fig. 1 and reference 49. Abbreviations are identified in the legend to Fig. 3, with the sources of strains indicated parenthetically.

Composition of the regions flanking the gene cluster. Since the PCR results with primers specific for purE and pepN showed that the fimbria gene clusters of all the strains tested appeared to be localized between purE and pepN in the H. influenzae chromosome, the 5' region of the gene cluster of nonencapsulated H. influenzae was analyzed by PCR with primers based on sequences of purE and hifA and the 3' region was analyzed by PCR with primers derived from hifC and pepN (Table 1). The results are presented schematically in Fig. 8. At the downstream end of the gene cluster (hifC to pepN), the PCR product of each of the nonencapsulated H. influenzae strains tested (seven from patients with chronic bronchitis and one from a patient with otitis media) had a length similar to that of the corresponding part (4,659 bp) of the fimbria gene cluster of H. influenzae 770235. Sequence analysis, starting from pepN, revealed that each of these nonencapsulated H. influenzae strains had a pur box in front of pepN; this box is normally found in front of purE. Such pur box duplication has also previously been found for H. influenzae type b strain 770235 (49). Each strain, except for one (248), had a duplication of 37 nucleotides at the 5' end of the pur box as well. Upstream of these pur sequences, 144-bp spaced short inverted repeats (GTAGGGTGGGCGTAAGCCCAC) were found in H. influenzae strain A930105. Similar repeats, but adjacent to each other, have previously been observed in the fimbria gene cluster of H. influenzae type b (49). Strains 6, 190, 602, 4564, and 4949 had one repeat, and strains 37 and 248 lacked both. A gene search of the H. influenzae Rd genome (9) revealed that this repeat occurs with 95% homology in four places of the genome outside the fimbria operon.

Analysis of the region between purE and pepN of strains negative for fimbria genes. As described above, the chromosomes of the H. influenzae strains examined contained either all of the fimbria genes from hifA to hifE or none of these fimbria genes. To determine whether the sequences found between purE and the 5' end of the fimbria gene cluster and between pepN and the 3' end of the cluster of a nonencapsulated H. influenzae strain were part of the gene cluster, the length and composition of DNA between purE and pepN were determined for fimbria-negative strains. By PCR with primers specific for purE and pepN, products of variable size longer than those in strains 760705 (H. influenzae type b) and Rd were obtained for 13 of 15 strains from chronic bronchitis patients, 5 of 5 strains from otitis media patients, and 3 of 5 strains from carriers (all strains lacked the fimbria gene cluster). The DNAs of PCR products obtained from seven fimbria-negative nonencapsulated strains were sequenced, and the sequences were compared with those of the nonfimbriate strains Rd and 760705. With primers specific for the flanking genes purE and pepN, PCR products were obtained for all strains. The compositions of some of the products are schematically represented in Fig. 8. The length of the insert varied from 256 to 1,098 bp. PCR products of 256 bp were found for four of nine strains, including strain Rd and H. influenzae type b strain 760705. This short product comprised the pur box and noncoding DNA. This short stretch of DNA adjacent to pepN is indicated in Fig. 8. At the purE side, the sequence was highly conserved for all strains until after the pur box. pur box duplication, which was common for fimbria gene cluster-positive strains, was also observed for two of the seven fimbria-negative nonencapsulated strains tested. The sequence between the two pur boxes consisted of an inverted repeat (strain 9814c) or a direct repeat (strain 724) with a sequence similar to that of the short repeats described above for the fimbria-positive strains. At the pepN site, sequence conservation started just before pepN. Noncoding DNA was found between pepN and the pur box in five of nine strains. These inserts had sequences similar to those of the inserts between purE and hifA in strains with the fimbria gene cluster. In strain 27, an extra stretch of DNA flanked by direct repeated DNA was inserted in the middle. None of these sequences aligned with the genome of strain Rd significantly.

Fimbria cluster diversity is independent of genomic diversity. To determine whether fimbria gene clusters occur in a subset of genetically clustered H. influenzae strains, H. influenzae strains were analyzed by RAPD based on random PCRs. Separate amplifications with primers ERIC2, HPO1, and HPO2 resulted in three banding patterns that were sufficient for clonal analysis. The results of the three PCR products are combined in Fig. 9 for all 18 of the strains included in Fig. 8 (fimbriate and nonfimbriate). CLUSTAL analysis of the banding patterns revealed the dendrogram depicted in Fig. 9. Fimbria gene cluster-positive strains and strains with distinct compositions of the region between purE and pepN were not clustered.

View larger version (24K): [in this window] [in a new window]   FIG. 9.   Dendrogram of the strains represented in Fig. 8 from a combination of the RAPD results obtained by using primers ERIC2, HPO1, and HPO2. Strain designations are indicated on the right. +, fimbria-positive strains; , fimbria-negative strains with inserts between purE and hifA. Abbreviations are identified in the legend to Fig. 3.

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

Presence of fimbria genes. LKP-4 fimbriae that recognize sialoganglioside receptors occur in H. influenzae type b and nonencapsulated H. influenzae strains (2, 5, 28-30, 40-42, 47) (this study). Here, we have shown that fimbria gene cluster-positive strains contain all known fimbria genes (hifA to hifE), since the DNAs of these strains hybridized with probes for the individual genes and gene-specific PCR products were obtained. PCRs of DNAs from positive strains with primers based on sequences of the genes flanking the gene cluster (purE and pepN) revealed products that were large enough to contain complete gene clusters in this region of the H. influenzae chromosome (Fig. 8). Obviously, truncation of any of the fimbria genes of the gene cluster is not very likely to occur. In addition, this result shows that a fimbria gene cluster is localized between purE and pepN in the chromosome of each of the strains positive for fimbria genes. We cannot exclude gene duplication of the fimbria gene cluster at another site in the chromosome, as was observed for H. influenzae biogroup aegyptius (33). However, this is unlikely, since the sums of the lengths of the fragments hybridizing with the fimbria gene cluster were rather similar for all strains and the hybridization patterns were rather simple (Fig. 2).

Fimbriae as virulence factors for H. influenzae. Independent of the clinical source, 18% of the nonencapsulated H. influenzae strains examined had fimbria genes (Table 2). Bakaletz et al. (2) found that almost all of their clinical isolates were fimbriate. Since the fimbriae on isolates varied in morphology, they may represent different types, as previously described (5). Combining their results with ours, we conclude that LKP-4 fimbriae represent only one class of fimbriae on nonencapsulated H. influenzae strains and that LKP-4 fimbriae are not disease-specific virulence factors. They may contribute to the establishment of bacteria at the mucosal site of the nasopharynx, but this may also be mediated by other adhesins, such as other fimbriae (5), high-molecular-weight proteins (3), and the Hap protein (38), which have previously been implicated in the adherence of nonencapsulated H. influenzae strains to epithelial cells.

Expression of fimbriae and promoter composition. None of the clinical isolates tested expressed fimbriae, as determined by hemagglutination. Since these strains were usually analyzed within five subcultures from the primary plates of patient materials, it is suggested that fimbriae are not expressed at the time the cultures were taken. Since Farley et al. found a lower proportion of fimbriate bacteria after mixed infection of nasopharyngeal organ cultures with fimbriate and nonfimbriate bacteria (8), H. influenzae may lose its ability to express fimbriae after it enters tissues. One of the characteristic features of infections in chronic bronchitis patients is the penetration of bacteria into tissues and establishment in subepithelial layers (14). In tissue, the expression of fimbriae likely results in the binding of bacteria to cells, thereby contributing to the clearance of bacteria (8, 54).

Noncoding intergenic regions inside the gene cluster. The length of the region between the promoter and hifB varied considerably among nonencapsulated H. influenzae strains (Fig. 8). The diversity in length of these DNA stretches was accompanied by strong sequence divergence (data not shown), most likely indicating that this part of the gene cluster has been exchanged between strains. In H. influenzae type b strain 770235, this part of the fimbria gene cluster is composed of multiple copies of repetitive extragenic palindromic sequences organized head to tail. In H. influenzae type b strain 770235, these structures are fimbria gene cluster specific and have previously been suggested to be important for the stabilization of mRNA by forming stem-loop structures (15, 49). In nonencapsulated strains, we found that the sequences of these regions diverged in most cases abruptly where the palindromes started in H. influenzae type b strain 770235, although they contained fragments of the corresponding sequences of H. influenzae type b strain 770235. The sequences seem to be specific for the fimbria gene cluster since no continuous homology was found with any other region of the H. influenzae genome (9) or any sequence in the GenBank, EMBL, DDBJ, and PDB databases. The extragenic sequences in nonencapsulated strains may influence the expression levels of fimbriae, since the expression levels in nonencapsulated H. influenzae strains appeared to be lower than that in H. influenzae type b (Fig. 6; Table 3).

Sequences flanking the gene cluster. Several regions flanking the gene cluster are reminiscent of recombinational events. (i) The duplication of the pur box and the adjacent 37 nucleotides that are always observed at the 3' end of the fimbria gene cluster is an indication that the fimbria gene cluster was inserted at this place in the chromosome by recombination through the pur box. This pur box duplication was observed for both H. influenzae type b and nonencapsulated strains. Since the G+C content of the fimbria gene cluster (39%) is similar to that of the H. influenzae genome (9, 20), the fimbria gene cluster probably has a Haemophilus origin. (ii) Short inverted repeats were observed adjacent to hifA and at the 5' ends of the fimbria gene clusters of nonencapsulated and H. influenzae type b strains. These repeats may form stem-loop structures, likely creating hairpins, which may be sites for recombination, possibly contributing to the strong diversity observed in hifA genes (Fig. 5). This type of exchange was not observed among persistent strains from chronic bronchitis patients but may have occurred in evolution by horizontal transfer.