Rosenau et al. described the adherence of 17 biotype IV H. influenzae cryptic genospecies strains to human epithelial cells (30). Twelve of the strains expressed peritrichous pilus-like structures on electron microscopy and two additional strains appeared piliated after enrichment with human red blood cells. The presence of pilus-like structures correlated with adherence to HeLa cells and only 3 of the 14 piliated strains agglutinated human red blood cells, suggesting that the pili on these strains differ functionally from the hemagglutinating, HifA- and HifE-containing pili that are present on respiratory H. influenzae (11, 13).

Adherence of respiratory H. influenzae strains to human epithelial cells is mediated through specific adhesins on the surface of the bacteria, of which the following six have been well described: hemagglutinating pili, Hia, HMW1 and -2, Hap, and P5 fimbriae (11, 28, 31). The surface structures on biotype IV H. influenzae isolates described by Rosenau et al. (30) as visualized by electron microscopy appeared to be most like hemagglutinating pili. These pili are composed of three elements: multiple subunits of the protein HifA (or pilin) polymerized into hollow shafts that protrude from the bacterial surface; HifD, which is located at the pilus tip; and HifE, which is also located at the pilus tip and possesses the epithelial cell binding domain (11). HifE mediates binding to shed buccal epithelial cells and to 16HBE14o⁻ bronchial cells, but not to human tracheal fibroblasts, human nasal tumor cells, A549 human alveolar epithelial carcinoma cells, HEp-2 laryngeal carcinoma cells, or HeLa human cervical carcinoma cells (13). Furthermore, HifE-mediated binding to epithelial cells is inhibited by the sialic acid-containing lactosylceramides GM1, GM2, GM3, and GD1a (34). Two additional proteins, HifB, which is the pilus chaperone, and HifC, which is the pilus usher protein, encoded by genes in the pilus gene cluster are required for assembly of functional pili (11).

The objective of this study was to explore similarities between the peritrichous pilus-like structures seen by electron microscopy of NTHI biotype IV cryptic genospecies strains and hemagglutinating HifA- and HifE-containing pili on respiratory H. influenzae isolates. To this end, adherence capabilities of six biotype IV clinical isolates to human red blood cells and to four epithelial cell types, two genital and two respiratory, were characterized. These strains were tested for the presence of hifA (the gene encoding the major structural protein [pilin] of hemagglutinating pili) and hifE (the gene encoding the HifE tip adhesin of hemagglutinating pili) of respiratory strains by PCR and for expression of HifA and HifE by Western blot analysis (11). The derived amino acid sequences of HifE proteins from the biotype IV strains were compared to those from respiratory strains. In addition, biotype IV wild-type strains were compared with those possessing mutations in hifA and hifB for the presence of pilus-like surface structures. Finally, the role of pili in mediating epithelial cell adherence was tested by comparing the adherence of the wild type and of pilus-deficient mutants to respiratory and genital cells.

Bacteria. Table 1 describes the bacterial strains used in this study. NTHI strains AAr176, AAr73, AAr49, AAr91, and Eagan were enriched for pilus-expressing (p⁺) variants as previously described (23). The H. influenzae biotype IV strains 6351, 785, 1595, 4162, 1216, and 597 were kindly donated by Timothy Murphy of the Buffalo Veteran's Administration Hospital, Buffalo, N.Y., and have been shown to possess the variant P6 phenotype that has been associated with the unique genospecies of H. influenzae strains that cause genital and neonatal infections (21, 30). H. influenzae biotype IV strains were cultured on Schaedler medium (Becton-Dickinson, Cockeysville, Md.) supplemented with lysed horse blood and 20 μg of NAD per ml in 5% CO₂ at 37°C overnight (19). All other H. influenzae strains were cultured on chocolate agar (Becton-Dickinson) in 5% CO₂ at 37°C overnight, transferred to Levinthal broth (brain heart infusion broth [Difco Laboratories, Cockeysville, Md.] supplemented with 100 μg of hemin per ml and 20 μg of NAD per ml), and grown in 5% CO₂ at 37°C overnight to stationary phase. Pilus-deficient NTHI strains carrying the kanamycin resistance gene were cultured in medium containing 25 μg of kanamycin (Sigma-Aldrich Co., St. Louis, Mo.) per ml of medium. Long-term storage of bacteria was in sterile skim milk at −70°C.

TABLE 1 H. influenzae strains used in this study

Mutated H. influenzae strains AAr176ΔA and 4162ΔA contain a kanamycin resistance cassette inserted into hifA, which encodes pilin, the major structural pilus protein. To construct these mutants, plasmid pBB16/K, which contains a kanamycin resistance gene inserted into the BglII site of hifA, was linearized with AvrII and used to transform NTHI strains AAr176 and 4162 made competent by the method of Herriott et al. (12, 15). Mutated H. influenzae strains AAr176ΔB and 1595ΔB contain a kanamycin resistance cassette inserted into hifB, which encodes the pilus chaperone gene. To construct these mutants, plasmid pABK15, which contains a kanamycin resistance gene inserted into the XhoI site within hifB, was linearized with KpnI and used to transform strains AAr176 and 1595 made competent by the method of Herriott et al. and St. Geme et al. (15, 33). Transformants were selected on medium containing 25 μg of kanamycin per ml. Successful allelic exchange was confirmed by Southern hybridization, probing with the kanamycin resistance cassette from pWW84 and hifA or hifB, as appropriate (20).

Epithelial cell lines. The human bronchial cell line 16HBE14o⁻ was a gift from Dieter Gruenert, University of California, San Francisco, and HEp-2 (human laryngeal cell line), HeLa (human cervical cell line), and ME180 (human cervical cell line) cells were obtained from the American Type Culture Collection (4). The epithelial cells were maintained in Eagles' minimum essential medium (16HBE14o⁻, HEp-2, HeLa) or McCoy's 5A medium supplemented with 10% fetal bovine serum, 1% l-glutamine, 1% penicillin, and 1% streptomycin (all from Gibco BRL, Gaithersburg, Md.). Cells were grown in 20 ml of medium in 75-cm³ tissue culture flasks (Costar Corp., Cambridge, Mass.) and cultured at 37°C and 5% CO₂. The confluent, adherent monolayers were released from the plastic surface after treatment with trypsin-EDTA (Gibco BRL) and diluted 1:5 every 5 days. For bacterial adherence assays, the cells were plated onto 96-well flat-bottom polystyrene plates (Costar Corp.) and grown overnight to confluency.

Hemagglutination and red blood cell enrichment. The erythrocyte selection technique to identify hemagglutinating, piliated variants was performed as previously described (23). Hemagglutination was tested in a semiquantative microtiter well assay as previously described, and positive hemagglutination was defined as a titer of ≥1:8 (23).

Adherence assay. Adherence assays were performed as previously described (13). Briefly, epithelial monolayers in 96-well tissue culture plates were fixed with 1% glutaraldehyde, washed with 0.2% Triton X-100, and blocked with phosphate-buffered saline (PBS) containing 2% bovine serum albumin. The epithelial monolayers were then incubated with 10⁷ CFU of biotinylated bacteria (Sulfo-NHS-Biotin; Pierce Chemical Company, Rockford, Ill.) for 1 h, washed, and incubated further with ExtrAvidin-peroxidase conjugate (Sigma-Aldrich). The wells were developed with the enzyme substrate o-phenylenediamine dihydrochloride (Sigma-Aldrich) and the intensities of the reactions were determined by measuring the A₄₉₀ on a Dynatech microplate reader (Dynatech Laboratories, Inc., Alexandria, Va.). Control wells contained epithelial cells without bacteria. Test wells were replicated eight times in each assay, and each assay was repeated six times.

Interference of pilus-mediated adherence by the presumed receptor analogue GM-1 was performed as previously described by our laboratory group (13). Comparisons of GM-1 interference of adherence, as measured by the A₄₉₀s of the immunoassays, were made using Student's two-tailed paired t test, calculated using Microsoft Excel 97 software.

To control for day-to-day variation in the assays that tested the nonpiliated phase variants and the pilus-deficient mutants, the raw optical density data were transformed to express the adherence of phase variants and mutants as a percentage of the wild-type adherence and analyzed by Student's two-tailed paired t test, using Systat, version 7.0.1, software (SPSS Inc., Chicago, Ill.).

Western immunoblot analysis. The presence of HifA or HifE was determined using Western immunoblot analysis of whole bacterial cells reacted with polyclonal antisera directed against HifA (R20) from H. influenzae type b (Hib) strain M43 and HifE (R46 [HifE₂]) from Hib strain Eagan performed as previously described (9, 18).

Genetic analyses and DNA constructs. Genomic DNA was isolated from H. influenzae using the Wizard genomic DNA purification kit (Promega, Madison, Wis.). The PCR amplification of hifA was performed using previously described primers with the following cycling parameters (3): the PCR mixture was first incubated for 1 min at 95°C and then for 35 cycles of 95°C for 1 min, 40°C for 1 min, and 72°C for 2 min, followed by a final elongation step for 3 min at 72°C in a Midel PTC-100 programmable thermal controller (MJ Research, Inc., Watertown, Mass.).

The PCR amplification of hifE was performed using primers from the 5′ end of the hifE gene from NTHI strain AAr49 (5′-ATGAAAACTCTAACAACATACGC) (GenBank accession no. AF045063) and from the 3′ end of the hifE gene from Hib strain Eagan (5′-TTATTGATATGACATTGTGAAAGTGG) (GenBank accession no. U13254). The primers were synthesized at the University of Michigan Biomedical Research Core Facility. The conditions for PCR amplification of hifE were the same as those used for hifA (see above). After amplification, samples were separated on 1% agarose gels and bands were visualized after staining with ethidium bromide (Sigma-Aldrich) and illumination by UV light. Molecular size markers were used to estimate PCR product sizes (1-kb ladder; Gibco BRL).

hifA genes from H. influenzae biotype IV strains 1595 and 4162 were cloned into plasmid pGEM-T Easy (Promega) according to the manufacturer's instructions. Escherichia coli JM109 (Promega) served as the host strain for the recombinant plasmids. Plasmid DNA from putative transformants was isolated using the Wizard Plus Minipreps DNA Purification System from Promega. EcoRI (Gibco BRL) digests were used to confirm DNA inserts in the isolated recombinant plasmids.

hifE genes from H. influenzae biotype IV strains 6351, 785, 1595, 4162, and 1216 were cloned into plasmid pCR2.1 using the Original TA Cloning kit from Invitrogen (San Diego, Calif.). E. coli DH5α (Gibco BRL) served as the host strain for the recombinant plasmids. Plasmid DNA from putative transformants was isolated with Qiagen Miniprep kits (Qiagen, Valencia, Calif.). EcoRI (Gibco BRL) digests were used to confirm DNA inserts in the isolated recombinant plasmids.

The cloned hifA and hifE genes were sequenced at the University of Michigan Medical School DNA Core Facility with Applied Biosystems models 373A and 3700 automated sequencers (Applied Biosystems, Inc., Foster City, Calif.). Sequencing primers were purchased from Stratagene Cloning Systems (La Jolla, Calif.) and synthesized at the University of Michigan Medical School DNA Synthesis Core Facility. The DNA and protein sequences were analyzed with Lasergene Biocomputing software from DNASTAR, Inc. (Madison, Wis.), and the Wisconsin Package, version 10.0, from the Genetics Computer Group (Madison, Wis.).

Transmission electron microscopy. H. influenzae strains M43 and AAr176 were cultured overnight in 10 ml of Levinthal broth at 37°C in 5% CO₂ and strains 1595 and 4162 were cultured overnight in supplemented Schaedler broth at 37°C in 5% CO₂. Bacterial cultures were washed once in PBS and then resuspended in PBS to an optical density at 610 nm of 0.4 to 0.5. A 20-μl drop of bacterial cell suspension was applied for 2 min to a 200-mesh formvar- and carbon-coated grid which had been previously glow discharged. Excess fluid was blotted from the grid, and then it was rinsed in double-distilled water. The grids were negatively stained with 20 μl of 1% phosphotungstic acid for 1 min. The grids were then blotted and air dried prior to visualization on a Philips CM 100 electron microscope at 60 kV.

Nucleotide sequence accession numbers. The GenBank accession numbers for the hifE DNA sequences and the derived protein sequences determined in this study are as follows: strain 6351, no. AF245359; strain 785, no. AF245360; strain 1595, no. AF245361; strain 4162, no. AF245362; and strain 1216, no. AF245363. The GenBank accession numbers for the hifA DNA sequences and the derived protein sequences determined in this study are as follows: strain 1595, no. AY034824; strain 4162, no. AY034825.

Hemagglutination and red blood cell enrichment. None of the six biotype IV NTHI strains tested in this study (strains 6351, 785, 1595, 4162, 1216, and 597) agglutinated human red blood cells. In addition, none of these strains could be enriched for hemagglutinating variants by repeated incubation with human red blood cells. These results suggest that red blood cell-adherent variants are not present in the biotype IV strains tested; in contrast, many respiratory strains can be enriched for variants expressing hemagglutinating pili (8).

Assessment of presence of hifA and hifE. The presence of hifA, which encodes the major structural subunit of hemagglutinating pili, was assessed by PCR amplification of genomic DNA from biotype IV strains using intragenic hifA primers (3). hifA amplicons of the appropriate size were identified from all six biotype IV strains. Analysis of the derived HifA amino acid sequences from strains 1595 and 4162 showed 77 and 63% identity, respectively, to the HifA sequence from H. influenzae strain Eagan (17). The HifA sequence from strains 1595 and 4162 showed 63 and 98% identity, respectively, to the GhfA sequence from biotype IV strain 26E (14).

The presence of hifE, which encodes the tip adhesin of hemagglutinating pili, was assessed by PCR amplification of genomic DNA from biotype IV strains using hifE intragenic primers from the type b strain Eagan and the nontypeable strain AAr49 (17, 18). hifE amplicons of the appropriate size were identified from five of the six biotype IV strains and were shown by nucleotide sequence analysis to contain hifE.

The derived amino acid sequences of HifE from the biotype IV strains were compared to those of H. influenzae strains of types a, b, and c and nontypeable respiratory H. influenzae strains, as well as to that of the genital biotype IV strain 26E described by Gousset et al. (14) (Fig. 1; Table 2). The HifE amino acid sequences of four of the five biotype IV strains described in this study were highly homologous, showing from 99 to 100% identity, and were 99 to 100% identical to HifEs of two H. influenzae type b strains, Eagan and AM30. In addition, the biotype IV HifE sequences showed between 47 and 71% identity with those from types a and c and the nontypeable strains tested (Table 2).

FIG. 1 Comparison of derived amino acid sequences of two representative HifE proteins from biotype IV strains 4162 and 1595. Identical residues throughout the HifE sequences are shown in the “Consensus” line along with conserved substitutions (more ...)

TABLE 2 Percent amino acid identities of HifE based on DNA sequences from various strains

On the other hand, the HifE amino acid sequence of the biotype IV isolate 4162 was only 55 to 56% similar to those of the other biotype IV strains but was 96% identical to GhfE, the HifE equivalent of biotype IV strain 26E; these sequences showed between 42 and 62% identity with those of the type a and c strains and the nontypeable strains tested (14).

The putative HifE leader sequences of all 18 H. influenzae strains whose amino acid sequences are available for analysis show three different patterns (29). The leader sequences from all five biotype IV HifE proteins identified in this study, as well as GhfE of biotype IV strain 26E, fall into a single pattern (Fig. 1) (14). All strains share the conserved tyrosine and glycine residues 2 and 14 amino acids, respectively, from the C terminus, which are common to all five proteins in the Hif complex (11).

Assessment of expression of pili. The expression of HifA (the major pilus structural subunit) was assessed by Western blot assay using an antiserum (R20) developed against denatured HifA (pilin) of Hib strain M43 (10). This antiserum binds to the approximately 23-kDa HifA band on all piliated Hib and NTHI respiratory strains tested (10). In addition, we assessed the expression of HifE (tip adhesin) by Western blot, using an antiserum (R46) developed against a truncated HifE, representing the C-terminal 40% of the protein, from Hib strain Eagan (18). This antiserum binds to the 37- to 43-kDa HifE band on all piliated Hib and NTHI respiratory strains tested (16). Results of these Western blot assays showed no reactivity with the appropriately sized bands from any of the six hifA and five hifE PCR-positive biotype IV NTHI strains tested. These results suggest either that HifA and HifE are not expressed by these strains or that HifA and HifE proteins of the biotype IV strains significantly differ immunologically from those of respiratory strains and do not bind the R20 and R46 antisera.

Transmission electron microscopy of control piliated H. influenzae strains M43 and AAr176 showed many long, thin pili extending from the bacterial surfaces (Fig. 2A and B). hifA and hifB mutants of strain AAr176 showed no pili (Fig. 2C and D). Neither biotype IV strain 1595 nor 4162 showed pilus-like structures on its surfaces (Fig. 2E and G). The hifB mutant of strain 1595 showed pilus-like structures that appear shorter and thicker than those seen on the piliated strains M43 and AAr176 (Fig. 2F). The hifA mutant of strain 4162 appeared identical to the wild-type strain, with no surface appendages (Fig. 2G and H).

FIG. 2 Transmission electron micrographs of H. influenzae strains stained with phosphotungstic acid. (A) Piliated H. influenzae strain AAr176. (B) Piliated H. influenzae strain M43. (C) H. influenzae strain AAr176 with insertionally inactivated hifA. (D) H. (more ...)

Adherence of biotype IV strains to human epithelial cells. Adherence of the type b strain Eagan p⁺, nontypeable strain AAr176 p⁺, strain Rd, and the six biotype IV strains to two respiratory cell lines (16HBE14o⁻ and HEp-2) and two genital cell lines (ME180 and HeLa) is shown in Fig. 3. Pilus-expressing organisms (strains Eagan p⁺ and AAr176 p⁺) adhered better than their nonpiliated (p⁻) phase variants to the respiratory cell line 16HBE14o⁻ (Eagan, P < 0.0003; AAr176, P < 0.0001), confirming data from our previous studies (13). Similar results were seen with adherence of piliated Eagan and AAr176 to the genital cell line ME180 (Eagan, P < 0.0004; AAr176, P < 0.007). Furthermore, nonpiliated strain Eagan adhered better than piliated Eagan to HEp-2 cells (P = 0.01), again confirming data from our previous studies (13). No difference was seen in the adherence of piliated and nonpiliated strain AAr176 to HEp-2 cells (P > 0.34). Similarly, no differences were seen in adherence of the piliated or nonpiliated respiratory H. influenzae strains Eagan and AAr176 to the genital cell line HeLa (Eagan, P > 0.66; AAr176, P > 0.43). H. influenzae strain Rd (which does not possess the genes for pili or the adhesins HMW1, HMW2, or Hap and does not express the adhesin Hia) showed low-level adherence to all cell lines tested (5, 28).

FIG. 3 Adherence of piliated (p+) and nonpiliated (p−) respiratory strains Eagan and AAr176, laboratory strain Rd, and the six biotype IV NTHI isolates, 6351, 785, 1595, 4162, 1216, and 597, to cell types 16HBE14o−, ME189, HEp-2, and (more ...)

All six biotype IV (genital) H. influenzae strains, including strain 597, which failed to show a hifE amplicon on PCR, adhered to all four cell lines tested at levels similar to that for the piliated respiratory strain Eagan (Fig. 3).

The presumed pilus receptor analog GM-1 significantly inhibited bacterial adherence to 16HBE14o⁻ respiratory cells of the piliated respiratory strains (Eagan, P < 0.002; AAr176, P < 0.004) and four of the five genital biotype IV strains (P = 0.01 to 0.05), including strain 597, which lacks hifE (Fig. 3) (34). Since the presence of hemagglutinating pili on biotype IV H. influenzae could not be confirmed by Western blot analysis and since strain 597 adherence was also inhibited by GM-1, these results suggest that GM-1 inhibits non-pilus-mediated binding of biotype IV strains to respiratory 16HBE14o⁻ cells. The binding of biotype IV strain 1595 to 16HBE14o⁻ cells was not inhibited by GM-1, suggesting that strain 1595 binds to these cells by a different mechanism from the other biotype IV strains. In addition, GM-1 significantly inhibited binding of piliated strain Eagan (P = 0.04), but not piliated strain AAr176 (P = 0.30) or any of the biotype IV strains (P = 0.09 to 0.85), to genital ME180 cells. GM-1 inhibition had no effect on the binding of any H. influenzae strains to HeLa cells (P > 0.2), confirming previous studies showing that pili do not mediate H. influenzae binding to HeLa cells (13).

Role of pili in adherence of biotype IV strains to human epithelial cells. To further assess the role of hemagglutinating pili in biotype IV NTHI binding to human epithelial cells, the adherence of both respiratory strains (type b Eagan and NTHI AAr176) and biotype IV strains was compared with that of nonpiliated mutants derived from insertional mutagenesis of hifA (which encodes the major pilus structural subunit) and hifB (which encodes the pilus chaperone). H. influenzae strains with mutations in both of these genes have been shown to lack pili and to exhibit reduced binding to human red blood cells (36). As seen in Fig. 4, the nonpiliated (p⁻) variant and the hifA and hifB mutants of AAr176 showed reduced adherence (64 to 87% of that of the piliated variant and wild-type strains) to the respiratory cell line 16HBE14o⁻ and to the genital cell line ME180. The hifA and hifB mutants of biotype IV strains 4162 and 1595 showed a minimal difference in 16HBE14o⁻ adherence (91 and 98%, respectively, of that of the wild-type strains). Although nonpiliated variants and hifA and hifB mutants of AAr176 showed increased adherence to HEp-2 cells (120 to 160% of that of the piliated variant) as seen in previous studies, biotype IV hifA and hifB mutants showed no difference in HEp-2 adherence (92 and 95%, respectively, of that of the wild-type strain) (13). Neither nonpiliated variants nor hifA and hifB mutants of any of the H. influenzae strains showed differences in adherence to HeLa cells (89 to 102% of that of piliated or wild-type strains).

FIG. 4 Adherence of piliated (p+) respiratory strain AAr176, its nonpiliated phase variant (p−) and hifA and hifB mutants, and the biotype IV 4162 and 1595 strains and their hifA and hifB mutants to cell types 16HBE14o−, ME189, HEp-2, (more ...)

Analysis of collapsed data showed that AAr176p⁻ and AAr176 hifA and hifB mutants showed decreased adherence (67.2 to 79.5% of that of the wild type) to 16HBE14o⁻ (respiratory origin) and ME180 (genital origin) cells, which exhibit pilus-mediated binding, and somewhat increased adherence (115 to 132.8%) to HEp-2 (respiratory origin) and HeLa (genital origin) cells, which exhibit pilus-independent binding. The hifA and hifB mutants of the biotype IV strains 4162 and 1595 showed a minimal decrease in binding to all cell types (91.4 to 95.3% of that of the wild type).

Adherence of H. influenzae to human epithelial cells is mediated by bacterial adhesins, including hemagglutinating pili, Hia, HMW1 and -2, Hap, and P5 (11, 28, 31). Among these, hemagglutinating pili have been well characterized structurally, functionally, and immunologically and their receptors (which share at least partial homology with the gangliosides GM1, GM2, GM3, and GD1a) appear to be restricted to certain epithelial cells, such as shed buccal epithelial cells and 16HBE14o⁻ cells (a human bronchial cell line), and as newly documented in the present study, to ME180 cells (a human cervical cell line) (11, 13, 34). On the other hand, pili do not appear to mediate adherence to human tracheal fibroblasts, human nasal cells, keratinocytes, A549 cells (human alveolar epithelial cell line), HEp-2 cells (a human laryngeal cell line), HeLa cells (human cervical cell line), Chang cells (a conjuctival cell line), or KB cells (an oropharyngeal cell line) (29, 32). Because the biotype IV cryptic genospecies NTHI strains are almost exclusively isolated from genital sites or from infected neonates (who presumably acquire it from vaginal passage during birth), these organisms appear to occupy an ecologic niche that is different from that for NTHI respiratory strains, which colonize the pharynx and nasopharynx. Based on previous reports that biotype IV isolates possess pili, we sought to characterize the adherence of these strains to respiratory and genital cells and to define the role of hemagglutinating pili in this adherence (14, 26, 30).

Our results show that genital and neonatal biotype IV NTHI strains adhere to epithelial cells of both respiratory and genital origin. Furthermore, the adherence of respiratory NTHI to bronchial cells (16HBE14o⁻) as well as to cervical cells (ME180) is pilus mediated, as shown by the adherence studies using both piliated and pilus-deficient mutant NTHI. On the other hand, the adherence of the genital and neonatal biotype IV NTHI to all of the four cell lines tested appears to be pilus independent.

The mechanism of the biotype IV NTHI adherence to epithelial cells is unclear. Presence or absence of the other NTHI adhesins on these strains has not been well documented, although Gousset et al. (14) showed that 16 of 19 biotype IV NTHI strains of the cryptic genospecies (including strain 1595 also examined in our study) exhibited the P5-fimbrial gene by PCR analysis. In our studies, the magnitude of adherence of the biotype IV NTHI strains to the four cell lines tested was considerably greater than the adherence of strain Rd, which has been shown by total genome analysis to possess the P5 fimbrial gene but not the genes for pili, HMW1, HMW2, or Hap and not to express the adhesin Hia (5, 28). Thus, an adhesin present on the biotype IV NTHI strains and not present on strain Rd most likely explains the biotype IV NTHI adherence.

Although five of the six biotype IV strains in this study possess hifA and hifE, these strains were not immunologically reactive with antisera raised against HifA and HifE from H. influenzae strain Eagan, suggesting that biotype IV strains do not express hemagglutinating pili under the conditions of our assays. Pili of respiratory H. influenzae isolates are often not expressed under routine laboratory conditions, but because pilus expression is phase variable, piliated (p⁺) variants within populations of nonpiliated (p⁻) bacteria may be identified and enriched for by their adherence to human red blood cells (2). It is likely that the HifA- and HifE-containing pili of the biotype IV strains were in the “off” phase and that we were not able to select for them because they do not bind to, and thus cannot be enriched by, human red blood cells. Theoretically, if HifA- and HifE-containing pili of biotype IV NTHI possess different binding specificities from those of hemagglutinating pili of respiratory isolates, piliated variants of the biotype IV isolates could be identified by enriching the strains with human cells that possess the specific ligands for their adhesins. The identity of such human cells is unclear. Neither of the two genital cell lines used in this study showed pilus-dependent H. influenzae binding.

The adhesive domain on HifE by which it binds to epithelial cells has not been defined, but the HifE proteins of four of the six biotype IV NTHI strains in this study show remarkable amino acid sequence homology to HifE of type b, biotype II strain Eagan. Furthermore, the five nucleotide differences between HifE of strain Eagan and the HifE proteins of the biotype IV strains occur in different positions within the biotype IV sequences.

Although biotype IV NTHI have been shown to have pilus-like structures by electron microscopy, our analysis failed to show such structures on the two biotype IV strains examined (26, 30). Curiously, the hifB mutant of strain 1595 showed surface pilus-like appendages that differed structurally from the hemagglutinating pili seen on our piliated strains AAr176 and M43. Rosenau et al. (30) reported these pili on 14 of 17 cryptic genospecies strains, including strain 1595, which is included in our study. Only 3 of these 14 piliated strains, however, exhibited hemagglutination. Although the actual nature and function of these pilus-like structures was not defined by Rosenau et al. (30), they appear not to be the hemagglutinating pili found on respiratory and non-biotype IV strains. We cannot explain this discrepancy readily, although the presence of the 1595 pilus-like structures on a strain possessing a mutation in the pilus chaperone suggests that expression of these structures may be influenced by unknown regulation-dependent events.

The results from our studies differ somewhat from those of Gousset et al. (14), who identified hifE by PCR in only 1 of 19 biotype IV NTHI isolates and hifA by Southern blot hybridization in only 6 of 19 biotype IV NTHI isolates. We identified hifE by PCR and sequence analysis from five of the six strains in our study, including strain 1595, from which Gousset et al. (14) did not amplify hifA, -D, or -E. The use of different PCR conditions may explain this discrepancy. Even among respiratory NTHI isolates, hifA and hifE may not be present in all strains; Geluk et al. (6) showed that 18% of respiratory strains hybridized to a pilus cluster probe and Read et al. (29) showed that 56% of strains hybridized to a hifC (which encodes for the pilus usher protein and is highly conserved) probe.

Musser et al. (22) and Quentin et al. (27) showed, by restriction fragment length polymorphism analysis of ribosomal DNA and multilocus enzyme electrophoresis, respectively, that the majority of biotype IV NTHI strains are closely related to each other and phylogenetically quite distinct, thus forming a cryptic genospecies. On the other hand, amino acid sequence analysis of the HifE proteins of biotype IV NTHI genital strains reveals two clusters. The HifE proteins of four of the biotype IV strains in this study (including 1595, a defined cryptic genospecies strain) show strong amino acid homology (99%) with HifE proteins of two type b H. influenzae strains, Eagan and 77-0235. The HifE protein of strain 4162 in this study showed only 56% homology with strain Eagan but showed 96% homology with the cryptic genospecies strain 26E described by Gousset et al. (14). Previous studies from our laboratory have shown that although the HifE proteins (tip adhesins) of strains Eagan and 77-0235 are similar, their HifA proteins (pilus major structural subunits) are quite different, showing only 79% homology and falling into two distinctly different restriction fragment length polymorphism patterns (3). These results corroborate earlier findings suggesting horizontal gene transfer among H. influenzae strains that may explain the genetic and immunological diversity of many of its surface proteins (7).

In summary, this study suggests that genital biotype IV NTHI strains which form a cryptic genospecies contain hifA and hifE genes that were not expressed under the conditions of our assays and bind to human epithelial cells in a different manner from respiratory H. influenzae strains expressing hemagglutinating pili. The function of these pilus-like structures is unclear, but from their structural character, they most likely serve as adhesins in some fashion. Similarly, the function of the hifA and hifE gene products of these strains is unclear. Mutants deficient in these proteins did not show reduced adherence to any of the four epithelial cell lines tested, most significantly the two cell lines that exhibit pilus-mediated binding. It remains possible that the hifE and hifA gene products play a role in biotype IV NTHI adherence of unique epithelial cell specificity but were not expressed under the conditions of our studies.

This work was supported by Public Health Service grant AI25630 from the National Institute of Allergy and Infectious Diseases.