Copy Number of Pilus Gene Clusters in Haemophilus influenzae and Variation in the hifE Pilin Gene

Abstract

Brazilian purpuric fever (BPF)-associated Haemophilus influenzae biogroup aegyptius strain F3031 contains two identical copies of a five gene cluster (hifA to hifE) encoding pili similar to well-characterized Hif fimbriae of H. influenzae type b. HifE, the putative pilus tip adhesin of F3031, shares only 40% amino acid sequence similarity with the same molecule from type b strains, whereas the other four proteins have 75 to 95% identity. To determine whether pilus cluster duplication and the hifE_F3031 allele were special features of BPF-associated bacteria, we analyzed a collection of H. influenzae strains by PCR with hifA- and hifE-specific oligonucleotides, by Southern hybridization with a hifC gene probe, and by nucleotide sequencing. The presence of two pilus clusters was limited to some H. influenzae biogroup aegyptius strains. The hifE_F3031 allele was limited to H. influenzae biogroup aegyptius. Two strains contained one copy of hifE_F3031 and one copy of a variant hifE allele. We determined the nucleotide sequences of four hifE genes from H. influenzae biogroup aegyptius and H. influenzae capsule serotypes a and c. The predicted proteins produced by these genes demonstrated only 35 to 70% identity to the three published HifE proteins from nontypeable H. influenzae, serotype b, and BPF strains. The C-terminal third of the molecules implicated in chaperone binding was the most highly conserved region. Three conserved domains in the otherwise highly variable N-terminal putative receptor-binding region of HifE were similar to conserved portions in the N terminus of Neisseria pilus adhesin PilC. We concluded that two pilus clusters and hifE_F3031 were not specific for BPF-causing H. influenzae, and we also identified portions of HifE possibly involved in binding mammalian cell receptors.

Haemophilus influenzae is a gram-negative bacterium that colonizes the human nasopharynx. Strains with the serotype b capsule are highly virulent in nonimmune children, whereas nonencapsulated (nontypeable H. influenzae [NTHI]) bacteria are an increasingly prevalent cause of localized respiratory tract infections in the adult population. Many H. influenzae strains express pili that enable the bacteria to agglutinate human erythrocytes (9). The primary role of the pili is believed to be initiation of contact between the bacterium and human cell during infection. Analysis of the chromosomal region containing the pilus determinants of H. influenzae serotype b (Hib) strains Eagan and M43 revealed five genes, hifA to hifE, necessary for pilus production (13, 16, 38, 40). The H. influenzae genes have sequence homology with, and have similarities in organization to, the Pap and type I pilus genes (8, 9, 37). HifA is the major pilus antigen, HifB is a periplasmic chaperone, HifC acts as an outer membrane assembly platform, and HifD is believed to be minor pilus component (33). Recent evidence suggests that the fifth gene in the cluster, hifE, encodes a pilus tip-associated adhesin responsible for binding to gangliosides on the surface of mammalian cells (3, 9). Mutations in hifE do not affect the stability of the other pilins but severely reduce assembly of the whole pilus (16, 17, 24, 38). For Hib strain Eagan, McCrea et al. (17) have shown that anti-HifE antiserum binds to pilus tips and blocks pilus-mediated hemagglutination. Conversely, van Ham et al. (36) reported that the major pilus subunit HifA and not the minor pilins HifD or HifE effect binding to erythrocytes and human oropharyngeal epithelial cells. These conflicting results might be due to the slight (20%) difference in primary sequences between the HifA proteins studied or because of differential expression of the recombinant H. influenzae genes in the Escherichia coli DH5α background used in the second study. Pilus expression by H. influenzae oscillates between phase off and phase on at a rate of about 10⁻⁴/generation (4). This phase variation is likely due to slip-strand mispairing in a poly(TA) dinucleotide tract separating the divergent, bidirectional hifA and hifBCDE promoters altering the spacing upstream of the transcriptional start sites (39).

In 1984, a unique and lethal condition caused by H. influenzae emerged in central Brazil. The disease called Brazilian purpuric fever (BPF), afflicted young children who had recovered earlier from conjunctivitis, causing an array of symptoms including acute fever, petichiae, and vascular collapse (1). The etiologic agent of BPF was found to be a clonal strain of H. influenzae biogroup aegyptius, nonencapsulated organisms in biotype III that had previously only been known to cause self-limited conjunctivitis. The BPF-associated reference strain, F3031, had several characteristics not encountered in other NTHI strains, such as endothelial cell toxicity, resistance to human serum, and the ability to cause bacteremia in infant rats (15, 20, 22).

One possible factor in the unusual pathogenicity of F3031 is the immunologically distinctive hemagglutinating pilus elaborated by the strain (41). The pilus appears to be necessary for invasion of primary epithelial cells by F3031 (5). Further, BPF clone strains isolated from systemic sources are piliated more often than is Hib (15). In a recent study we cloned and characterized genetically the F3031 pilus determinants (24). F3031 has two independently phase-variable fimbrial clusters, one of which, hif1, is in a chromosomal location not previously described for H. influenzae (Fig. 1). Therefore, pilus production occurs when one or both sets of genes are in the on phase, perhaps explaining the abundant piliation of the strain. We originally designated the F3031 genes hafA to hafE (24), but because they are functionally similar to the H. influenzae hif cluster, we will henceforth refer to them as hif_F3031 alleles. While the BPF-associated HifA to HifD gene products share 72 to 96% identity with Hib proteins, the HifE_F3031 putative pilus tip adhesin has only 40% identity with its Hib homolog.

FIG. 1.

Schematic diagram of hif1 and hif2 clusters of BPF-associated strain F3031 (24). The five pilus genes of each cluster are indicated by open arrows, while the flanking H. influenzae genes are indicated by gray arrows. hif1 and hif2 are flanked by direct repeats homologous to Rd genome coordinates (6) 1220363 to 1220394 and 1682304 to 1682362, respectively.

The purpose of this study at its outset was to determine whether the possession of duplicated pilus clusters and the distinctive HifE adhesin were unique features of BPF-associated H. influenzae strains. To further this aim we examined a collection of H. influenzae biogroup aegyptius, NTHI, and capsulated H. influenzae by Southern DNA hybridization, PCR, and sequencing of variant hifE genes.

MATERIALS AND METHODS

Bacteria.

The H. influenzae strains used in this study are described in Table 1. All were grown for 16 to 20 h on chocolate agar plates (Becton Dickinson, Cockeysville, Md.) at 37°C in a 5% CO₂ atmosphere.

TABLE 1.

H. influenzae strains used in this study

Straina	Location of original isolate	Description	Sourceb
H. influenzae biogroup aegyptius
F3031	Brazil	BPF reference strain (1)	CDC
F6422	Australia	From patient with BPF-like symptoms (23)	CDC
F4931	Brazil	“Pradopolis” strain (23)	CDC
F4933	Brazil	“Pradopolis” strain (23)	CDC
F2066	Brazil	Non-BPF-associated isolatec	CDC
F3118	Brazil	Non-BPF-associated isolatec	CDC
F3331	Brazil	Non-BPF-associated isolatec	CDC
43794	Texas		ATCC
43800	Texas		ATCC
43806	Texas		ATCC
NTHI
F8835	Connecticut	Invasive disease isolate	CDC
GA2188	Georgia	Invasive disease isolate	EIP
GA2326	Georgia	Invasive disease isolate	EIP
GA4752	Georgia	Invasive disease isolate	EIP
GA5445	Georgia	Invasive disease isolate	EIP
H. influenzae
Serotype a
GA1790	Georgia	Invasive disease isolate	EIP
GA2078	Georgia	Invasive disease isolate	EIP
GA4774	Georgia	Invasive disease isolate	EIP
Serotype b
GA1812	Georgia	Invasive disease isolate	EIP
1007	Texas	Nasopharyngeal isolate	CDC
Serotype c 9007			ATCC
Serotype d
9008			ATCC
Rd (ATCC 51907)		Capsule negative	ATCC
Serotype e GA2284	Georgia	Invasive disease isolate	EIP
Serotype f
GA840	Georgia	Invasive disease isolate	EIP
GA4090	Georgia	Invasive disease isolate	EIP
GA4913	Georgia	Invasive disease isolate	EIP

^aAll bacteria tested negative for human type O-positive hemagglutination except F2066, F3031, F3331, F6422, and 43800. 

^bATCC, American Type Culture Collection, Rockville, Md.; CDC, Centers for Disease Control and Prevention, Atlanta, Ga.; EIP, Georgia Emerging Infections Program, Atlanta, Ga. 

^cIsolated concurrently with outbreak (1). 

PCR.

Amplifications were run for 35 cycles at an annealing temperature of 55°C with Promega (Madison, Wis.) reagents. Oligonucleotide DNA primers (Table 2) were synthesized by the Emory University Microchemical Facility. We used 15 μM Mg²⁺ and 30 pmol of each primer. To prepare templates, single colonies of bacterial culture grown overnight on chocolate agar plates were resuspended in 50 μl of sterile water and boiled for 5 min. In a standard PCR mixture of 50 μl, 2 μl of template was used.

TABLE 2.

Oligonucleotides used in PCR amplifications

Oligonucleotide	Sequence	Location of binding
HA1	5′-GGT AAG GTT GTT GAG AAT ACT TG	hifAa
HE1	5′-CCA CTG CAA CTT TTC AAG TGC	Near 3′ terminus of hifDa
PMB	5′-GGG CGA ATT GCA AGA TAT GTT G	pmbA gene flanking hifA1
PUR	5′-CGT GGC GTA GAA TAG CAT AG	purE gene flanking hifA2
HI1153	5′-CGA GCA TGA ACA TCA TTC TCT G	HI1153 hypothetical ORF flanking hifE1
PEP	5′-CTG TGA CCG TAA AAT CTG GTT G	pepN gene flanking hifE2
HE2	5′-TAT TGA TAT GAC ATT GTG AAA GTG G	Near 3′ terminus of hifEa
HC1	5′-CCT CTG ATG ATA GAA TGC TTG	hifCa
HC2	5′-CGT CCT CCA GCT AAT TGA TAA C	hifCa

^aPrimer complementary to F3031, Hib strain 770231, and NTHI 86-025 hif gene sequences (accession numbers U54780, Z33502, and U19730, respectively). 

Southern hybridization.

H. influenzae DNA was prepared from cultures grown overnight on chocolate agar plates by using the DNA/RNA Isolation Kit (United States Biochemicals). Digested DNA was transferred to nylon membranes (31) and hybridized to digoxigenin-labeled DNA probes with the Genius Nonradioactive Detection (Boehringer Mannheim, Indianapolis, Ind.) reagents and protocols. The hybridizations with labeled probe were carried out overnight at 65°C. The final washes of the membrane were at 65°C in 0.5× SSC buffer (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate) (31).

Restriction fragment length polymorphism (RFLP)-PCR.

PCR fragments containing hifE1 and hifE2 (products of reactions with HE1-HI1153 or HE1-PUR pairs [Table 2]) were cut from 0.7% agarose gels, melted at 70°C in 2 volumes of sterile H₂O, and used as templates for a nested PCR with primers HE1 and HE2. The resulting PCR products were all 1.3 kb in length from a conserved region at the 3′ end of hifD to a conserved region at the 3′ end of hifE. These DNA fragments were digested separately with ClaI, Sau3AI, RsaI, and DraI restriction enzymes (New England Biolabs, Beverly, Mass.), each of which cut zero to six times.

Sequencing.

PCR amplicons containing hifE2 genes of strains ATCC9007, GA2078, and F2066 were generated with the HE1 and PEP primers (Table 2). The hifE1 gene of H. influenzae biogroup aegyptius F4931 was amplified with HE1 and HI1153. The four genes were sequenced with a set of specifically generated primers by dideoxy dye terminator cycle sequencing with an ABI Prism fluorescence sequencing kit on an ABI Prism 377 sequencer (Perkin Elmer). DNA sequence alignments were computed with ABI Prism software.

Hemagglutination.

H. influenzae strains were tested for their ability to agglutinate human O-positive erythrocytes by a microplate assay described previously (24).

Nucleotide sequence accession numbers.

The hifE2₉₀₀₇, hifE2_GA2078, hifE1_F4931, and hifE2_F2066 nucleotide sequences were deposited in GenBank under accession numbers AF026299 to AF026302, respectively.

RESULTS

Hybridization of the hifC probe with an H. influenzae strain set.

A panel of 27 H. influenzae strains (Table 1) was probed with a conserved hifC gene fragment in order to determine the presence of pilus gene complexes. The probe was a 296-bp digoxigenin-labeled PCR fragment containing a portion of the hifC_F3031 locus, generated with primers HC1 and HC2 (Table 2). The hifC gene was chosen because the encoded protein is essential for pilus production (40) and the nucleotide sequence is highly conserved between H. influenzae pilus clusters (9); for example, Hib strain Eagan hifC has 96% similarity to hifC_F3031. The H. influenzae set included BPF type strain F3031; H. influenzae biogroup aegyptius F6422, a strain associated with a BPF-like outbreak in Australia; two H. influenzae biogroup aegyptius “Pradopolis” strains, which are indistinguishable from F3031 by enzyme isotyping but do not express pilus; and several non-BPF-associated H. influenzae biogroup aegyptius strains from Brazil and Texas. Also included were representative isolates of the six H. influenzae capsule serotypes and four nontypeable strains from the Georgia Emerging Infections Surveillance Program (formerly the Metropolitan Atlanta Active Surveillance Program). H. influenzae Rd was included as a control. The genome of Rd has been sequenced completely and does not contain pilus genes (6).

Southern hybridizations of the hifC_F3031 probe with H. influenzae BglII chromosomal digests are depicted in Fig. 2. Because the probe was relatively small (296 bp) and did not contain a BglII recognition site and because sites for this enzyme are relatively uncommon in the Haemophilus genome, it was likely that two bands in the Southern hybridization represented distinct hifC alleles rather than two restriction fragments of the same gene hybridizing with the probe. Fifty-six percent (15/27) of the strains contained at least one hybridizing restriction fragment. Seven of the 10 H. influenzae biogroup aegyptius strains had two hybridizing fragments. Only 1 of 10 H. influenzae biogroup aegyptius strains, ATCC 43806, isolated in Texas, did not hybridize to the probe. In contrast, only 2 of 5 NTHI strains and 4 of 12 typeable strains (from a, b, and c capsule serotype strains) contained hifC homologs. It is notable that many strains carry hif genes but do not agglutinate erythrocytes (Table 1, footnote a), suggesting that, like F3031 and Hib strain 770325 (24, 39), pilus expression by these bacteria is phase variable.

FIG. 2.

Southern hybridization of the set of H. influenzae strains with hifC-specific probe (see Materials and Methods for details). Chromosomal DNA was digested with BglII. The positions of DNA molecular size markers (in kilobases) are shown on the left. Strains with two fragments hybridizing to the hifC probe were F3031, F6422 (double band), F4931, F4933, F2066, F3118, and ATCC 43794 (all H. influenzae biogroup aegyptius strains). H. influenzae strains with one hybridizing BglII fragment are biogroup aegyptius strains F3331 and ATCC 43800, NTHI strains F8835 and GA2188, serotype a strains GA2078 and GA4774, serotype b strain 1007, and serotype c strain ATCC 9007. The biogroup aegyptius strain F6422 and serotype f strains GA084, GA4090 and GA4913 had extra hybridizing DNA fragments, but because they appeared after extended exposure of the film, they were not deemed significant.

PCR analysis of Haemophilus gene clusters.

Next, we determined the chromosome location of the hif genes in the 27 H. influenzae strains on the basis of the known positions of the hif1 and hif2 clusters of F3031 (Fig. 1). The F3031 clusters are inserted in intergenic regions between pairs of genes, purE and pepN (hif2) and pmbA and HI1153 (hif1), respectively, that are contiguous in the Rd genome (24). We subjected the H. influenzae strain set to PCR with primers complementary to the ends of the F3031 hif1 and hif2 gene clusters (Table 2). These PCRs produced amplicons from conserved portions of hifA and hifE (with primers HA1 and HE1) to conserved regions flanking the hif1 (with primers PMB and HI1153) and hif2 (with primers PUR and PEP) clusters. Table 3 shows the results of the PCR analysis. In seven of the 27 strains, all of them H. influenzae biogroup aegyptius, PCR were successful with both hifA1- and hifE1-specific primers, indicating the presence of a complete five gene hif1 cluster located between the pmbA and HI1153 open reading frames (ORFs). In addition, H. influenzae biogroup aegyptius strain F3118 produces hifE1 and hifE2 amplicons, but neither the hifA1 nor the hifA2 amplification product is produced. One possible reason for this is that the HA1 primer does not bind to sites in the hifA_F3118 genes. Fifteen of 27 strains amplified with the HA1-PUR (hifA2-specific) and HE1-PEP (hifE2-specific) primer pairs, indicating the presence of an entire hif2 cluster. With a few exceptions (see Discussion), the number of intact pilus gene clusters deduced from the PCRs matched the number of bands hybridizing to the hifC gene probes (Fig. 2). Strains believed to carry two hif gene clusters from the Southern hybridization described previously were amplified successfully with both hif1 and hif2 primer sets, whereas strains with only one BglII fragment hybridizing to hifC produced amplification products with either the hif1 or hif2 primer sets. Thus there is no evidence that any H. influenzae strain contains pilus genes situated in any location other than hif1 or hif2.

TABLE 3.

PCR fragment sizes generated by primer pairs specific for hifA1, hifE1, hifA2, and hifE2

Strain	Typea	Size (kb) of PCR amplicons forb:				No. of hifC copiesc
		hifA1 (PMB/HA1)	hifE1 (HE1/HI1153)	hifA2 (PUR/HA1)	hifE2 (HE1/PEP)
F3031	BPF	0.95	1.50	1.10	1.60	2
F6422	BPF	0.40	0	0.70	1.50	2
F4931	“Pradopolis”	0.95	1.50	1.10	1.60	2
F4933	“Pradopolis”	0.95	1.50	1.10	1.60	2
F2066	Aegyptius	0.95	1.50	0.75	1.70	2
F3118	Aegyptius	0	1.50	0	1.30	2
F3331	Aegyptius	0.95	1.50	0	0	1
43794	Aegyptius	0.95	1.50	1.20	1.80	2
43800	Aegyptius	0.95	1.50	0	0	1
43806	Aegyptius	0	0	1.10	0	0
F8835	NTHI	0.95	0	1.30	1.60	1
GA2188	NTHI	0	0	1.35	1.60	1
GA2326	NTHI	0	0	0	0	0
GA4752	NTHI	0	0	0	0	0
GA5445	NTHI	0	0	0	0	0
GA1790	Serotype a	0	0	0	0	0
GA2078	Serotype a	0.95	0	1.30	1.50	1
GA4774	Serotype a	0	0	1.35	1.50	1
GA1812	Serotype b	0	0	0	0	0
1007	Serotype b	0	0	0.70	1.35	1
9007	Serotype c	0.95	0	1.25	1.40	1
9008	Serotype d	0	0	0	0	0
Rd	Serotype d	0	0	0	0	0
GA2284	Serotype e	0	0	0	0	0
GA840	Serotype f	0	0	0.75	0.45	0
GA4090	Serotype f	0	0	0.75	0.45	0
GA4913	Serotype f	0	0	0.75	0.45	0

^aBPF, BPF-associated H. influenzae biogroup aegyptius; “Pradopolis”, H. influenzae biogroup aegyptius strain; Aegyptius, H. influenzae biogroup aegyptius. 

^bThe primer pairs specific for each gene are given in parentheses (see Table 2). A size of 0 indicates that no product was detected. 

^cNumber of copies of hifC gene from hybridization data (Fig. 2). 

Significant variation was noted in fragment lengths produced by amplification with primer sets specific for the hifA2 and hifE2 genes (Table 3). The size of the HA1-PUR amplicon varied from 0.70 to 1.35 kb; the HE1-PEP amplicon varied from 1.30 to 1.80 kb (not counting the type f strains; see Discussion). As described below, nested PCR showed that the size of hifE was always 1.3 kb (Fig. 3), suggesting that DNA fragment size variation is in the hifE-to-pepN intergenic region. Except for those of F6422, the hifA1 and hifE1 amplicon sizes were uniformly 0.95 and 1.50 kb, in significant contrast to the variation found in the hif2 cluster.

FIG. 3.

RFLP of hifE genes. (A) Fourteen PCR amplicons generated with oligonucleotide primers HE1 and HI1153 (hifE1) or HE1 and PEP (hifE2). (B) These DNA fragments were excised from the gel and used as templates for nested PCR with HE1 and HE2. The uniformly sized hifE amplicons were digested with four restriction enzymes. (C) ClaI digestion of the amplicons. The sizes of DNA markers (in kilobases) are shown on the left.

Variation in hifE genes examined by RFLP-PCR.

We developed a simple RFLP-PCR procedure to sample the genetic variation of hifE in our Haemophilus strain collection (Fig. 3). Based on digestion of nested hifE PCR products with four restriction enzymes, there was a distinct group of alleles from H. influenzae biogroup aegyptius strains that were indistinguishable from, or very similar to, the duplicated hifE_F3031 genes. The H. influenzae biogroup aegyptius genes with identical restriction patterns were hifE1_F2066, hifE2_F4931 and hifE2_F4933, while strain 43794 hifE genes differed by only one restriction site. Of the aegyptius strains with duplicated pilus gene clusters (Table 3), F3031, F3118, and 43794 contained identical hifE1 and hifE2 alleles. However, non-BPF-associated strain F2066 and “Pradopolis” strains F4931 and F4933 had one hifE copy indistinguishable from hifE_F3031 but the other was genetically quite distinct. For F2066, variation from hifE_F3031 was noted in the hifE2 gene whereas in the “Pradopolis” strains variation occurred in hifE1. Aside from the conserved hifE_F3031 allele, no other hifE genes shared identical or nearly identical RFLPs.

Sequence comparisons of hifE genes.

Two H. influenzae biogroup aegyptius strains with divergent hifE genes (hifE_F4931 and hifE_F2066) were selected for sequence analysis. Both had apparently identical copies of the hifE_F3031 allele within a second gene cluster. In addition, PCR products containing hifE2_GA2078 (serotype a) and hifE2₉₀₀₇ (serotype c) were sequenced, as pilus genes from these H. influenzae capsule serotypes had not been described previously. The nucleotide and predicted HifE amino acid sequences of the four genes were aligned with sequences of three hifE genes already in GenBank from strain F3031 (accession number U54780 [24]), Hib strain 770235 (accession number U19761 [38]), and NTHI strain 86-025 (accession number U19730) (Table 4). The amino acid distance matrix was used to create a dendrogram of HifE relatedness by the neighbor joining method (30) (data not shown). The phylogenetic analysis grouped the HifE proteins into three classes consisting of sequences from F3031 (BPF), 9007 (serotype c), and F4931 (H. influenzae biogroup aegyptius “Pradopolis”) in the first group; GA2078 (serotype a), F2066 (non-BPF H. influenzae biogroup aegyptius), and 86-025 (NTHI) in the second group; and Hib strain 770235 in a group of its own. The HifE sequence of Hib did not have more than 54% similarity with any other encoding nucleotide sequence or 45% similarity with any other protein sequence (Table 4). Significantly, hifE1_F4931 and hifE2_F2066 were only 77% and 44% similar, respectively, to hifE_F3031 despite having an apparently identical hifE gene on a second pilus gene cluster. The F2066 HifE protein had much more sequence similarity with HifE from NTHI and H. influenzae type a than with the two other H. influenzae biogroup aegyptius protein sequences (Table 4).

TABLE 4.

Percentages of nucleotide and amino acid similarities of hifE genes and encoded proteins

Protein	% similaritya
	F2066 HifE2 (aegyptius)	GA2078 HifE2 (serotype a)	86-025 HifE2 (NTHI)	770325 HifE2 (serotype b)	F3031 HifE (BPF)	F4931 HifE1 (“Pradopolis”)	9007 HifE2 (serotype c)
F2066 HifE2		69.6	65.2	38.7	36.0	38.6	36.5
GA2078 HifE2	76.5		66.6	45.5	36.3	38.8	34.4
86-025 HifE2	73.9	81.4		43.9	35.1	37.0	36.3
770325 HifE2	53.6	52.3	53.0		39.5	41.4	43.0
F3031 HifE	44.3	44.8	47.6	46.8		66.3	57.9
F4931 HifE1	44.2	45.3	43.9	43.3	77.1		70.2
9007 HifE2	46.5	47.2	49.9	51.6	69.0	80.2

^aValues in top right corner are percent similarities for each pair of amino acid sequences. Values in the bottom left corner are the percent similarities for the nucleotide sequences of each gene. Strain abbreviations are given in parentheses (see Table 3, footnote a). Values for the two groups of three proteins that share significantly higher percentages of similarity in amino acid and encoding gene sequences are in bold type. 

Alignment of the HifE sequences (Fig. 4) demonstrated similarities in the C-terminal region, including 14 identical amino acids at the end of each molecule. The C terminus of HifE is homologous to those of HifA and HifD and of other pilins from Pap and type I pili (9, 16). This region is required for binding of the pilus-specific periplasmic chaperone, which is necessary for correct assembly of the pilus fibers (33, 42). In contrast to the C terminus, the N-terminal two-thirds of the proteins had little conserved sequence. Particularly striking were the predicted signal peptides (Fig. 4), which are distinctive to each of the three groups of HifE arranged by overall sequence similarity. Despite extensive sequence diversity, the N-terminal portion included a series of short, highly conserved amino acid motifs (Fig. 4). A search of the SWISSPROT database revealed that these conserved motifs in the N terminus of HifE aligned with nearly identical spacing to conserved domains in the N terminus of PilC, the putative Neisseria pilus-associated adhesin (23) (Fig. 5). This result raises the possibility that the conserved motifs might be necessary for recognition of a mammalian cell receptor common to Neisseria and Haemophilus. The HifE motifs did not match the motifs of any other proteins in the database, including HifA and HifD sequences. No other regions of the PilC proteins, with an average size of about 1,040 amino acids, showed significant homology to HifE.

FIG. 4.

Alignment of H. influenzae HifE amino acid sequences. The amino acid sequences of Hib strain 770235 and BPF-associated strain F3031 were published by van Ham et al. (39) and Read et al. (24), respectively. The sequence of NTHI strain 86-025 was obtained from GenBank (accession no. U19730). Other sequences were reported in this study. Residues conserved between the seven proteins are shown in the last line each block (consensus). Dashes indicated gaps introduced by the alignment program. The highly conserved C-terminal region of the HifE proteins following the last cytosine residue is shaded. Predicted signal peptidase cleavage sites are illustrated with an arrow (↑). The three groups of variant signal peptides are boxed separately. Three conserved domains in the N terminus of the proteins with similarities to the Neisseria PilC sequence are underlined. Residues in these domains similar to the PilC consensus are shown in bold type.

FIG. 5.

Comparison of HifE domains with Neisseria PilC proteins. The alignment of PilC proteins is taken from the study by Rahman et al. (23). A portion of the N terminus of each of six PilC proteins from residues 124 to 270 is aligned against the three HifE conserved regions underlined in Fig. 4. PilC residues in bold are conserved within all six proteins and share similarity to the HifE domain. In two cases phenylalanine residues (f) are typed in lower case where they match tyrosine residues in PilC. N. meningitidis (NM) FAM120 PilC1, FAM120 PilC2, and A1493 PilC (24, 29) have been assigned EMBL accession numbers Y13020, Y13021, and Z54202, respectively. N. gonorrhoeae (NG) MS11 PilC1, MS11 PilC2 and GC-640 PilC (13, 24, 28) have been assigned EMBL accession numbers Z50180, Z49120, and Z54202, respectively.

DISCUSSION

Several findings about H. influenzae pili emerge from the data presented in this paper. First, the presence of two complete hif gene complexes and the hifE_F3031 allele are common in H. influenzae biogroup aegyptius and not just limited to BPF-associated strains. Second, many H. influenzae strains contain partial sets of hif genes, or do not appear to have any genes, at the hif1 and hif2 chromosome location. In this study, a broad range of H. influenzae strains, including biogroup aegyptius, NTHI, and capsule serotypes a, b and c, were found to carry complete hif pilus clusters. Similar host ranges were reported in previous surveys with Hib or NTHI hifA probes (2, 7, 8, 14). A third finding is that HifE protein sequences can be classed into at least three subfamilies, with members sharing at most 50% identity with proteins from other groups and with each subfamily having a specific N-terminal signal sequence. Finally, we identified conserved motifs in the N-terminal portion of HifE that match conserved domains in the N terminus of the neisserial pilin PilC.

Previous work (24, 35) showed that Brazilian BPF isolates contained two sets of pilus genes. As abundant piliation is a feature of BPF-associated H. influenzae isolated during the systemic phase of infection (1), we speculated that the duplicated pilus configuration was an important adaptation to pathogenesis. This survey indicates that several non-BPF-associated H. influenzae biogroup aegyptius strains, as well as BPF-associated strain F3031, carry two complete pilus complexes. Besides aegyptius strains, no other group of H. influenzae tested contained two entire sets of hif genes. It seems, therefore, that two pilus complexes are a common feature of H. influenzae biogroup aegyptius but not specific to BPF-causing biogroup aegyptius. The pilus cluster duplication event might have occurred in an ancestor common to many H. influenzae biogroup aegyptius isolates. Identifying evolutionary relationships between H. influenzae bacteria containing hif1 might provide information on the genetic determinants responsible for the enhanced pathogenesis of the BPF-associated strain. The two pilus complexes might also be an adaptation to conjunctival colonization or infection. If piliation is important for BPF infection, then the regulation of expression of the hif genes in addition to pilus gene copy number is important. It is notable that “Pradopolis” strains F4931 and F4933, related to F3031 by enzyme isotyping and shown by this study to contain two complete clusters, did not elaborate hemagglutinating pilus under laboratory conditions (22).

Some H. influenzae strains in this study appear to contain hifA or hifE homologs in truncated copies of the fimbrial cluster as well as complete pilus complexes. Strains F8835 (NTHI), GA2078 (serotype a) and ATCC 9007 (serotype c) are templates for the hifA1-specific primer pair but not for the hifE1 primers. These strains contain only one hifC homolog (Fig. 2) nd appear to have a complete hif2 complex due to amplification with hifA2 and hifE2 primers, hence the possibility they contain partial clusters with at least the hifA1 gene but not hifC1. Australian H. influenzae biogroup aegyptius strain F6422 appears to have two hifC genes by Southern analysis (Fig. 2), but the hifA1-specific primers produce a smaller amplicon than do other strains and hifE1-specific PCR gives a negative result. Perhaps the hifE1-specific primers have too little homology to bind, or alternatively, F6422 contains an incomplete cluster in the hif1 location that includes portions of the hifA and hifC genes. Likewise, H. influenzae biogroup aegyptius strain ATCC 43806 might have an incomplete copy of the hif2 genes as hifA2 but not hifE2 PCR products are amplified. Lastly, serotype f strains GA840, GA4090, and GA4913 do not hybridize to the hifC_F3031 probe but produce amplicons with hifA2- and hifE2-specific primers (Table 3). The small size of the hifE2 PCR product (0.45 kb) suggests a deletion in this gene. These incomplete H. influenzae pilin gene clusters may be nonfunctional remnants, or they might have one of two intriguing possible roles. The genes might express pilin subunits that can be assembled into the fimbriae encoded by the other cluster to promote variation in the surface-exposed protein. Alternatively, the pilin genes in the abbreviated pilus cluster might effect antigenic variation by recombining with the expressed loci of the second complex in a manner analogous to the silent pilin genes of Neisseria (26).

We used a simple RFLP-PCR method in order to establish relationships in the 21 hifE alleles identified by PCR. Because only a relatively small number of restriction sites were used for typing, the technique was more useful for detecting genes with very similar nucleotide sequences than for enabling accurate estimations of genetic distance. One striking result was the conservation of the hifE_F3031 allele among BPF- and non-BPF-associated H. influenzae biogroup aegyptius strains, indicating again that the F3031 pilus is not a characteristic unique to strains causing BPF. However, apart from the group similar to hifE_F3031, no other pairs of hifE genes showed significant similarity by this crude test. This variability was reflected in the nucleotide sequences of four hifE genes determined in this work and three already in the GenBank database. No gene had more than 80% similarity to another, and the majority shared only 40 to 55% identity (Table 4). Classification of the predicted amino acid sequences by relatedness revealed that the HifE from Hib is the sole member of one of three groups that shared less than 50% overall similarity. The other groups both included HifE from serotypes a and c and NTHI. It is interesting that H. influenzae biogroup aegyptius proteins are grouped with those of H. influenzae type c, as isolates of this capsule serotype have been aligned phylogenetically with the Brazilian BPF clone (18).

H. influenzae biogroup aegyptius strains F2066 and F4931 appear to have one copy of hifE_F3031 and a second highly divergent hifE allele, with similarities to genes from H. influenzae capsule serotypes a and c. In contrast, biogroup aegyptius strains F3031, F3118, and 43794 have copies of hifE that are indistinguishable by RFLP-PCR. Amino acid sequence divergence in HifE2_F2066 and HifE1_F4931 does not appear to have originated from localized recombination events between hifE genes as was the case, for example, for the H. influenzae genes encoding type 1 immunoglobulin A1 proteases (21). Inspection of aligned hifE_F3031, hifE1_F4931, and hifE2_F2066 revealed nucleotide substitutions, deletions, and insertions distributed more or less randomly through the divergent 5′ two-thirds of the sequences rather than concentrated in any specific portion of the genes (data not shown).

By comparing the divergent HifE sequences (Fig. 4), we identified conserved domains that might contribute to the function of the protein. In PapG and other related fimbrial adhesins, the N-terminal portion is necessary for binding to mammalian cell receptors (11, 32). The N terminus of HifE consists of a series of short, highly conserved domains interrupted by regions with little similarity (Fig. 4). That the HifE domains play some part in the recognition of mammalian cell receptors is suggested by their conspicuous similarity with portions of the Neisseria PilC protein (Fig. 5). PilC is a minor component of the type IV pilus that confers epithelial cell binding proficiency (19). Neisseria species commonly carry two genes, pilC1 and pilC2. Both copies function in pilus biogenesis, but only pilC1 is necessary for adhesion of N. meningitidis to certain human endothelial and epithelial cell lines (19). The exact role of PilC is still in question; the protein might act as a tip-associated adhesin (27) or, alternatively, might be located in the outer membrane and involved in processing of the pilus to an adhesive form (23).

This study provides interesting insights into the role of pilus in the biology of H. influenzae. Pili allow the bacterium to attach to certain epithelial cells (10, 29, 34), but it is unclear whether they are necessary for nasopharyngeal colonization (9). As colonization is vital for persistence of H. influenzae, the pilus might be a highly specialized structure, useful for exploiting a limited number of specific environments. Current information on H. influenzae pilus genetics adds weight to this idea. Fimbrial expression is phase variable, which implies that, in certain microenvironments at least, nonpiliated organisms survive in their human host. Also, the results of this study suggest that a significant portion of H. influenzae strains contain no hif-related pilus genes at all. Possibly there is flux in the H. influenzae population, with strains deleting pilus genes when there is no selective pressure and reacquiring them through natural transformation. Insertion of pilus clusters has been inferred from sequencing data: hif1 and hif2 genes are flanked by pairs of genes that are contiguous in the pilus-negative Rd strain and the junctions contain extended direct repeats and multiple copies of an H. influenzae repetitive extragenic palindrome-like element (24, 25, 38). This work also documents extensive variation of the hifE putative adhesin gene sequences. There are two likely reasons for this variability. First, genetic changes might generate variable epitopes on the surface-exposed protein, allowing H. influenzae to avoid the host immune response mounted against the pilus. The second explanation for variation of HifE is to allow attachment of the pilus to novel mammalian cell receptors. A structure-function relationship for HifE-mediated adhesion during infection could explain why certain hifE alleles are highly conserved among isolates from different locations and times, such as hifE_F3031 and hifE_Hib. Identification of conserved and variable domains in the N terminus of HifE and similarities to another putative adhesin in Neisseria is an important first step in dissecting the relationship between the primary sequence of the protein and function as an adhesin in disease-causing isolates of H. influenzae.