Immunologic and Structural Relationships of the Minor Pilus Subunits among Haemophilus influenzae Isolates

Abstract

Two proteins, HifD and HifE, have been identified as structural components of Haemophilus influenzae pili. Both are localized at the pilus tip, and HifE appears to mediate pilus adherence to host cells. In this study we examined the immunologic and structural diversity of these pilus subunits among type b H. influenzae (Hib) and nontypeable H. influenzae (NTHI) strains. Western immunoblot analysis revealed that antibodies directed against the C terminus of HifD and HifE from Hib strain Eagan bound to HifD and HifE proteins, respectively, of all piliated Hib and NTHI strains tested. Whole-cell enzyme-linked immunosorbent assays showed that antibodies specific for native HifD or HifE of strain Eagan also bound to all piliated Hib strains but did not bind to the piliated NTHI strains. Antibodies against HifE of strain Eagan inhibited the binding of Hib to human erythrocytes but did not inhibit the binding of NTHI strains. Restriction fragment length polymorphism (RFLP) analysis was used to determine strain-to-strain structural differences within hifD and hifE genes, either by PCR or by nucleotide sequence analysis. DNA and derived amino acid sequence analyses of HifD and HifE confirmed the uniqueness of the RFLP types. The hifD and hifE genes of all type b strains showed identical restriction patterns. Analysis of hifD and hifE genes from the NTHI strains, however, revealed seven unique RFLP patterns, suggesting that these genes encode proteins with diverse primary structures. These results indicate that HifD and HifE are immunologically and structurally similar among the Hib strains but vary among the NTHI strains.

Haemophilus influenzae is a gram-negative bacillus that causes a variety of infections in humans. In nonimmune children, H. influenzae possessing the type b capsule (Hib) causes bacteremia and subsequent serious invasive infections, including septic arthritis, cellulitis, and meningitis. Extensive use of the conjugated type b vaccine in children has significantly reduced the incidence of serious Hib infections in the United States (1). Although immunologically normal individuals rarely are infected with H. influenzae possessing any of the other capsular types, a or c to f, they may develop localized respiratory infections, such as pneumonia, bronchitis, otitis media, or sinusitis, caused by nonencapsulated, nontypeable H. influenzae (NTHI).

The first step in the pathogenesis of both local respiratory and invasive H. influenzae infections is asymptomatic colonization of the nasopharynx. Both encapsulated and nonencapsulated strains express adhesive pili that mediate H. influenzae binding to human erythrocytes (hemagglutination) (8) and to specific human epithelial cells (13, 31), thus promoting respiratory tract colonization (19, 34).

H. influenzae pili from both typeable and NTHI strains are polymeric structures that contain a major structural subunit called pilin, or HifA (14) and two minor subunits, HifD and HifE, which are localized at the distal ends of pili (21, 29). The ability of antibodies directed against HifE to inhibit hemagglutination suggests that HifE contains the pilus binding domain and mediates pilus adherence (21, 29). The function of HifD is unknown. The decreased pilus expression in hifD mutant strains (20, 32) and the presence of HifD at the pilus tip, however, suggest that HifD acts as an initiator of pilus assembly in a manner analogous to the well-described fibrillar tip proteins of P and type 1 pili (29) or as a linker that binds HifE to the assembled pili.

Genes encoding HifD and HifE are tandemly arranged in and cotranscribed within a gene cluster responsible for pilus expression (20, 32). This gene cluster also contains genes encoding HifA, the major pilus structural component; HifB, a periplasmic pilus subunit chaperone protein (29); and HifC, an outer membrane protein proposed to direct the ordered assembly of pilus subunits from HifB in the periplasm to the final pilus structure on the cell surface (33).

H. influenzae pili show strain-to-strain immunologic diversity, and 14 different pilus serotypes have been described (2). The antibodies generated by immunization with pili bind to conformation-dependent epitopes (8, 11) on the polymerized HifA subunits (29). The diversity of these epitopes is reflected in the variable primary structures of the HifA subunits. To date, pilin (HifA) genes from 19 typeable H. influenzae and NTHI strains have been cloned and sequenced, and the deduced sequences have between 59 and 100% amino acid identity (4–6, 25, 28).

In contrast to HifA, the immunologic and structural relationships of the minor pilus components (HifD and HifE) among various H. influenzae strains have not been well described. In this study, we used immunologic and genetic analyses to assess the relatedness of the HifD and HifE proteins among Hib and NTHI strains. We found that HifD and HifE proteins were very similar among Hib strains but comparatively diverse among NTHI strains. In addition, the binding of all Hib strains to erythrocytes (as measured by hemagglutination) was strongly inhibited by polyclonal antibodies directed against native HifE protein but not by antibodies directed against HifD.

MATERIALS AND METHODS

Bacterial strains and growth media.

Strains AAr149, AAr60, and AAr64 were obtained from the University of Michigan Hospital diagnostic microbiology laboratories. All other H. influenzae strains used in this study have been described previously (2, 8, 9, 12). For the immunologic assays, 12 Hib and 13 NTHI test strains were analyzed and compared to piliated (p⁺) and nonpiliated (p⁻) variants of the control Hib strain Eagan. All 12 Hib test strains were piliated, and nonpiliated phase variants were available for study from 8 of them (Table 1). Of the 13 NTHI strains, 11 were piliated, and nonpiliated variants were available for study from all 13 (Table 1). As previously described, a piliated phenotype of H. influenzae is defined by a hemagglutination titer greater than 1:4 and by the presence of pilin (HifA) on Western immunoblot analysis with anti-pilin antiserum (8).

TABLE 1.

Piliated variants and hifD and hifE RFLP patterns of the H. influenzae strains used in this study

Hib strain	RFLP pattern	Source or reference		NTHI strain	RFLP pattern	Source or reference
E1a	i	11		AAr39	iii	7
AA13	i	11		AAr45	iii	7
AA61	i	11		AAr49	iv	7
AAr2	i	11		AAr60	iii	This study
AAr7	i	11		AAr64	i	This study
AAr68	i	11		AAr73	ii	7
AAr149	i	This study		AAr91	iii	7
C54	i	11		AAr100	iv	7
D9	i	11		AAr160	ii	7
M9	i	11		AAr169	ii	7
M43	i	11		AAr176	ii	7
R9	i	11		M37	v	8
SL4	i	11		Mr13	NAa	7
				1712	vi	7
				86-0295	vii	2
				F3031	viii	24

^aNA, not available. 

H. influenzae strains were grown on Levinthal agar containing brain heart infusion base (Difco, Detroit, Mich.) supplemented with lysed horse blood and NAD (22). Escherichia coli DH5α was used as a recombinant host strain and was grown in Luria broth or agar containing 100 μg of carbenicillin per ml. The growth and induction conditions used for recombinant E. coli containing expression constructs have been described previously (21).

Genetic analyses and DNA constructs.

PCR was used to amplify the tandemly linked hifD and hifE genes from genomic H. influenzae DNA. The primers used for PCR were chosen from analysis of the Eagan pilus gene cluster (20, 33) and represented sequences just 5′ of the hifD gene (5′-CCCCCTGCAGTTGATTTGGATAATGC) and within the hifE gene at the 3′ end (including the stop sequence) (5′CCCCTGCAGTTATTGATATGACATTG). The primers were synthesized at the University of Michigan Biomedical Research Core Facility and constructed to contain PstI sites. Deep Vent DNA polymerase (New England Biolabs, Inc., Beverly, Mass.) and deoxynucleotides (Pharmacia Biotech, Piscataway, N.J.) were used as recommended by the manufacturer. PCR products were analyzed on 1% agarose gels to ascertain their size and purity. DraI and AluI restriction enzymes were used for restriction fragment length polymorphism (RFLP) analysis of the PCR products on 1.2% agarose gels. Representative PCR products from unique RFLP groups (see Results) were cloned into the pGEM-T cloning vector (Promega, Madison, Wis.), and DNA sequence analysis was performed at the University of Michigan Biomedical Research Core Facility by primer walking. Nucleotide and amino acid sequences were analyzed with Lasergene Biocomputing software for the Macintosh (DNASTAR, Inc., Madison, Wis.) and the Wisconsin Package version 9.0 (Genetics Computer Group [GCG], Madison, Wis.).

To generate a hifD gene product for use as an antigen in antiserum production, the hifD gene of strain Eagan was PCR amplified and cloned into a pMAL-cR1 expression vector (New England Biolabs). The hifD gene was first PCR amplified with oligonucleotides containing PstI sites (forward, 5′CCCCTGCAGCTGCCTGCTTATGCCG; reverse, 5′CCCCTGCAGGTTATACTGCACTTG), and the resulting product was ligated directly into the pCR-Script SK(+) cloning vector (Stratagene, La Jolla, Calif.). The DNA insert of this intermediate construct was removed by digestion with PstI and subsequently ligated into pMAL-cRI cut with PstI. This construct was designated pMED₁ and contains a malE-hifD gene fusion that encodes maltose binding protein and 84% of the predicted mature HifD protein at the C terminus.

Generation of the hifE gene product and truncated hifD and hifE gene products has been described previously (21).

Antisera.

The gene product of pMED₁ (designated HifD₁) was expressed, affinity purified, and used to raise antiserum as described previously (21). Additional antisera used in this study have been described previously (21). Briefly, HifD₂ and HifE₂ antisera were raised against truncated hifD and hifE gene products, respectively, of Hib strain Eagan; they represent approximately 40% of each mature protein at the C terminus. HifD₂ and HifE₂ antibodies bind to their respective HifD and HifE pilus subunit proteins in a Western immunoblot assay but do not bind to mature pili on the piliated, homologous Hib strain Eagan in an enzyme-linked immunosorbent assay (ELISA). HifE₁ antiserum was made against a recombinant HifE protein representing 99% of the mature HifE pilus subunit. HifE₁ antibodies bind to native HifE on pili of the homologous Hib strain Eagan (as shown by immunogold electron microscopy) but do not bind to linear epitopes on denatured pilus proteins.

Immunologic analyses.

Western immunoblot analysis (12) of Hi cell lysates was used to determine the immunologic cross-reactivity of denatured HifD and HifE subunits among H. influenzae strains. Whole-cell ELISA (10) was used to assess the immunologic cross-reactivity of native HifD and HifE proteins. HifD₁ and HifE₁ antisera, diluted 1:500 with phosphate-buffered saline (PBS), were used in these assays. The ability of the antibodies in each antiserum to bind to a given strain was assessed by dividing the optical density measurements at 560 nm (OD₅₆₀) from piliated phase variants by those from homologous nonpiliated phase variants (OD₅₆₀ p⁺/OD₅₆₀ p⁻). For piliated Hib strains without a nonpiliated phase variant (Hib strains C54, D9, M9, R9), comparisons to the averaged absorbance values of all nonpiliated Hib strains within an experiment were made. All the strains were assayed two or three times, and mean values are reported.

Hemagglutination inhibition assays.

The ability of antibodies to block pilus-mediated H. influenzae binding to erythrocytes was measured by hemagglutination inhibition assays as described previously (21). Briefly, H. influenzae bacterial suspensions were preincubated for 1 h at room temperature in PBS or with HifD₁ or HifE₁ antiserum diluted 1:10 in PBS. The bacteria were then centrifuged and resuspended in the original volume of PBS, and standard hemagglutination assays were performed with twofold serial dilutions of the treated bacteria (17).

Immunoelectron microscopy.

Transmission electron microscopy of antibody-treated H. influenzae was performed as previously described (20).

Statistical analysis.

The t tests on unpaired samples and χ² test were performed with the SYSTAT 7 for Windows program. P < 0.05 was considered to represent significant differences.

Nucleotide sequence accession numbers.

The hifD and hifE nucleotide sequences reported in this study have been submitted to the GenBank data bank with accession no. AF045062 (AAr73), AF045063 (AAr49), and AF045064 (AAr91).

RESULTS

Immunologic conservation of epitopes on denatured HifD and HifE.

The immunologic relationships of HifD and HifE expressed by Hib and NTHI strains were initially examined by a Western immunoblot assay with antisera specific for truncated HifD and HifE subunits of Hib strain Eaganp⁺ (HifD₂ and HifE₂ antisera, respectively) (21). These antisera bind to epitopes on denatured pilus subunit proteins.

The HifD₂ antibodies bound to a ∼22-kDa band in the piliated Eagan (homologous) strain and to each of the other 12 piliated Hib strains (Fig. 1, panel A1). These bands were not present in nonpiliated Eagan strains (panel A1) or in eight nonpiliated phase variant Hib strains (data not shown). The HifD₂ antibodies also bound to all 11 piliated NTHI strains, with reactive bands ranging between ∼20.5 and 22 kDa. These reactive bands were not present in nonpiliated NTHI strains (panels A2 and A3).

FIG. 1.

Western immunoblot reactivity of HifD₂ and HifE₂ antisera to H. influenzae strains. (A) Reactivity of HifD₂ antisera; (B) reactivity of HifE₂ antisera. Hib strains Eaganp⁺ and Eaganp⁻ are control strains (lanes 1 and 2, respectively, of each panel). Hib strains in panels A1 and B1 are as follows: lane 3, AA13p⁺; lane 4, AA61p⁺; lane 5, AAr2p⁺; lane 6, AAr7p⁺; lane 7, AAr68p⁺; lane 8, AAr149p⁺; lane 9, C54p⁺; lane 10, D9p⁺; lane 11, M9p⁺; lane 12, M43p⁺; lane 13, R9p⁺. NTHI strains are as follows: in panel A2, lane 3, AAr64p⁺; lane 4, AAr64p⁻; lane 5, AAr73p⁺; lane 6, AAr73p⁻; lane 7, AAr169p⁺; lane 8, AAr169p⁻; lane 9, AAr45p⁺; lane 10, AAr45p⁻; lane 11, AAr91p⁺; lane 12, AAr91p⁻; in panel A3, lane 3, AAr49p⁺; lane 4, AAr49p⁻; lane 5, AAr100p⁺; lane 6, AAr100p⁻; lane 7, Mr13p⁺; lane 8, Mr13p⁻; lane 9, M37p⁺; lane 10, M37p⁻. NTHI strains in panel B2 are as follows: lane 3, AAr64p⁺; lane 4, AAr64p⁻, lane 5, AAr73p⁺; lane 6, AAr73p⁻; lane 7, AAr91p⁺; lane 8, AAr91p⁻; lane 9, AAr100p⁺; lane 10, AAr100p⁻; lane 11, M37p⁺; lane 12, M37p⁻. Molecular mass markers are shown on the left of each panel and the pilus subunit names are shown on the right.

The HifE₂ antibodies bound to ∼43-kDa bands from the piliated Eagan (homologous) strain as well as from all other piliated Hib strains (Fig. 1, panel B1). These antibodies also bound to bands with molecular masses ranging from ∼37 to 43 kDa in the piliated NTHI strains (panel B2 [contains representative NTHI strains]). These immunoreactive molecules were not present in nonpiliated Hib strains (data not shown) or in nonpiliated NTHI phase variant strains (panel B2). The Western immunoblot results therefore suggest that both HifD and HifE subunits are expressed by the piliated Hib and NTHI strains used in this study. In addition, the results show that common linear epitopes within HifD and HifE are revealed when the subunit proteins from a variety of H. influenzae strains are denatured.

Immunologic diversity of native HifD and HifE between H. influenzae strains.

HifD₁ antiserum was produced in this study by inoculating rabbits with an affinity-purified, recombinant HifD protein representing 84% of the mature HifD subunit of piliated Hib strain Eagan. In whole-cell ELISA, the polyclonal HifD₁ antiserum showed greater reactivity (P < 0.001) to piliated Eagan (the homologous strain) than to nonpiliated Eagan or to a nonpiliated hifD insertion mutant (E114) (20) (OD₅₆₀ = 0.199 ± 0.006, 0.071 ± 0.007, and 0.074 ± 0.007 [mean ± standard deviation], respectively). Preimmune antiserum showed minimal reactivity to both piliated and nonpiliated Eagan (OD₅₆₀ = 0.080 ± 0.003 and 0.079 ± 0.005, respectively [P > 0.1]). Immunoelectron microscopy confirmed that HifD₁ antiserum reacted specifically with pili of Eagan (data not shown) and that antibodies to HifD bound predominantly to the pilus tips. These results suggest that HifD₁ antibodies specifically bind to native HifD in the pili of Hib strain Eagan and corroborate similar findings by St. Geme et al. (29).

To examine the possibility that the native minor pilus subunits, HifD and HifE, possess common epitopes among H. influenzae strains, we performed whole-cell ELISA to assess the binding of HifD₁ and HifE₁ antisera to piliated Hib and NTHI. In these assays, HifD₁ antiserum showed greater reactivity to the 12 piliated Hib isolates than to their nonpiliated phase variants (OD₅₆₀ = 0.369 and 0.220 [mean], respectively [P < 0.001]). This antiserum, however, showed no significant difference in reactivity between piliated and nonpiliated variants of nine NTHI strains (OD₅₆₀ = 0.602 and 0.596 [mean] respectively [P = 0.602]); this moderate level of binding is nonspecific and independent of pili. These data are further presented in Table 2, which shows the ratio of the OD₅₆₀ values of piliated variants compared to nonpiliated variants, i.e., OD₅₆₀ p⁺/OD₅₆₀ p⁻, where values significantly greater than 1 suggest specific binding to pili.

TABLE 2.

Reactivity of HifD and HifE antisera with H. influenzae strains^a

Hib strain	Reactivity of antiserum toa:			NTHI strain	Reactivity of antiserum toa:
	HifD1	HifE1			HifD1	HifE1
E1a control	1.805	2.013		AAr45	0.926	0.925
AA13	1.360	3.146		AAr49	1.007	0.820
AA61	1.180	3.104		AAr64	NDb	2.649
AAr2	1.546	3.462		AAr73	0.981	0.878
AAr7	1.626	3.010		AAr91	1.002	0.921
AAr68	2.332	4.522		AAr100	0.936	0.823
AAr149	1.685	4.056		AAr160	1.067	0.862
C54	1.806	3.057		AAr169	1.034	0.952
D9	2.448	4.675		AAr176	ND	0.590
M9	2.448	5.216		M37	1.095	1.041
M43	1.407	1.421		Mr13	1.037	1.106
R9	1.462	2.159
SL4	2.007	4.467

^aReactivity is expressed as OD₅₆₀ of piliated variant/OD₅₆₀ of nonpiliated variant (see the text). 

^bND, not done. 

Similarly, the HifE₁ antiserum, directed against 99% of recombinant HifE protein and previously shown to bind to the native HifE subunit of Eagan pili (21), showed greater reactivity with piliated Hib isolates than with the nonpiliated phase variants (OD₅₆₀ = 0.452 and 0.132 [mean] respectively [P < 0.001]). Anti-HifE₁, however, showed similar, low-level reactivity with both piliated and nonpiliated variants of NTHI (OD₅₆₀ = 0.231 and 0.230 [mean], respectively [P = 0.938]), suggesting nonspecific binding of this antiserum that is independent of pili. Thus, HifEs of Hib strains appear to be immunologically similar and HifEs of NTHI strains are immunologically dissimilar to those of Hib strains.

Table 2 shows that the results from NTHI strain AAr64 are discordant from those obtained with the other strains. Analysis of the outer membrane proteins of this strain by sodium dodecyl sulfate-polyacrylamide gel electrophoresis revealed a profile more like Hib strains than like NTHI strains (15). Thus, strain AAr64 appears to be a Hib strain that no longer produces capsule. These data suggest that immunologically conserved epitopes are present on both HifD and HifE among Hib strains but that these epitopes are absent or not detectable on NTHI strains.

Anti-HifD and -HifE inhibition of hemagglutination.

Previous studies demonstrated that HifE₁ antibodies block pilus-mediated H. influenzae binding to erythrocytes, suggesting that the receptor binding domain is located on HifE (21). To extend this observation, we tested the ability of both the HifD₁ and HifE₁ antisera to inhibit pilus-mediated hemagglutination by the Hib and NTHI strains used in this study.

The HifD₁ antiserum did not impair agglutination of human erythrocytes by the homologous strain Eagan p⁺ or by the other Hib strains (Table 3) (P = 0.082). In addition, HifD₁ antiserum did not inhibit hemagglutination by the piliated NTHI strains (P > 0.343). In contrast, the HifE₁ antiserum significantly reduced the hemagglutination titer of piliated Hib strains (P < 0.001) (Table 3) but did not reduce the hemagglutination of piliated NTHI strains (P = 0.343); again, strain AAr64, which has an outer membrane protein profile similar to that of Hib strains, is the exception. These results suggest that HifE, rather than HifD, mediates the hemagglutinating property of H. influenzae strains, and they further support the conclusion that HifE subunits among the Hib strains used in this study are immunologically related.

TABLE 3.

Effects of HifD₁ and HifE₁ antisera on H. influenzae hemagglutination titers

Hib strain	Titer with antiserum to:				NTHI strain	Titer with antiserum to:
	None	HifD1	HifE1			None	HifD1	HifE1
E1a control	1:16	1:16	<1:1
AA13	1:8	1:4	<1:1		AAr45	1:32	1:32	1:32
AA61	1:16	1:16	<1:1		AAr49	1:16	1:16	1:16
AAr2	1:8	1:8	<1:1		AAr64	1:16	1:16	<1:1
AAr7	1:32	1:16	1:1		AAr73	1:16	1:16	1:16
AAr68	1:8	1:8	<1:1		AAr91	1:32	1:32	1:32
AAr149	1:8	1:8	<1:1		AAr100	1:16	1:16	1:16
C54	1:16	1:16	<1:1		AAr160	1:4	1:4	1:4
D9	1:16	1:8	<1:1		AAr169	1:16	1:16	1:16
M9	1:16	1:16	<1:1		AAr176	NDa	ND	ND
M43	1:16	1:16	<1:1		M37	1:16	1:16	1:16
R9	1:32	1:32	<1:1		Mr13	1:16	1:16	1:16
SL4	1:16	1:16	<1:1

^aND, not done. 

RFLP analysis of hifD-hifE PCR products.

DNA sequence variations within a gene dictate differences in the primary structure of the translated protein. Therefore, we initially assessed the structural variations of HifD and HifE subunits among H. influenzae isolates by comparing the RFLP characteristics of PCR products containing the tandem hifD and hifE genes.

PCR products were obtained from all 12 Hib strains and 12 of the 13 NTHI strains (Table 1). All the PCR products were approximately 2 kb, the size expected from published hifD and hifE sequences (20, 25, 32). RFLP analysis with two different restriction enzymes (AluI and DraI) revealed identical restriction patterns in the PCR products of the Hib Eagan control strain, all 12 Hib strains, and 1 NTHI strain (AAr64) (Fig. 2), and this pattern was designated RFLP pattern i (Table 1). Seven additional RFLP patterns (designated ii to viii) were identified among the other 11 NTHI PCR products and from published HifD and HifE nucleotide sequences (Table 1). These results suggest that the nucleotide sequences encoding HifD and HifE are similar among Hib strains but more diverse among NTHI strains.

FIG. 2.

HifD/E patterns of selected H. influenzae strains. DNA in lanes 1 to 5 was digested with AluI, and DNA in lanes 7 to 11 was digested with DraI. Lanes: 1 and 7, DNA from strain Eagan; 2 and 8, DNA from AAr169; 3 and 9, DNA from AAr45; 4 and 10, DNA from AAr49; 5 and 11, DNA from 1712; 6 lambda phage DNA digested with PstI as a size marker.

Previous work from our laboratory defined RFLP groups among nucleotide sequences encoding HifA, the major pilus subunit protein, from a variety of H. influenzae strains, including those used in this study (4). Analysis of the HifA and HifD/E RFLP patterns of these strains failed to show good correlation (χ² = 5.231; 0.20 < P < 0.30). Thus, the HifA and HifD/E RFLP patterns were not coordinately distributed in these strains.

Comparative analysis of the primary structures of Hib and NTHI minor subunits.

The HifD and HifE amino acid sequences have been deduced from the cloned genes of two Hib isolates Eagan (20) and AM30 (32). The DNA sequences encoding these proteins are 99.9% identical (there is a 1-bp difference between the hifE genes), and their amino acid sequences are 100% identical (Fig. 3 and 4). These observations corroborate the RFLP results for the Hib strains (above), suggesting that the HifD and HifE pilus components of Hib strains are structurally similar, if not identical.

FIG. 3.

Amino acid sequence analysis of HifD. The consensus amino acids are 100% conserved across strains. The consensus was performed with the Pileup program of the Wisconsin Package, version 9.0, from GCG.

FIG. 4.

Amino acid sequence analysis of HifE. The consensus amino acids are 100% conserved across strains. The consensus was performed with the Pileup program of the Wisconsin Package, version 9.0, from GCG.

In addition, the HifD and HifE amino acid sequences have been reported for two NTHI strains, 86-0295 (the prototype strain for Lkp1 pili [2]) (GenBank accession no. U19730) and the Brazilian purpuric fever strain F3031 (25). To further detail the structural differences of the minor pilus components possessed by NTHI strains, the hifD- and hifE-containing PCR products from strains AAr73, AAr91, and AAr49 (representing RFLP patterns ii, iii, and iv, respectively) were cloned and the DNA sequences of the inserts were determined. Two open reading frames with DNA sequence similarity to the hifD and hifE genes of strain Eagan were found within each insert. HifD and HifE amino acid sequences were derived from each set of DNA sequence and compared to previously published HifD and HifE sequences (Fig. 3 and 4). The number of residues in each protein, the lengths of the proposed signal sequences, and the predicted molecular mass and pI for each proposed mature protein are shown in Table 4.

TABLE 4.

Structural properties of H. influenzae HifD and HifE

Subunit and strain	No. of residues	No. of amino acids in signal sequence	Molecular mass (kDa)a	pIa
HifD
E1a	216	18	20.6	8.69
AAr73	209	18	20.1	9.25
AAr91	209	18	20.0	9.28
AAr49	203	18	19.5	8.19
86-0295	205	18	19.6	9.20
F3031	221	18	21.3	8.87
HifE
E1a	435	29	45.5	9.26
AAr73	431	26	45.1	9.17
AAr91	437	30	44.8	9.02
AAr49	434	29	45.1	8.89
86-0295	437	30	44.8	9.27
F3031	430	31	44.7	9.34

^aCalculations are for the predicted mature protein based on derived amino acid sequences, using Lasergene Biocomputing software for the Macintosh from DNASTAR, Inc. 

Table 4 shows that HifD proteins from a number of H. influenzae strains and one H. influenzae biogroup aegyptius strain are similar in length (203 to 221 residues), predicted molecular mass (19.5 to 21.3 kDa), and predicted pI (8.19 to 9.25). Pairwise comparisons between the predicted mature HifD proteins (Table 5) demonstrate that their amino acid sequences are 79 to 92% identical. The HifE proteins of the same strains are also similar in length (430 to 437 residues), predicted molecular mass (44.7 to 45.5 kDa), and predicted pI (8.89 to 9.34). The amino acid sequences of the HifE proteins have 43 to 69% identity among H. influenzae strains (Table 5). Further analysis of the HifE amino acid sequences revealed that the N-terminal two-thirds of the proteins are more diverse than their C-terminal one-third (overall average identity, 38.3 and 70.71%, respectively).

TABLE 5.

Pairwise comparisons^a of H. influenzae HifD and HifE

Subunit and strain	% amino acid identity for strain:
	AAr73	AAr91	AAr49	86-0295	F3031
HifD
E1a	84	80	79	81	92
AAr73		86	80	80	85
AAr91			74	70	79
AAr49				84	79
86-0295					82
HifE
E1a	50	54	68	52	49
AAr73		43	48	42	63
AAr91			52	69	45
AAr49				53	46
86-0295					43

^aAmino acid comparisons were performed with the BestFit program of the Wisconsin Package, version 9.0, from GCG. 

Comparison of the HifE sequences with those of other known bacterial proteins reveals a region (residues 155 to 171 in Fig. 3) with 58% amino acid identity to PilC of both Neisseria meningitidis and N. gonorrhoeae, a protein that is important in the biogenesis of Neisseria pili and appears to contain an epithelial cell binding domain (16, 26).

DISCUSSION

Antigenic diversity is a common theme among surface molecules of H. influenzae (7a). Accordingly, previous studies have shown that antibodies directed against assembled pili (2, 12) bind to the assembled pili of some, but certainly not all, H. influenzae strains. Recent molecular studies (4, 6) have demonstrated that the primary structure of HifA subunits from diverse H. influenzae isolates, both Hib and NTHI strains, vary significantly (59 to 100% amino acid identity). In contrast, our results demonstrate that among Hib strains, minor pilus components are immunologically and structurally similar; antibodies against HifD and HifE of Hib strain Eagan bound to the pili of all Hib strains analyzed in our study, and the hifE and hifD genes of all 13 Hib strains fell within the same RFLP pattern. Furthermore, the HifD and HifE amino acid sequences of the two Hib strains sequenced (Eagan and M43) were identical, even though their HifA sequences are only 80% identical (4).

Among NTHI strains, however, the minor pilus components, HifD and HifE, showed considerable immunologic and structural diversity. Antibodies against HifD and HifE of Hib strain Eagan did not bind well to any of the NTHI strains, and their deduced amino acid sequences showed only moderate similarity, with the exception of strain AAr64, the Hib strain that has lost its capsule.

The differences between minor pilus subunits from either Hib or NTHI strains is further magnified by the distant geographic relationships of the isolates. For instance, 13 Hib strains isolated from patients in seven different states (12) yielded hifD/hifE PCR products with identical RFLP patterns whereas 10 NTHI strains isolated from patients seen at the University of Michigan (8) possessed hifD and hifE genes of three different RFLP groups. This observation may reflect evolutionary patterns of H. influenzae, since Hib strains appear to be more clonally related than NTHI strains (24).

Although anti-HifD₁ and anti-HifE₁ bound to assembled pilus subunits of Hib strains but not of NTHI strains, as seen by ELISA, Western immunoblot analysis revealed conserved epitopes within HifD and HifE among both Hib and NTHI strains. These experiments were performed with HifD₂ and HifE₂ antisera that were made against truncated recombinant HifD and HifE proteins and most probably recognize linear epitopes within the denatured proteins. These antibodies, however, did not bind to assembled pili of the homologous Hib strain Eagan (21), suggesting that the conserved epitopes present in denatured HifD and HifE are not surface exposed or are conformationally altered when the subunits are assembled within the pilus. Similar results have also been observed in immunologic studies of HifA subunits of Hib and NTHI strains (8, 12).

The HifE₁ antibodies, which bound to native HifE of all Hib strains tested, effectively blocked pilus-mediated binding of the Hib strains to erythrocytes, as demonstrated by hemagglutination inhibition. In contrast, the HifD₁ antibodies, which also bound to native HifD of all Hib strains tested, did not significantly inhibit hemagglutination, suggesting that HifE, rather than HifD, mediates receptor binding. An alternative explanation for this observation is that the HifD₁ antiserum contains a limited repertoire of high-affinity antibodies, thus compromising the ability to detect hemagglutination inhibition.

Sequence analysis revealed that the HifD proteins of the strains studied were more similar to each other (79 to 92% identical) than were the HifE proteins (43 to 69% identical). The function of HifD is unknown, but its location at the pilus tip (29) and the failure of hifD mutants to express pili in the face of pilin expression (20) suggest that HifD may play a role in pilus assembly. As such, structural diversity of this protein may be limited by functional constraints, which would explain the high level of conservation of HifD. Similarly, the relatively low amino acid conservation seen among HifE proteins may reflect fewer functional constraints. Of note, both the higher amino acid conservation in the C-terminal one-third of HifE than in the N-terminal two-thirds and its homology to the C termini of both HifA and HifE are consistent with the role of this region in binding to HifB, the pilus chaperone (27). The conserved tyrosine and glycine, 2 and 14 residues, respectively, from the C terminus, are common in pilus structural components of other gram-negative organisms (20).

The discordance of the hifA and hifD/E RFLP patterns suggest “mixing and matching” of these genes, which are located within the same gene cluster. This is reminiscent of the poor correlation of IgA protease serotype with biotype (18, 23), again suggesting mixing and matching of IgA protease genes among strains with different genetic backgrounds. A possible mechanism that explains this mixing and matching is horizontal gene transfer (3), in which DNA from the environment is taken up by H. influenzae by natural transformation and incorporated into the bacterial genome by homologous recombination, similar to the mechanism that explains the mosaic-like genes of H. influenzae and Neisseria spp. Recent data showing high heterogenicity in the hifA and hifB intergenic regions of eight H. influenzae strains provide supporting evidence of horizontal gene transfer in the hif gene cluster (7).

We have previously demonstrated that pili of NTHI strains bind to the same erythrocyte receptors as do pili from Hib strains (8). The observation that Hib and NTHI pilus adhesins are functionally similar suggests that they possess common structural properties or motifs that mediate binding. Data from the present study, however, argue for the existence of complex and interactive, rather than simple, structural properties of HifE that mediate adherence. Examination of the HifE sequences reveals very few highly conserved peptide regions that would serve as candidate adhesive domains. Rather than linear peptide regions, these adhesive domains may be assembled from widely separated residues that are brought into continuity by protein folding. Furthermore, since HifE₁ antiserum does not inhibit the hemagglutination of NTHI strains, the polyclonal antibodies contained in this antiserum most probably do not recognize a putative adherence motif common to both Hib and NTHI strains. This may be due to loss of the binding motif in the recombinant HifE antigen, to a binding motif that is nonimmunogenic, or to the presence of a structurally modified but functionally similar binding motif within NTHI adhesins. The amino acid sequence similarity, although limited, between HifE and PilC is of interest. Since PilC appears to bind to the human complement regulatory protein MCP, or CD46, and HifE binds to sialic acid-containing lactosylceramide-like molecules (30), the significance of this similarity is unclear.

In conclusion, H. influenzae strains express minor pilus components that appear to be immunologically and structurally conserved among Hib strains but heterologous among NTHI strains. Further analysis of the HifE proteins among these strains may reveal regions that function as receptor binding domains.