Nontypeable Haemophilus influenzae in Carriage and Disease: A Difference in IgA1 Protease Activity Levels

Context Nontypeable Haemophilus influenzae strains form part of the normal flora of the human upper respiratory tract but are also implicated in a wide range of diseases. Infections caused by nontypeable H influenzae are major health and socioeconomic burdens. No single bacterial trait has been associated with disease as opposed to colonization.

Objectives To compare IgA1 protease activity in nontypeable H influenzae strains isolated from patients with symptomatic Haemophilus infection (sputum, cerebrospinal fluid, blood, or normally sterile tissue) vs strains from throat swabs of asymptomatic carriers and to compare iga gene carriage and variability in nontypeable H influenzae strains.

Design and Setting Retrospective study of 63 strains (44 clinical and 19 carriage) collected between 1991 and 2000 and maintained at the Public Health Laboratory, Gwynedd General Hospital, Bangor, Wales.

Main Outcome Measures Levels of IgA1 protease activity produced by carriage strains and clinical isolates from symptomatic patients; the determination of the size and sequence of a variable region of the iga gene.

Results Bacterial IgA1 protease activity was significantly higher (P<.001) in strains isolated from sputum, blood, cerebrospinal fluid, or normally sterile tissue of symptomatic individuals (median, 155 mU; interquartile range [IQR], 80-220 mU; mean, 169 mU; 95% confidence interval [CI], 126-211 mU) than in those isolated from throat swabs of asymptomatic carriers (median, 30 mU; IQR, 15-90 mU; mean, 56 mU; 95% CI, 26-86 mU; assayed on secretory IgA). The iga gene was detected in 97% of all strains examined. Variations in the sizes and sequences of part of the iga genes were also apparent. This variable region encodes a polypeptide linker connecting the protease domain to the β-core autotranslocator, a porelike structure required for secretion of the protease.

Conclusions These findings reveal the importance of iga gene variability and expression levels in the establishment of disease phenotype. They identify nontypeable H influenzae IgA1 protease as a virulence factor and as a potential target for the development of new strategies to fight these important pathogens.

Haemophilus influenzae causes a number of diseases in humans, including both localized and systemic infections.¹ Early work established that not all isolates of this gram-negative bacterium possess the same pathogenic potential. Lower respiratory tract infections caused by nontypeable H influenzae (NTHi) strains are responsible for significant mortality in both infants and children in developing countries.² They also represent a major cause of morbidity in both developed and developing countries. In general, carriers of NTHi are healthy but occasionally develop localized acute respiratory tract infections (eg, otitis media, sinusitis, pneumonia, and conjunctivitis). In addition, NTHi is associated with exacerbations of underlying lung disease (eg, chronic bronchitis, bronchiectasis, and cystic fibrosis). Less frequently, it has been reported to cause septicemia, endocarditis, epiglottitis, septic arthritis, and meningitis, illnesses more usually associated with H influenzae type b (Hib).³ Also, NTHi is reported as causal in female genital tract infection and postpartum and neonatal infection (including septicemia).⁴ Brazilian purpuric fever is the most serious disease exclusively caused by strains of NTHi.⁵

Attempts to identify a particular virulence factor responsible for NTHi pathogenesis have so far been unsuccessful. Evidence suggests that NTHi relies on more than 1 specific factor or mechanism to colonize the human nasopharyngeal tract and then, on occasion, to invade other sites and cause localized infections.⁶

Immune exclusion is the most important defense mechanism involved in protection of mucosal membranes.⁷ This is an essentially mechanical process in which secretory IgA (sIgA) agglutinates the colonizing bacteria, resulting in steric hindrance of the adhesin-epithelial receptor interaction. Once agglutinated, bacterial complexes, entrapped in mucus, are expelled by mucociliary clearance. The key element in this defense mechanism, IgA1 is the most abundant antibody associated with human upper respiratory tract mucosal surfaces.⁸

Isolates of NTHi produce specific enzymes able to cleave human IgA1.⁹ Bacterial IgA1 proteases (IgA1-specific postproline endopeptidases) are extracellular enzymes able to cleave human IgA1 exclusively within the hinge region of the α-heavy chain, thus separating the antigen recognition fragments (Fab) and constant region (Fc).¹⁰ This cleavage results in decoupling of the recognition function (mediated by Fab fragments) from effector functions (mediated by the Fc fragment) and ultimately in impedance of agglutination and mechanical clearance. Released Fab fragments still have the capacity to bind cognate antigen,¹¹ resulting in the masking of specific epitopes on the bacterial surface and prevention of subsequent recognition by intact antibodies.

The link between IgA1 protease production by bacterial mucosal pathogens and the key immunologic role of IgA1 suggests these enzymes may be virulence determinants in bacterial pathogenesis.¹² The literature contains conflicting reports on the importance of IgA1 proteases.^13,14 To date, no quantitative studies have been reported comparing levels of IgA1 protease in carriage and disease isolates of NTHi. Studies involving quantitative variations in any virulence factors are rare. The aim of this retrospective study was to compare the levels of IgA1 protease activity in individual strains of NTHi isolated from cases of clinical infection with those obtained from throat swabs of healthy individuals.

Strains of NTHi (44 clinical and 19 carriage, collected between 1991 and 2000) were randomly chosen from the collection maintained at the Public Health Laboratory, Gwynedd General Hospital, Bangor, Wales. Clinical strains (n = 44), isolated throughout Wales, were defined as those isolated from symptomatic patients diagnosed as having current infections, such as pneumonia (from blood samples), exacerbation of chronic obstructive pulmonary disease, bacteremia, meningitis, chest infection, subdural abscess, and ear infection. They included samples from purulent sputum (n = 20), blood (n = 18), cerebrospinal fluid (n = 3), and other normally sterile sites (n = 3). These strains were collected as a part of an ongoing epidemiologic survey begun in 1988 to evaluate the patterns of invasive disease before and after introduction of Hib vaccination. Carriage strains were defined as those isolated from throat swabs of healthy volunteer laboratory staff (n = 11) or preschool-aged children attending 4 local crèches (n = 8) with no recent history of respiratory disease. The strains used in this study are presented in Table 1 and Table 2. Strain 2509 is a nontypeable strain of unknown origin also isolated in this laboratory. Strains HK368, HK61, HK393, HK715, and Rd (all capsulated, except nontypeable HK61), used for comparative purposes, were described previously.^15-17

Clinical samples were isolated according to standard laboratory procedures. The sputum samples were microscopically examined for the presence of polymorphonuclear leukocytes and epithelial cells to establish their suitability before further analysis. A sample was considered bronchopulmonary in origin and suitable for culture if it contained less then 10 oropharyngeal squamous epithelial cells and 25 or more polymorphonuclear leukocytes per low-power magnification (×10) field. Samples of purulent or mucopurulent sputum (mucoid and salivary samples were excluded), submitted from patients with acute chest infection or exacerbation of chronic obstructive pulmonary airways disease, were vortexed with an equal volume of sterile phosphate-buffered saline, and the resultant emulsion was further diluted 1:100 with peptone water. Ten-microliter aliquots of each dilution (ie, 1:2 and 1:200) were cultured on blood agar and heated blood (chocolate) agar, containing 10 mg/L of bacitracin to inhibit normal gram-positive flora. Body fluids (subdural, cerebrospinal fluid, and vitreous humor) were centrifuged for 5 minutes at 3000 rpm and the deposit subcultured onto blood and chocolate agars. Blood culture isolates were derived from 5 to 10 mL of venous blood, inoculated into 50 to 100 mL of tryptone soya broth, and incubated for up to 5 days before subculture. All plates were incubated at 37°C in an atmosphere enhanced with 5% carbon dioxide for up to 48 hours. Throat swabs from healthy volunteers were inoculated onto chocolate agar containing 10 mg/L of bacitracin and incubated for 48 hours under the same conditions.

Nutritional requirements, hemolytic activity, porphyrin production, and serotyping were performed according to standard laboratory methods. A commercial identification system (API NH; BioMerieux Ltd, Marcy l'Etoile, France) was used to confirm the identity of the strains. This system includes detection of indole, urease, and ornithine decarboxylase production from which the biotype may be derived.¹⁸

IgA1 protease activities were measured using an enzyme-linked immunosorbent assay (ELISA) originally described by Reinholdt and Kilian¹⁹ and modified as previously described.²⁰ Briefly, a 10-mL culture was started by inoculating a loopful of bacteria from chocolate agar plates. Bacterial cultures were grown until late log phase, and absorbance at 550 nm (A₅₅₀) of each culture was monitored. A secondary 10-mL culture was started using 0.1 mL of the primary culture as an inoculum. This culture was grown to mid log phase (0.4-0.5 A₅₅₀). Four milliliters of each bacterial culture was subsequently filtered through a 0.22-mm disposable filter unit (Pall Gelman Laboratory Inc, Ann Arbor, Mich) and the filtrate used directly in the ELISA.

IgA1 protease substrates (purified IgA1, 3.2 µg/mL; Calbiochem, San Diego, Calif) or purified human colostrum IgA (3.2 µg/mL; Sigma, St Louis, Mo) were bound to the surface of a polystyrene microtitration plate (Immulon 2; Dynex Technologies, Chantilly, Va) through their Fab domain using rabbit anti–human λ light chain antibody (Dako, Glostrup, Denmark). Immobilized IgA substrates were exposed to filtered culture supernatants (150 µL) for 3 hours. Incubation of bound substrates with bacterial culture supernatants results in the release and loss of the Fcα region on washing and the retention of the Fab fragment. Loss of Fcα was detected indirectly through a reduced binding of peroxidase-conjugated rabbit anti–human Fcα antibody (Dako) as assayed with the chromogenic substrate o-phenylenediamine dihydrochloride (Sigma). The difference in absorbance at 490 nm (ΔA_490nm) between the undigested substrate molecules in control wells and the well containing bacterial supernatants was recorded for each bacterial isolate. Relative activity was calculated by dividing the ΔA_490nm value by the absorbance of the same culture (measured at 550 nm). One unit of enzyme activity was defined as the amount of enzyme able to effect a change in optical density of 1 absorbance unit (at 490 nm) in 1 hour at 37°C. Each bacterial strain was assayed in triplicate for both IgA1 and sIgA.

The nonparametric Mann-Whitney U test was used to analyze data, and the null hypothesis assumes that both groups have the same capacity to produce IgA1 protease.

We used the polymerase chain reaction (PCR) to examine strains for presence of the iga protease gene and to assess the genetic variability of this locus in individual isolates. Four sets of primers were used to identify 3 separate domains of the iga gene in each strain examined. All primers were derived from the published sequence of the H influenzae HK368 IgA1 protease gene (GenBank sequence database accession No. M87492)¹⁵: linker domain (LF, GTTCCACCACCTGCGCCTGCTAC and LR1, GTTTTCTCTGTTTCTACTTTAGC or LR2, GTTATATTGCCCCTCGTTATTCAT), mature protease (PF1, ACGCCGTGAAGACTACTATATG and PR1, CTCGTTGTTGATATGGTTCAT), and β-core domain (CF1, GCAGAATTCAAAGCACAATTTGTTGCA and CR1, TTATAACGTTAATTCAAACAGGCTT). The primers were designed to bind to positions 3334-3356, 3595-3617, 4123-4146, 741-762, 2182-2202, 4051-4077, and 4889-4913, respectively, of the HK368 sequence. CR1 was designed to overlap with the sequence immediately downstream from the iga gene. Genomic DNA was prepared by suspending a loopful of bacteria from a single plate derived from a single colony in 0.5 mL of 10-mM Tris hydrochloride (pH 8) buffer and boiling for 10 minutes. Amplification was performed using 1 U of AmpliTaqDNA polymerase (Perkin Elmer, Norwalk, Conn) and buffer supplied by the manufacturer supplemented with 250 µM of each deoxynucleoside triphosphate, 100 ng of each primer, and 5 µL of crude genomic DNA in a total volume of 50 µL. Negative controls were included using PCR mixtures lacking either bacterial DNA or primers. The PCR products were analyzed by electrophoresis of 10 mL of the amplification mixture on 1% agarose gel and detected by staining with ethidium bromide. Amplification of Haemophilus 16S ribosomal RNA genes was performed as a control for DNA concentration and quality for each strain before IgA1 protease gene domain detection using previously published primer sequences.²¹

Fragment sizing was performed by constructing calibration curves by plotting the distance traveled on the gel in millimeters against log₁₀ number of base pairs (bp) for each DNA size standard fragment. The size of the variable region fragment, in bp, was then calculated from the relative distances of migration through the gel. The PCR fragment sequences were determined using the dye termination method with an ABI Taq FS sequencing kit (Applied Biosystems, Cheshire, England) analyzed on an ABI 373 A Stretch automated sequencing machine (Applied Biosystems). DNA sequence analysis was performed using MacVector and AssemblyLIGN (International Biotechnologies, Cambridge, England) software packages.

Strains were accepted as H influenzae if they were gram-negative bacilli, hemin, and nicotine adenine dinucleotide dependent, nonhemolytic on horse blood agar, non–carbon dioxide requiring, and unable to produce porphyrin from δ-aminolevulinic acid. Serotype was determined by agglutination with anticapsular-type antiserum. The characteristics of the strains used in this study are presented in Table 1 and Table 2. Strains were further characterized using the API NH kit (BioMerieux) (data not shown). These results suggested that the isolated strains represent a diverse group, as expected for NTHi, and were not clonally related as are Hib strains.²²

The levels of IgA1 protease activity detected in individual strains are presented in Table 1 and Table 2 and a summary appears in Figure 1. Analysis of these data revealed a wide variation in detectable levels of enzyme activity. For example, a 60-fold higher level of enzyme activity was detected in clinical strain 77412 compared with carriage strain C12. IgA1 protease activity was detected in 39 of the 44 clinical strains. Using a myeloma-derived IgA1 substrate, high levels of protease activity were observed for the clinical isolates (median, 155 mU; mean, 169 mU [95% confidence interval (CI), 126-211 mU]). Similar results were obtained using sIgA substrate (median, 150 mU; mean, 166 mU [95% CI, 127-206 mU]). Measurable activity was detected in 15 of 19 carriage strains using either substrate (median, 30 mU; mean, 56 mU [95% CI, 26-86 mU]; median, 30 mU; mean, 49 mU [95% CI, 23-76 mU]). The differences between the levels of IgA1 protease activity between the clinical and carriage strains is statistically significant at the P< .001 level for both substrates using the Mann-Whitney U test. All protease-negative supernatants were incubated with substrate for extended periods (up to 12 hours) to detect low levels of activity but still failed to show detectable activity.

The presence of an iga gene was detected in 61 (97%) of the 63 strains examined (all clinical and 17 of 19 carriage), based on the positive amplification of at least 1 iga gene fragment (protease, linker, or β-core) using 4 different primer sets. Results of individual amplifications for each strain are expressed as an overall positive or negative score in Table 1 and Table 2. A positive signal for β-core domain was obtained in 86% (54 of 63) of strains and for protease domain in 67% (42 of 63) of strains. The PCR products corresponding to the linker region were detected in 43 of 44 clinical isolates but only 4 of 19 carriage isolates.

The β-core domain and protease domain fragments from the 63 strains had minimal variation in size, as judged by agarose gel electrophoresis (865 and 1464 bp ± 5%). Eleven clinical strains were used to more accurately demonstrate the size variability of the amplified linker fragment. The sizes of the PCR products of the amplified variable fragments using the same set of primers (LF1 and LR1) were between 200 and 1000 bp (210, 200, 440, 460, 235, 1005, 460, 480, 420, 613, and 231 bp in length for isolate numbers 1428, 1958, 2005, 5220, 6338, 6350, 7244, 7693, 8304, 8625, and 77688, respectively). The sequenced linker region iga PCR products of H influenzae strains 6338, 77688, 8625, and 2509 were composed of 235, 231, 613, and 749 nucleotides, respectively (GenBank database accession Nos. AF274859, AF274860, AF274861, and AF274862). Linker region sequences of iga genes from HK368, HK393, HK715, Rd, and HK61 were available from GenBank (accession Nos. X64357, M87490, M87489, U32779, and M87491, respectively). Percentage similarities between nucleotide sequences were generated using CLUSTAL W and results are presented in Table 3.²³ Comparison with available H influenzae linker domain sequences revealed varying levels of homology between equivalent domains of different strains (59%-100% identity on nucleotide level). Sequence similarities were much higher among encapsulated strains (96%-100%) compared with NTHi strains (59%-92%).

In this article, we provide quantitative evidence that suggests that IgA1 protease activity is an important virulence factor in NTHi-mediated infections. This finding parallels our previous report²⁰ showing increased levels of IgA1 protease activity in invasive strains of Neisseria meningitidis. Taken together, these 2 results indicate that IgA1 protease contributes significantly to the pathogenic potential of these bacteria.

The first step toward investigation of the potential role of IgA1 protease in infections caused by NTHi was to verify the presence of iga genes in all isolates by PCR amplification. Failure to detect a PCR product in some strains could be explained by the absence of a suitable complementary nucleotide sequence caused by a mutation abolishing the priming reaction. The high percentages of strains with positive signals for β-core and protease domains, together with the uniform size of the amplified fragments, illustrates the conserved nature of these domains. The PCR fragments corresponding to the linker region in contrast were detected in almost all of the clinical isolates but in only 4 of the 19 carriage strains, and exhibited the widest variation in size. We further characterized 4 iga protease genes by sequencing their linker domain PCR products and found much lower sequence similarities than among encapsulated strains. This result supports previous observations regarding the clonal population structure of encapsulated strains in contrast to the existence of extensive genetic diversity among NTHi strains.²² In particular, our results underline previous findings regarding the limited antigenic and genetic diversity among Hib IgA1 proteases and more apparent antigenic heterogeneity of NTHi IgA1 proteases.^24,25 Routine biotyping results (data not shown), together with the PCR and sequencing data, confirm the relatively wide genetic variability of the samples used in this study.

Quantitative ELISA was used to assay IgA1 protease activity. All strains belonging to each of the 2 groups were examined in triplicate for IgA1 protease activity using 2 different substrates (monoclonal IgA1 and polyclonal sIgA). This guarded against the potential problem of the presence of enzyme-neutralizing antibodies, although this complication did not materialize. Two main observations were made: (1) the level of detectable IgA1 protease activity varies between individual strains (10-600 mU) and (2) strains isolated in association with clinical NTHi infections demonstrated significantly higher levels of IgA1 protease activity than colonizing strains (median values of 155 mU vs 30 mU for IgA1 and 150 mU vs 30 mU for sIgA). These findings, including the significant difference in protease activity for the 2 groups (P< .001), suggest that IgA1 protease is an important factor in NTHi pathogenesis.

In this article, we have shown that NTHi strains differ remarkably in their IgA1 protease activities and in the size and nucleotide sequence of the linker domain of their iga genes. However, it will be of interest to see whether these observations hold true for other populations of NTHi from different geographic regions. Indeed, the difference in IgA1 protease activity was previously reported for 3 NTHi strains, all clinical isolates.²⁶ What molecular mechanisms underlie this variability and what are the consequences of this variability for bacterial pathogenesis? Haemophilus is a naturally competent genus able to incorporate exogenous DNA via transformation. The observed variations in IgA1 protease activity levels could be explained by polymorphisms in the iga gene or promoter.^27,28 These polymorphisms arise during natural transformation and subsequent homologous recombination of DNA fragments containing the IgA1 protease gene. New variants of bacterial IgA1 protease with increased activity could result from this process, providing the mutated strain with survival advantages (eg, avoidance of immune surveillance in the hostile mucosal membrane environment).

Earlier indirect evidence implicating IgA1 protease in pathogenesis has been reported; nonpathogenic species of Neisseria and Haemophilus lack IgA1 protease activity.^9,29 The products of IgA1 cleavage have been found in the cerebrospinal fluid of patients with bacterial meningitis, in vaginal washings from gonorrhea cases, and in other secretions from individuals infected with known IgA1 protease–producing bacteria.¹² Animal model and tissue culture experiments have been cited for dismissal of IgA1 proteases as virulence factors. However, animal models of infection caused by IgA1 protease–producing bacteria are unsatisfactory because of the exclusive specificity of the enzyme for human IgA1.¹² Animal IgA1 does not possess the same hinge region and cannot be cleaved by NTHi IgA1 protease. Experiments based on infection of organ cultures with wild-type and isogenic iga mutants again do not mimic the situation in vivo on human mucosal membranes because they do not produce or contain human IgA1 or sIgA strain-specific antibodies for bacteria used in the experiment.^30,31

Recently, new roles for bacterial IgA1 protease have been identified. Human lysosome-associated membrane protein, a heavily glycosylated protein forming the protective lining of terminal phagolysosomes, is cleaved by Neisseria gonorrhoeae IgA1 protease.^32,33 In addition, the participation of N gonorrhoeae IgA1 protease in transepithelial trafficking³⁴ and the ability of a fragment of the iga gene product (the α-peptide) to migrate into the nucleus of epithelial cells were also reported.³⁵ It was suggested that this protein might be directly involved in regulation of host cell functions via interaction with nuclear DNA. Thus, bacterial IgA1 proteases could contribute to bacterial adaptive fitness not only via avoidance of immune exclusion, but also via intracellular survival and possible regulation of host cell functions. The recent results showed that IgA1 protease is a potent inducer of proinflammatory cytokines, further underlining the importance of this enzyme in pathogenesis.³⁶ Thus, high-level IgA1 protease production could assist NTHi in avoiding the immune response by aiding longer or more extensive colonization of the host. The presence of larger numbers of organisms, possibly for extended periods, would presumably increase the likelihood of pathogenic consequences being displayed, such as localized inflammation or penetration of sterile sites. Possession of an enzyme with enhanced activity could clearly contribute to the pathogenic potential of a particular bacterial strain.

Our increased appreciation of the roles of IgA1 protease in pathogenicity suggests that the enzyme may be a target for development of vaccines against NTHi infection. Alternatively, enzyme inhibitors could be developed to interfere with the protease. Elucidation of the exact molecular mechanisms governing the interaction of IgA1 protease–producing bacteria with the host will enhance our capacity to develop new strategies for fighting these important pathogens.