INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Haemophilus influenzae is an important cause of human disease worldwide, with serotype b (Hib) capsulated strains causing invasive bacteremic infections such as meningitis, epiglottitis, septicemia, and septic arthritis, particularly in infants. Strains lacking a capsule (HiNT) are well established as etiologic agents of otitis media and lower respiratory tract infections in children and account for millions of deaths among children in developing countries (20). The availability of Hib conjugate vaccines has dramatically reduced the incidence of invasive disease in Western Europe and North America (2, 5, 30). As the incidence of Hib decreases, focus in this public health problem has turned to other capsular types (a and c to f) and noncapsulate strains (8, 36).

H. influenzae strains have been traditionally classified by determination of the biotype and capsular serotype. Such methods are subject to phenotypic variations and do not give information on the clonal origin of the strains. More discriminatory methods, such as outer membrane protein analysis, lipopolysaccharide profiles, and multilocus enzyme electrophoresis, have been used to study the epidemiology and pathogenesis of H. influenzae infections (1, 11, 14, 23). Studies of the pathobiology of Haemophilus indicate that marked differences in virulence occur among strains (31). However, associations between specific diseases and virulence determinants are sometimes difficult to establish.

Genome variation of H. influenzae has been evaluated by applying several molecular biology techniques, including analysis of DNA restriction fragment length polymorphisms, PCR with arbitrary primers (randomly amplified polymorphic DNA), and rRNA gene restriction patterns (4, 28). The last method has a low discriminatory capability, and the first two give results in complex patterns. Binary data output and computer-assisted cluster analysis are of additional value.

Interspersed repetitive DNA sequences have been described for eubacteria. In 1992, de Bruijn (7) examined the distribution of the enterobacterial repetitive intergenic consensus (ERIC) sequences in the genomes of a number of gram-negative isolates. ERIC sequences are highly conserved at the nucleotide sequence level, but their chromosomal locations differ between species or strains (7). ERIC sequences are 126 bp long and appear to be restricted to transcribed regions of the genome, either in intergenic regions of polycistronic operons or in untranslated regions upstream or downstream of open reading frames (18). These elements have been successfully used for molecular typing purposes. By use of PCR differences in band sizes which represent polymorphisms in the distances between repetitive elements of different genomes, ERIC-PCR allows for the identification of interstrain genotypic diversity and has the potential to differentiate pathovars (17, 18, 34, 35).

In H. influenzae and other bacterial species, some of the genes encoding pathogenicity determinants have been shown to contain contiguous repetitive DNA that appeared to be related to adaptive virulence (12, 13, 25, 33).

This study reports an analysis of ERIC variability distribution in genomic DNAs from H. influenzae strains isolated from sick Mexican children and from healthy children who were carriers. The aim of the study was to ascertain whether restricted variability among isolates was related to the clinical origin. The data on levels of similarity shown here reflect the extent of variability among H. influenzae strains. The results revealed little variability among clinical H. influenzae strains, providing additional evidence of clonality. The full richness of diversity of nontypeable H. influenzae is discussed in relation to some possible genetic mechanisms.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Bacterial strains. This study was based on a collection of 92 H. influenzae strains isolated from Mexican children between 1990 and 1997. These included 24 nontypeable and 48 serotype b strains recovered during episodes of different diseases. A further 20 nontypeable strains which had been isolated from throat swabs obtained from healthy child carriers were also included. Clinical isolates were part of collections from the Instituto Nacional de Diagnóstico, Referencia y Estudios Epidemiológicos (INDRE), and the Hospital Infantil de México "Federico Gómez" (HIM), both located in Mexico City. With the exception of seven strains which were obtained from children in the Mexican states of Michoacan and Morelos, all strains were recovered in the metropolitan area of Mexico City.

DNA extraction. All Haemophilus strains were incubated overnight at 37°C on brain heart infusion agar plates supplemented with 2.5% Fildes fluid (Difco Laboratories). For DNA extraction, bacterial growth was scraped from the plates and suspended in 1 ml of phosphate-buffered saline (pH 7.0). DNA was isolated from this bacterial suspension using the guanidium isothiocyanate protocol as described by Boom et al. (3). DNA extracts were suspended in 100 µl of Tris-EDTA (10 mM Tris-HCl, 0.01 mM EDTA [pH 8.0]). The concentration was adjusted fluorometrically at excitation and emission wavelengths of 365 and 460 nm, respectively, using the Hoechst 33258 dye and a DyNA Quant 200-115 v minifluorometer (Hoefer Pharmacia Biotech Inc., San Francisco, Calif.). Aliquots were stored at 4°C for further analysis.

ERIC-PCR and amplification conditions. The ERIC-PCR was optimized for template, deoxyribonucleoside triphosphates, magnesium ion concentrations, and primers by a modified Taguchi method, based on orthogonal arrays as described by Cobb and Clarkson (6). The optimized amplification reactions were performed in 50-µl volumes containing 10 mM Tris-HCl (pH 9.0), 50 mM KCl, and 2.5 mM MgCl (Sigma Chemical Co., St. Louis, Mo.). Each reaction mixture included 35 pmol of each primer (ERIC 1 and ERIC 2; 3'CACTTAGGGGTCCTCGAATGTA5' and 5'AAGTAAGTGACTGGGGTGAGCG3', respectively) (Lakeside, Mannheim Boeringer Biochemicals, San Francisco, Calif.), deoxyribonucleoside triphosphates to a final concentration of 0.2 mM each (Ultrapure dNTP Set; Pharmacia, Biotech, Piscataway, N.J.), 50 ng of DNA sample, and 0.5 U of Taq DNA polymerase (Gibco, BRL Life Technologies, Inc., Gaithersburg, Md.). A negative control, consisting of the same reaction mixture but with no DNA template, was included in each amplification procedure. E. coli genomic DNA obtained using the same extraction protocol was used as a positive control. The PCR was initiated with a 5-min denaturation period at 94°C, followed by 35 cycles of denaturation (1 min at 94°C), annealing (1 min at 40°C), and enzymatic chain extension (4 min at 74°C) with a final extension at 74°C for 10 min. Amplifications were performed in an automated thermal cycler (DNA Thermal Cycler 480; Perkin-Elmer).

Electrophoretic patterns (Eps). An 8-µl portion of each amplified PCR product was loaded in gel slots and separated by electrophoresis at 10°C and 5 V/cm for 6 to 7 h on 1.4% agarose gels (Ultrapure; Gibco, BRL Life Technologies, Inc.) of 24 by 20 cm in 1× TBE buffer (10 mM Tris base, 50 mM boric acid, 2 mM EDTA [pH 8.0]). Gels were stained with 0.5 µg of ethidium bromide per ml in 1× TBE for 20 to 35 min and were visualized and photographed under UV transillumination for 45 to 60 s on Polaroid type 55 film.

Cluster analysis and statistics. The cluster analysis of the 92 strains was conducted on the basis of the characteristics of the fingerprints generated. Based on the data for presence or absence of 13 different DNA fragments in the fingerprints of the 92 strains of H. influenzae, a binary data matrix was created. Overall similarity between pairs of strains was calculated from the binary data matrix using the simple matching coefficient (29). The resulting similarity matrix was used as the input data for a cluster analysis by the unweighted pair-group method with arithmetic averages (UPGMA) (27). The goodness of the clustering method was assessed by calculating the cophenetic correlation coefficient (r). The Numerical Taxonomy and Multivariate Analysis System version 2.0 was used to carry out these analyses (26). The comparison of the distributions in the phenogram of Eps associated with groups of diseases and carriers was analyzed by the Wilcoxon sum rank test.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

General characteristics and diversity of the Eps. PCR with primers ERIC I and ERIC II produced multiple fragments of DNA in sizes ranging between 0.43 and 2.11 kb. A total of 69 Eps were found among the 92 H. influenzae strains studied. Only eight of the Eps were found more than once, indicating a high variability among the H. influenzae strains. Sixty-one of the 69 Eps (88%) were represented by a single strain. Table 1 shows the data obtained in terms of Ep, assigned reference number, fragment sizes, anatomic source, serotype, and biotype for each analyzed strain.

Relationships among genotypes. The phenogram shown in Fig. 1 represents the similarity relationships between the 69 Eps found. As can be seen from this phenogram, two major groups were formed at a similarity level of 0.59. They are identified in Fig. 1 as I and II. Two outliers were discriminated from the rest of the strains at the lowest value of similarity, namely, 0.57. Cluster I includes two smaller clusters, a and b, which are formed at a level of similarity of 0.64. Cluster II also includes two groups, d and c, which were formed at a similarity level of 0.61. Forty-eight strains were found within the large cluster, designated subgroup Ia. Interestingly, 95% of these (46 of 48) were Hib and HiNT strains isolated from sick children. This was the subgroup in which the most similar pairs of strains were found; many of them were grouped into five clusters, each one including between four and six strains which were identical (they grouped at a similarity level of 1.0). Seventy-nine percent (19 of 24) of the strains distributed within these five clusters were isolated from patients with meningitis.

View larger version (25K): [in this window] [in a new window]   FIG. 1.   Similarity relationships between 92 H. influenzae isolates. The phenogram of fingerprints was generated from a simple-matching similarity matrix by means of the unweighted pair-group average clustering method. A similarity of 1.0 indicates 100% identity between strains. The cophenetic correlation coefficient (r) was 7.33. , Hib strains recovered from meningitis; , Hib and noncapsulated isolates from diseases other than meningitis; , noncapsulated isolates from healthy carriers.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

This study is the first practical application of ERIC-PCR and clustering analysis to assess the variability between clinical Hib isolates from children. Clinical Hib strains were included in this study due to their interest from any perspective, since serotype b is very seldom found in North America and Western Europe following the introduction of conjugate vaccines into the Expanded Programme of Immunization.

Noncapsulated strains of H. influenzae from cases of chronic obstructive pulmonary disease and cystic fibrosis were analyzed previously by van Belkum et al. (32) using ERIC-PCR, with the data revealing marked differences between isolates. Those results are positively correlated with the data in this study, which also shows large genetic differences at this DNA level between most of the nontypeable isolates. Differences in the molecular sizes ranging from 430 to 2,110 bp in the present study (using a total primer amount of 70 pmol), in comparison with those previously published by van Belkum et al., where fragments of 100 to 2,000 bp (using a total primer amount of 100 pmol) were obtained, could be due to differences in technical procedures (see Results).

The results of the present study show that some H. influenzae strains grouped according to the type of infection (meningitis strains from group Ia versus other diseases isolates from groups Ia and Ib) and, to some extent, the typeable b character (Tables 3 and 4). By using ERIC-PCR, many of the type b strains causing meningitis, isolated from unrelated hosts, were shown to be identical. The clone concept postulates that bacterial populations are arrays of lineages that maintain their genetic identity for extended periods of time (24). Bacteria of the same (or nearly the same) genotype can be isolated from samples from different geographic regions or at different times. Our data showed the recovery of Ep 1 to Ep 5 in the Mexico City area and the occurrence of Ep 1, Ep 3, Ep 4, and Ep 5 in two different Mexican states (Michoacan and Morelos). Although a reduced number of strains isolated in localities other than the Mexico City area were included, these strains were recovered during a 7-year period. The observations for this data set agree with the hypothesis of clonal persistence for Hib strains, in concordance with other lines of evidence obtained by using multilocus enzyme electrophoresis (22, 23). Examination by ERIC-PCR of additional isolates from disease episodes on different continents or intracontinentaly and for more than 7 years will provide additional insights.

Musser et al. (23) reported that type b strains of electrophoretic types (ETs) 12.5 and 12.8, commonly associated with outer membrane protein (OMP) subtype 1, and ET 21.8, which occurred in 83% of strains of OMP subtype 1c, were nonranodmly associated with different types of invasive disease. Strains of ET 21.8 OMP subtype 1c caused proportionally more meningitis and less of other invasive disease types than did those of ET 12.5 and ET 12.8. This observation was interpreted as showing a true difference in the virulence of isolates expressing ET 21.8 OMP subtype 1c. Considering that properties analyzed by multilocus enzyme electrophoresis, OMP typing, and ERIC-PCR tend to be distributed in restricted sets of strains, we hypothesized that ET 21.8 OMP subtype 1c are marking clones especially successful in invading the CSF. However, the high occurrence of Ep 1 to Ep 5 in strains recovered from cases of meningitis does not mean that these, per se, have a role in virulence or organotropism.

Capsule-deficient mutants of type b strains arise with high frequency. It would be interesting to see where clinical capsulated strains other than b strains and capsule-deficient mutants fall in the phenogram. A small number of nontypeable isolates expressed the same Ep as type b strains. The analyzed nontypeable isolates may represent spontaneous b mutants.

In clonal populations, such as those of Hib, strong nonrandom associations between genetic loci (linkage disequilibrium) are generated (15, 22, 23). The results obtained for Hib meningitis strains are probably showing some linkage disequilibrium between some virulence factors and ERIC elements. Further theoretical and experimental bases are necessary to determine whether meningitis isolates present specific characteristics related to ERIC motifs and if these in turn may be recognized as having some involvement in the pathogenesis of Hib meningitis strains. Additional analysis may contribute to better understanding of the study results. Work on direct sequencing of ERIC-PCR-generated amplification fragments is in progress in this laboratory. It is important to mention that other type b meningitis strains were also spread within other groups (Ib and IIc).

Studies of noncapsulated H. influenzae isolates by OMP and lipopolysaccharide (LPS) profiles revealed that nontypeable strains isolated from patients with different disease types are, as a group, significantly more variable than those Hib strains recovered from invasive disease sources. The latter were represented by a restricted number of OMP and LPS Eps (21). The distribution of noncapsulated isolates from bronchial secretions, nasopharynx, sputum, genital tract, middle ear, eye, and suppurative arthritis, as shown in the phenogram, was represented by a variety of single Eps. In this way, any restricted distribution related to clinical origin was observed among noncapsulated strains. In addition, these patterns were not shared with patterns of carrier isolates. The repeated findings on the strong genetic diversity of noncapsulated H. influenzae, which are consistent with the results described here using ERIC-PCR, allowed some authors to make hypothetical predictions suggesting the possibility of a nonclonal population for nontypeable H. influenzae (22). However, Fusté et al. (9), who recently surveyed the extent of clonality within nontypeable strains, showed a basic clonal structure with little possibility of recombination. Mutations in the H. influenzae genome probably contribute to the strong variability of noncapsulated isolates, as observed using ERIC-PCR. It is possible that the error rate of bacterial DNA synthesis is increased during persistent infections, such as cystic fibrosis and chronic bronchitis, and during carriage in healthy individuals.

Recently a hypothesis on association between repetitive DNA in the bacterial genome and virulence potential has been described and proved. A characteristic of H. influenzae is phase variation in its fimbriae and antigenic variation in LPS (10). Intragenomic variation mechanisms, mediated by high-frequency mutations in H. influenzae repetitive sequences, are responsible for modulated expression of such surface molecules (10, 16; E. R. Moxon, Abstr. 38th Intersci. Conf. Antimicrob. Agents Chemother., abstr. 96, 1998). Repeat variability in H. influenzae related to phenotypic switching as a consequence of host selective pressures is well documented (19, 37, 38). It would be worth analyzing the possible variation of ERIC-PCR Eps due to immunological or physiological selection imposed by the individual patient during the infectious process.

This study is the first to identify correlations between Eps determined by ERIC-PCR and distinctive pathogenicities among H. influenzae strains. The data presented here can be utilized to generate additional hypotheses regarding the pathogenicity and epidemiology of H. influenzae.