ABSTRACT
Study of the epidemiology of invasive infections caused by encapsulated Haemophilus influenzae has been complicated by the poor sensitivity and specificity of the serologic assays used to identify specific capsular polysaccharides. The population structure of these bacteria is highly clonal, however, and serotype is highly correlated with other genetic characteristics. We sought to determine if alleles of the highly conserved phosphoglucose isomerase (pgi) gene correspond to the serotypes of encapsulated H. influenzae strains. pgi alleles of 52 well-characterized encapsulated H. influenzae isolates were amplified by PCR, sequenced, and compared to one another and to additional previously reported H. influenzae pgi alleles. Overall, 83% of the strains possessed pgi alleles associated with the major serotype a, b, e, and f clonotypes that cause the most invasive disease in the United States. Six strains (four type a and two type f) had unusual pgi alleles, which suggested that these strains belonged to less common clonotypes of encapsulated bacteria or were actually nontypeable strains. pgi genotyping may provide a simple and stable surrogate for capsular serotyping. Further studies correlating pgi typing with the expression of capsule are likely to increase our understanding of the epidemiology and pathogenesis of these infections.
Encapsulated Haemophilus influenzae strains are distinguished from nontypeable strains by the expression of one of six capsular polysaccharides (a to f) (22). Whereas nontypeable H. influenzae (NTHi) strains are common causes of non-life-threatening respiratory infections, encapsulated H. influenzae strains are associated with invasive disease, particularly of infants and immunocompromised persons. H. influenzae type b (Hib) was formerly the most common cause of serious infections in young children and caused approximately 95% of invasive Haemophilus disease. The use of Hib polysaccharide and oligosaccharide protein conjugate vaccines has dramatically decreased the incidence of Hib disease in North America in the last 2 decades (14, 24).
The incidence of serious infection caused by non-type b encapsulated H. influenzae is currently much lower than that of Hib infections in the prevaccine era; however, for several reasons, public health experts recommend continued surveillance of invasive disease caused by these bacteria (21). First, invasive infections due to non-type b encapsulated H. influenzae have increased in relative importance with the decline of Hib disease (7, 18, 21). Active surveillance studies report that the incidence of invasive non-type b encapsulated H. influenzae infections is currently approximately 0.3 case per 100,000 people, with a patient fatality rate of 3.2 to 18.7% (2). Also, outbreaks of invasive infection caused by Hia in healthy young children have occurred in the intermountain West and in the southwestern United States, which suggests that the incidence of non-Hib infections may be increasing in some geographic areas (1).
Our understanding of the epidemiology of encapsulated H. influenzae infections has been hampered by the poor sensitivity and specificity of the slide agglutination tests most commonly used to identify the serotypes of non-Hib strains. Some of these bacteria exhibit nonspecific agglutination, and strains expressing limited amounts of capsule may not be identifiable (19, 23, 25). Also, the encapsulation locus (cap) of H. influenzae is flanked by directly repeated IS1016 insertional sequences (11, 12). In the absence of selective pressure to maintain capsule expression, irreversible loss of these genes may occur spontaneously. This may occur rapidly both in vivo and in vitro, presumably by homologous recombination (9). Molecular techniques for determining serotype, such as cap genotyping and amplification of serotype-specific capsular synthesis genes, are more sensitive than traditional slide agglutination assays, but they cannot accurately identify strains that have lost cap genes (3, 5, 10). In contemporary surveillance studies, the capsulation status of 16.3 to 26.3% of the H. influenzae strains that cause invasive disease is not known (23). A recent study attempted to verify the serotypes of clinical non-Hib isolates by slide agglutination and cap-based molecular techniques (19). Despite the fact that these strains were contemporary isolates which had been passaged in the laboratory a minimal number of times, only 69.0% of the serotypes determined by referring laboratories could be confirmed by conventional serologic assays, and only 77.6% could be confirmed by molecular methods (19).
Analysis of the population structure of large numbers of encapsulated H. influenzae strains demonstrates that the population structure of these bacteria is highly clonal, with infrequent recombination (15-17). These data are corroborated by an analysis of recent invasive U.S. isolates of non-type b encapsulated H. influenzae, which have very limited genetic diversity (20). Using restriction digest pattern (RDP) typing, it was determined that Hia strains belong to two major clonotypes and Hif and Hie strains each belong to a single group of bacteria that are genetically closely related. The nucleotide sequences of highly conserved housekeeping genes have been used to identify the genetic relationships of groups of bacteria with a high degree of sensitivity (4, 13). In particular, phosphoglucose isomerase (pgi) alleles are highly congruent with other genes studied and we reasoned that the determination of pgi genotypes might be a useful surrogate for H. influenzae classification systems based on the cap locus. To test this hypothesis, we determined the pgi genotypes of a well-characterized collection of invasive encapsulated H. influenzae strains.
MATERIALS AND METHODS
Bacterial strains.Invasive encapsulated H. influenzae strains were obtained as previously described (19). These strains had been isolated from blood or cerebrospinal fluid and referred to various state public health departments for serotyping. Laboratory strains R421 (type a), Eagan (type b), and Rd (type d) were obtained from A. Smith (University of Missouri, Columbia) and J. St. Geme III (Washington University School of Medicine). Bacteria were grown on chocolate II agar (Edge Biological, Memphis, Tenn.) or in brain heart infusion broth supplemented with 10 μg of hemin (Sigma Chemical Co., St. Louis, Mo.) per ml and 2 μg of β-NAD (Sigma) per ml. Frozen stocks were stored at −80°C in 100% skim milk.
pgi amplification.Genomic DNA from each bacterial strain was prepared as previously described (19). An internal fragment of the pgi gene was amplified from genomic DNA by PCR by using sense primer pgiUP (nucleotides 295 to 314: 5′ GGTGAAAAAATCAATCGTAC 3′) and antisense primer pgiDN (nucleotides 865 to 884: 5′ ATTGAAAGACCAATAGCTGA 3′) (4). The reaction mixture included 1 μg of genomic DNA, a 0.001 mM concentration of each deoxynucleoside triphosphate, 0.005 mM MgCl2, and 50 pmol of each primer in the manufacturer's supplied buffer (Expand High Fidelity PCR system; Roche Molecular Biochemicals, Mannheim, Germany). In some cases, amplification reactions were performed by using a single bacterial colony from an agar plate as a substrate. Each of the 34 amplification cycles consisted of denaturation for 1 min at 94°C, annealing for 1 min at 53°C, and elongation for 1 min 30 s at 72°C. Both strands of the amplification products were sequenced by using amplification primers and an ABI Prism 3700 DNA analyzer (Applied Biosystems, Appelera, Foster City, Calif.).
Comparison of pgi alleles.The overall homology of pgi alleles was determined by comparing 491-bp partial pgi sequences of each bacterial strain. To determine how these sequences were related to those of H. influenzae strains previously characterized by other investigators, we also compared these sequences to the large collection of H. influenzae pgi alleles entered into the Multi Locus Sequence Typing (MLST) database (http://www.mlst.net ). Homology was graphically represented by generating a phylogenetic tree from a distance matrix calculated by the method of Jukes and Cantor and constructed by using the unweighted pair group method within the GrowTree program (SeqWeb version 2.0; Accelrys, Madison, Wis.).
RESULTS
pgi alleles.A total of 18 Hia, 1 Hib, 1 Hid, 10 Hie, and 22 Hif strains were genotyped. The Hia strains had five pgi alleles, of which three were previously unreported. The Hib strain, Eagan, possessed the same pgi allele as previously reported and was identical to that of previously reported strains b7717 and b6107 (4). The Hid strain, Rd, also had the same pgi allele as previously reported and was identical to that of strains Drm118 and d6137. All Hie strains shared a single pgi allele, identical to that of reference strains e6169, e6181, and e7066. Hif strains had three pgi alleles, of which two were novel.
Congruence of pgi alleles with serotype.The pgi alleles of all but nine strains were shared with previously identified H. influenzae pgi alleles (Fig. 1). Strain AT01, originally identified as Hia, shared a pgi allele with the nontypeable strain nt1158. Strain NC01 was reported to be type a but shared a pgi allele with nontypeable strain nt477. Strain UT06 was identified by the referring laboratory as Hia but shared a pgi allele with the Hib strains Eagan, b6107, and b7717. cap genotyping confirmed that this strain was actually Hib (data not shown). Strain OK04 was identified by the referring laboratory as Hie and had a unique pulsed-field gel electrophoresis RDP and pgi allele. Strains MO01 and MO02, which were identified by the referring laboratory as Hia, shared a previously unidentified pgi allele but had unique RDPs. The capsular serotype and genotype of each of these three strains could not be confirmed by serology or genotyping (19). The serotypes of strains CO05 and IA05 (putative Hif) also could not be confirmed, and these strains possessed novel pgi alleles.
Similarity between H. influenzaepgi alleles. A 491-bp fragment of the pgi coding sequence was amplified by PCR and sequenced. Distance measures between pgi alleles of each encapsulated H. influenzae strain used in this study (bold type) and selected H. influenzae pgi alleles entered into the MLST database (plain type) were calculated, and a phylogenetic tree was derived from a distance matrix by using the unweighted pair group method with arithmetic averages (4). The serotype of each strain is indicated by a lowercase letter. nt, nontypeable (nonencapsulated) strain; nk, strain that reacted with polyclonal anti-capsular polysaccharide antisera but not with antisera against individual capsule types. The numbers of nucleotide differences between pgi alleles are represented horizontally, and identical alleles are indicated by vertical lines.
Strain NC05 reacted with polyclonal sera against all H. influenzae capsular types but not with individual antisera directed against specific capsular polysaccharides. Its pgi allele was also unique.
DISCUSSION
Various nonserologic methods to identify and classify encapsulated H. influenzae have been proposed.cap genotyping is more sensitive than slide agglutination but requires considerable time, specialized equipment, and expertise to interpret (19). Strategies involving PCR amplification and nucleic acid sequencing of serotype-specific cap genes, although also complex, have the potential to be incorporated into automated systems requiring little hands-on time or sophisticated analysis. Unfortunately, both of these techniques depend on the persistence of an intact cap locus. These genes, because of their flanking repetitive elements, may not be stable over time. To identify an alternative molecular target for rapid and reliable identification of H. influenzae subpopulations, we initially investigated a number of genes that have been useful in previous studies of clinical H. influenzae isolates. Sequences of 16S rRNA genes have been used to rapidly detect and identify a variety of bacterial pathogens (6). Genotyping using alleles of infB, the gene that encodes translation initiation factor 2, has also proved useful in distinguishing Haemophilus species from related bacteria (8). Unfortunately, we found that neither of these techniques was sensitive enough to identify distinct serotypes of H. influenzae (data not shown). MLST uses the nucleotide sequences of seven housekeeping genes as surrogates for selected marker enzymes with electrophoretic mobility (13). Recent studies suggest that MLST is comparable to multilocus enzyme electrophoresis in its ability to discriminate between genetically related groups of H. influenzae (4, 17). Because of these features, we reasoned that pgi genotypes may provide sufficient discrimination to identify serotypes and major subgroups of serotypes. Comparison of the results of restriction digest typing of the pgi alleles of our collection of invasive, non-type b encapsulated H. influenzae strains to the results of pgi genotyping suggests that a distinct pgi allele corresponds to each major restriction digest clonotype (20). The two major Hia clonotypes in our strain collection are represented by two pgi alleles. Strains belonging to the predominant Hie and Hif clonotypes each share a single pgi allele. The MLST database contains strains from diverse geographic areas isolated over the course of several decades. The observation that reference strains within this database share pgi alleles with our strains suggests that these alleles are fairly stable and that pgi classification may be useful in different geographic locations and for the foreseeable future.
In this study, 13 of 18 isolates designated as Hia possessed a pgi allele corresponding to one of the two major Hia clonotypes. Of the five strains that did not belong to these clonotypes, two strains (MO01 and MO02) could not be typed serologically and failed to hybridize with a cap probe, which suggests that they may not have been encapsulated strains or that they represent a minor Hia clonotype that has lost capsulation genes. Strain UT06 was originally classified as an Hia strain; however, its cap genotype, RDP, and pgi allele type indicate that it is actually type b. Hia strain TN01 could not be serotyped or genotyped using the cap probe. Its pgi allele is identical to that of NTHi strain 1158; therefore, it is possible that TN01 is also a NTHi strain. Hia strain NC01 could not be serotyped by slide agglutination but had an a(T) cap genotype (10). This strain shares a pgi allele with the MLST database strain nt477. This raises the possibility that nt477 may also be an Hia strain that lost capsulation genes or, conversely, that NC01 represents the rare situation in which an NTHi strain acquired capsulation genes from a type a donor.
Correlation of the putative serotype and pgi allele type is more straightforward for Hie strains, which all share a common pgi allele. Of note, two of these strains had lost the cap locus and could not be serotyped or cap genotyped. Only two Hif strains did not possess the predominant type f pgi allele. Both strains CO05 and IA05 had unique pgi alleles and unique RDP types (19). Genomic DNA from CO05 failed to hybridize with a cap probe, but aberrant hybridization was observed with DNA from strain IA05, suggesting that it, at some point in the past, possessed the cap locus. Clearly, pgi typing of larger numbers of encapsulated and nontypeable strains and correlation of pgi alleles with the cap locus will be required to clarify the genetic origins of strains possessing unusual alleles. Because of the simplicity and effectiveness of pgi typing, prospective studies of larger numbers of isolates are warranted.
pgi genotyping appears to provide a more stable marker for the genetic origin of an H. influenzae strain than capsule phenotyping. The rapid loss of capsulation genes may make it impossible to distinguish encapsulated bacteria from NTHi or to determine the capsular type if serotyping is not performed expediently. Despite the fact that most of the strains examined in this study were recent clinical isolates and were passaged a minimum number of times, we were able to confirm the serotypes of only 65% of Hia strains, 57% of Hie strains, and 86% of Hif strains by slide agglutination or agarose immunodiffusion assays (19). In our hands, pgi typing was more sensitive and specific than serologic or capsulation gene-based strategies for identifying encapsulated strains belonging to the major invasive clonotypes. The current lack of a “gold standard” makes it difficult to estimate the overall effectiveness of this technique. It can be argued that the polysaccharide capsule is an important virulence factor and that attempts should be made to determine whether significant clinical isolates are encapsulated. However, as noted previously, these serologic assays are not always easily interpreted and the capsulation status of a significant proportion of strains cannot be determined (23). It is possible that a combination of techniques, such as testing reactivity with polyclonal sera against all capsular polysaccharides followed by pgi genotyping to identify the origin of a particular strain, may be most helpful in better understanding the epidemiology of H. influenzae infections.
In summary, we have demonstrated that pgi genotyping may identify the genetic background of encapsulated H. influenzae and, for the majority of clinical isolates, provide a good surrogate for conventional serologic assays. Further prospective studies correlating H. influenzae pgi alleles with the expression of capsule are likely to improve the accuracy of epidemiological investigations and advance our understanding of Haemophilus infections.
ACKNOWLEDGMENTS
We thank the state public health departments and other investigators for providing us with bacterial isolates and A. Smith and E. Tuomanen for their helpful comments.
This study was supported by Cancer Center Support CORE CA21765 and the American Lebanese Syrian Associated Charities. The MLST website (http://www.mlst.net ) was developed by Man-Suen Chan and funded by the Wellcome Trust. E.E.A. is an Established Investigator of the American Heart Association.
FOOTNOTES
- Received 1 August 2002.
- Returned for modification 16 September 2002.
- Accepted 24 January 2003.
- Copyright © 2003 American Society for Microbiology
