• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of jcmPermissionsJournals.ASM.orgJournalJCM ArticleJournal InfoAuthorsReviewers
J Clin Microbiol. Dec 2006; 44(12): 4316–4325.
Published online Sep 27, 2006. doi:  10.1128/JCM.01331-06
PMCID: PMC1698427

Identification of New Genetic Regions More Prevalent in Nontypeable Haemophilus influenzae Otitis Media Strains than in Throat Strains[down-pointing small open triangle]

Abstract

Nontypeable (NT) Haemophilus influenzae strains cause significant respiratory illness and are isolated from up to half of middle ear aspirates from children with acute otitis media. Previous studies have identified two genes, lic2B and hmwA, that are associated with NT H. influenzae strains isolated from the middle ears of children with otitis media but that are not associated with NT H. influenzae strains isolated from the throats of healthy children, suggesting that they may play a role in virulence in otitis media. In this study, genomic subtraction was used to identify additional genetic regions unique to middle ear strains. The genome of NT H. influenzae middle ear strain G622 was subtracted from that of NT H. influenzae throat strain 23221, and the resultant gene regions unique to the middle ear strain were identified. Subsequently, the relative prevalence of the middle ear-specific gene regions among a large panel of otitis media and throat strains was determined by dot blot hybridization. By this approach, nine genetic regions were found to be significantly more prevalent in otitis media strains. Classification tree analysis of lic2B, hmwA, and the nine new potential otitis media virulence genes revealed two H. influenzae pathotypes associated with otitis media.

Otitis media is an infection of the middle ear that results in middle ear effusion, fever, irritability, and inflammation of the tympanic membrane and is the most common bacterial infection among infants and young children (25). Nontypeable (NT) Haemophilus influenzae strains are isolated from up to half of middle ear aspirates from children with acute otitis media (14). Furthermore, since the introduction of the Streptococcus pneumoniae heptavalent conjugate vaccine, a significant reduction in otitis media due to S. pneumoniae strains and an increase in otitis media due to NT H. influenzae strains (15) have been reported. Thus, currently, NT H. influenzae is the causative agent of otitis media in 56 to 57% of children in vaccinated communities (3, 9).

The first step in the pathogenesis of H. influenzae otitis media is bacterial colonization of the respiratory tract (33), and asymptomatic carriage has been found in from approximately 25% up to a range of 77 to 84% (4, 16, 19, 22, 48, 50) of healthy children. Through mechanisms not completely understood, NT H. influenzae isolates living in the pharynx ascend via the eustachian tube into the middle ear space, where they initiate an inflammatory response that leads to the disease acute otitis media.

The closely related species Haemophilus haemolyticus also resides in the human nasopharynx; both organisms require factor V (NAD) and factor X (heme) for in vitro cultivation. H. haemolyticus, which has not been associated with clinical disease, is distinguished from H. influenzae by its capacity to hemolyze horse red blood cells (24). Recently, variant H. haemolyticus strains that fail to hemolyze red blood cells have been isolated from pharyngeal and sputum samples (5, 28) and may be erroneously designated H. influenzae.

The development of otitis media has been associated with several epidemiologic factors, including genetic predisposition, preceding viral respiratory infection, day care center attendance, a lack of breast-feeding, and young age (2, 10, 51). While these factors have been well studied, the specific bacterial virulence factors important in the development of otitis media have been less well defined. Potentially important otitis media virulence factors of NT H. influenzae are the high-molecular-weight adhesins (47), as the hmw genes are significantly more prevalent among middle ear NT H. influenzae isolates than among NT H. influenzae isolates from the throats of healthy children (13).

A second potential otitis media virulence gene, lic2B, of NT H. influenzae isolates was previously reported by us (37). This lipooligosaccharide biosynthesis gene was identified by a molecular epidemiologic analysis consisting of genomic subtraction to isolate genetic regions present in middle ear NT H. influenzae isolate G622 but absent in sequenced laboratory strain H. influenzae Rd. Subsequently, the prevalence of the middle ear strain G622-specific regions among a set of NT H. influenzae middle ear and throat isolates was determined by dot blot hybridization. This approach, however, was unable to detect genes present in H. influenzae strain Rd that might be important in the pathogenesis of otitis media. Here we report on an additional genomic subtraction of middle ear strain G622 with driver strain NT H. influenzae 23221, which was isolated from the throat of a healthy child, and report on the prevalences of the genes represented by the subtraction products among a panel of both disease-associated and commensal H. influenzae strains. This analysis identified nine new genetic regions potentially important in the pathogenesis of otitis media caused by NT H. influenzae.

MATERIALS AND METHODS

H. influenzae strains.

The bacterial strain collections used in this study included H. influenzae strain Rd (18), 121 middle ear NT H. influenzae isolates, 173 throat isolates from healthy children attending day care, and 39 H. influenzae type b invasive strains. These isolates were defined as H. influenzae on the basis of colonial morphology during growth on chocolate agar with bacitracin, the requirement for X and V factors, porphyrin negativity, and a lack of hemolysis of horse red blood cells (8, 17, 24). With the exception of H. influenzae strain Rd, the isolates used in this study were collected from sites in Minnesota (20); Ann Arbor, Mich. (48); Battle Creek, Mich. (48); and Bardstown, Ky. (26), between 1980 and 2001. As described in detail later, NT H. influenzae middle ear strain G622 (26) was used as the tester and NT H. influenzae throat strain 23221 (48) was used as the driver in genomic subtraction.

P6 protein immunoblot assay.

While H. influenzae and H. haemolyticus both express P6 proteins, the 7F3 antibody (kindly donated by Timothy Murphy from the Buffalo Veteran's Administration Hospital, Buffalo, N.Y.) is specific for epitopes present on the P6 protein of H. influenzae and absent on the P6 protein of H. haemolyticus and was used to differentiate H. influenzae and nonhemolytic H. haemolyticus by an immunoblot assay based on a previously described method (34). Briefly, a 2-μl suspension of each H. influenzae isolate was dotted onto a nitrocellulose membrane and the membrane was allowed to dry. The membrane was incubated in BLOTTO (5% nonfat dry milk in sterile water) for 1 h at room temperature to block reactive sites. After the membrane was washed with phosphate-buffered saline (PBS), it was incubated overnight at room temperature in antibody diluted 1:1 in BLOTTO. The nitrocellulose was washed with PBS and incubated for an hour in goat anti-mouse immunoglobulin G (A-3562; Sigma) diluted 1:1,000 in BLOTTO. The immunoblots were developed with nitroblue tetrazolium-5-bromo-4 chloro-3-indolylphosphate (Pierce) color developer.

Haemophilus strains that failed to bind to the P6-specific antibody but that required the X and V factors for growth and that did not hemolyze horse erythrocytes were designated probable nonhemolytic Haemophilus haemolyticus.

DNA isolation and dot blot hybridization analysis for iga.

More than 90% of H. influenzae isolates possess the iga gene, which encodes the enzyme immunoglobulin A1 protease (24). The presence of iga was detected by the dot blot hybridization method described previously (13). Briefly, an iga-specific DNA probe was amplified from H. influenzae type b strain Eagan by PCR, as described previously (53), with β-core domain-specific primers (primer BF1, GCAGAATTCAAAGCACAATTTGTTGCA; primer BR1, TTATAACGTTAATTCAACAGGCTT) derived from the published sequence (39) of the H. influenzae HK368 iga gene (GenBank accession no. M87492). The probes were fluorescein labeled by using an enhanced chemifluorescence random prime labeling system (Amersham). Crude DNA was isolated from the H. influenzae lysates, and the DNA concentrations were standardized by spectrophotometry. DNA samples were then transferred onto nylon membranes by using a Bio-Dot microfiltration apparatus (Bio-Rad), which generated a fixed array of DNA dots (8 by 12 dots). The dot blots were hybridized to the fluorescein-labeled DNA probes under stringent conditions by using the fluorescein-based enhanced chemifluorescence detection system (Amersham). The signal intensity of each dot was detected by using the Storm system from Molecular Dynamics (Sunnyvale, Calif.) and is reported as a percentage of the positive control signal after correction for the background signal. The positive controls used for PCR analysis included pepN, a peptidase-encoding gene found in all H. influenzae strains studied (13), and 16S rRNA sequences (27).

Haemophilus strains that did not hybridize with the iga probe but that required the X and V factors for growth and that did not hemolyze horse erythrocytes were designated probable nonhemolytic H. haemolyticus.

Differential cloning by subtraction PCR.

Subtractive hybridization is a technique used to isolate regions of genomic DNA present in a bacterial strain with a characteristic of interest (the tester) but absent in a strain without that characteristic (the driver). For subtractive hybridization experiments in this study, H. influenzae middle ear strain G622 was used as the tester strain. This strain was chosen from a collection of 17 middle ear strains (48) because it demonstrated a large number of restriction fragment bands held in common with other middle ear strains, as determined by pulsed-field gel electrophoresis. Furthermore, strain G622 had been used as a tester in an earlier subtraction procedure with strain Rd as the driver (37). To maximize the number of new genetic regions detected, in this study, strain 23221, which was originally identified as an NT H. influenzae isolate from the throat of a healthy child, was chosen as the driver. This strain has a ribotype pattern very different from that seen with strain Rd (37), and subsequent analysis has shown that it is a nonhemolytic H. haemolyticus isolate.

A commercial subtractive hybridization kit (PCR-select; Clontech, Palo Alto, Calif.) was used to identify gene fragments specific to the tester strain through differential cloning by previously described methods (37, 55).

Determination of tester and driver specificities.

To confirm the specificity of the tester, nested PCR products were blotted onto nylon membranes, hybridized separately to AluI-digested tester (otitis media) and driver (throat) genomic DNA, labeled, and detected as described above.

PCR product prevalence among H. influenzae isolates.

The presence or absence of each DNA fragment identified by the subtraction PCR technique (sPCR) was determined by genomic dot blot hybridization, as described previously (37, 56). Initial screening was done by using filters containing six control spots (two spots of NT H. influenzae strain G622, the middle ear [tester] strain; two spots of 23221, the throat [driver] strain; one spot of Rd; and one spot as a no-DNA control), 46 NT H. influenzae middle ear isolates, and 33 NT H. influenzae isolates from the throats of healthy children, all of which were randomly placed on the filter. Subsequent analysis showed that 10 of the 33 NT H. influenzae isolates from the throats of healthy children were nonhemolytic H. haemolyticus isolates and were removed from the analysis.

Each filter was hybridized with unique middle ear (tester)-specific sPCR fragments labeled as described above. Dot blot images were analyzed with ImageQuaNT, version 5.0 (Molecular Dynamics), and the signal was expressed as a percentage of the signal obtained for strain G622 (the middle ear [tester] strain used as a positive control) after correction for the background. All isolates were screened independently with each sPCR probe at least twice. To define negative versus positive hybridization of the probes, the results from each duplicate filter were graphed against each other on a scatter plot, and discrepancies were resolved by Southern blot hybridization (37, 42).

Sequence analysis.

DNA sequence analysis of the tester-specific sPCR fragments was performed at the University of Michigan Molecular Biology Core facility on an Applied Biosystems model 3700 DNA sequencer. Software packages from DNASTAR (Madison, Wis.) were used for primer design, DNA sequence comparison, and analysis. Nucleotide sequence searches and comparison used the National Center for Biotechnology Information website (http://www.ncbi.nlm.nih.gov).

Data analysis.

Excel (Microsoft) software was used for data entry. Prevalence ratios (PRs) were calculated as the ratio of the proportion of isolates containing the gene of interest among the middle ear strains to the proportion containing that gene among the strains isolated from the throats of healthy children. A χ2 analysis was used to determine the co-occurrence of paired newly identified and prior candidate middle ear strain virulence genes. The odds ratio (OR) and 95% confidence intervals (CIs) for each paired genetic region were obtained to estimate the magnitude of the association. A P value obtained by use of the Bonferroni correction for multiple comparisons was calculated to determine the level of significance between each paired association. A P value of ≤0.05 was considered significant. All statistical analyses, except the classification tree analysis described in the next paragraph, were performed by using SAS software (version 9.1).

To identify pathotypes (or clusters of genes that predict virulence) for the isolates on the basis of genetic region patterns, we used a classification tree analysis (6). The goal of this analysis was to partition the population of isolates into subgroups that are relatively homogeneous according to the absence or the presence of combinations of genetic regions. The classification tree analysis can be broken down into three main steps. First, the data are partitioned into homogeneous subgroups on the basis of the impurity index, resulting in a classification tree. A finding that the terminal nodes of the resultant tree are less than approximately 5% of the total sample size suggests that the initial partitioning was too selective and the resultant tree would not be a good predictor for future data. Second, in this situation pruning may be used to reduce the number of nodes to make the tree more representative of the data and give them a better predictive power for new sample data. Third, the predictive ability of the constructed tree may be tested and the degree of misclassification may be identified. The last step, however, requires an independent data set and was not done for this initial analysis. The deviance at each node is also shown (see Fig. Fig.3).3). The deviance is an estimate of the misclassification error at each node of the tree. Smaller values of the deviance indicate a better fit of the model to the data. All classification tree analyses were performed by using rpart, R, version 3.1-29 (49).

FIG. 3.
Classification tree for middle ear isolates (ME) versus throat isolates (TH) obtained by using 11 genetic regions: lic2B, hmwA (sJPX140), sJPX120, sJPX124, and sJPX132 as individual factors and sJPX84, sJPX101, sJPX147, sJPX161, sJPX163, and sJPX176 as ...

Nucleotide sequence accession numbers.

The sequences for the 20 sPCR fragments presented in Table Table11 have been submitted to GenBank and have accession numbers AY828203 through AY828225, respectively.

TABLE 1.
Dot blot hybridization screening of 46 NT H. influenzae isolates from the middle ears of patients with otitis media and 23 NT H. influenzae isolates from the throats of healthy children probed with 20 unique sPCR fragments

RESULTS

Taxonomy of Haemophilus strains.

Identification of the NT H. influenzae strains was based on colonial morphology on chocolate agar, the lack of a capsule, the requirement for the X and V factors for growth, and a failure to lyse horse erythrocytes. Recently, it has been discovered that some isolates identified as NT H. influenzae strains are actually much more similar to H. haemolyticus, although they are nonhemolytic (5, 28), and can be identified by their failure to react to P6 7F3 monoclonal antibodies and/or to hybridize to an iga-specific DNA probe. Thirty-six of 173 isolates (20.8%) from the throats of healthy children (including driver strain 23221) originally identified as NT H. influenzae turned out to be nonhemolytic H. haemolyticus. Since over 20% of the NT H. influenzae isolates from the throats of healthy children but none of the middle ear isolates were nonhemolytic H. haemolyticus, these isolates were removed from our screening analyses to prevent bias in the results. Although we had not originally planned to do so, it is not uncommon to use closely related species as drivers in subtraction approaches (for example, subtractions of Mycobacterium tuberculosis with Mycobacterium bovis as the driver). The main disadvantage of this approach is that it increases the number of sPCR fragments to be screened, but as we had already done our initial screening prior to the discovery that strain 23221 was actually H. haemolyticus, it made no sense to do a new subtraction with a true NT H. influenzae strain.

Subtraction hybridization results.

From the subtractive hybridization procedure with the tester strain (otitis media NT H. influenzae isolate G622) and driver strain (throat NT H. influenzae isolate 23221 from a healthy child) described above, 188 sPCR fragments were identified. Of the 188 clones obtained, 17 had multiple PCR bands, as determined by agarose gel electrophoresis; and the remaining 171 clones, which had insert sizes ranging from 150 bp to 1.5 kb, were selected for further analysis. One clone (clone sJPX101) contained a cloning artifact that resulted in the amplification of vector sequences when the nested primers were used. To obtain regions lacking vector sequences, PCR was initially done with primers M13 and T7, followed by nested primer PCR.

To determine which clones were found in middle ear strain G622 but not in throat strain 23221 (i.e., to determine which clones were tester specific), the remaining 171 sPCR fragments were then separately hybridized to genomic DNA from the tester and the driver strains labeled with fluorescein. Due to the small region of genomic DNA represented on a given sPCR fragment compared to the larger size of the labeled whole-genome DNA probe, this hybridization was not highly specific; and some of the sPCR fragments initially determined to be tester (middle ear) strain-specific by this procedure were later also found to be present in the driver (throat) strain, as detailed later. This misclassification does not negatively affect the ultimate results of the analysis, as it merely results in the identification of additional probes to be used in the prevalence screening assays. The resultant 59 sPCR fragments that were initially deemed tester (middle ear) specific were then labeled and used as probes in hybridization assays to determine their prevalence among the H. influenzae genomic DNAs arrayed on duplicate filters, as described in Material and Methods.

Some probes resulted in unambiguous distinctions between probe-positive and probe-negative strains (Fig. (Fig.1A),1A), while other probes gave ambiguous results (Fig. (Fig.1B)1B) when duplicate blot results were compared. To determine whether the genetic regions represented by the sPCR fragments were truly present in these ambiguous isolates, they were further tested by Southern blot hybridization (Fig. (Fig.1C).1C). For example, isolates probed with sPCR fragment sJPX84, whose intensity was less than 50% of that for the positive control in the scatter plot (Fig. (Fig.1B),1B), were true negatives when they were tested by Southern blotting (Fig. (Fig.1C),1C), while isolates that had intensities greater than 50% of that for the positive control in the scatter plot (Fig. (Fig.1B)1B) were true positives when they were tested by Southern blotting (Fig. (Fig.1C1C).

FIG. 1.
Scatter plots of 95 Haemophilus influenzae isolates for probes sJPX101 (A) and sJPX84 (B) (hybridization results are expressed as a percentage of the signal intensity of the positive controls). Spots: 1, duplicate driver strain 23221; 2, strain 22125; ...

Thirty-seven of the 59 sPCR fragments hybridized to over 95% of all NT H. influenzae strains tested and, thus, were not useful in distinguishing factors more common among ear strains than among throat strains. The remaining 22 sPCR fragments tested were found in some isolates but not others and, thus, proved useful in comparing their differential distributions among NT H. influenzae populations. Table Table11 shows the distributions for 20 of these 22 sPCR fragments (sequence analysis showed that two sets of the sPCR fragments were identical, and so only one of each is included in Table Table1)1) among middle ear isolates compared to those among throat isolates from healthy children. The PRs of 8 of the 20 sPCR fragments varied from 0 to 1.0, with 4 fragments (PRs = 0.06 to 0.69) occurring more often among the throat isolates than among the middle ear isolates. Three of the sPCR fragments (PRs = 0) were found only in throat strains. Of the 12 probes that were more prevalent among middle ear isolates than among throat isolates, 3 hybridized to the driver (throat strain) DNA and, thus, were not truly middle ear tester strain specific; nevertheless, they were more prevalent among ear strains than throat strains.

Expanded prevalence ratios of subtraction fragments.

As the number of isolates from the middle ears and the throats of healthy children tested in the initial analysis described above was too small for optimal determination of middle ear isolate specificity, we used the 12 sPCR fragments with PR values of 1.02 and higher to probe 75 additional NT H. influenzae middle ear isolates, 114 additional NT H. influenzae throat isolates, and 39 H. influenzae type b invasive isolates. Ten of the probes tested were significantly associated with middle ear isolates when they were hybridized to this larger screening panel (Table (Table2).2). Interestingly, while sJPX170 was more frequently found in middle ear isolates than in throat isolates (although the difference was not statistically significantly), it was found significantly less often in H. influenzae type b isolates than in throat isolates of healthy children. Most of the sPCR fragments that were significantly more prevalent among middle ear isolates in this larger panel of strains were also significantly more prevalent in type b isolates, the exception being sJPX140, which was not present in any of the type b isolates.

TABLE 2.
Dot blot hybridization screening of 121 NT H. influenzae isolates from the middle ears of patients with otitis media, 137 NT H. influenzae isolates from the throats of healthy children, and 39 H. influenzae type b isolates probed with 12 sPCR fragments ...

Sequence analysis revealed that 9 of the 10 middle ear strain-specific sPCR fragments matched DNA sequences in fully sequenced H. influenzae strain Rd (Fig. (Fig.2)2) and were randomly distributed across the chromosome (18). Of these, sJPX101 matched a sequence found within a large inverted duplication separated by over 126 kb of DNA found during the complete genomic sequencing of Rd (18). The sJPX140 sequence was included in hmwA, a gene encoding a high-molecular-weight adhesin that has already been associated with otitis media (13, 47), thus validating the use of our approach to find potential virulence genes of middle ear isolates.

FIG. 2.
Locations of nine subtraction sPCR fragments on the Haemophilus influenzae Rd KW20 map. Solid triangles, tester-specific sPCR fragments present significantly more often among isolates from middle ears than among isolates from healthy throats; hollow circles ...

Gene co-occurrence and classification tree analysis.

The co-occurrence of the new putative otitis media virulence genes identified in this study (Table (Table3)3) was determined by a χ2 analysis for paired genes. The initial results focusing on the nine newly identified genetic regions showed that the presence of 28 paired genetic regions co-occurred at a statistically significant level. Four of these pairs (sJPX84-sJPX161, sJPX84-sJPX163, sJPX132-sJPX176, and sJPX161-sJPX163) had a cell frequency of zero during the analysis because the presence of one region was 100% correlated with the presence of its pair. A correction of 1.0 was added in every cell of the tables that contained a zero to avoid an overestimated odds ratio. Although the magnitude of association for these three pairs was thus underestimated, the associations were still statistically significant.

TABLE 3.
Pairwise associations of selected sJPX probes and lic2B among a panel of 121 middle ear NT H. influenzae isolates and 137 NT H. influenzae isolates from the throats of healthy childrena

In addition to pairwise analysis, each genetic region was tested for its co-occurrence with two well-characterized virulence genes associated with otitis media pathogenesis, lic2B and hmwA. sPCR fragment sJPX140 resides within hmwA and thus serves as a marker for the presence of this virulence gene. A statistically significant association was observed between lic2B and sJPX84, sJPX120, sJPX124, sJPX140 (hmwA), sJPX163, and sJPX176 (Table (Table3).3). Similarly, a statistically significant association was observed between hmwA and sJPX84, sJPX120, sJPX124, sJPX132, sJPX161, sJPX163, and sJPX176. For the comparisons of both lic2B and hmwA, a correction of 1.0 was again added to cells with a frequency of zero for 7 of these 19 pairs.

Using classification tree analysis, we explored the presence of all possible combinations of the nine newly identified genetic regions plus lic2B and hmwA and observed the combinations with the highest power to discriminate between middle ear strains and throat isolates from healthy children. These gene combinations, termed pathotypes, are predicted to contribute to the pathogenesis of otitis media through unique pathways.

A classification tree with two branches was constructed; all terminal nodes approximated 5% of the original sample size, thus negating the need for tree pruning. Two pathotypes were identified in the tree, with sJPX163 playing a strong role in both pathotypes. The presence of sJPX163 and hmwA had a high power to discriminate between middle ear and throat isolates. Given the high correlation found with six of the subtraction fragments (sJPX84, sJPX101, sJPX147, sJPX161, sJPX163, and sJPX176), it seemed likely that sJPX163 might be a marker for this whole genetic constellation in the classification tree. In fact, when the classification tree analysis was rerun without sJPX163, sJPX84 replaced it at the top of the tree. Next, we constructed a new variable, sJPXcombined, which represented the presence of all six of these genetic regions, and reran the analysis. The resulting classification tree (Fig. (Fig.3)3) showed that the whole set of these six genetic regions is the primary discriminator for the pathogenesis of otitis media. In the sJPXcombined-positive isolates that lacked hmwA, the presence of lic2B also gave a strong power to discriminate between middle ear and throat isolates.

DISCUSSION

Recent studies have demonstrated the dynamic nature of NT H. influenzae asymptomatic infection of the respiratory mucosa, characterized by the carriage of multiple NT H. influenzae strains at any one time (21, 29, 35, 46, 50) and rapid bacterial turnover (12, 16, 43, 50). Studies from our laboratory demonstrated that 43% of the cultures of throat swab specimens from healthy children attending day care contained two or more genetically distinct NT H. influenzae strains (range, zero to five strains) (17, 48). Furthermore, the week-to-week rate of turnover of NT H. influenzae strains (as defined by the pulsed-field gel electrophoresis genotype) was 62% (48). Thus, asymptomatic H. influenzae throat infection provides a diverse pool of organisms from which organisms infecting the middle ear emerge.

In order to survive in various environments, H. influenzae appears to rely on the selection and clonal expansion of selected strains rather than on the complex regulatory systems that are seen with other bacteria (54). Among H. influenzae strains, nontypeable strains, which cause the majority of respiratory tract infections, are much more genetically diverse than strains possessing the type b capsule, which cause the majority of bacteremic, invasive infections (36, 38, 41). To create the genetic diversity that ensures its survival, H. influenzae possess several molecular mechanisms that facilitate gene exchange. These bacteria are naturally competent and take up from the environment DNA fragments that possess the 9-nucleotide uptake signal sequence 5′AAGTGCGGT-3′, which is present at 1,471 copies in the H. influenzae strain Rd genome (44, 45). Following a transformation event, the acquired DNA is inserted into the H. influenzae chromosome through homologous recombination. Horizontal transfer of genetic material into H. influenzae also appears to occur by phage-mediated transduction, as many genetically diverse regions of the H. influenzae chromosome contain or are flanked by phage-related sequences (1, 32, 37).

While individual fitness characteristics of bacteria foster the survival of individual organisms, the population dynamics of bacteria describe fitness characteristics that foster the survival of the group. In short, the sum of the specialized fitnesses of individual bacteria ensures the survival of the population in various environments. Thus, the gene products required for bacterial survival in one environmental niche may not be required in another niche. Over time and under the influence of natural selection, the gene contents of organisms from the same species living in different niches are altered to reflect the necessity of certain gene products and the dispensability of others. Bacterial factors that are highly diverse are the most susceptible to this process of selection.

Thus, natural genetic variation can be exploited to identify potential virulence genes that are more prevalent in bacteria cultured from certain environments than from others or that are more prevalent in certain infectious processes than others. In this study, we used epidemiologic analyses to identify specific H. influenzae genes that have been preserved among specific bacterial populations. Specifically, we used subtractive hybridization to identify candidate gene regions that are unique to one otitis media NT H. influenzae strain and not present a strain from the throat of a healthy child. Furthermore, we used population prevalence analyses to identify genes that have been retained by otitis media NT H. influenzae isolates but that are not present in throat strains.

A previous study from our laboratory (37) used techniques similar to those described here and identified a strong association of lic2B, which encodes a glycosyltransferase active in modifying lipooligosaccharide of H. influenzae, with otitis media. Another study from our laboratory (13) identified the association of hmw, which encodes a nonpilus adhesin of H. influenzae, with otitis media. In addition, the hif locus, which encodes the hemagglutinating pili of H. influenzae, was more prevalent among throat strains than among middle ear strains; and hia, which encodes an autotransporter protein that also serves as an H. influenzae adhesin, showed no differential prevalence among these populations of H. influenzae strains (13).

In the present study, using genomic subtraction of NT H. influenzae middle ear strain G622 with the driver strain NT H. influenzae strain 23221, which was isolated from the throat of a healthy child, followed by analysis of gene prevalence among a large panel of middle ear and throat NT H. influenzae strains, we have newly identified nine genetic regions that are potentially involved in the pathogenesis of acute otitis media and confirmed the involvement of an adhesin gene region. The nine newly identified genetic regions may represent metabolic pathways necessary for NT H. influenzae survival in a unique environment, such as the middle ear as opposed to the throat. Identification of sJPX161 (representative of HI0113, which is homologous to hemR, which encodes a hemin receptor) and sJPX163 (representative of HI0661, which is homologous to hgpB, which encodes a hemoglobin-haptoglobin binding protein) (31) as being significantly more common among otitis media isolates suggests that iron metabolism may be important in the survival of NT H. influenzae in the middle ear space. Alternatively, the new genetic regions that we identified may represent virulence genes that contribute to the pathogenesis of infection and/or disease in that environment.

Seven of the nine newly identified genetic regions that were significantly more prevalent among middle ear strains than throat strains were also significantly more prevalent among type b isolates, which cause bloodstream and invasive infections, than among NT H. influenzae isolates that asymptomatically colonize the pharynx. This observation suggests that these seven genetic regions may be associated with the pathogenesis of infection and disease, irrespective of the infection site. In addition, sJPX140 was absent from 100% of the type b strains. The nucleotide sequences of sJPX140 are present in hmwA, which encodes the high-molecular-weight adhesin of NT H. influenzae and which has previously been found in nontypeable strains, uniformly absent in type b strains, and variably present in non-type b encapsulated strains (40). A previous report from our laboratory described the increased prevalence of hmw among middle ear strains compared to that among strains isolated from the throats of healthy children (13). Furthermore, sJPX170 was found in 39.4% of throat strains and 45.5% of middle ear strains but in only 1 (2.6%) of the type b strains. This fragment has no homology with other genes contained in the National Center for Biotechnology Information GenBank database, so the function of its gene product remains unknown.

Six of the newly identified genetic regions (sJPX163, sJPX84, sJPX161, sJPX176, sJPX147, and sJPX101) were present in 98.4% to 100% of the 121 middle ear isolates tested, suggesting that these regions may be very important to the pathogenesis of otitis media in the middle ear. However, the high prevalence of the six genetic regions in the throat isolates (73 to 92%) suggests that factors other than these six genetic regions may additionally be important for the pathogenesis of otitis media.

Pairwise analysis of the nine newly identified genetic regions, plus the two previously identified otitis media virulence genes lic2B and hmwA, revealed 41 statistically significant associations (Table (Table3).3). We hypothesize that certain combinations of these genetic regions may have an important role in the pathogenesis of otitis media. These regions may act as markers for pathogenic strains or as actual virulence genes that code proteins synergistically involved in the pathogenesis of otitis media. It seems plausible that different combinations of genes may allow different “pathotypes” of H. influenzae to cause otitis media. On the basis of our classification tree analysis (Fig. (Fig.3),3), we identified at least two definable otitis media pathotypes among NT H. influenzae strains: pathotype 1 isolates contain sJPX84, sJPX101, sJPX147, sJPX161, sJPX163, sJPX176, and hmwA. Pathotype 2 isolates contain sJPX84, sJPX101, sJPX147, sJPX161, sJPX163, and sJPX176; lack hmwA; but have lic2B. Although the classification tree, as constructed, provides a good fit to our data, the analysis is limited by a relatively small sample size. Thus, this classification tree serves as an initial attempt to identify the pathotypes of isolates that cause otitis media. With an increase in sample size and the addition of future putative virulent gene regions, we will be able to gain a better understanding of the pathogenesis of otitis media and to identify in greater depth the pathotypes of isolates that cause otitis media.

The present study identifies the presence or the absence of specific genes or gene regions and provides no information on the association of allelic differences (if they exist) among these genes with otitis media. Such studies are in progress. Other studies have documented allelic differences in hmw, which encodes the high-molecular-weight adhesin (7, 13, 52) represented by sJPX140, and possibly in hgpB, which encodes a hemoglobin and hemoglobin-haptoglobin binding protein (30), represented by sJPX163. Furthermore, our findings do not address the important role of the phase-variable expression of virulence genes in the pathogenesis of otitis media. Our current model holds that natural selection and clonal expansion provide a diverse population of bacteria that can occupy various environmental niches, and phase variation provides a rapid response mechanism that allows the bacteria to adapt to quickly changing environments. Interestingly, two of the sPCR fragments found in this study to be more prevalent among otitis media strains represent genes that appear to be phase variable; sJPX163 represents hgpB, whose coding region contains multiple tetranucleotide (CCAA) repeats (31); and sJPX140 represents hmwA, whose upstream region contains a number of septanucleotide (ATCTTTC) repeats (11, 52).

Another observation from this study was the relatively high frequency of nonhemolytic H. haemolyticus isolates among the isolates from healthy throats. We examined our collections of traditionally defined H. influenzae isolates and found that more than 20% of the strains isolated from the throats of healthy children did not contain the iga gene or react with the P6 monoclonal antibody. These findings, together with the dependence of these strains on both the X and the V growth factors and their lack of hemolytic activity, suggest that they are nonhemolytic variants of H. haemolyticus (5, 28). These observations are further supported by previous studies showing discrepant phylogenetic and taxonomic relationships between a small number of H. influenzae and H. haemolyticus isolates (23, 27). More precise relationships between H. influenzae and H. haemolyticus are unclear, however, because H. haemolyticus is infrequently distinguished from H. influenzae among the isolates of the normal throat flora and has never been implicated in human disease; thus, the organism has historically received little attention. Further studies to determine the potential roles of these variant strains in human health and disease seem warranted.

Acknowledgments

We thank Timothy Murphy for supplying us with P6 monoclonal antibody 7F3 and his helpful discussions about Haemophilus influenzae variant isolates and Kirk W. McCrea for both technical assistance and helpful discussions.

This study was supported by R01 award DC05840 to J.R.G. from the National Institute on Deafness and Other Communication Disorders.

Footnotes

[down-pointing small open triangle]Published ahead of print on 27 September 2006.

REFERENCES

1. Bergman, N. H., and B. J. Akerley. 2003. Position-based scanning for comparative genomics and identification of genetic islands in Haemophilus influenzae type b. Infect. Immun. 71:1098-1108. [PMC free article] [PubMed]
2. Berman, S. 1995. Current concepts: otitis media in children. N. Engl. J. Med. 332:1560-1565. [PubMed]
3. Block, S. L., J. Hedrick, C. J. Harrison, R. Tyler, A. Smith, R. Findlay, and E. Keegan. 2004. Community-wide vaccination with the heptavalent pneumococcal conjugate significantly alters the microbiology of acute otitis media. Pediatr. Infect. Dis. J. 23:829-833. [PubMed]
4. Bou, R., A. Dominguez, D. Fontanals, I. Sanfeliu, I. Pons, J. Renau, V. Pineda, E. Lobera, C. Latorre, M. Majo, and L. Salleras. 2000. Prevalence of Haemophilus influenzae pharyngeal carriers in the school population of Catalonia. Eur. J. Epidemiol. 16:521-526. [PubMed]
5. Brauer, A. L., A. J. Lesse, S. Sethi, X. Cai, M. Kilian, and T. F. Murphy. 2006. Haemophilus haemolyticus: a human respiratory tract commensal to be distinguished from Haemophilus influenzae. Abstr. 106th Gen. Meet. Am. Soc. Microbiol. 2006. American Society for Microbiology, Washington, D.C. [PubMed]
6. Breiman, L., J. H. Friedman, R. A. Olshen, and C. J. Stone. 1984. Classification and regression trees. Wadsworth, Belmont, CA.
7. Buscher, A. Z., K. Burmeister, S. J. Barenkamp, and J. W. St. Geme III. 2004. Evolutionary and functional relationships among the nontypeable Haemophilus influenzae HMW family of adhesins. J. Bacteriol. 186:4209-4217. [PMC free article] [PubMed]
8. Campos, J. M. 1999. Haemophilus, p. 604-613. In P. R. Murray, E. Baron, M. A. Fallen, F. C. Tenover, and R. H. Yolken (ed.), Manual of clinical microbiology, 7th ed. ASM Press, Washington, D.C.
9. Casey, J. R., and M. E. Pichichero. 2004. Changes in frequency and pathogens causing acute otitis media in 1995-2003. Pediatr. Infect. Dis. J. 23:824-828. [PubMed]
10. Casselbrant, M. L., E. M. Mandel, P. A. Fall, H. E. Rockette, M. Kurs-Lasky, C. D. Bluestone, and R. E. Ferrell. 1999. The heritability of otitis media: a twin and triplet study. JAMA 282:2125-2130. [PubMed]
11. Dawid, S., S. J. Barenkamp, and J. W. St. Geme III. 1999. Variation in expression of the Haemophilus influenzae HMW adhesins: a prokaryotic system reminiscent of eukaryotes. Proc. Natl. Acad. Sci. USA 96:1077-1082. [PMC free article] [PubMed]
12. Dhooge, I., M. Vaneechoutte, G. Claeys, G. Verschraegen, and P. Van Cauwenberge. 2000. Turnover of Haemophilus influenzae isolates in otitis-prone children. Int. J. Pediatr. Otorhinolaryngol. 54:7-12. [PubMed]
13. Ecevit, I. Z., K. W. McCrea, M. M. Pettigrew, A. Sen, C. F. Marrs, and J. R. Gilsdorf. 2004. Prevalence of hifBC, hmw1A, hmw2A, hmwC, and hia genes in Haemophilus influenzae isolates. J. Clin. Microbiol. 42:3065-3072. [PMC free article] [PubMed]
14. Eskola, J., and T. Kilpi. 2000. Potential of bacterial vaccines in the prevention of acute otitis media. Pediatr. Infect. Dis. J. 19S:S72-S78. [PubMed]
15. Eskola, J., T. Kilpi, A. Palmu, J. Jokinen, J. Haapakoski, E. Herva, A. Takala, H. Kayhty, P. Karma, R. Kohberger, G. Siber, P. H. Makelä, and The Finnish Otitis Media Study Group. 2001. Efficacy of a pneumococcal conjugate vaccine against acute otitis media. N. Engl. J. Med. 344:403-409. [PubMed]
16. Faden, H., L. Duffy, A. Williams, D. A. Krystofik, and J. Wolf. 1995. Epidemiology of nasopharyngeal colonization with nontypeable Haemophilus influenzae in the first 2 years of life. J. Infect. Dis. 172:132-135. [PubMed]
17. Farjo, R. S., B. Foxman, M. J. Patel, L. Zhang, M. M. Pettigrew, S. I. McCoy, C. F. Marrs, and J. R. Gilsdorf. 2004. Diversity and sharing of Haemophilus influenzae strains colonizing healthy children attending daycare centers. Pediatr. Infect. Dis. J. 23:41-46. [PubMed]
18. Fleischmann, R. D., M. D. Adams, O. White, R. A. Clayton, E. F. Kirkness, A. R. Kerlavage, C. J. Bult, J. F. Tomb, B. A. Dougherty, J. M. Merrick, K. McKenney, G. Sutton, W. FitzHugh, C. Fields, J. D. Gocayne, J. Scott, R. Shirley, L.-I. Liu, A. Glodek, J. M. Kelley, J. F. Weidman, C. A. Phillips, T. Spriggs, E. Hedblom, M. D. Cotton, T. R. Utterback, M. C. Hanna, D. T. Nguyen, D. M. Saudek, R. C. Brandon, L. D. Fine, J. L. Fritchman, J. L. Fuhrmann, N. S. M. Geoghagen, C. L. Gnehm, L. A. McDonald, K. V. Small, C. M. Fraser, H. O. Smith, and J. C. Venter. 1995. Whole-genome random sequencing and assembly of Haemophilus influenzae Rd. Science 269:496-512. [PubMed]
19. Fontanals, D., R. Bou, I. Pons, I. Sanfeliu, A. Dominguez, V. Pineda, J. Renau, C. Munoz, C. Latorre, and F. Sanches. 2000. Prevalence of Haemophilus influenzae carriers in the Catalan preschool population. Eur. J. Clin. Microbiol. Infect. Dis. 19:301-304. [PubMed]
20. Gilsdorf, J. R., H. Y. Chang, K. W. McCrea, and L. O. Bakaletz. 1992. Comparison of hemagglutinating pili of Haemophilus influenzae type b with similar structures of nontypeable H. influenzae. Infect. Immun. 60:374-379. [PMC free article] [PubMed]
21. Gratten, M., J. Montgomery, G. Gerega, H. Gratten, H. Siwi, A. Poli, and G. Koki. 1989. Multiple colonization of the upper respiratory tract of Papua New Guinea children with Haemophilus influenzae and Streptococcus pneumoniae. Southeast Asian J. Trop. Med. Public Health 20:501-509. [PubMed]
22. Harabuchi, Y., H. Faden, N. Yamanaka, L. Duffy, J. Wolf, and D. Krystofik. 1994. Nasopharyngeal colonization with nontypeable Haemophilus influenzae and recurrent otitis media. J. Infect. Dis. 170:862-866. [PubMed]
23. Hedegaard, J., H. Okkels, B. Bruun, M. Kilian, K. K. Mortensen, and N. Nørskov-Lauritsen. 2001. Phylogeny of the genus Haemophilus as determined by comparison of partial infB sequences. Microbiology 147:2599-2609. [PubMed]
24. Kilian, M. 2005. Genus III. Haemophilus Winslow, Broadhurst, Buchanan, Krumwiede, Rogers and Smith 1917, 561AL. In Bergey's manual of systematic bacteriology, 2nd ed. Springer-Verlag, New York, N.Y.
25. Klein, J. O. 1997. Role of nontypeable Haemophilus influenzae in pediatric respiratory tract infections. Pediatr. Infect. Dis. J. 16:S5-S8. [PubMed]
26. Krasan, G. P., D. Cutter, S. L. Block, and J. W. St. Geme III. 1999. Adhesin expression in matched nasopharyngeal and middle ear isolates of nontypeable Haemophilus influenzae from children with acute otitis media. Infect. Immun. 67:449-454. [PMC free article] [PubMed]
27. Lu, J. J., C. L. Perng, S. Y. Lee, and C. C. Wan. 2000. Use of PCR with universal primers and restriction endonuclease digestions for detection and identification of common bacterial pathogens in cerebrospinal fluid. J. Clin. Microbiol. 38:2076-2080. [PMC free article] [PubMed]
28. McCrea, K. W., J. Xie, D. Mukundan, M. Patel, C. F. Marrs, and J. R. Gilsdorf. 2006. Phylogenetic relationships between Haemophilus influenzae and Haemophilus haemolyticus. Abstr. 106th Annu. Meet. Am. Soc. Microbiol. 2006. American Society for Microbiology, Washington, D.C.
29. Moller, L. V., A. G. Regelink, H. Grasselier, J. E. Dankert-Roelse, J. Dankert, and L. van Alphen. 1995. Multiple Haemophilus influenzae strains and strain variants coexist in the respiratory tract of patients with cystic fibrosis. J Infect. Dis. 172:1388-1392. [PubMed]
30. Morton, D. J., L. L. Madore, A. Smith, T. M. VanWagoner, T. W. Seale, P. W. Whitby, and T. L. Stull. 2005. The heme-binding lipoprotein (HbpA) of Haemophilus influenzae: role in heme utilization. FEMS Microbiol. Lett. 253:193-199. [PubMed]
31. Morton, D. J., and T. L. Stull. 1999. Distribution of a family of Haemophilus influenzae genes containing CCAA nucleotide repeating units. FEMS Microbiol. Lett. 174:303-309. [PubMed]
32. Munson, R. S., Jr., A. Harrison, A. Gillaspy, W. C. Ray, M. Carson, D. Armbruster, J. Gipson, M. Gipson, L. Johnson, L. Lewis, D. W. Dyer, and L. O. Bakaletz. 2004. Partial analysis of the genomes of two nontypeable Haemophilus influenzae otitis media isolates. Infect. Immun. 72:3002-3010. [PMC free article] [PubMed]
33. Murphy, T. F., J. M. Bernstein, D. M. Dryja, A. A. Campagnari, and M. A. Apicella. 1987. Outer membrane protein and lipooligosaccharide analysis of paired nasopharyngeal and middle ear isolates in otitis media due to nontypable Haemophilus influenzae: pathogenic and epidemiologic observations. J. Infect. Dis. 156:723-731. [PubMed]
34. Murphy, T. F., C. Kirkham, and D. J. Sikkema. 1992. Neonatal, urogenital isolates of biotype 4 nontypeable Haemophilus influenzae express a variant P6 outer membrane protein molecule. Infect. Immun. 60:2016-2022. [PMC free article] [PubMed]
35. Murphy, T. F., S. Sethi, K. L. Klingman, A. B. Brueggemann, and G. V. Doern. 1999. Simultaneous respiratory tract colonization by multiple strains of nontypeable Haemophilus influenzae in chronic obstructive pulmonary disease: implications for antibiotic therapy. J. Infect. Dis. 180:404-409. [PubMed]
36. Musser, J. M., S. J. Barenkamp, D. M. Granoff, and R. K. Selander. 1986. Genetic relationships of serologically nontypable and serotype b strains of Haemophilus influenzae. Infect. Immun. 52:183-191. [PMC free article] [PubMed]
37. Pettigrew, M. M., B. Foxman, C. F. Marrs, and J. R. Gilsdorf. 2002. Use of genomic subtraction to identify the LOS biosynthesis gene lic2b as a putative virulence factor in strains of nontypeable Haemophilus influenzae that cause otitis media. Infect. Immun. 70:3551-3556. [PMC free article] [PubMed]
38. Porras, O., D. A. Caugant, B. Gray, T. Lagergard, B. R. Levin, and C. Svanborg-Eden. 1986. Difference in structure between type b and nontypable Haemophilus influenzae populations. Infect. Immun. 53:79-89. [PMC free article] [PubMed]
39. Poulsen, K., J. Reinholdt, and M. Kilian. 1992. A comparative genetic study of serologically distinct Haemophilus influenzae type 1 immunoglobulin A1 proteases. J. Bacteriol. 174:2913-2921. [PMC free article] [PubMed]
40. Rodriguez, C. A., V. Avadhanula, A. Z. Buscher, A. L. Smith, J. W. St. Geme III, and E. E. Adderson. 2003. Prevalence and distribution of adhesins in invasive non-type b encapsulated Haemophilus influenzae. Infect. Immun. 71:1635-1642. [PMC free article] [PubMed]
41. Saito, M., A. Umeda, and S. Yoshida. 1999. Subtyping of Haemophilus influenzae strains by pulsed-field gel electrophoresis. J. Clin. Microbiol. 37:2142-2147. [PMC free article] [PubMed]
42. Sambrook, J., E. F. Fritsch, and T. Maniatis. 1989. Molecular cloning: a laboratory manual, 2nd ed. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, N.Y.
43. Samuelson, A., A. Freijd, J. Jonasson, and A. A. Lindberg. 1995. Turnover of nonencapsulated Haemophilus influenzae in the nasopharynges of otitis-prone children. J. Clin. Microbiol. 33:2027-2031. [PMC free article] [PubMed]
44. Smith, H. O., M. L. Gwinn, and S. L. Salzberg. 1999. DNA uptake signal sequences in naturally transformable bacteria. Res. Microbiol. 150:603-616. [PubMed]
45. Smith, H. O., J.-F. Tomb, B. A. Dougherty, R. D. Fleischmann, and J. C. Venter. 1995. Frequency and distribution of DNA uptake signal sequences in the Haemophilus influenzae Rd genome. Science 269:538-540. [PubMed]
46. Smith-Vaughan, H. C., A. J. Leach, T. M. Shelby-James, K. Kemp, D. J. Kemp, and J. D. Mathews. 1996. Carriage of multiple ribotypes of non-encapsulated Haemophilus influenzae in aboriginal infants with otitis media. Epidemiol. Infect. 116:177-183. [PMC free article] [PubMed]
47. St. Geme, J. W., III, S. Falkow, and S. J. Barenkamp. 1993. High-molecular-weight proteins of nontypable Haemophilus influenzae mediate attachment to human epithelial cells. Proc. Natl. Acad. Sci. USA 90:2875-2879. [PMC free article] [PubMed]
48. St. Sauver, J., C. F. Marrs, B. Foxman, P. Somsel, R. Madera, and J. R. Gilsdorf. 2000. Risk factors for otitis media and carriage of multiple strains of Haemophilus influenzae and Streptococcus pneumoniae. Emerg. Infect. Dis. 6:622-630. [PMC free article] [PubMed]
49. Therneau, T. M., and B. Atkinson. 2006. rpart: recursive partitioning. [Online.] http://mayoresearch.mayo.edu/mayo/research/biostat/splusfunctions.cfm.
50. Trottier, S., K. Stenberg, and C. Svanborg-Eden. 1989. Turnover of nontypable Haemophilus influenzae in the nasopharynges of healthy children. J. Clin. Microbiol. 27:2175-2179. [PMC free article] [PubMed]
51. Uhari, M., K. Mantysaari, and M. Niemela. 1996. A meta-analytic review of the risk factors for acute otitis media. Clin. Infect. Dis. 22:1079-1083. [PubMed]
52. van Schilfgaarde, M., P. van Ulsen, P. Eijk, M. Brand, M. Stam, J. Kouame, L. van Alphen, and J. Dankert. 2000. Characterization of adherence of nontypeable Haemophilus influenzae to human epithelial cells. Infect. Immun. 68:4658-4665. [PMC free article] [PubMed]
53. Vitovski, S., K. T. Dunkin, A. J. Howard, and J. R. Sayers. 2002. Nontypeable Haemophilus influenzae in carriage and disease. JAMA 287:1699-1705. [PubMed]
54. Weiser, J. 2000. The generation of diversity by Haemophilus influenzae. Trends Microbiol. 8:433-435. [PubMed]
55. Xie, J., L. Zhang, B. Foxman, and C. F. Marrs. 2006. Molecular epidemiologic identification of Escherichia coli genes potentially involved in movement from the intestinal tract to the vagina and bladder. J. Clin. Microbiol. 44:2434-2441. [PMC free article] [PubMed]
56. Zhang, L., B. W. Gillespie, C. F. Marrs, and B. Foxman. 2001. Optimization of a fluorescent-based phosphor imaging dot blot DNA hybridization assay to assess E. coli virulence gene profiles. J. Microbiol. Methods 44:225-233. [PubMed]

Articles from Journal of Clinical Microbiology are provided here courtesy of American Society for Microbiology (ASM)
PubReader format: click here to try

Formats:

Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...

Links

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...