Microbiological characterization of Streptococcus pneumoniae and non-typeable Haemophilus influenzae isolates as primary causes of acute otitis media in Bulgarian children before the introduction of conjugate vaccines

Background Pneumococcal and Haemophilus influenzae type b (Hib) vaccines were introduced in our national immunisation program in April 2010. The aims of this retrospective, laboratory-based study were to determine the serotypes and antibiotic resistance of Streptococcus pneumoniae and H. influenzae isolates from middle ear fluid (MEF) collected before the introduction of immunization. Methods S. pneumoniae (n = 128) and H. influenzae (n = 40) strains isolated from MEF of children with AOM between 1994 and 2011 were studied. MICs were determined by a microdilution assay. Serotyping of S. pneumoniae was done by Quellung method and PCR capsular typing was used for H. influenzae. Macrolide resistance genes were detected by PCR for erythromycin resistant S. pneumoniae (ERSP). DNA sequencing of ftsI gene was performed for ampicillin nonsusceptible H. influenzae. Results The most common serotypes found among children with pneumococcal AOM were 19 F (20.3%), 6B (15.6%), and 19A (10.9%). The potential coverage rates by the PCV7, PCV10 and PCV13 of children aged < 5 years were 63.6%, 66.4% and 85.5%, respectively. Reduced susceptibility to oral penicillin was seen in 68.1%; resistance to erythromycin was 46.9%. We found erm(B) gene in 56.7% of the ERSP, mef(E) gene in 25%; 15% harbored both genes erm(B) + mef(E) and 3.3% had mutations of L4 ribosomal protein. Of the 40 H. influenzae isolates 97.5% were nontypeable. Nonsusceptibility to ampicillin occurred in 25%. Ampicillin resistance groups were: β-lactamase-positive ampicillin resistant (BLPAR) strains (10%), β-lactamase-negative ampicillin resistant (BLNAR) strains (12.5%) and β-lactamase-positive amoxicillin-clavulanate resistant (BLPACR) strains (2.5%). Among BLNAR and BLPACR most of the isolates (5/6) belonged to group II, defined by the Asn526Lys substitution. Conclusions The levels of antibiotic resistance among S. pneumoniae and H. influenzae causing severe AOM in children are high in our settings. The existence of multidrug-resistant S. pneumoniae serotype 19A is of particular concern. The rate of BLNAR and BLPACR strains among H. influenzae isolates was 15%.


Background
Acute otitis media (AOM) is the most common infection in children and the most common reason for which antibiotics are prescribed in this age group [1]. Currently, Streptococcus pneumoniae along with nontypeable Haemophilus influenzae are the major pathogens of AOM cases [2,3]. In a study of bacterial etiology of AOM between 2009 and 2010 in Bulgaria, the most frequently isolated pathogen was S. pneumoniae (47.6%), followed by H. influenzae (27.6%) [4].
Vaccination against pneumococcal diseases with PCV10 was introduced in Bulgaria in our national immunization program (NIP) for the first time in April, 2010. PCV7 was not used in Bulgaria prior to PCV10 introduction. Additionally, H. influenzae type b (Hib) conjugate vaccine was also included for routine immunization of infants as part of a pentavalent vaccine since 1 st April, 2010. Both conjugate vaccines are free of charge.
The prevalence of antibiotic-resistant S. pneumoniae has increased globally before the pre-vaccine era. The pattern of changes in antimicrobial susceptibility varies among serotypes and geographic regions [9]. Prior to PCV7 introduction, the major problem with antimicrobial resistance in S. pneumoniae worldwide was nonsusceptibility to penicillin, macrolides and multidrug resistance [10].
Resistance to β-lactams in H. influenzae, especially to ampicillin/amoxicillin, ranging from 3 to 65% in some countries is an increasing problem [11]. Although, ampicillin resistance is predominantly mediated by the βlactamase production, alterations in penicillin-binding proteins (PBPs), mainly PBP 3, due to ftsI amino acid substitutions occur in some parts of the world [12]. In addition, β-lactamase positive, amoxicillin-clavulanate -resistant strains have been described due to the combination of both mechanisms [13].
The objectives of our study were to determine serotypes and antibiotic susceptibility of S. pneumoniae and H. influenzae isolates from middle ear fluid (MEF) cultures collected before the introduction of conjugate vaccines in the NIP. The presence of macrolide resistance determinants in S. pneumoniae isolates and the prevalence of mechanisms of ampicillin resistance in H. influenzae strains were also studied.

Patients and isolates
Pediartic (< 14 years of age) S. pneumoniae (n = 128) and H. influenzae (n = 40) isolates from middle ear fluid cultures obtained from children with the clinical diagnosis "acute otitis" complicated with otorrhea were examined. The AOM episodes were confirmed by an otorrinolaryngologist or pediatrician. The strains were isolated from several microbiological laboratories throughout Bulgaria between 1994 and 2011 and isolates had been sent in our department on a voluntary basis through a passive, laboratory-based study for retrospective analysis. Only one episode of AOM or one isolate of any of the two pathogens was included per patient, except for one child from whom both were detected from the same sample (mixed infection). Serotypes of S. pneumoniae were distributed per age group (≤ 2 years, 2-5 years, and >5-14 years); 90% of H. influenzae strains were isolated from children aged less than 5 years. None of the children with AOM had been immunized with pneumococcal or Hib conjugate vaccines. Five pneumococcal strains isolated in 2011 were also included because they were from children which had not received vaccines.
S. pneumoniae strains were confirmed with both methods -optochin susceptibility test and bile solubility. H. influenzae isolates were identified by Gram-stain morphology, growth requirements for factor V, X, and porphyrin (A.L.A. disks, Remel, Lenexa, Kansas, USA), biochemical tests according to standard methods and the presence of iga protease gene determined by conventional PCR [14,15].

Antimicrobial susceptibility test
The antibiotic susceptibilities were determined by the broth microdilution method according to guidelines of Clinical and Laboratory Standards Institute (CLSI) [16]. Minimal inhibitory concentrations (MICs) were performed on Microtiter plates (Sensititre, Trek Diagnostic Systems Ltd., UK). STR6F MIC plate was inoculated with cationadjusted Mueller-Hinton broth and 5% lysed horse blood for S. pneumoniae, and for H. influenzae HPB-Haemophilus/Streptococcus pneumoniae plate was inoculated with Haemophilus test broth. Final bacterial density was 5 × 10 5 CFU/ml and the plates were incubated at 35°C for 20-24 h in 5% CO 2 . The MICs were interpreted according to CLSI, 2010 [16]. All H. influenzae strains were tested for β-lactamase production by the chromogenic cephalosporin test with nitrocefin as substrate. Standard quality control strains were S. pneumoniae ATCC 49619 and H. influenzae ATCC 49247. Strains resistant to ≥ 3 classes of antibiotics were defined as multidrug resistant (MDR).

Serotyping
Serogrouping of S. pneumoniae was performed using the latex agglutination method (Pneumotest-Latex kit, Statens Serum Institut, Copenhagen, Denmark). Serotyping was performed later with the capsular Neufeld Quellung method by using some common factor antisera (SSI, Copenhagen, Denmark) [17]. H. influenzae strains were serotyped by a PCR capsule typing method, detecting the capsule-producing gene -bexA. The algorithm used for detection of the cap genes and for determining capsular serotypes was that proposed by Falla et al. [18].

Detection of macrolide resistance determinants in S. pneumoniae
The presence of erm(B) and mef(A) was evaluated concurrently in a duplex reaction by PCR for all erythromycinresistant S. pneumoniae (ERSP) isolates. The primer sets were as described by Sutcliffe et al. [19]. Gene amplification was performed using a Techgene-thermal cycler (Techne, England). Mef(A) and mef(E) were then distinguished by PCR restriction fragment length polymorphism analysis of the 348-bp mef amplicon with BamHI restriction enzyme (New England Biolabs), which has no restriction site in mef(E) and one in mef(A), generating two fragments of 284 and 64 bp [20]. Mutations in ribosomal protein L4 were determined as described elsewhere [21].

DNA sequencing analysis of ftsI gene in H. influenzae
The DNA fragment encoding the transpeptidase region of PBP 3 was amplified by PCR from strains which were nonsusceptible to ampicillin (MICs of ≥ 2 μg/ml). Amplification of ftsI gene was done using primers and PCR conditions described previously [22]. Nucleotide and amino acid sequences were analysed using DNAMAN 4.11 Software (Lynnon BioSoft, Canada). Deduced amino acid sequences were compared with the reference sequence obtained from the H. influenzae Rd strain. Strain ATCC 49247 was included as an H. influenzae BLNAR control strain.
We classified H. influenzae strains according to their ampicillin susceptibility, β-lactamase production, and on the basis of the DNA sequencing results obtained from ampicillin-nonsusceptible isolates. Four main classes were determined: β-lactamase negative, ampicillin-susceptible strains with MIC of ≤ 1 μg/ml (BLNAS); strains producing β-lactamase, and without mutations in the ftsI gene (BLPAR); β-lactamase negative, ampicillin-nonsusceptible strains (MICs of ≥ 2 μg/ml) which have amino acid substitutions in the ftsI gene (BLNAR); and isolates that have both mechanisms (β-lactamase production and amino acid substitutions in the ftsI gene)(BLPACR).

Serotype distribution among S. pneumoniae and H. influenzae
Serotypes of 128 S. pneumoniae isolates obtained from MEF are presented in Table 1. The strains are divided in three age groups: children aged < 2 years (50.0%), children aged >2 to 5 years (35.9%), and children aged > 5 y -14 y Non-vaccine serotypes Coverage by serotypes included in the current PCV was evaluated and the calculations were performed without accounting for possible cross-protection. The potential coverage rates by the PCV7, PCV10 and PCV13 for children aged < 5 years were 63.6%, 66.4% and 85.5%, respectively.

Antimicrobial susceptibility
Antibiotic susceptibility data for 128 S. pneumoniae isolates are presented in Table 2. Reduced susceptibility to penicillin (oral penicillin V) was seen in 68.1% of the strains. This included strains (28.1%) with high level of resistance to penicillin (MIC ≥ 2 mg/L). The rate of resistance to macrolides (erythromycin) was 46.9%. Susceptibility results to amoxicillin shows 6.2% intermediately resistant strains and 4.7% fully resistant isolates. In a pre-vaccine era the percentage of multidrug-resistant strains was 53.9% (69/128). In order to observe any changes in antibiotic resistance over the time, all pneumococcal strains were divided in two periods depending on the year when the strain was isolated: 1994-2005 and the more representative pre-vaccine period 2006-2011 (Table 2). During the second period we revealed increased resistance to erythromycin from 36.7% to 53.2%, and to clindamycin from 22.4% to 40.5%, respectively. Non-susceptibility rates to amoxicillin, cefuroxime sodium and ceftriaxone were found in 8.2%, 28.6% and 4.1%, respectively in the first period, and were increased up to 12.7%, 41.8% and 8.9%, respectively in the second period 2006-2011.
A total of 40 H. influenzae strains were tested for susceptibility to antibiotics. High-nonsusceptibility rates were found to trimethoprim-sulfamethoxazole -35.0% (14/40), and to ampicillin -25.0% (10/40). Of the 10 ampicillin nonsusceptible H. influenzae strains, 9 were fully resistant to ampicillin (MIC ≥ 4 mg/L) and one was intermediately susceptible (MIC = 2 mg/L) isolate (Table 3). Ceftriaxone and levofloxacin were the most active agents tested, and all isolates were fully susceptible to these agents.

Ampicillin-resistant mechanisms and amino acid substitutions of PBP 3 among MEF H. influenzae isolates
The total collection of 40 H. influenzae was grouped according to our definition of ampicillin resistance classes: BLNAS strains (n = 30; 75.0%), BLPAR (n = 4; 10.0%) isolates without mutations in the ftsI gene, BLNAR strains (n = 5; 12.5%) with amino acid substitutions in the ftsI gene, and BLPACR (n = 1; 2.5%) strain that was β-lactamase positive, and had mutations in the critical region of ftsI gene with amino acid substitutions in PBP 3. Table 3 shows these ampicillin-resistance classes according to their β-lactamase production, susceptibilities to ampicillin and amoxicillin-clavulanic acid, and amino acid substitutions which present in ampicillin nonsusceptible strains. Of note, among six strains with amino acid substitutions in PBP 3 (5 BLNAR and one BLPACR), three were resistant to amoxicillin-clavulanic acid, and three strains were susceptible according to CLSI breakpoints. The nucleotide sequences of the ftsI gene were determined for 10 ampicillin-nonsusceptible H. influenzae strains (MICs ≥ 2 μg/ml) and their deduced amino acid sequences were compared with those of the ampicillinsusceptible Rd strain. Five β-lactamase nonproducing strains had significant amino acid substitutions in the ftsI gene, thereby we categorizing them as BLNAR strains. Among remaining 5 β-lactamase positive strains, one isolate had amino acid substitutions in PBP 3. According to the classification of Dabernat et al. [22] different mutation patterns were identified among BLNAR and BLPACR strains. The Arg517His substitution that defines group I was observed in one ampicillin nonsusceptible H. influenzae. Most isolates (n = 5) belonged to group II, defined by the Asn526Lys substitution, and Table 3 Distribution of ampicillin resistance classes of paediatric middle ear fluids H. influenzae isolates according to their β-lactamase production, susceptibilities to ampicillin and amoxiclav, and amino acid substitutions present in ampicillin nonsusceptible strains with mutations in ftsI gene  Table 4 Genotype distribution of 60 erythromycin-resistant S. pneumoniae (ERSP) isolates from middle ear fluids: association with serotypes and co-resistance to penicillin

Discussion
This is a retrospective study about serotype distribution and antibiotic resistance of MEF S. pneumoniae and H. influenzae isolates collected from children before introduction of PCV10 and Hib conjugate vaccines nationwide. Microbiological diagnosis of AOM is not done on a routine basis in Bulgaria and this mild disease is underrepresented in our data set. Also, the tympanocentesis is not a common practice in our country. Thus, bacterial cultures studied here most likely represent cases of severe otitus media or patients with a complicated AOM. The analysis included 128 S. pneumoniae isolates, of which 86% occurring in children aged ≤ 5 years. We found a relatively high rate of serotype 19F infections (20.3%), which is in accordance with the serotype distribution observed in the studies of AOM in German children [8], and the Finland [7]. Serotypes 6B and 19A were next most prominent, each comprising about 16% and 11%, respectively of the overall data set. The coverage calculations for the serotypes included in PCV7 without cross-protection was 63.6% in children with AOM aged < 5 years, which is almost the same as was the serotype coverage previously reported for invasive pneumococcal disease (IPD) for the same age group in Bulgaria [23]. The PCV10 increase the coverage by 2.8% of pneumococcal AOM (66.4%). The PCV13, which includes three serotypes in addition to the PCV10, would increase the coverage to 85.5%, mainly due to serotype 19A followed by serotypes 3 and 6A (Table 1). However the cross-protection of serotype 6B towards serotype 6A was not included in calculations for our study. We found 14 serotype 19A isolates, 13 of which were multidrug resistant.
The results showed a high rate of reduced oral penicillin nonsusceptibility (68%) and a very high rate of ERSP isolates (47%) among MEF strains. These rates are more than twice as high as previously recorded with IPD isolates in Bulgaria studied during the same year period [23]. These high resistance rates may be due to clonal spread of certain MDR pneumococcal serotypes such as 6B, 19 F and 19A. This resistance reflects the complexity of many of the severe otitis media episodes which included therapy failures. Notably, we observed that the percentages of macrolide, clindamycin, amoxicillin, cefuroxime sodium and ceftriaxone nonsusceptibility increased among S. pneumoniae strains isolated during the last 6 years of the study period. The high antibiotic consumption for the treatment of AOM represents the major pressure for resistance selection [24].
Overall, macrolide resistance in S. pneumoniae now is mainly due to erm(B) gene ( 71.7%), including 15% of isolates with a dual macrolide resistance mechanism (Table 4). This pattern of macrolide resistance determinants in the last years replaced the efflux mechanism of macrolide resistance (mef genotype), which was predominant among ERSP in Bulgaria before 2005 [25]. Macrolide-resistant S. pneumoniae strains were found to be co-resistant to oral penicillin (Table 4), and to other compounds, such as tetracycline and trimethoprimsulfamethoxazole (data not shown). Thus, the prevalence of MLS B genotype may be due to clonal spread of certain MDR pneumococcal serotypes as 6B, 19 F and 19A. Interestingly, we observed multi-resistant serotype 19A in many years ago. For the first time, this MDR serotype was found in 1995 with L4 mutations as the mechanism of macrolide resistance [21]. Since 2000 and after, erm (B) genotype was found in serotype 19A isolates (nine strains). Dual macrolide resistance mechanisms were observed recently, between 2008 and 2010 in two serotype 19A strains (Table 1 and Table 4). Both the emergence of a new clone or capsular switching might be associated with these changes in our 19A serotype isolates. A considerable increase in the disease caused by serotype 19A has been recorded after the introduction of PCV7 in some countries [24,26]. Of note, MDR serotype 19A isolates have also been identified in Israel and Finland before the large-scale use of conjugate vaccines as was the case in our study [27,28].
Our findings of 97.5% nontypeable H. influenzae strains isolated from AOM are in accordance with the literature [11,29]. Nontypeable H. influenzae are the major etiologic agents of respiratory tract infections in our country as was shown before [30]. Ampicillin/ amoxicillin nonsusceptible H. influenzae were found to be a quarter of our MEF isolates. Among BLNAR and BLPACR most of the isolates (5/6) belonged to Dabernat group II, defined by the Asn526Lys substitution. A high incidence of BLNAR H. influenzae strains has been reported in Japan, Spain and France [11,12,22]. The emergence of BLNAR and BLPACR strains among MEF H. influenzae isolates may have implications for the treatment of AOM infections, because amoxicillin and amoxicillin-clavulanate are the most common antibiotics used in the treatment of AOM in our country.
The main limitations of our study are: 1) it covers a long period of time, and the power is low; 2) isolates were sent by microbiology labs in hospitals, and we do not have clinical information on the type of AOM; 3) as isolates came from hospitals, they are not representative of S. pneumoniae and H. influenzae isolates causing otitis in children attending primary medical centers; 4) the numbers of S. pneumoniae and H. influenzae strains are small, because only a portion of children with AOM in Bulgaria are tested. Despite these limitations, we believe that they may represent a rough estimate of the serotype distribution and antibiotic resistance among a group of children with complicated AOM. This information will be invaluable when planning vaccination strategies to protect against otitis media.
In conclusion, the levels of antibiotic resistance among S. pneumoniae and H. influenzae causing severe otitis media in children are high in our settings. This has important therapeutic implications because such antibioticresistant isolates compromise the choice of macrolides and amoxicillin as treatment of AOM episodes in children. The existence of MDR S. pneumoniae serotype 19A before the active pneumococcal immunization is of particular concern. The rate of 15% BLNAR and BLPACR strains among H. influenzae MEF isolates may have implications for the treatment of AOM. Ongoing surveillance is needed for both pathogens in order to observe any changes in serotypes prevalence and antibiotic resistance after introduction of conjugate vaccines in our country.