Characterization of Nontypeable and Atypical Streptococcus pneumoniae Pediatric Isolates from 1994 to 2010

Abstract

Streptococcus pneumoniae is a major cause of bacteremia, meningitis, pneumonia, sinusitis, and acute otitis media in children. Although optochin susceptibility, bile solubility, and Quellung testing are the standards for identifying and differentiating pneumococci, there are several reports of nontypeable pneumococci that give inconsistent results with one or more of these tests. We characterized 52 isolates previously labeled as nontypeable pneumococci. Microbiological methods included repeating the Quellung reaction using a new and expanded group of antisera, optochin susceptibility and bile solubility tests, and automated Vitek 2 identification. Molecular methods included PCR detection of ply and psaA genes, multilocus sequence typing (MLST), 16S rRNA gene sequencing, and pyrosequencing. Of the 52 isolates, 38 (73%) were optochin susceptible, were psaA and ply positive, and could be serotyped by the Quellung reaction. The remaining 14 isolates, isolated from patients with otitis media (n = 6), bacteremia (n = 6), meningitis (n = 1), and pneumonia (n = 1), underwent further analysis. Three of these 14 isolates were nontypeable due to autoagglutination but were pneumococci by all tests and represented pneumococcal sequence types previously recognized by MLST. The 11 remaining isolates were optochin resistant, and 6 of these were bile soluble. Three of 11 were both psaA and ply positive and clustered with pneumococci by MLST (2 were bile soluble); 8 lacked psaA (5 ply positive, 4 bile soluble) and likely belonged to other Streptococcus species. In conclusion, few isolates were truly nontypeable by Quellung reaction, and MLST and the presence of psaA proved useful in distinguishing between atypical pneumococci and other streptococcal species.

INTRODUCTION

Streptococcus pneumoniae isolates are conventionally differentiated from closely related streptococcal species by optochin susceptibility and bile solubility testing (20). Typical pneumococci also have a capsular polysaccharide that produces a type-specific immune response and serves as the basis for pneumococcal serotyping. Currently, 92 serotypes have been identified (2, 8, 14). Although it is a widely used and reliable test, the accuracy of the Quellung reaction can vary depending on particular batches of antisera. Over the past decade, there have been improvements in the antisera, allowing enhanced identification of various serotypes—particularly those of pool B, which identifies serotypes 6A, 6B, 6C, and 6D (7).

Although optochin susceptibility, bile solubility, and Quellung testing are the standards for identifying and differentiating pneumococci, atypical pneumococci frequently give inconsistent results (20). Nonserotypeable (nontypeable) pneumococci are S. pneumoniae isolates that cannot be identified with the capsular polysaccharide typing sera. Nontypeable pneumococci can be divided into isolates that lack capsule genes, those that have capsule genes but are phenotypically unencapsulated, and those that phenotypically resemble S. pneumoniae but differ genetically from pneumococci (6).

Genotypic techniques can be utilized to help define the pneumococcus. Multilocus sequence typing (MLST) utilizes sequences from seven housekeeping loci and is useful in characterizing the isolate when the typing sera are not sufficient. Phylogenetic trees based on concatenated sequences of MLST loci can at least generally distinguish whether an isolate is truly S. pneumoniae (5). Furthermore, pneumococci can often be distinguished from nonpneumococci by amplification of the genes encoding pneumolysin (ply) and pneumococcal surface antigen A (psaA). The presence of psaA in particular has been found to correspond with encapsulation in strains and has been suggested to be the best indicator for genuine pneumococci (19).

Our objective was to characterize S. pneumoniae isolates previously classified as nontypeable from Texas Children's Hospital, 1994 to 2010, by using microbiological and molecular typing methods. We repeated Quellung testing using current antiserum pools and repeated optochin susceptibility and bile solubility tests. MLST was used to determine genetic relatedness by creating phylogenetic trees incorporating serotypeable and nontypeable pneumococci sequences available from the MLST website (www.mlst.net). Amplification of ply and psaA was used to further characterize strains, particularly those that were optochin resistant.

MATERIALS AND METHODS

Isolates.

S. pneumoniae isolates were obtained from the Clinical Microbiology Laboratory at Texas Children's Hospital between 1994 and 2010. The isolates were originally identified as pneumococci by the Clinical Microbiology Department at the time of isolation and were collected by the Infections Diseases laboratory as part of a surveillance study on Streptococcus pneumoniae, approved by the institutional review board at Baylor College of Medicine (9). The isolates were determined to be nontypeable by the Quellung reaction at the time of collection and were stored in horse blood at −80°C in the Infectious Diseases Laboratory until further use.

Optochin susceptibility.

Isolates were grown on tryptic soy agar plates with 5% sheep blood. The optochin disk test (BBL Taxo discs; Becton Dickinson Company, Sparks, MD) was performed according to the manufacturer's instructions. Zone sizes were measured, and isolates with zones of >14 mm were considered optochin susceptible. For isolates with a few colonies growing within an apparent zone, colonies from within the zone and from outside the zone were replated, isolated, and independently retested for optochin susceptibility.

Bile solubility.

Isolates were tested using the tube suspension method (12) and a 10% sodium deoxycholate suspension (Remel, Lenexa, KS) according to the manufacturer's instructions. Isolates with partial clearing were considered nonsoluble.

DNA preparation and PCR.

Isolates were grown on tryptic soy agar plates with 5% sheep blood. DNA was extracted on a QIAcube using a blood and tissue kit and the Gram-positive bacterial lysis protocol (Qiagen, Valencia, CA) as recommended. PCR amplification of the genes for pneumolysin (ply) and pneumococcal surface antigen A (psaA) was done using previously described primers and methods (19).

Multilocus sequence typing.

The housekeeping genes aroE, gdh, gki, recP, spi, xpt, and ddl were amplified using primers and conditions described at www.mlst.net. PCR products were sequenced at Beckman Coulter Genomics, Beverly, MA. Allele and ST designations were obtained at www.mlst.net. For isolates that had one or more unrecognized alleles, dendrograms were created using concatenated sequences to show the relationship to known pneumococcal sequence types (http://spneumoniae.mlst.net/sql/concatenate/isitaddto.asp). Neighbor-joining trees were constructed using MEGA5 software (18) and combined, concatenated sequences of aro, gdh, gki, recP, spi, and xpt. A set of reference sequences from both genuine, serotypeable S. pneumoniae and nontypeable, atypical S. pneumoniae, available at the MLST website, were downloaded and compared to the sequences of a subset of pediatric isolates.

Vitek identification, pyrosequencing, and 16S sequencing.

Isolates that were optochin resistant were further reclassified using the GP identification card on the Vitek 2 platform, software version 4.01, in the Clinical Microbiology Laboratory at Texas Children's Hospital. Pyrosequencing of three hypervariable regions (V1, V3, and V6) of the 16S rRNA gene was performed by the Molecular Microbiology Laboratory at Texas Children's Hospital as described previously (11). In short, bacterial DNA was isolated from culture, amplification was performed by utilizing universal primers targeting the conserved regions directly flanking the variable regions, and pyrosequencing targeted the variable region sequence for bacterial identification. The resulting sequence was then searched against the Ribosomal Database Project (RDP) maintained by Michigan State University (4). The complete 16S rRNA gene was also amplified by well-characterized universal primers (8F and 1541R), and the amplification products were submitted to Seqwright DNA Technology Services (Houston, TX) for Sanger sequencing. The resulting sequences were searched against RDP and GenBank.

RESULTS

Fifty-two isolates from Texas Children's Hospital (TCH) previously described as nontypeable pneumococci were characterized by optochin susceptibility testing, Quellung testing, PCR amplification of psaA and ply, and MLST (Table 1). Of these, 38 were serotyped by the Quellung reaction with the expanded group of antisera (group Ia), leaving 14 isolates for further study. The 14 isolates represented 0.5% of the total number of pneumococcal isolates collected from TCH within the time period. These were further characterized by bile solubility assay and by bacterial identification (Vitek automated system, pyrosequencing, and 16S rRNA gene analysis). All 52 isolates were classified into three groups (Table 1).

Table 1.

Summary of results

Group	Description	No. of isolates	No. positive for:
			Quellung reactiona	Optochin susceptibility	Bile solubility	psaAb	plyb
Ia	Typeable	38	38	38	NDc	38	38
Ib	Nontypeable (clump)	3	0	3	3	3	3
II	Nontypeable/atypical	3	0	0	2	3	3
IIIa	Nonpneumococci	5	0	0	3	0	5
IIIb	Nonpneumococci	3	0	0	1	0	0

^aCapsule swelling by the Quellung reaction.

^bDetermined by PCR.

^cND, not done.

Group I consisted of 41 strains that were optochin susceptible and psaA and ply positive. Among the 38 strains that were serotyped by Quellung reaction (group Ia) significant serotypes were 6B (3 isolates), 6C (16 isolates), and 38 (5 isolates). Other serotypes identified were 4, 9V, 10, 12, 14, 16, 18C, 19A, 19F, 38, and pool G (including serotypes 29, 35, and 42). MLST profiles for the typeable isolates separated the isolates into 24 sequence types. Group Ib contained 3 isolates that were optochin susceptible, bile soluble, and psaA and ply positive but not serotypeable due to autoagglutination (Table 2). All three isolates were obtained from children with recurrent otitis media. Their respective MLST profiles were ST2011, ST344, and ST448.

Table 2.

Nontypeable isolates with (group Ib) and without (group II) MLST designations

Isolate	Disease	Oa	Bb	PCR result for:		MLST result forc:								Putative serotyped
				psaA	ply	aro	gdh	gki	recP	spi	xpt	ddl	ST
Group Ib
28	Otitis	S	+	+	+	1	5	4	5	5	3	101	2011	6/NT
23	Otitis	S	+	+	+	8	37	9	29	2	12	53	344	NT/19F/6A
25	Otitis	S	+	+	+	8	5	2	27	2	11	71	448	NT
Group II
24	Otitis	R	+	+	+	5	60	1	-	6	3	-	ND	NA
27	Bacteremia	R	+	+	+	8	-	14	2	17	4	-	ND	NA
33	Otitis	R	−	+	+	-	-	14	-	6	-	-	ND	NA

^aOptochin susceptibility test. R, resistant; S, susceptible.

^bBile solubility test. +, soluble; −, insoluble.

^cST, sequence type; ND, not determined (unrecognized alleles or STs were not submitted to the MLST database for new designations, since the isolates were likely not S. pneumoniae), -, allele differed from alleles in the MLST database by 1 to 6%.

^dSerotypes previously associated with the ST in the MLST database (www.mlst.net). NT, nontypeable; NA, not applicable.

Group II consisted of three isolates, obtained from patients with otitis (2) and bacteremia (1), that were optochin resistant but carried the psaA and ply genes. Two of the three isolates were bile soluble by the method used. The isolates were omninegative by Quellung testing (Table 2) and represented new MLST profiles with two or more new alleles. The concatenated sequences were compared to reference sets in the MLST database by neighbor-joining analysis (Fig. 1). Isolates 24 and 27 clustered within the serotypeable “genuine pneumococcus” cluster, while isolate 33 did not. Analysis by Vitek identified isolate 24 as S. pneumoniae; isolates 27 and 33 were grouped as Streptococcus mitis/oralis. 16S rRNA gene Sanger sequencing and pyrosequencing methods did not distinguish well between streptococcal species, but the results were in accordance with Vitek and MLST results.

Fig 1.

Evolutionary relationships inferred using the neighbor-joining method (16). Isolates with TCH numbers are Texas Children's Hospital isolates (this study). Nine of 11 atypical isolates from group II and III were included; isolates 36 and 47 (group IIIB) did not produce all the necessary PCR products. ST, sequence type. All isolates with ST numbers are defined as authentic S. pneumoniae isolates. These were obtained from the MLST database (www.mlst.net). NT, nontypeable. Isolates with NT numbers were obtained from the MLST database and described as atypical “related to pneumococci” isolates (5).

Group III isolates were optochin resistant and psaA negative (Table 3); five were ply positive (group IIIa), and three lacked the ply gene (group IIIb). Four of the eight isolates were bile soluble. The isolates were obtained from patients with bacteremia (5), meningitis (1), pneumonia (1), and otitis (1). All group III isolates were negative by Quellung testing using the omni-sera. Six of eight isolates yielded PCR products for all seven MLST loci, but the majority of the sequences represented new alleles that were 94 to 99% identical to alleles in the MLST database. Characterization by Vitek classified the isolates as S. mitis/oralis species or Streptococcus parasanguinis. This was supported by 16S rRNA gene Sanger sequencing and pyrosequencing. Neighbor-joining tree analysis sorted all but one (isolate 50) of the group III isolates into a group separate from the typeable pneumococci.

Table 3.

Nonpneumococci without MLST designations

Isolate	Disease	Oa	Bb	PCR result for:		MLST result forc:								Identification by:
				psaA	ply	aro	gdh	gki	recP	spi	xpt	ddl	ST	Pyrosequencing	16Sd	Vitek
Group IIIa
17	Bacteremia	R	−	−	+	-	-	14	-	-	-	-	NT3	Streptococcus spp.	S. mitis	S. mitis/oralis
34	Meningitis	R	−	−	+	-	-	-	-	-	4	-	NT	Streptococcus spp.	S. mitis	S. mitis/oralis
42	Bacteremia	R	+	−	+	-	-	-	-	-	-	229	NT	Streptococcus spp.	S. mitis	S. mitis/oralis
50	Bacteremia	R	+	−	+	10	5	4	5	-	10	18	NT	Streptococcus spp.	S. mitis	S. mitis/oralis
53	Bacteremia	R	+	−	+	-	-	1	-	68	198	260	NT	Streptococcus spp.	S. mitis	S. mitis/oralis
Group IIIb
35	Otitis	R	−	−	−	-	-	-	-	-	-	-	NT	S. mitis	S. mitis	S. mitis/oralis
36	Pneumonia	R	+	−	−	-	13	14	4	—	4	14	NT	Streptococcus spp.	S. parasanguinis	S. parasanguinis
47	Bacteremia	R	−	−	−	—	-	-	4	94	239	-	NT	Streptococcus spp.	S. mitis	S. mitis/oralis

^aOptochin susceptibility test. R, resistant; S, susceptible.

^bBile solubility test. +, soluble; −, insoluble.

^cST, sequence type; NT, nontypeable (unrecognized alleles or STs were not submitted to the MLST database for new designations since the isolates were likely not S. pneumoniae); -, allele differed from alleles in the MLST database by 1 to 6%; —, no PCR product.

^d16S rRNA gene sequencing best hits (alignments were nearly identical between several Streptococcus spp. for each isolate).

DISCUSSION

Between 1993 and 2010, a subset of isolates from our pneumococcal collection have been designated nontypeable by the Quellung method. Several recent articles have reported findings of nontypeable S. pneumoniae and have highlighted the occasional difficulty in distinguishing S. pneumoniae isolates from closely related species, including S. mitis and Streptococcus pseudopneumoniae (1, 17, 20). We initiated this study to gain a better understanding of the phenotypic and genotypic characteristics of our nontypeable S. pneumoniae isolates.

Among 52 isolates previously described as nontypeable, 38 were in fact typeable when reanalyzed. Twenty of these were 6A, 6B, and 6C and likely could be identified because of the new pool B antisera. The 6C isolates reacted weakly with the omni-serum (weak swelling) and had a delayed reaction with pool B, while the 6C-specific serum caused a strong, immediate reaction, with both clumping and swelling. Other explanations of why previously nontypeable isolates were typeable upon reculture include passage of isolates on agar plates, differences in the batch quality and age of antisera, and human error.

Optochin susceptibility and bile solubility are the preferred tests to distinguish S. pneumoniae from closely related species, although reports have highlighted occasional contradictory results using these highly discriminatory tests (1, 13, 20). We used these tests and molecular methods to characterize our nontypeable isolates and found that only a handful were considered true nontypeables. Three isolates (group Ib) were pneumococci by both phenotypic and genotypic tests and could be considered “mechanically nontypeable” because of autoagglutination. Their respective MLST profiles (ST2011, ST344, and ST448) had previously been associated with nontypeable isolates in the MLST database; ST2011 was also associated with serotype 6, and ST344 was associated with 19F and 6A. Three optochin-resistant isolates from groups II and III clustered with genuine, serotypeable pneumococci by neighbor-joining analysis. The classification of group II isolates is ambiguous based on the methods used. While phenotypically distinct from true pneumococci, they maintain genotypic identity with regard to psaA, ply, and phylogenetic tree analysis. By MLST, these cluster with typeable S. pneumoniae, while the Vitek analysis identified two of three as S. mitis/oralis. All three were optochin resistant and omni-negative by the Quellung reaction. Although optochin susceptibility is often used as a benchmark of pneumococcal identification, the occurrence of optochin-resistant isolates is not extremely rare (3). These isolates still maintain capsular psaA and ply. However, the ply gene has been observed in other closely related streptococcal groups such as S. mitis (20) and is a poor marker for distinguishing S. pneumoniae, while psaA has been suggested to be the best identifier of true pneumococci independent of optochin susceptibility status (19). When all the findings are combined, it seems appropriate to classify these isolates as atypical pneumococci. Since nontypeable isolates have been associated with higher transformation rates than encapsulated strains (15), it is possible that the downregulation of capsule expression led to conditions that allowed for these isolates to become further divergent. The isolates in group III were optochin resistant, psaA negative, and Quellung test negative. The divergence observed by both molecular and microbiological methods suggests that, while they have certain similarities to genuine pneumococci, these isolates may be more accurately classified as nonpneumococcal streptococcal species. 16S sequencing and pyrosequencing were not useful in distinguishing between species, a finding previously discussed by Verhelst et al. (19) Earlier studies estimated a <60% homogeneity between species (10), suggesting that whole-genome sequencing would be necessary to determine where the atypical isolates fit systematically. All 14 isolates representing nontypeable/atypical or nonpneumococcal species were isolated between 1995 and 2004; current laboratory techniques at our hospital may not identify similar isolates as S. pneumoniae.

In conclusion, new antisera for the Quellung reaction greatly reduced the number of pneumococci designated nontypeable at Texas Children's Hospital. Optochin susceptibility testing was specific for classic pneumococci in our study and discriminated between typical and atypical isolates. Likewise, bile solubility testing was helpful, although not completely consistent with the other tests performed. The presence of the psaA gene and MLST together with neighbor-joining analysis proved useful in further discriminating between nontypeable and atypical pneumococci and could distinguish these from nonpneumococci. Regardless of the exact nomenclature, it is noteworthy that the atypical isolates in this study were pathogenic, causing invasive disease that required clinical management and classification. Nonvaccine serotypes have increased in prevalence among the pediatric population since pneumococcal conjugate vaccine implementation; atypical and divergent pneumococci may find a similar opportunity, and their contribution to disease should not be overlooked.