Serotyping Streptococcus pneumoniae by Multiplex PCR

Abstract

The capsule is a major virulence factor of pneumococci, and it was shown that some capsular variants are associated with antimicrobial resistance and certain types of disease. Moreover, pneumococcal capsular typing has received renewed interest since the availability of conjugate vaccines, which include serotypes frequently associated with pediatric disease. Our aim was to develop a simple, reliable, and economical method for detecting epidemiologically important serotypes present in the proposed 11-valent conjugate vaccine. We designed primers based on the sequences available for the capsular types 1, 3, 4, 6B, 14, 18C, 19F, 19A, and 23F and combined them into seven multiplex PCRs. The method involves streamlined DNA template preparation and agarose gel electrophoresis to analyze the amplification products. A total of 446 pneumococci selected from among isolates colonizing the nasopharynx of children attending day care centers in Lisbon, Portugal, were typed both by conventional immunological techniques and by multiplex PCR. Capsular types identified by the PCR method invariably produced results concordant with the conventional serotyping technique. Even when the method presented does not fully type an isolate, the PCR data can guide the experimenter when using immunological serotyping. Multiplex PCR for the analysis of pneumococci provides an accurate, expeditious, and cost-effective way of reducing the number of strains that have to be serotyped by conventional immunological techniques.

Streptococcus pneumoniae is a major worldwide causative agent of morbidity and mortality among young children and the aged (3). The ability of pneumococci to cause disease is directly related to the production of a capsule, a polysaccharide structure external to the cell wall that provides resistance to phagocytosis and promotes evasion of the host immune system by the bacteria (1). Pneumococci can produce at least 90 immunologically distinct capsules that differ in chemical structure (7). Importantly, despite the large variety of capsular types, only a small fraction of these cause most cases of invasive disease (16), and an even more limited number of serotypes and serogroups is most often associated with pediatric disease (serotypes 4 and 14 and serogroups 6, 7, 9, 18, 19, and 23), although the rank of serotypes may vary somewhat with the geographic region and time period (26).

The management of pneumococcal disease has become more difficult because of the rapid increase of antimicrobial resistance. Interestingly, the vast majority of antibiotic-resistant strains of S. pneumoniae express the relatively limited number of pediatric serotypes incorporated in the conjugate vaccines currently in use and development (3). The same serotypes recovered both from pediatric and adult infections are also most frequently identified among both drug-susceptible and drug-resistant strains that colonize healthy children attending day care (23). It is generally agreed that the use of an effective pneumococcal vaccine during infancy could significantly reduce morbidity and mortality associated with pneumococcal infections among young children (12). A 7-valent antipneumococcal vaccine is already licensed in several countries and has shown promising results. Nine- and 11-valent vaccines are also under evaluation (12).

To optimize the development of future conjugate vaccines and to evaluate their efficacy, it is necessary to understand the serogroup-specific epidemiology of pneumococci and their associated disease types (25). Continuous monitoring of S. pneumoniae serotypes is essential since it has been shown that the incidence of types responsible for invasive disease can change over time (1).

In this work we describe a rapid, simple, and cost-effective multiplex PCR-based method to type pneumococci and reduce the number of strains that have to be serotyped by using the standard capsular reaction test and discuss other potential applications of the methodology.

MATERIALS AND METHODS

Bacterial strains.

Forty pneumococcal strains representing different serotypes and serogroups constituted our control collection. The serotypes and groups represented in this collection and the strain designations (in parentheses) are as follows: 1 (SSISP 1/4), 2 (D39S), 3 (ATCC 6303), 4 (TIGR4), 5 (AR314), 6A (PnDCC2513), 6B (ATCC 6326), 7F (PnDCC2636), 8 (SSISP 8/3), 9V (ATCC 10368), 9A (SSISP 9A/1), 9L (SSISP 9L/4), 9N (SSISP 9N/2), 10A (SP96), 11F (SSISP 11F/2), 11A (K41), 11B (SSISP 11B/2), 11C (SSISP 11C/1), 11D (SSISP 11D/1), 12F (SSISP 12F/3), 14 (ATCC 6314), 15F (SSISP 15F/3), 15A (SSISP 15A/2), 15B (SSISP 15B/2), 15C (SSISP 15C/2), 17F (SSISP 17F/1), 18F (SSISP 18F/1), 18A (SSISP 18A/2), 18B (SSISP 18B/2), 18C (SSISP 18C/1), 19F (OP5248), 19A (SSISP 19A/5), 19B (SSISP 19B/2), 19C (SSISP 19C/2), 20 (SSISP 20/4), 22F (AR459), 23F (ATCC 6323), 23A (ATCC 10346), 23B (ATCC 10364), and 33F (PnDCC3012). The Statens Serum Institut (SSI; Copenhagen, Denmark) and the American Type Culture Collection (Manassas, Va.) supplied the SSI and ATCC strains, respectively. All other strains belonged to the pneumococcal collection of the Laboratory of Molecular Genetics, Instituto de Tecnologia Química e Biológica, Universidade Nova de Lisboa (ITQB/UNL), Oeiras, Portugal.

A total of 446 pneumococcal isolates from the Laboratory of Molecular Genetics ITQB/UNL collection were studied. The samples were recovered from the nasopharynx of asymptomatic children attending day care centers in Lisbon and Oeiras, Portugal, during 2001 by using a previously described protocol (4). A single colony from each swab, identified as pneumococcus according to standard tests (α-hemolysis, colony morphology, optochin susceptibility, and bile solubility) (17), was picked for further analysis. Confirmed pneumococcal isolates were then typed by both immunological and molecular techniques.

Immunological serotyping.

Conventional serotyping was performed by the standard capsular reaction test by using the chessboard system (27) and specific antisera. Briefly, a pure bacterial culture suspension was mixed with group-specific and type-specific antisera (SSI), and the reaction was considered positive when swelling of the capsule or agglutination occurred, as seen under the phase-contrast microscope.

Molecular capsular typing. (i) Oligonucleotide primers.

We designed a set of 27 primers that were divided into three classes. Class one primers were designed to serve as internal control and are represented by the primer pair cpsA-f and cpsA-r, which targets cpsA, a highly conserved gene that exists in all capsular loci thus far characterized (19).

The 18 specific primers in the second class were designed to target genes specific for serotypes 1, 3, 4, 6B, 14, 18C, 19F, 19A, and 23F, comprising eight of the components of the 11-valent vaccine (serotype 19A is not included in this vaccine formulation). The three additional capsular determinants for serotypes 5, 7F, and 9V also present in this vaccine had not been sequenced at the time this method was developed; the sequence for serotype 9V was recently published (29). Based on Southern hybridization results (14, 19, 22) and sequence homologies, the primers in this second class were designed to target different genes which were as follows: gene wzy for serotypes 1 (21), 4 (11), 6B (11), 14 (13), and 19F (6) (initially named cpsH and cpsI for serotypes 14 and 19F, respectively); gene capB for serotype 3 (2); genes wciY and gct for serotype 18C (11); gene cpsK for serotype 19A (19); and gene cpsG for serotype 23F (22).

The third class of primers was designed to identify genes common to certain sets of serotypes. In designing this third group the aim was to use a minimal number of primer pairs that could separate sets of serotypes due to differential amplification. Southern hybridization results (19, 22) indicate that cpsB is a highly conserved pneumococcal gene, whereas cpsC can be divided into two distinct classes (19). Based on these results, one forward (cpsB-f) and two reverse primers (cpsC-r1 and cpsC-r2) were designed to target the genes cpsB and cpsC, respectively. Pneumococcal serotypes can also be differentiated by the presence of rhamnose in the capsule structure. From the available sequences, we knew that serotypes 1, 6B, 18C, 19F, 19A, and 23F have the required genes involved in the synthesis of this sugar. Hence, two primers (19FcpsO-f and 19FcpsO-r) were designed to target a conserved region of the gene cpsO. A final primer pair (19FcpsB-f and 19FcpsB-r) was included in an attempt to allow a further separation of the serotypes for which cps cluster sequences were not available.

To design the primers, we aligned all the available relevant gene sequences by using the ClustalX software (28). To find homologous and heterologous regions, we analyzed the alignments by using the GeneDoc software (K. B. Nicholas and H. B. Nicholas, Jr., GeneDoc: a tool for editing and annotating multiple sequence alignments, distributed by the author). Using the Vector NTI software (Informax Inc., Frederick, Md.), single primers were analyzed for dimer and hairpin loop formation as well as for hybridization in any of the available capsular loci. Every primer pair present in the same reaction was also tested for primer-primer interactions. The sequences, target sites, and expected product sizes are shown in Table 1. The oligonucleotides were obtained from MWG Biotech (Ebersberg, Germany) and from Invitrogen Life Technologies (Barcelona, Spain).

TABLE 1.

Oligonucleotide primers used in the study

Primera	Sequence (5′-3′)b	GenBank accession no.	Start positionc	Product size (bp)	Product name
cpsA-f	GGT GTT CTC TAT CCT TGT CAG CTC TGT GTC GCT C	AF057294d	2277	657	Control
cpsA-r	GTG TGA ATG GTC GAA TCA ACT CTA TAA ATG CC		2902
1wzy-f	GGA GAC TAC TAA ATT GTA ATA CTA ACA CAG CG	Z83335	10809	99	L
1wzy-r	CAA GGA TGA ATA AAG TAA ACA TAT AAT CTC		10878
3capB-f	TTG TTT TTT GTC TTT ATT CTT ATT CGT TGG	Z47210	7174	818	D
3capB-r	TAC TGA GAA CCT TCT GCC CAC CTT AGT TGC		7962
4wzy-f	CTG TTA CTT GTT CTG GAC TCT CGT TAA TTG G	AF316639	9558	430	G
4wzy-r	GCC CAC TCC TGT TAA AAT CCT ACC CGC ATT G		9957
6Bwzy-f	CGA CGT AAC AAA GAA CTA GGT GCt GAA AC	AF316640	10074	220	J
6Bwzy-r	AAG TAT ATA ACC ACG CTG TAA AAC TCT GAC		10264
14cpsH-f	GTC tGT TTA TTC TAT ATA CAA AGA GGC TCC	X85787	7701	268	H
14cpsH-r	GCA TTG CtA CAA TCG CTA TaC TAG ATa TGC		7939
18CwciY-f	GCA TCt GTA CAG TGT GCT AAT TGG ATT GAA G	AF316642	15411	354	I
18Cgct-r	CTT TAA CAT CTG ACT TTT TCT GTT CCC AAC		15735
19FcpsI-f	CAC CTA ATT TTA ATA CTG AGG TTA AGA TTG C	U09239	7808	408	M
19FcpsI-r	CAT AGG CtA TCA GAA TTT TAA TAA TAT CTT GC		8184
19AcpsK-f	GTT AGT CCT GTT TTA GAT TTA TTT GGT GaT GT	AF094575	12118	478	F
19AcpsK-r	GAG CAG TCA ATA AGA TGA GAC GAT aGT TAG		12566
23FcpsG-f	GTA ACA GTT GCT GTA GAG GGA ATT GGC TTT TC	AF057294	8927	384	K
23FcpsG-r	CAC AAC ACC TAA CAC aCG ATG GCT ATA TGA TTC		9278
19FcpsB-f	CGA ACC ATT GTC TCT ACC TCT CAC	U09239	1718	301	E
19FcpsB-r	CAA TTA CTG GCG TGA TTC CC		1999
19FcpsO-f	TAG AGA TGA TTT TAA TTA CAG GGG CAA ATG	U09239	14131	814	C
19FcpsO-r	CAA GTT GGA ATA ACA AAT CCA GTA GCT TTG		14915
cpsB-f	CGG AAG AGA AGA TAG CAG AAA ACT TTC TTC	AF057294d	3639
		U09239d	1767
cpsC-r1	ACT CAA TCA AAA GAA CAG CAA TTA CTG TTA	AF057294d	4796	1,187e	A
cpsC-r2	CTC GAT CAT TAA CTG AAA CAG AGA CAA TAC G	U09239d	2716	980f	B

^aThe primer names were given according to the serotype or serogroup and the corresponding target gene.

^bLower case letters indicate mismatches to the target sequences introduced to facilitate the combination of various primers in the same reaction mixture.

^cBase positions of the corresponding GenBank sequences at which primer sequences start.

^dGenBank accession numbers are given as reference. The corresponding primers were designed based on all the cps gene cluster sequences available at the time. Sequences of the gene of interest were aligned, and primers were designed to match conserved regions.

^ePrimers cpsB-f and cpsC-r1 generate a product of 1,187 bp.

^fPrimers cpsB-f and cpsC-r2 generate a product of 980 bp.

(ii) Multiplex reaction assembly.

The design of the primers reflects our initial idea of how to determine the capsular type of an isolate by using a minimal number of PCRs. Two types of multiplex reactions were developed, taking into account the size of the PCR products and the compatibility between primers (the optimized primer conditions are summarized in Table 2). First, the group reaction, which used the third class of primers in order to perform an initial screening of the samples, classified the pneumococcal isolates. Related serotypes were thus grouped based on the amplification of common genes. Next, depending on the amplification pattern obtained, a second reaction (out of six possibilities) was done (Table 3). These were named specific reactions, and in them, we used the second class of primers, i.e., primers that are serotype specific. With this latter reaction we wanted to distinguish the related serotypes within each set determined by the group reaction. The scheme of the multiplex reactions proposed for the sorting out of pneumococcal serotypes is summarized in Fig. 1.

TABLE 2.

Conditions used in multiplex reactions

Reaction	Primer(s)	Primer concn (μM)a	MgCl2 concn (mM)
Group	cpsA-f + cpsA-r	5	2.4
	19FcpsO-f + 19FcpsO-r	10
	cpsB-f	100
	cpsC-r1	50
	cpsC-r2	5
Specific 1	cpsA-f + cpsA-r	25	0.7
	3capB-f + 3capB-r	50
	19FcpsB-f + 19FcpsB-r	25
Specific 2	cpsA-f + cpsA-r	35	1
	19AcpsK-f + 19AcpsK-r	25
Specific 3	cpsA-f + cpsA-r	50	1
	4wzy-f + 4wzy-r	100
	14cpsH-f + 14cpsH-r	50
Specific 4	cpsA-f + cpsA-r	25	1
	6Bwzy-f + 6Bwzy-r	50
	23FcpsG-f + 23FcpsG-r	25
	19AcpsK-f + 19AcpsK-r	25
	18CwciY-f + 18Cgct-r	50
Specific 5	cpsA-f + cpsA-r	5	1
	1wzy-f + 1wzy-r	50
	18CwciY-f + 18Cgct-r	100
	19Fwzy-f + 19Fwzy-r	50
Specific 6	cpsA-f + cpsA-r	25	1
	18CwciY-f + 18Cgct-r	50

^aFinal concentration in reaction mixture for each primer.

TABLE 3.

Summary of control collection serotyping by multiplex PCR

Group reaction pattern	Corresponding serotype(s)	Specific reaction no.	Specific PCR identification(s)a
G.1	3, 5, 10A, 20	1	3
G.2	8, 9V, 9A, 11F, 11A, 11B, 11C, 11D, 12F, 15A, 19A, 33F	2	19A
G.3	4, 9L, 9N, 14, 15B, 15C	3	4, 14
G.4	2, 6A, 6B, 15F, 17F, 18F, 19A, 22F, 23F, 23B	4	6b, 18b, 19A, 23F
G.5	1, 7F, 18B, 18C, 19F, 19B, 19C, 23A	5	1, 18b, 19F
G.6	18A	6	18b

^aThe primers that constitute the specific reactions allow an identification to the serotype or serogroup level.

^bThe specific reaction primers are serogroup specific and do not allow the distinction of the various serotypes within the same serogroup.

FIG. 1.

Multiplex reaction scheme. GR, group reaction. G.1 through G.6, patterns obtained with the group reaction. SR1 through SR6, specific reactions 1 through 6. A through M, name of the PCR products as defined in Table 1. Serotypes in bold are those included in the proposed 11-valent vaccine.

(iii) DNA preparation.

Before the PCR, isolated colonies from freshly grown bacterial cultures were picked with a sterile tip and briefly immersed in the multiplex reaction mix.

(iv) PCR conditions.

Multiplex PCR was done in a 10-μl volume with 1× PCR buffer (10 mM Tris-HCl [pH 9.0], 50 mM KCl, 0.1% Triton X-100, 0.01% [wt/vol] gelatin), 150 μM concentrations of each deoxynucleoside triphosphate, 0.2 U of Super Tth DNA polymerase (HT Biotechnology LTD, Cambridge, United Kingdom), MgCl₂, and primers as indicated in Table 2. Thermocycling was done in a Biometra T-1 thermocycler apparatus (Whatman Biometra, Göttingen, Germany) with the following conditions: 94°C for 4 min; 30 amplification cycles of 94°C for 45 s, 60°C for 45 s, and 72°C for 90 s; and a final extension step at 72°C for 5 min.

(v) PCR analysis.

The total volume of the PCR mixtures was analyzed by electrophoresis on 2% Seakem LE agarose gels (BMA, Rockland, Maine) in 0.5× TBE buffer (44.5 mM Tris, 44.5 mM Boric acid, 1 mM EDTA [pH 8.3]; Bio-Rad, Munich, Germany) at 5.2 V/cm for 2 h. Gels were stained in a 0.1-μg ml⁻¹ ethidium bromide solution and photographed by standard procedures, and the amplification results were visually analyzed. The sizes of the PCR products were estimated by comparison with a molecular size standard (100-bp ladder; Amersham Biosciences, Little Chalfont, Buckinghamshire, United Kingdom).

(vi) Sensitivity of multiplex PCR.

To evaluate the sensitivity of the technique developed, both the minimum and maximum numbers of cells that can be detected by this method were determined. Strains expressing different serotypes were grown at 37°C for 6 to 7 h on C+Y medium (15). The cells were washed and suspended in 1× phosphate-buffered saline solution (137 mM NaCl, 2.7 mM KCl, 4.3 mM Na₂HPO₄ · 7H₂O, 1.4 mM KH₂PO₄ [pH 7.3]), and turbidity was adjusted to McFarland standards (0.5 through 5) in both 1× phosphate-buffered saline solution (for plating and cell counting) and ultrapure water (for PCR). When determining the sensitivity of the method, the McFarland dilutions of the cells suspended in ultrapure water were boiled for 3 min, cooled on ice, and stored at −70°C until use. To assess the minimum number of cells, 10-fold serial dilutions of the 0.5 McFarland standard bacterial suspensions were done. The number of CFU per milliliter was determined by plating appropriate dilutions in tryptic soy agar supplemented with 5% sheep blood and incubating them overnight at 37°C in a 5% CO₂ atmosphere. At least three independent experiments were done to establish the sensitivity of the PCRs.

RESULTS

Multiplex reaction assembly.

With the group reaction we could distinguish six groups according to the differential amplification of target genes (Fig. 2 and Table 3). If pattern G.1, G.2, G.3, G.4, G.5, or G.6 was observed upon testing an isolate with the group reaction, we followed up with capsular type classification by using specific reaction 1, 2, 3, 4, 5, or 6, respectively (Fig. 3 and Table 3.). All reaction mixtures developed were tested with each isolate of the control collection in order to (i) define the different capsular types yielding the same amplification pattern in the group reaction and (ii) to ascertain the specificity of the primers designed for the detection of distinct capsular types.

FIG. 2.

Group reaction. Lanes represent the different patterns generated by the group reaction as indicated. Lane M, 100-bp ladder molecular size marker. The arrows indicate the names of the PCR products as defined in Table 1.

FIG. 3.

Specific reactions. (A) Specific reaction 1. Lane 1, serotype 3 (ATCC 6303); lane 2, serotype 5 (AR314). (B) Specific reaction 2. Lane 3, serotype 19A (SSISP 19A). (C) Specific reaction 3. Lane 4, serotype 4 (TIGR4); lane 5, serotype 14 (ATCC 6314). (D) Specific reaction 4. Lane 6, serotype 6B (ATCC 6326); lane 7, serotype 23F (ATCC 6323); lane 8, serotype 19A (SSISP 19A); lane 9, 18F (SSISP 18F/1). (E) Specific reaction 5. Lane 10, serotype 1 (SSISP 1/4); lane 11, serotype 18C (SSISP 18C/1); lane 12, serotype 19F (OP5248). (F) Specific reaction 6. Lane 13, serotype 18A (SSISP 18A/2). Lanes M, 100-bp ladder molecular size markers. The arrow indicates the internal control product.

As indicated in Table 3, all the primers that constitute the specific reactions and allow a specific PCR identification yielded products only when strains expressing the corresponding serotypes were used. Concerning serogroups 6 and 18, although we could not distinguish between the different serotypes of the same serogroup, the two sets of primers initially designed for serotypes 6B and 18C proved to be serogroup specific. In specific reaction 1, we included primers 19FcpsB-f and 19FcpsB-r, which are not type specific. Still, as these primers yielded amplification products with serotypes 5 and 10A, they were included to further discriminate the capsular types of pattern G.1 and facilitate the typing of isolates presenting this pattern by traditional methods.

Sensitivity of multiplex PCR.

The detection limit of the multiplex PCR assay for pure bacterial cultures varied between 20 and 50 CFU, depending on the serotype (Fig. 4). The use of bacterial suspensions greater than 4 McFarland standards (equivalent to 1.6 × 10⁵ to 4 × 10⁵ CFU) as the DNA template inhibited the PCR (data not shown).

FIG. 4.

Sensitivity of multiplex PCR. Numbers above the lanes indicate numbers of CFU per 10-μl reaction mixture. Lane M, 100-bp ladder molecular size marker. The arrow indicates the internal control product.

Isolates.

We analyzed 446 individual isolates by both PCR capsular typing and conventional immunological techniques with a double-blind procedure. Of the 446 isolates, 294 (65.9%) could be typed by the multiplex reaction; the remaining 152 strains expressed serotypes for which no specific primers were included in the multiplex scheme. All 294 isolates typed to the serotype or serogroup level by PCR (Table 4) agreed fully with the results obtained by conventional immunological methods. Of these, the 104 isolates (35.4%) belonging to serogroup 6 were further divided by agglutination with type-specific sera into types 6A (18%) and 6B (17.3%), and the 15 isolates (5.1%) belonging to serogroup 18 were further divided into types 18C (4.4%) and 18A (0.7%). Twenty-three isolates did not yield any PCR products, despite repeated attempts. The group reaction identified the remaining 129 isolates to the pattern level. These were further characterized into 17 distinct groups by conventional antibody serotyping, none of which were expected to be identified by our PCR capsular typing methodology (Table 5). Three of the 17 different groups identified, totaling 23 distinct isolates, were typed as belonging to pools D, E, and G, according to the chessboard system of conventional serotyping (Table 5). Whereas isolates typed to pools D and E were only related to a particular pattern each (G.5 and G.4, respectively), pool G isolates were equally distributed between patterns G.2 and G.3. This is not unexpected, since pool G includes five different serotypes (27) (Table 5).

TABLE 4.

Isolates typed by multiplex PCR to the serotype or serogroup level (n = 294)

Serotype or serogroup	No. of isolates	% of isolates
6	104	35.4
19F	58	19.7
23F	51	17.3
14	29	9.9
19A	29	9.9
18	15	5.1
3	6	2.0
4	2	0.7

TABLE 5.

Conventional serotyping of isolates typed by multiplex PCR to the pattern level (n = 129)

Serotype or serogroupa	No. of isolates	% of isolates
11A	52	40.3
Pool G	14	10.9
9V	10	7.8
15B	8	6.2
10	7	5.4
Pool D	6	4.7
33F	6	4.7
9N	4	3.1
15A	4	3.1
23A	4	3.1
Pool E	3	2.3
12	3	2.3
17	3	2.3
23B	2	1.6
15C	1	0.8
9L	1	0.8
15F	1	0.8

^aPool G includes serogroups 29, 34, 35, 42, and 47; pool D includes serogroups 16, 36, and 37; and pool E includes serogroups 21 and 39.

DISCUSSION

The increasing difficulty in the management of pneumococcal disease and the adoption of preventative measures against infection by S. pneumoniae has led to a renewed interest in pneumococcal capsular typing techniques. This study shows that multiplex PCR, with primers targeted to cps genes, is a cost-effective and expeditious method for the capsular typing of large numbers of pneumococcal isolates and is capable of reducing the quantity of pneumococci that have to be serotyped by immunological techniques. Moreover, the multiplex PCR method accurately detects the majority of serotypes and serogroups frequently isolated from young children, allowing the characterization of the colonization patterns before and after vaccination.

The existence of 90 different pneumococcal capsular types (7) renders difficult the development of new capsular typing methods based on genetic techniques. As our interest was to detect the most frequent types isolated from children, we based our method on the serotypes included in the proposed 11-valent conjugate vaccine, i.e., serotypes 1, 3, 4, 5, 6B, 7F, 9V, 14, 18C, 19F, and 23F (12). At the time this work was initiated, all the corresponding loci responsible for the synthesis of these capsular polysaccharides were known except for types 5, 7F, and 9V. Thus, with the method described here, typing of isolates expressing these particular polysaccharides is based on the amplification of fragments that are also amplified in other serogroups. The rest of the capsular type determinations included in the multiplex PCR method are based on type-specific products (Table 3). In its present form, the multiplex typing scheme does not distinguish serotype 6A from 6B and serotype 18B from 18C, whereas all remaining serotypes included in the 11-valent vaccine are distinguished from the other members of the same serogroup included in the control collection (Table 3).

The cpsA-f-cpsA-r primer pair included in all multiplex reactions target the cpsA gene, a ubiquitous component of all capsular loci of S. pneumoniae; this pair of primers was designed as an internal control for the efficiency of the PCRs. However, of the 446 studied isolates, 23 (5.2%) yielded no amplification, even with these primers. Since all our primers target genes belonging to cps loci, the lack of any amplification may result from pneumococcal isolates that either have distinct or mutated capsular genes or no cps locus at all. Serotyping these isolates with antisera yielded no positive reactions, indicating that these could be rough pneumococci. The proportion of such nontypeable isolates was similar to that obtained by Fenoll et al. (3.7%) by using a dot blot assay for serotyping pneumococci (5).

The multiplex typing scheme correctly identified, at least to the serogroup level, 65.9% of the 446 isolates analyzed. Of these, 59.4% represented types (3, 4, 14, 19F, and 23F) or groups (6 and 18) that are included in the 11-valent antipneumococcal vaccine.

The decision to include primers to detect type 19A, not included in the above-mentioned vaccine, resulted from our own experience in the laboratory. Previous studies had shown it to be a serotype commonly found among carriers, and indeed, it accounted for 6.5% of the 446 isolates analyzed. Serotype 9V accounted for 7.8% (n = 10) of the isolates identified to the pattern level (n = 129) and 2.2% of the total 446 pneumococci. No representatives of serotypes 1, 5, and 7F (all present in the proposed formulation of the 11-valent vaccine) were detected among the 446 nasopharyngeal isolates tested in our study, either by multiplex PCR or by conventional serotyping (Tables 4 and 5).

Of the 294 isolates typed by the specific reactions (Table 4), 277 (94.2%) yielded results that were in agreement with the preliminary capsular typing by the group reaction. The remaining 17 isolates that gave discrepant results were found to be type 19A by the capsular reaction test. This was not expected, as serotype 19A was initially found associated with pattern G.4 instead of the G.2 obtained for these isolates. We then confirmed that we could type them with specific reaction 4, which includes primers to detect serotype 19A. Specific reaction 4 is done on samples presenting pattern G.4 (Table 3), which differs from G.2 only by the product generated by primers 19FcpsO-f and 19FcpsO-r (Fig. 2) targeted to a gene involved in the synthesis of rhamnose (20). The serotype 19A capsule has been described as having rhamnose in its chemical structure (20), and the only complete 19A-cps locus sequence available (GenBank accession no. AF094575) (19) has all the genes required for the synthesis of this sugar. It is possible that distinct 19A strains have DNA polymorphisms explaining the amplification pattern G.2 instead of the expected G.4 that includes the PCR product of the gene involved in rhamnose biosynthesis. Indeed, analysis of all 19A isolates by pulsed-field gel electrophoresis revealed 10 distinct restriction patterns, of which 4 are associated with pattern G.2 and 6 are associated with pattern G.4 (data not shown). Specific reaction 2 (Table 2) was thus developed to allow the detection of 19A isolates with pattern G.2 (Table 3).

The reactions developed were optimized to yield reproducible and unambiguous results and to allow the use of this method as a routine typing technique. In order to minimize costs, several parameters were evaluated for their impacts on the yield and specificity of the reactions. An initial economic assessment of the method indicated that the major cost was the DNA polymerase. Both AmpliTaqGold (Perkin-Elmer Applied Biosystems, Warrington, Cheshire, United Kingdom) and Super Tth were tested, and the results obtained were indistinguishable (data not shown). However, upon testing the latter, we were able not only to cutback on enzyme units and primer amounts per reaction but also to reduce the total reaction volume to 10 μl while still producing the same results as with 100- and 20-μl reaction volumes (data not shown). Template preparation was also simplified, as the extraction of highly purified DNA would be incompatible with the analysis of large numbers of samples in routine assays. Both the boiled cells and tip-pick procedures described in Materials and Methods could be done with bacteria growing on the surface of blood agar plates, bypassing the need for culture in liquid medium and of laborious and time-consuming DNA purification protocols. Depending on the serotypes, the sensitivity of the method varied between 20 to 50 CFU (Fig. 4). Along with this minimum, there was also a maximum number of cells (1.6 × 10⁵ to 4 × 10⁵ CFU) that allowed amplification by PCR (results not shown). The use of boiled bacterial suspensions might be the reason for this latter limit. Similarly, to an excess of purified DNA template, accumulation of cell debris probably inhibits the PCR either by trapping increasing amounts of DNA or by binding the Mg²⁺ required for the polymerase activity or by doing both (8, 18).

There are many advantages in using the presented method to type pneumococci. First, it involves techniques that many microbiology laboratories can easily implement: (i) simple and fast DNA template preparation, (ii) a minimal number of PCRs, and (iii) analysis of the amplification products by agarose gel electrophoresis. Second, it is a more reliable method than its immunological counterpart in that the interpretation of the results is not subjective and needs no highly specialized expertise. Third, it also has revealed to be more expeditious as well: a single technician may analyze up to 96 samples a day. Moreover, there is the possibility of expanding the method to cover the detection of more capsular types. Based on the recent publication of the complete sequence of a 9V capsular locus (29) and the public release of partial sequences of capsular loci 6A and 6B by Griffiths and Hall (GenBank accession no. AF246898, AY078347, AY078342, AY078343, AY078344, AY078341, AY078339, AY078345, AY078340, and AY078346), we designed new primers to allow the identification of these serotypes and the expansion of our typing scheme. Finally, we estimated the costs associated with each method to achieve comparable results, i.e., correct serotyping of types and groups 1, 3, 4, 6, 14, 18, 19F, 19A, and 23F. The average price for serotyping a sample by the capsular reaction with SSI antisera in our laboratory is $28.90 per sample, whereas typing by our method costs about $1.80 per sample (no setup or labor costs were included in the calculations).

The isolation of colonies from a nasopharyngeal sample is a biased procedure. By picking a single colony there is a high probability that it represents the most prevalent serotype in the bacterial population. The more colonies we pick, the more likely it is that the diversity of the pneumococcal nasopharyngeal flora is better resolved. However, the costs and labor involved with multiple colony isolation would make this procedure impossible to deploy in a large scale. This is especially problematic given that for many nasopharyngeal carriage studies to achieve statistical significance, sample sizes of ≥100 pneumococcal carriers are required (10). As Huebner et al. report, if the less common serotype represents only 5% of the total pneumococcal population, 59 colonies from each specimen would need to be serotyped to have a 95% probability of picking the second pneumococcal type (10). The need to develop a method that would allow a feasible analysis of minority strains in a pneumococcal population is confirmed by studies on the carriage of multiple pneumococcal capsular types (9, 10, 24). Preliminary results point to the possibility of using a modified version of our technique to analyze heterogeneous bacterial samples, which would not only eliminate the need to isolate multiple colonies but also give a more realistic perspective on the coexistence of different serotypes in the nasopharynx.

This study has shown that, in its present format, our method is of valuable use in microbiology laboratories, in that it provides a fast and cost-effective way of analyzing large numbers of samples, allows the detection of the most frequent serotypes that colonize children, and reduces the number of strains that have to be serotyped by conventional immunological techniques. Although the method presented may not fully type an isolate, the information gathered can guide the experimenter when serotyping these isolates by using the traditional antiserum method. In addition, it allows the use of PCR, a technique widely established in microbiology laboratories, avoiding the necessity of developing specific expertise and using specific serological reagents. Moreover, the possibility to use this method as a qualitative assay to evaluate the true composition of a possibly diverse nasopharyngeal population of pneumococci should increase its usefulness as a new capsular typing technique.