Identification of Alpha-Hemolytic Streptococci by Pyrosequencing the 16S rRNA Gene and by Use of VITEK 2

Abstract

Alpha-hemolytic streptococci are very difficult to identify by phenotypic methods. In this study, a pyrosequencing method for the identification of streptococcal species based on two variable regions of the 16S rRNA gene is described. Almost all studied streptococcal species (n = 51) represented by their type strains could be differentiated except for some closely related species of the Streptococcus bovis or S. salivarius group. The pyrosequencing results of alpha-hemolytic streptococci isolated from blood (n = 99) or from the normal pharyngeal microbiota (n = 25) were compared to the results obtained by the VITEK 2 with GP card (bioMérieux, Marcy l'Etoile, France). As expected, the results of the two methods did not completely agree, but 93 (75.0%) of the isolates assigned to the same streptococcal group by both methods and 57 (46.0%) reached consistent results at the species level. However, 10 strains remained unidentified by VITEK 2, and 4 isolates could not be assigned to any streptococcal group by pyrosequencing. Identification of members of the S. mitis and S. sanguinis groups proved difficult for both methods. Furthermore, the pyrosequencing analysis revealed great sequence variation, since only 43 (32.3%) of the 133 isolates analyzed by pyrosequencing had sequences identical to a type strain. The variation was greatest in the pharyngeal isolates, slightly lower in the blood culture isolates, and nonexistent in invasive pneumococcal isolates (n = 17) that all had the S. pneumoniae type strain sequence. The resolution of the results obtained by the two methods is impeded by the lack of a proper gold standard.

The genus Streptococcus comprises a heterogeneous group of bacteria, some of which are major pathogens and some are members of the normal microbiota of humans and animals. Identification of streptococcal species has traditionally been based on hemolysis on blood agar, biochemical tests, and serological grouping. During the last decade, numerous molecular methods, such as ribotyping (25), amplified ribosomal restriction analysis (32), PCR-based methodology (10), oligonucleotide probes (1, 20), and sequencing of different targets (5, 6, 22, 23, 36, 37) have been developed to identify streptococcal species more rapidly and accurately.

Pyrosequencing is a sequencing by synthesis method, in which incorporation of nucleotides is enzymatically transformed to release light which is detected and presented as a peak histogram, a pyrogram (26, 28). The main advantage of pyrosequencing is its rapidity and lower price compared to conventional sequencing. Furthermore, since the dispensation order can be designed individually for each sample, verification of suspected mutations is simple. Also, heterogeneous sequences are quite easily determined, if the possible sequence combinations are known (11). By pyrosequencing fairly short sequences only can be determined, but usually the sequences of around 60 bases obtainable with suitable dispensation orders are sufficient for carefully designed applications. In the field of microbiology, pyrosequencing has already been applied to the identification of bacteria (18, 21, 27) and heterogeneous 23S rRNA gene sequences conferring antibiotic resistance (11, 34), for example. Pyrosequencing of two regions of the RNase P RNA gene, rnpB, has been proposed for the identification of streptococci (15).

Sequencing of the 16S rRNA gene is widely used in the identification of bacteria. Although the discriminative power of the 16S rRNA has recently been argued to be to too low for the identification of streptococci and other closely related bacteria (14, 15), it has been found effective in identification of clinically relevant cocci including viridans streptococci (3, 13). Determining the whole 16S rRNA gene sequence may not be necessary since the variable regions contain most of the variation between bacterial species. Indeed, the majority of the variation between 16S rRNA gene sequences of streptococcal species is found in the 5′ terminal part of the gene (13), especially in the regions used in the present study: the v1 and v2 region corresponding to bases 63 to 92 and 193 to 222 of the S. pneumoniae R6 16S rRNA (NC_003098), respectively. In our previous study, the v1 region of alpha-hemolytic commensal streptococcal isolates was determined by pyrosequencing and compared to the sequences of selected type strains. Although the variation in the v1 region was extensive, some of the studied type strains had identical v1 sequences (33) and, consequently, analysis of the v2 region was added to the pyrosequencing assay.

VITEK 2 is an instrument for automated phenotypic identification of medically relevant bacteria and yeasts (bioMérieux, Marcy l'Etoile, France). The new colorimetric GP card for identification of gram-positive bacteria consists of 43 biochemical tests that are monitored up to 8 h. The database contains 33 streptococcal species including two “slashline species”: S. mitis-S. oralis and S. lutetiensis-S. bovis, six subspecies belonging to S. constellatus, S. dysgalactiae, or S. equi and S. suis biotypes I and II. Previous studies with VITEK 2 have included only small numbers of viridans streptococci (9, 15). The purpose of the present study was to compare pyrosequencing of the v1 and the v2 regions of the 16S rRNA gene and the VITEK 2 with a colorimetric GP card for identification of alpha-hemolytic streptococci.

MATERIALS AND METHODS

Bacterial strains.

Nonpneumococcal alpha-hemolytic blood culture isolates (n = 102) collected in Turku University Hospital derived from 96 patients between 1993 and 2004 were analyzed. Thirty-six strains were single isolations from immunocompromised patients, while sixty-six were from patients with two to eight blood cultures growing alpha-hemolytic streptococci. Eighty-four of the strains were from blood cultures growing only streptococci, and eighteen were from blood with multibacterial growth. Multiple isolates from a single patient were only included if they had been originally identified as separate species, if they had different antimicrobial susceptibility profiles, or if the interval between their isolation was at least 2 weeks. The strains had been originally identified on the basis of alpha-hemolysis and resistance to optochin or using either API 20Strep or VITEK 2 with a fluorogenic identification card (both from bioMérieux) and were stored at −70° until the present study. For comparison, 39 alpha-hemolytic isolates from the normal pharyngeal microbiota of six elderly persons (33) and 17 invasive S. pneumoniae clinical isolates collected in Finland were included. The S. pneumoniae isolates had typical colony morphology and hemolysis, and they were optochin susceptible. The isolates were grown on blood agar and incubated at 35°C in an atmosphere of 5% CO₂. Three blood culture isolates were identified as Enterococcus, Granulicatella, or Lactobacillus, respectively, based on the pyrosequencing results and database searches. By VITEK 2, these isolates were identified as Enterococcus avium, Gemella morbillorum, and an unidentified organism, respectively. Similarly, one Granulicatella and four Enterococcus isolates were found among the pharyngeal isolates. Unfortunately, only 25 pharyngeal streptococcal isolates were viable for the VITEK analysis. Consequently, 99 blood culture isolates and 25 pharyngeal isolates were analyzed by both methods, and 9 pharyngeal isolates were analyzed only by pyrosequencing.

Streptococcal type strains (n = 59) and three additional well-characterized strains (Table 1) were cultured according to the recommendations of the bacterial strain depositors. In addition to the subspecies listed in Table 1, the type strains of S. bovis, S. caprinus and S. infantarius subsp. coli were considered as subspecies of S. equinus, S. gallolyticus subsp. gallolyticus and S. lutetiensis, respectively (24, 31, 35). Consequently, 51 species were represented by the 59 type strains listed in Table 1. In addition to the species listed in Table 1, the 16S rRNA gene sequences of S. castoreus, S. equi subsp. ruminatorum, S. gallinaceus, S. oligofermentans, and S. pseudopneumoniae type strains were retrieved from the GenBank and added to the local sequence database.

TABLE 1.

Streptococcal type or reference strains analyzed in this study^a

Strainb	Pyrosequencing		VITEK result
	v1+v2 sequencec	GenBank no.d	Species	Confidence levele
Beta-hemolytic streptococci
S. agalactiaef LMG 14694T	1	DQ303183	S. agalactiae	VG
S. canisf LMG 15890T	2	AB002483	S. canis	VG
S. didelphis CCUG 45419T	3	DQ303185	S. dysgalactiae subsp. equisimilis	Acc
S dysgalactiae subsp. dysgalactiaef LMG 15885T	4	EF151154	S. dysgalactiae subsp. dysgalactiae or S. dysgalactiae subsp. equisimilis	LD
S. dysgalactiae subsp. equisimilisf CCUG 36637T	5	AB008926	S. dysgalactiae subsp. equisimilis	Exc
S. equi subsp. equif LMG 15886T	6	DQ303186	S. equi subsp. equi	VG
S. equi subsp. zooepidemicusf LMG 16030T	7	EF151157	S. equi subsp. zooepidemicus	Exc
S. iniae CCUG 27303T	8	EF151159	S. porcinus	VG
S. phocae LMG 16735T	9	AJ621053	S. dysgalactiae subsp. equisimilis	Acc
S. porcinusf LMG 15980 T	10	AB002523	S. porcinus	VG
S. pyogenesf ATCC 700294, CCUG 4207T	11	NC_007297	S. pyogenes	Exc
Non-beta-hemolytic streptococci
S. pneumoniaef ATCC 49619, DSM 11867, LMG 14545T	12	AY281082	S. pneumoniae	VG
S. suisf CCUG 7984T	13	AB002525	S. suis I or S. suis II	LD
S. bovis group
S. bovish LMG 8518T	14	AB002482	S. lutetiensis/S. bovis	Exc
S. equinusf LMG 14897T	15	AF104116	S. equinus	Exc
S. gallolyticus subsp. gallolyticusg CCUG 35224T	16	AF104114	S. gallolyticus	Exc
S. caprinusg LMG 15572T	16	AF104114	S. gallolyticus or S. hyointestinalis	LD
S. gallolyticus subsp. macedonicusg LMG 18488T	17	Z94012	S. lutetiensis/S. bovis	VG
S. gallolyticus subsp. pasteurianusg CCUG 46150T	18	AJ297216	S. pasteurianus	Exc
S. infantarius subsp. coli CCUG 47831T	14	AF429763	S. pasteurianus	Good
S. infantarius subsp. infantariusf CCUG 43820T	19	AF429762	S. infantarius	Exc
S. lutetiensish,i CCUG 46149T	14	AF429763	S. infantarius subsp. coli	VG
Viridans streptococci
S. mutans group
S. criceti LMG 14508T	20	AJ420198	S. mutans	Exc
S. downeif CCUG 24890T	21	AY188350	S. sobrinus	Exc
S. ferus LMG 16520T	22	AY058218	S. mutans	Acc
S. hyovaginalis LMG 14710T	23	EF151158	S. uberis	Acc
S. macacae LMG 15097T	24	AY188351	S. alactolyticus	Acc
S. mutansf CCUG 17824T	25	NC_004350	S. mutans	Exc
S. ovisf,j LMG 19174T	26	Y17358	S. ovis	Exc
S. ratti CCUG 27642T	27	AJ420201	S. thoraltensis or S. mutans	LD
S. sobrinusf CCUG 25735T	28	AY188349	S. sobrinus	Exc
S. salivarius group
S. alactolyticusf CCUG 27297T	29	AF201899	S. alactolyticus	VG
S. hyointestinalisf LMG 14579T	30	AF201898	S. hyointestinalis	VG
S. salivariusf DSM 20560T	31	AY188352	S. equinus	Exc
S. thermophilusf CCUG 21957T	31	NC_006448	ND	-
S. vestibularisf DSM 5636T	31	AY188353	S. vestibularis	VG
S. anginosus group
S. anginosusf DSM 20563T	32	AF104678	S. anginosus	VG
S. constellatus subsp. constellatusf DSM 20575T	33	AF104676	S. constellatus subsp. constellatus	Exc
S. constellatus subsp. pharyngisf CCUG 46377T	33	AY309095	S. constellatus subsp. pharyngis	Exc
S. intermediusf DSM 20573T	34	AF104671	S. intermedius	Exc
S. sinensisj CCUG #0059#T	35	AF432856	S. constellatus subsp. pharyngis or S. anginosus	LD
S. sanguinis group
S. gordoniif DSM 6777T	36	AY485606	S. gordonii	Exc
S. sanguinisf DSM 20567T	37	AF003928	S. sanguinis or S. cristatus	LD
S. parasanguinisf DSM 6778T	38	DQ303191	S. parasanguinis or S. suis II	LD
S. mitis group
S. australisj CCUG 45919T	39	AY485604	S. mitis or S. oralis	VG
S. cristatusf DSM 8249T	40	AY584476	S. cristatus	Exc
S. entericusj CCUG 44616T	41	EF151156	S. salivarius	Exc
S. infantis LMG 18720T	42	AB008315	S. mitis or S. oralis	Exc
S. mitish DSM 12643T	43	AF003929	S. mitis or S. oralis	Exc
S. oralish DSM 20627T	44	AY485602	S. mitis or S. oralis	VG
S. orisratti CCUG 43577T	45	AF124350	S. hyointestinalis	Acc
S. peroris LMG 18719T	46	AB008314	S. mitis or S. oralis	VG
Unusual streptococcus species
S. acidominimus LMG 17755T	47	EF151153	unidentified organism
S. parauberis LMG 12174T	48	AF284579	S. thoraltensis	Exc
S. pleomorphusi CCUG 11733T	49	M23730	ND
S. pluranimaliumf LMG 14177T	50	EF151155	Granulicatella adiacens	VG
S. thoraltensisf LMG 13593T	51	Y09007	S. thoraltensis	Exc
S. uberisf LMG 9465T	52	AB023573	S. thoraltensis	Acc
S. urinalis LMG 19649T	53	AJ131965	S. dysgalactiae subsp. equisimilis	Acc

^aATCC, American Type Culture Collection, Manassas, VA; CCUG, Culture Collection, University of Göteborg, Göteborg, Sweden; DSM, Deutsche Sammlung von Mikroorganismen GmbH, Braunschweig, Germany; LMG, Laboratorium voor Mikrobiologie, Universiteit Ghent, Ghent, Belgium. ND, not determined.

^bThe species are grouped based on the review of Facklam (8). Only the type strains were analyzed by the VITEK 2.

^cThe different sequences of the combined 30-bp v1 and v2 region are indicated by numbers: species with identical sequences have the same sequence numbers.

^dThe GenBank accession number from which the correct v1 and v2 sequences can be found.

^eThe confidence level is a measure of the strength of the VITEK analysis in the descending order Exc (excellent), VG (very good), good, Acc (acceptable), and LD (low discrimination). If VITEK 2 analysis resulted in low discrimination between two species and an additional Voges-Proskauer test proposed by the device could discriminate the species, the final result is in boldface.

^fSpecies included in the VITEK GP database.

^gS. gallolyticus is included in the VITEK GP database, but its subspecies are not differentiated except for S. gallolyticus subsp. pasteurianus, which is identified as S. pasteurianus.

^hS. bovis and S. lutetiensis, as well as S. mitis and S. oralis, are identified as “slashline taxa” by the VITEK GP card.

ⁱNot grouped in the review of Facklam (8).

^jGrouping uncertain based on the review of Facklam (8).

Template preparation and PCR amplification.

The PCR templates were prepared by dissolving bacterial colonies in water and inactivating the solution at 95°C for 10 min. The biotinylated PCR product covering the v1 and v2 regions was generated using primers BioStrepV1For (5′-Bio-AGTTTGATCCTGGCTCAGGACG-3′) and StrepV2Rev (5′-CAACTAGCTAATACAACGCAGGTC-3′). The 50- to 100-μl PCRs contained 0.3 μM concentrations of primers, 0.015 U of AmpliTaqGold DNA polymerase/μl, 1× Geneamp PCR Gold buffer, 1.5 mM MgCl₂ (Applied Biosystems, Foster City, CA), 0.2 mM concentrations of deoxynucleoside triphosphates (Amersham Biosciences, Piscataway, NJ), and template at 1/10 of the reaction volume. Thermal cycling consisting of initial denaturation at 95°C for 10 min, followed by 40 cycles of 15 s at 95°C, 15 s at 57°C, and 30 s at 72°C was performed by using a PTC-200 thermal cycler (MJ Research, Waltham, MA). The presence of the approximately 250-bp PCR product was verified on an agarose gel.

Pyrosequencing.

Pyrosequencing was performed by using streptavidin-coated Sepharose beads (Amersham Biosciences), an PSQ 96 MA instrument, a vacuum prep workstation, and PyroGold SQA reagents (Biotage AB, Uppsala, Sweden) according to the instructions of the manufacturer. A portion (20 μl) of the PCR product and 15 pmol of the sequencing primer StrepV1RevV2 (5′-CCTACGCGTTACTCACCCGTTC-3′) or StrepV2RevV2 (5′-ACTAGCTAATACAACGCAGGTCCA-3′) was used in each sequencing reaction. To save reagents, 6 μl of enzymes and substrates were added manually to the reaction wells. The dispensation order for the v1 region was GCGACTC10(ACTG), and for the v2 region it was TCTACAGTCGA10(CGTA). The sequences obtained by pyrosequencing were handled by using a BioEdit sequence editor (12). Thereafter, 30-base v1 and v2 sequences were concatenated and compared to the corresponding sequences of type strains by using the Identifire 1.0.5.0 software (Biotage). The software gave the isolates a score which was 100 if the sequences of the isolate and a type strain were identical. One mismatch in the 60-base sequences generated a score of 96.9, and a score of ≥96.9 was considered an acceptable identification.

Partial 16S rRNA gene sequencing.

To verify some of the pyrosequencing results distinct from the type strains, approximately 1,000 bp of the 16S rRNA gene of nine pharyngeal isolates (Table 3) was sequenced. The PCR primers fd1mod2 (5′-AGAGTTTGATCMTGGCTCAG-3′) and rp2 and the sequencing primers 357f, 357r, 533f, 533r, and 907r (17) were used. For the S. acidominimus, S. dysgalactiae subsp. dysgalactiae, S. pluranimalium, S. entericus, S. equi subsp. zooepidemicus, S. hyovaginalis, and S. iniae type strains a sequence identical to the sequences determined by pyrosequencing could not be found in the GenBank, and their 5′ terminal part of the 16S rRNA gene was determined using sequencing primer 533r. The PCR products were purified with the High-Pure PCR product purification kit (F. Hoffmann-La Roche, Ltd., Basel, Switzerland) and sequenced by using the ABI Prism BigDye Terminator v3.0 (Applied Biosystems). The sequences were processed with Vector NTI advance 10.1.1 software (Invitrogen Corp., Carlsbad, CA), and the pharyngeal sequences were subjected to an NCBI BLAST search (http://www.ncbi.nlm.nih.gov/BLAST/). The sequences were in agreement with the ones obtained by pyrosequencing, good correlation between the pyrosequencing result and the BLAST search result was found, and even heterogeneous peaks could be detected at the positions that agreed with the pyrosequencing results.

TABLE 3.

Pyrosequencing and VITEK 2 results of the pharyngeal normal microbiota isolates

Group and no. of strains	Pyrosequencing resulta		VITEK 2 result
	Species	Scoreb	Species	Confidence levelc (n)
S. mitis group
2	S. mitis	100d	NDe	— (2)
4	S. mitis	96.9	S. intermedius	Exc (1)
			S. mitis/S. oralis	Exc (1) or VG (2)
1	S. infantis	96.9	ND	— (1)
1	S. pneumoniae, S. pseudopneumoniae, or S. mitis	96.9	ND	— (1)
1	S. cristatus, S. infantis or S. peroris	93.7	S. mitis/S. oralis	Exc (1)
1	S. infantis	93.7	S. mitis/S. oralis	Exc (1)
1	S. pneumoniae, S. pseudopneumoniae or S. mitis	93.7d	ND	— (1)
1	S. oralis	90.5	S. mitis/S. oralis	VG (1)
1	S. infantis	84.2d	S. parasanguinis	VG (1)
1	S. australis	87.4d	ND	— (1)
1	S. mitis or S. oralis (v2)	100f	S. salivarius	Exc (1)
1	S. australis (v2)	100d,f	S. constellatus subsp. pharyngis	Good (1)
2	S. mitis, S. oralis, or S. infantis (v2)	93.7d,f	S. mitis/S. oralis	VG (1) or Good (1)
S. sanguinis group
1	S. parasanguinis	100	S. parasanguinis	Exc (1)
2	S. gordonii	96.9	S. gordonii	Exc (1)
			ND	— (1)
1	S. sanguinis	96.9	ND	— (1)
1	S. sanguinis	90.5	S. mitis/S. oralis	Acc (1)
5	S. parasanguinis	93.7	S. parasanguinis	VG (2) or Exc (3)
1	S. parasanguinis	90.5d	Inconclusive identification	— (1)
2	S. parasanguinis (v2)	93.7d,f	S. parasanguinis	Exc (1) or Good (1)
S. salivarius group
1	S. salivarius or S. vestibularis	100	ND	— (1)
1	S. salivarius or S. vestibularis	96.9d	S. salivarius	Acc (1)
Undefined group
1	S. gallinaceus	90.5	S. mitis/S. oralis	Exc (1)

^aThe isolates have been listed according to the pyrosequencing result.

^bThe score is a measure of the relatedness of the studied sequence to the type strain sequence determined by the Identifire software. Score 100 is given to sequences that are identical to each other.

^cConfidence level is a measure of the strength of the VITEK analysis in the descending order Exc (excellent), VG (very good), good, Acc (acceptable), and LD (low discrimination).

^dOne of the isolates with this pyrosequencing result was sequenced by conventional sequencing. The GenBank accession numbers of the sequences are EF151144 to EF151152.

^eND, not determined.

^fThe pyrograms of the v1 region of these isolates indicated heterogeneous 16S rDNA alleles, and the isolates were tentatively identified using the v2 region.

VITEK 2 analysis.

The streptococcal type strains and alpha-hemolytic isolates were analyzed by the VITEK 2 system using the GP card and WSVT2-R04.01 software (bioMérieux) according to the instructions of the manufacturer. The results were interpreted so that the isolates belonging to a species not included in the VITEK database were satisfactorily identified if the result belonged to the same streptococcal group with the known species or the pyrosequencing result. If VITEK 2 analysis resulted in low discrimination between two species and an additional Voges-Proskauer test proposed by the device resulted in the same species or group with the pyrosequencing result, the results were regarded as concordant. VITEK 2 proposed some other additional tests for the resolution of “low-discrimination” results, but unfortunately they were not available in our laboratory, and thus such results remained unresolved.

RESULTS

Identification of streptococcal type strains.

When the 30-base v1 and v2 sequences were combined, 46 of the 51 studied species could be differentiated, and two sequence combinations were shared by two or more species (Table 1). The first was found in the S. bovis group members S. bovis and S. lutetiensis, and the second in the S. salivarius group members S. salivarius, S. thermophilus, and S. vestibularis. However, 53 different sequence combinations were found among the 59 studied type strains since some studied subspecies also contained slightly different sequences (Table 1). For example, the sequence of the S. equinus type strain differed from the sequence of the S. bovis type strain by one nucleotide, revealing small intraspecies variation. Small sequence differences were found also between the different S. gallolyticus subspecies. However, the sequences of the analyzed S. pneumoniae and S. pyogenes reference strains were identical to the sequences of the type strains (Table 1), and all of the different alleles of the streptococcal genomic sequences available in the GenBank were identical to their respective type strains in the studied regions. The smallest difference between distinguishable type strains in the combined v1 and v2 region was one nucleotide, which was detected between three species pairs of the S. bovis group, whereas the smallest difference between species of the S. mitis group was two nucleotides. Except for seven type strains, a type strain sequence containing the v1 and v2 sequences identical to the ones obtained by pyrosequencing could be found in GenBank. However, careful searching was required to find the accurate one among the many sequence versions available for many streptococcal type strains. The GenBank entries containing the sequences identical to the pyrosequencing results are presented in Table 2.

TABLE 2.

Pyrosequencing and VITEK 2 results of the alpha-hemolytic blood culture isolates

Group and no.	Pyrosequencing resulta		VITEK 2 result		Unresolved (n)d
	Species	Scoreb	Species	Confidence levelc (n)
S. anginosus group
7	S. anginosus	100	S. anginosus	Exc (4)
			S. anginosus or S. gordonii	LD (2)
			Inconclusive identification	— (1)	1
3	S. anginosus	96.9	S. anginosus	Exc (1)
			S. anginosus or S. gordonii	LD (2)
6	S. intermedius	100	S. intermedius	Exc (4)
			S. intermedius or S. constellatus subsp. pharyngis	LD (1)
			S. constellatus subsp. pharyngis	Good (1)
3	S. constellatus	100	Inconclusive identification	— (2)	2
			S. pyogenes, S. constellatus subsp. constellatus, or S. dysgalactiae subsp. equisimilis	LD (1)
2	S. constellatus or S. intermedius	96.9	S. constellatus subsp. constellatus	Exc (1)
			S. constellatus subsp. constellatus, S. gordonii, or S. anginosus	LD (1)	1
S. sanguinis group
7	S. sanguinis	100	S. sanguinis	Exc (6) or Good (1)
2	S. sanguinis	96.9	S. sanguinis	Exc (2)
5	S. sanguinis	93.7	S. mitis/S. oralis	Exc (2), VG (1), or Good (1)	4
			S. mitis/S. oralis or S. constellatus subsp. pharyngis	LD (1)	1
4	S. sanguinis	90.5	S. mitis/S. oralis	Exc (2), VG (1), or Good (1)	4
1	S. gordonii	100	S. gordonii	Exc (1)
1	S. gordonii	87.4	S. gordonii	Exc (1)
S. salivarius group
5	S. salivarius or S. vestibularis	100	S. salivarius	Exc (3)
			S. infantarius	VG (1)	1
			Unidentified organism	— (1)	1
2	S. salivarius or S. vestibularis	96.9	S. salivarius	Exc (2)
S. mitis group
4	S. mitis	100	S. mitis/S. oralis	Exc (1), VG (1), Good (1), or Acc (1)
1	S. oralis	100	S. mitis/S. oralis	Exc (1)
24	S. mitis, S. pneumoniae, or S. pseudopneumoniae	96.9	S. mitis/S. oralis	Exc (10), VG (5), or Good (3)
			S. pluranimalium or S. mitis/S. oralis	LD (1)	1
			S. mitis/S. oralis or S. constellatus susbp. pharyngis	LD (1)
			Unidentified organism	— (4)	4
6	S. infantis	93.7	S. mitis/S. oralis	Exc (2), VG (2), or Good (1)
			Unidentified organism	— (1)	1
1	S. oralis	93.7	S. mitis/S. oralis	VG (1)
1	S. oralis	90.5	S. mitis/S. oralis	Exc (1)
1	S. peroris	93.7	S. mitis/S. oralis	Exc (1)
3	S. mitis or S. oralis (v2)e	100	S. mitis/S. oralis	Exc (1) VG (1), Good (1)
S. bovis group
2	S. gallolyticus subsp. gallolyticus	100	S. gallolyticus	Exc (2)
2	S. bovis or S. lutetiensis	100	S. infantarius	VG (1), Exc (1)
1	S. gallolyticus subsp. pasteurinanus	100	S. pasteurianus	Exc (1)
2	S. bovis or S. lutetiensis	96.9	S. infantarius	Exc (1)
			S. gallolyticus, S. lutetiensis/bovis or S. hyointestinalis	LD (1)
Undefined group
2	S. oralis or S. gallinaceus (v1)e	100	S. parasanguinis	VG (1)	1
			S. parasanguinis or S. mitis/S. oralis	LD (1)	1
1	Both regions heterogeneousf		S. mitis/S. oralis	Exc (1)	1

^aThe isolates have been listed according to the pyrosequencing result.

^bThe score is a measure of the relatedness of the studied sequence to the type strain sequence determined by the Identifire software. Score 100 is given to sequences that are identical to each other.

^cConfidence level is a measure of the strength of the VITEK analysis in the descending order Exc (excellent), VG (very good), good, Acc (acceptable), and LD (low discrimination). If VITEK 2 analysis resulted in low discrimination between two species and an additional Voges-Proskauer test proposed by the device could discriminate the species, the final result is in boldface.

^dThe number of isolates having discrepant pyrosequencing and VITEK 2 results at the group level.

^eThe isolates with a heterogeneous sequence in either region were identified using the other region.

^fOne isolate had heterogeneous sequences in both regions, and hence it could be identified only as Streptococcus by pyrosequencing.

With the VITEK 2 system, 30 (81.1%) of the 37 type strains represented in the instrument's database were correctly identified to the species level. In three additional strains the correct result was among two possibilities between which VITEK 2 was unable to discriminate. Twenty-one type strains represented a species not included in the VITEK database. Sixteen of these were members of viridans groups, and thirteen (81.3%) were designated to the correct group (Table 1). The misidentifications mainly concerned uncommon species, but the S. salivarius type strain was repeatedly identified as S. equinus.

Identification of alpha-hemolytic streptococcal blood culture isolates.

The invasive S. pneumoniae (n = 17) isolates contained the S. pneumoniae-specific sequences in both studied regions, and their species designation got verification. In contrast, only 39 (39.4%) of the 99 blood culture isolates had a sequence identical to some of the streptococcal type strains and could be unambiguously assigned to a species by pyrosequencing (Table 2). The most prevalent species were S. anginosus (n = 7), S. sanguinis (n = 7), S. intermedius (n = 6), and S. salivarius or S. vestibularis (n = 5). Thirty (79.5%) of these isolates were identified as the same species with VITEK 2, and thirty-four (87.2%) were identified at least to the same streptococcal group by VITEK 2. Notably, there were no group-level discrepancies in this group of strains. However, four (10.3%) of these isolates could not be conclusively identified by VITEK 2.

Fifty-four (54.5%) blood culture isolates could be designated to the most probable streptococcal species by pyrosequencing (Table 2), but often the result was not clear since the sequence differences between the closest and the second closest type strain were minimal. This was discernible especially with isolates with sequences closest to that of a member of the S. mitis group. For example, the most prevalent v1 and v2 sequence combination (n = 24) among the blood culture isolates was a sequence differing by one nucleotide from the type strain sequences of S. mitis, S. pseudopneumoniae, and S. pneumoniae. Interestingly, this sequence was found in S. sanguinis ATCC 49296 (AY281086). By the VITEK 2 system, 19 (79.2%) of these isolates were identified as S. mitis/S. oralis with at least a good confidence level, 1 was identified as S. mitis/S. oralis or a third species with low discrimination, and 4 isolates could not be identified (Table 2). Thus, these isolates most likely belong to the S. mitis group. In the S. sanguinis group, nine strains with one difference at the maximum to the S. sanguinis type strain sequence were identified as S. sanguinis also by VITEK 2, and the nine strains with lower sequence similarities were identified as S. mitis/S. oralis by VITEK 2. The latter ones support the combining of the S. mitis and S. sanguinis groups (30).

In addition, pyrograms indicating heterogeneous 16S rRNA alleles were detected in six blood culture isolates (Table 2 and Fig. 1). The other region was readable in five of them, and they could be tentatively identified using it. The results of one region were generally concordant with the VITEK results. One isolate identified as S. mitis/S. oralis by the VITEK had heterogeneous sequences in both regions, but it could, however, be ascertained to be Streptococcus also by pyrosequencing as the sequences were readable at least 10 bases from the beginning of the sequencing reaction.

FIG. 1.

Pyrogram of the v1 region obtained with a blood culture isolate carrying apparently at least two different v1 sequences. (a) The sequence is unambiguous (GCAACTCATCC) until the 12th dispensation, where apparently only some of the alleles contain the A nucleotide. After the 16th dispensation, interpretation of this pyrogram is hard. (b) Pneumococcal sequence using the same dispensation order. The height of the peaks indicating single nucleotide incorporations remain at approximately the same level in this pyrogram, indicating homogeneous 16S rRNA alleles. The pneumococcal v1 sequence is GCAACTCATCCAGAGAAGCAAGCTCCTCCTT.

Excluding the strains with a pyrosequencing score of <96.9% for all type strains, VITEK 2 and pyrosequencing achieved over 85% congruence in all streptococcal groups. One strain with v1 and v2 sequences identical to the S. salivarius type strain was identified as S. infantarius, which may be included in either the S. bovis or the S. salivarius groups according to its bile esculin reaction (8). Only three isolates with sequences identical to those of S. gallolyticus species were included, and they were correctly identified by VITEK 2.

Identification of streptococcal pharyngeal isolates.

Of the 34 pharyngeal isolates, 4 (11.8%) could be unambiguously identified by pyrosequencing (Table 3), 2 could be identified as S. mitis, 1 could be identified as S. parasanguinis, and 1 could be identified as S. salivarius or S. vestibularis. Of these, only the S. parasanguinis isolate was viable for the VITEK analysis and was identified as S. parasanguinis also by VITEK. Of the 18 isolates with one or two mismatches to the closest type strain by pyrosequencing, 13 were viable for the VITEK analysis. These isolates were assigned to the S. mitis (n = 6), S. sanguinis (n = 6), or S. salivarius (n = 1) group by pyrosequencing and were identified to the same streptococcal group with VITEK except for one isolate. Five isolates with lower Identifire scores were variably identified by using VITEK as well. Six pharyngeal isolates had apparently heterogeneous 16S rDNA alleles in the v1 region (Fig. 1 and Table 3), and four of them were subcultured several times in order to purify these potentially mixed isolates. However, the pyrograms of these isolates remained changeless after the subculturing.

Overall agreement of the results.

Ninety-three (75.0%) of the isolates were identified at least to the same streptococcal group by both methods, and fifty-seven (46.0%) reached consistent results at the species level (Table 4). Of the 31 isolates with inconsistent results, 12 (36.4%) had low pyrosequencing scores (<96.9) or had heterogeneous sequences (n = 4 [12.1%]) and 11 (34.4%) isolates could not be conclusively identified by VITEK. In addition, two isolates could not be reliably identified by either method, and two isolates had discrepant results, although both results seemed to be of high confidence. Importantly, the confidence levels or pyrosequencing scores of the incongruent results were often low (Tables 2 to 4), whereas 48 of the 62 (77.4%) VITEK results with excellent confidence levels were congruent with pyrosequencing results at the species level when S. mitis/oralis was considered as a counterpart of the unresolved pyrosequencing result S. pseudopneumoniae/S. pneumoniae/S. mitis. (Tables 2 to 3).

TABLE 4.

Agreement between the identifications of pyrosequencing and VITEK 2

Pyrosequencing score (n)	VITEK confidence levela	No. (%) concordant		No. (%) discordant
		Species level	Group level
≥96.9 to one species (43)	Exc or VG	29		1
	Good or Acc	3	1
	LD (resolved)	5
	LD (unresolved)	0	1
	Unidentified	0		3
	Subtotal	37 (86.0)	2 (4.65)	4 (9.30)
≥96.9 to two or three species (38)	Exc or VG	5	19	1
	Good or Acc	1	3
	LD (resolved)		2
	LD (unresolved)			2
	Unidentified			5
	Subtotal	6 (15.8)	24 (63.2)	8 (21.1)
<96.9 to all species (43)	Exc or VG	12	8	11
	Good or Acc	2	2	4
	LD (unresolved)			1
	LD (resolved)			1
	Unidentified			2
	Subtotal	14 (32.6)	10 (23.3)	19 (44.2)
All isolates (124)	Total	57 (46.0)	36 (29.0)	31 (25.0)

^aConfidence level is a measure of the strength of the VITEK analysis in the descending order Exc (excellent), VG (very good), good, Acc (acceptable), and LD (low discrimination).

DISCUSSION

The taxonomy of especially alpha-hemolytic streptococci is complex, and new species are constantly described (8). Furthermore, a proper gold standard for the identification of streptococci is not available, and therefore we had to compare the pyrosequencing and VITEK 2 results to each other in the present study. The results of the present study illustrate the difficulty of identifying alpha-hemolytic streptococci and the poor knowledge of streptococci of the human normal microbiota. However, the identification of alpha-hemolytic streptococci is becoming more and more important as the amount of immunocompromised patients is increasing. In addition to discrimination between S. pneumoniae and other species of the S. mitis group, the frequent isolation of S. mitis group members from patients with hematologic malignancies and the increasing penicillin resistance of these strains may make their identification a relevant issue (7, 16). Species of the S. anginosus group may have differing clinical importance (4), and high accuracy in the identification of S. bovis group isolates would be important with regard to the association between the isolation of S. gallolyticus subsp. gallolyticus or S. gallolyticus subsp. pasteurianus from blood culture and malignant or premalignant colon lesions (29, 31). However, group-level identification of alpha-hemolytic streptococci may be sufficient for clinical purposes in most cases.

Since the majority of the variation between the 16S rRNA gene sequences of streptococcal species is located close to the 5′ end of the gene, the discriminatory power of the v1 and the v2 region is close to that of the full-length 16S rRNA gene sequence. This was clearly seen also in the present study since the database search results with the longer sequences were in agreement with the pyrosequencing results. Furthermore, the 30-nucleotide sequences of the v1 and v2 regions differentiated the type strains quite well, since only very closely related species belonging to the S. salivarius or S. bovis group had identical v1+v2 sequences. The difficulty in differentiating these species is not surprising since they have highly homologous 16S rRNA genes, as well as rnpB, rpoB, and tuf sequences (2, 6, 22, 23, 32, 37). On the other hand, the rnpB pyrosequencing method could not differentiate S. anginosus and S. constellatus (15).

In the first published evaluation of the colorimetric GP card, all of the 18 alpha-hemolytic isolates were correctly identified (9). Innings et al. (15) compared pyrosequencing of the rnpB gene and VITEK 2 with the fluorimetric GPC card in the identification of 113 blood culture isolates, 44 of which were nonpneumococcal alpha-hemolytic streptococcal isolates. In that study, 36 (81.8%) of the 44 isolates could be unambiguously identified by the pyrosequencing method, and the pyrosequencing and VITEK results of 34 (77.3%) isolates were concordant at least at the group level. Of the 10 isolates with nonconformant results, 9 (90.0%) were isolates that could not be identified by using VITEK. In the present study, the pyrosequencing and VITEK 2 results of 124 streptococcal isolates were compared, and 93 (75.0%) isolates reached concordant results at least at the group level. Most of the discrepant results were found in the S. mitis and S. sanguinis groups, and many of the isolates had low pyrosequencing scores, indicating variation from the type strain. The species belonging to the S. mitis and S. sanguinis groups are difficult to differentiate, and they are often regarded as a single group (30). In contrast, 30 (83.3%) of the 36 isolates belonging to other groups by both methods were concordantly identified, and five of the six discordant results were due to the low quality of the VITEK result. Consequently, the quality controls of both methods seem to be quite reliable, and the results with lower quality can be used for group-level identifications at best.

Most of the sequences obtained in the present study were available in GenBank, but since the species epithets of other than type strains may be incorrect, only the streptococcal type strains can be reliably used for species identification (14). The higher rate of low pyrosequencing scores in our study compared to the study of Innings et al. is also most likely due to the fact that only type strain sequences were included in our sequence database, whereas other reference strains were also included in the database of Innings et al. This illustrates the need for careful study of streptococci to determine the amount of intraspecies sequence variation, which is almost impossible before the species designations of reference strains are ascertained (19). On the other hand, since the smallest difference between the combined v1 and v2 sequences of several streptococcal species belonging to the S. bovis group was only one nucleotide, there may be members of yet-undescribed species among the isolates of the present study. The high degree of variation in streptococcal 16S rRNA gene sequences detected in the present study also indicates that bacteria have a continuum of characters. Consequently, the 16S rRNA gene is unable to give every isolate a clear definition, and the same applies to any species definition.

The pyrosequencing method used in the present study is reliable, user-friendly, and rapid: the whole procedure from a streptococcal colony on an agar plate to the 16S rRNA gene sequence can be carried out during a single working day. Very similar results were obtained with this method compared to the previously published rnpB pyrosequencing method (15). One advantage of the present method is the ability of the primers to also amplify other than streptococcal DNA, whereas the rnpB method failed to amplify the DNA of eight streptococcal species (15). Therefore, presumptive identification before pyrosequencing analysis by this method is sufficient. The pyrosequencing method could also be useful in clinical microbiology laboratory since the major pathogens had specific v1 and v2 sequence combinations. Most importantly, S. pneumoniae had a specific sequence, and no intraspecies variation could be found among the studied pneumococcal isolates or among available genomic sequences.