GyrB Polymorphisms Accurately Assign Invasive Viridans Group Streptococcal Species

Abstract

Viridans group streptococci (VGS) are a heterogeneous group of medically important bacteria that cannot be accurately assigned to a particular species using conventional phenotypic methods. Although multilocus sequence analysis (MLSA) is considered the gold standard for VGS species-level identification, MLSA is not yet feasible in the clinical setting. Conversely, molecular methods, such as sodA and 16S rRNA gene sequencing, are clinically practical but not sufficiently accurate for VGS species-level identification. Here, we present data regarding the use of an ∼400-nucleotide internal fragment of the gene encoding DNA gyrase subunit B (GyrB) for VGS species-level identification. MLSA, internal gyrB, sodA, full-length, and 5′ 16S gene sequences were used to characterize 102 unique VGS blood isolates collected from 2011 to 2012. When using the MLSA species assignment as a reference, full-length and 5′ partial 16S gene and sodA sequence analyses failed to correctly assign all strains to a species. Precise species determination was particularly problematic for Streptococcus mitis and Streptococcus oralis isolates. However, the internal gyrB fragment allowed for accurate species designations for all 102 strains. We validated these findings using 54 VGS strains for which MLSA, 16S gene, sodA, and gyrB data are available at the NCBI, showing that gyrB is superior to 16S gene and sodA sequence analyses for VGS species identification. We also observed that specific polymorphisms in the 133-amino acid sequence of the internal GyrB fragment can be used to identify invasive VGS species. Thus, the GyrB amino acid sequence may offer a more practical and accurate method for classifying invasive VGS strains to the species level.

INTRODUCTION

Viridans group streptococci (VGS) are a genetically heterogeneous group of commensal bacteria that cause a wide range of infections in humans, particularly in persons with hematologic disease or neutropenia as a result of chemotherapy (1, 2). Although >25 VGS species have been described, rapid and reliable identification to the species level eludes this group of organisms (3, 4). Accurate species identification is pertinent not only for epidemiological purposes but has potential implications for the clinical management of VGS infections. For example, recent studies show that Streptococcus mitis strains cause a higher incidence of severe clinical disease and have higher rates of resistance to penicillin and fluoroquinolones than do other VGS species (2, 4–6). Additionally, other VGS species have predilections to cause distinct types of infections, such as endocarditis by Streptococcus sanguinis and gastrointestinal infections by Streptococcus anginosus (2, 6).

Routine tests and commercial kits used in clinical practice based on the phenotypic and biochemical traits of individual species, such as API Rapid ID32 Strep, Vitek 2, and Streptogram, have only a 30% to 80% accuracy rate for VGS, depending on the species (7–10). The inherent problems with these approaches lie in the variability of traits within VGS species, poor reproducibility, subjectivity of interpretation, and the lack of availability of up-to-date phenotypic information in the databases on more recently described or reevaluated species (7, 8, 10). Sequence analysis of the 16S rRNA gene has been utilized for VGS species-level identification; however, there is no agreement on a universal cutoff for species delineation (11). In addition, comprehensive 16S databases, such as GenBank, RDP-II, MicroSeq, Ridom, SmartGene IDNS, and 16SpathDB, lack many medically important VGS species, or the species of the reference VGS strains may be misidentified themselves (12–16).

Other molecular-based identification methods for VGS species assignment have employed single-gene targets, such as rpoA, rpoB, rnpB, rodA, sodA, and gdh, with varied success (8, 17–24). Although a single-gene approach is practical, problems with single-gene techniques in VGS stem from a highly diverse population undergoing extensive rates of inter- and intraspecies homologous recombination and horizontal gene transfer, which minimizes the utility of identification procedures based on any one particular gene tested to date (3, 8, 9, 25–30).

Currently, only one methodology has proven successful for differentiating VGS isolates to the species level: multilocus sequence analysis (MLSA). MLSA assigns VGS strains to a species by comparing the concatenated partial sequences of seven housekeeping genes to a database of VGS strains, including type strains (3). Although this approach produces well-resolved species clusters, MLSA is not feasible in a routine clinical microbiology setting due to the time, workload, expense, and difficulty in reliably amplifying all the genes (10, 31). Therefore, there is still a need for a simple and dependable method for identifying VGS strains to the species level.

In a previous work, we sequenced the quinolone resistance-determining regions (QRDRs) of four genes (gyrA, gyrB, parC, and parE) in order to better understand the molecular basis for fluoroquinolone resistance in VGS causing bacteremia in cancer patients (32). We noted that an amino acid change in GyrB seemed to distinguish S. oralis strains from S. mitis strains (32). In this study, we performed systematic analyses of 102 unique VGS bloodstream isolates, using the gyrB nucleotide and amino acid sequences, in an attempt to categorize them into individual species. We compared these data with those from MLSA, 16S gene, and sodA sequence analyses. Our results suggest that the internal amino acid sequence of GyrB offers a practical and effective method for assigning VGS strains to the species level.

MATERIALS AND METHODS

Bacterial strains, culture conditions, and determination of VGS species.

One hundred two VGS bloodstream isolates were collected from unique patients treated at the MD Anderson Cancer Center (Houston, TX, USA) from 15 April 2011 to 31 December 2012. The bacterial isolates were identified as viridans group streptococci based on a Gram-positive reaction, coccoid morphology in chains, negative catalase test, and alpha-hemolysis on blood agar. Pneumococci and enterococci were excluded by routine biochemical tests (optochin, bile solubility, bacitracin susceptibility, leucine aminopeptidase [LAP] and pyrrolidonyl arylamidase [PYR] tests) (4). The strains were grown in a nutrient-rich medium (Todd-Hewitt broth with 0.2% yeast extract [THY]) at 37°C with 5% CO₂, and the genomic DNA was isolated using the Qiagen DNeasy kit. VGS species determination was performed using the gold standard approach of MLSA, as described by Bishop et al. (3). We previously determined the specific species identities of the 102 VGS isolates from cancer patients in a separate study (32).

Sequencing of the 16S gene and sodA, gyrB, and phylogenetic analyses.

The full-length 16S gene was sequenced using the universal primers 8F and 1492R, as previously described (29, 33, 34). The region of the 16S gene used for phylogenetic analyses and species identification was nucleotides 50 to 1457, due to ambiguous reads at the 5′ and 3′ ends for a small number of strains. Additionally, the 5′ end of the 16S gene was also sequenced for comparison using the 10 to 30 and 501 to 522 primers, as previously described (35). The 5′ region of 16S used for phylogenetic analyses, and species identification was from nucleotides 50 to 522. The sodA sequence used for phylogenetic analysis was the same sequence fragment used for MLSA (base pairs 46 to 423), as described previously (3). Amplification and sequencing of the gyrB fragment for each of the VGS strains were done using previously published primers and the protocol of Maeda et al. (36). The region of gyrB used for phylogenetic analyses and species identification was nucleotides 1113 to 1512, corresponding to amino acids 371 to 503. Following alignment with Clustal W, the sequences were analyzed in MEGA version 5.2 to create radial trees using the neighbor-joining statistical method and the maximum likelihood composite model. The robustness of the nodes was evaluated via bootstrapping (1,000 replicates). Percent identity matrices created using Geneious software version 6.1 were analyzed to determine the inter- and intraspecies nucleotide identity ranges using the 16S gene, sodA, and gyrB sequences (Table 1) (37).

TABLE 1.

Intra- and interspecies variation in sodA, 16S rRNA, and gyrB among 102 viridans group streptococci bloodstream isolates from cancer patients

Isolates by sequence analysis type (n)	Intraspecies identity range (%)	Lowest interspecies identity (%)	Highest interspecies identity (%)
16S rRNA
S. mitis (61)	96.6–100	S. anginosus (92.1)	S. oralis (99.9)
S. oralis (16)	97.0–100	S. anginosus (92.2)	S. mitis (99.9)
S. parasanguinis (11)	98.9–100	S. anginosus (92.8)	S. australis (97.8)
S. infantis (4)	99.4–99.9	S. anginosus (92.8)	S. australis (99.3)
S. sanguinis (4)	99.6–99.9	S. vestibularis and S. salivarius (95.7)	S. oralis (97.9)
S. salivarius (2)	99.9	S. anginosus, S. oralis, and S. infantis (93.5)	S. vestibularis (99.8)
S. vestibularis (2)	100	S. oralis (93.5)	S. salivarius (99.8)
S. australis (1)	NAa	S. anginosus (93.0)	S. infantis (99.3)
S. anginosus (1)	NA	S. mitis (92.1)	S. vestibularis (93.6)
sodA
S. mitis	91.3–100	S. anginosus (73.3)	S. oralis (100)
S. oralis	92.9–100	S. vestibularis (75.4)	S. mitis (100)
S. parasanguinis	96.8–100	S. salivarius and S. vestibularis (74.6)	S. mitis (90.2)
S. infantis	93.4–100	S. anginosus (75.1)	S. mitis (95.8)
S. sanguinis	96.3–98.7	S. vestibularis (74.1)	S. parasanguinis (83.3)
S. salivarius	100	S. anginosus (73.3)	S. vestibularis (90.2)
S. vestibularis	100	S. sanguinis (74.1)	S. salivarius (90.2)
S. australis	NA	S. vestibularis (77.5)	S. infantis (94.4)
S. anginosus	NA	S. salivarius (73.3)	S. sanguinis (79.4)
gyrB
S. mitis	88.3–100	S. sanguinis (75.1)	S. oralis (93.5)
S. oralis	92.8–100	S. sanguinis (75.6)	S. mitis (93.5)
S. parasanguinis	88.6–100	S. oralis and S. sanguinis (76.4)	S. australis (85.1)
S. infantis	86.6–100	S. sanguinis (76.1)	S. mitis (86.6)
S. sanguinis	94.5–96.5	S. mitis (75.1)	S. anginosus (83.3)
S. salivarius	97.8	S. parasanguinis and S. mitis (77.6)	S. vestibularis (95.8)
S. vestibularis	100	S. oralis (77.4)	S. salivarius (95.8)
S. australis	NA	S. sanguinis (76.6)	S. mitis and S. infantis (85.3)
S. anginosus	NA	S. mitis (77.1)	S. sanguinis (83.3)

^aNA, not applicable.

Validation of analyses using VGS strains available at the NCBI.

The 54 VGS NCBI strains (17 of which were either type strains, strains deposited in the ATCC database, or strains that had whole genomes available) were chosen based on the availability of their full-length 16S gene, sodA, gyrB and MLSA sequences. If all four were not present in their entirety, those NCBI VGS strains were excluded from the study. The multilocus sequencing types for the NCBI strains were determined by extracting the seven housekeeping gene fragments (map, pfl, ppaC, pyk, rpoB, sodA, and tuf) from the genomic sequences from the NCBI to generate a single concatenated 3,063-bp sequence, followed by constructing a phylogenetic tree that included all concatenated MLSA sequences from VGS isolates within this study, in addition to strains previously examined for species determination (3). Phylogenetic tree construction was performed in MEGA, as described above. The strains were assigned to a species based on their grouping with well-defined strains whose species were previously determined via MLSA (3).

The 54 VGS NCBI strains represented 10 different species known to cause invasive disease in humans, including four S. anginosus, one S. australis, four S. constellatus (a species not present in our 102 VGS isolates), two S. infantis, nine S. mitis, six S. oralis, six S. parasanguinis, five S. salivarius, 15 S. sanguinis, and two S. vestibularis strains. Analyses of the 16S gene, sodA, and gyrB sequences and corresponding phylogenetic analyses were performed for data derived from the NCBI, as described for our invasive VGS isolates.

RESULTS

Limitations of the 16S rRNA gene in resolving VGS species within the mitis and salivarius groups.

16S gene sequencing is the most widely used single-gene approach for determining bacterial species, but previous investigations have shown its limited utility for classifying VGS strains, particularly those of the S. mitis and S. oralis species (7, 8, 33). However, these are the two predominant VGS species isolated from oncology patients (2, 4–6). Because prior studies of 16S gene sequencing did not include MLSA data as a standard for species designation, it remains a possibility that the previous problems reported for 16S analyses in VGS species assignment might be attributed to inaccurate reference species identification. When using MLSA data as a benchmark in the analysis of our 102 invasive VGS isolates, we confirmed that the 16S gene sequences (50 to 1457 bp) of S. mitis and S. oralis strains could be up to 99.9% identical (Table 1). Importantly, the intraspecies identities for S. mitis and S. oralis strains could be lower than the highest interspecies identities found for these species, invalidating the use of the 16S gene for accurately assigning strains to these two species (Table 1). Many publications suggest a universal cutoff of ≥98% identity of the 16S gene sequences to determine the correct species (11, 16, 38). In our data set, this cutoff would not work for S. mitis, S. oralis, S. infantis, S. salivarius, or S. australis, as the highest interspecies identity found is >98% for the strains assigned to these species by MLSA (Table 1).

When phylogenetic analyses were performed using the 16S gene, only S. sanguinis and S. parasanguinis isolates were split as separate species clusters with relatively high bootstrap values (>97) (Fig. 1A). The S. infantis isolates and an S. australis isolate branched together, and S. salivarius and S. vestibularis isolates also branched together. Moreover, multiple S. oralis isolates branched with S. mitis isolates. The bootstrap values for many of the main branches were low (<60), emphasizing the difficulties of using 16S gene sequences to correctly assign VGS strains to particular species.

FIG 1.

Phylogenetic analyses using the full-length 16S rRNA, sodA fragment, and gyrB fragment sequences from 102 VGS bloodstream isolates isolated from cancer patients. Phylogenetic trees were created for each gene, as described in Materials and Methods. The robustness of the nodes was evaluated via bootstrapping (1,000 replicates). (A) Phylogenetic analysis using the 16S rRNA sequence from nucleotides 50 to 1457. (B) Phylogenetic analysis using the sodA sequence fragments from nucleotides 46 to 423. (C) Phylogenetic analysis using the gyrB sequence from nucleotides 1113 to 1512. (A to C) Strains were assigned to particular VGS species based on MLSA and given a color, as noted in the key. The scale bar indicates genetic distance. If strains were genetically identical at a locus, they appear as one circle (two overlapping circles). The numbers represent the bootstrap values of >80 at each major branch point.

The majority of the variation between the 16S gene sequences of different species is found in the 5′ region (v1 region), and, therefore, sequence analysis of the 5′ region is often used for bacterial species identification (e.g., MicroSeq 500 PerkinElmer Applied Biosystems) (5, 38). For this reason, we decided to compare 16S gene 5′ sequencing to the traditional 16S gene sequencing (using universal primers 8F and 1492R) in order to determine their relative efficacies for VGS species assignment compared to that of the MLSA gold standard.

Overall, we found that using the region 50 to 522 bp of the 16S gene gave lower bootstrap values and showed a greater divergence of isolates within the same species (i.e., S. oralis isolates are present on four different branches) compared to the results of full-length 16S gene sequencing (see Fig. S1 in the supplemental material). In addition, there was more intermingling of species when only the 5′ 16S gene sequence was analyzed, as S. oralis isolates were found on the same major branches as the S. mitis, S. parasanguinis, and S. sanguinis isolates. Therefore, the full-length 16S gene sequence gives better resolution than the 5′ region alone. However, neither form of 16S gene sequence analysis is ideal for VGS species assignment.

sodA sequence analysis is an unreliable method for species identification of mitis group streptococci.

The housekeeping gene sodA has frequently been studied as a genotypic method for identifying VGS strains to the species level and has shown to be one of the most effective single-gene analyses, correctly identifying most VGS strains approximately 95% of the time (8–10, 19, 23). However, shortcomings in the ability of sodA sequencing to accurately resolve some S. mitis and S. oralis strains have been reported (3, 8–10, 19). Thus, we next sought to determine how accurate sodA sequence analysis would be for our strains. Similar to 16S gene sequence analyses, sodA sequence analysis displayed limitations in identifying VGS species, as the highest interspecies identities were greater than some of the intraspecies identities between isolates of the S. mitis, S. oralis, and S. infantis species (Table 1).

Phylogenetic trees using sodA sequence analysis effectively delineated strains belonging to the species S. sanguinis, S. vestibularis, S. salivarius, and S. parasanguinis, with bootstrap values of >98 (Fig. 1B). However, sodA sequence analysis seems to be problematic for the precisely identifying isolates of S. oralis, S. mitis, S. australis, and S. infantis. In our data set, S. australis and S. infantis isolates were clustered on the same branch, and two S. mitis isolates branched with S. oralis isolates when phylogenetic analyses were performed using sodA. Although sodA is included in the MLSA scheme, this shows how the overall species designation can be different from a single-allele species designation. Indeed, when the sodA alleles of these strains were entered into the MLSA database (www.emlsa.net), they were assigned S. oralis as their species, according to that allele (data not shown), despite the overall MLSA species designation of S. mitis. This phenomenon has been reported by Bishop et al. (3) and is most likely due to homologous recombination with the sodA gene between strains of different species. In accordance with the nucleotide data, phylogenetic analysis indicated that using the corresponding amino acid sequence of SodA also failed to accurately assign VGS isolates to their MLSA designated species (data not shown).

Using gyrB nucleotide and amino acid sequence analyses as a novel method to identify viridans group streptococci.

Given that neither the full-length 16S gene nor sodA sequences precisely matched our invasive VGS strains to the MLSA-assigned species, we next determined the efficacy of using the gyrB sequence. The region of gyrB used for comparative analyses was from nucleotides 1113 to 1512, which corresponds to the quinolone resistance-determining region (QRDR). When nucleotide identity matrices were analyzed, we determined again for S. mitis, S. oralis, and S. infantis that the highest interspecies identity might be more than some intraspecies identities (Table 1). Importantly, however, none of the strains exhibited >96% nucleotide interspecies identity when using the gyrB nucleotide sequence, compared to analyses with the 16S gene, which had up to 99.9% interspecies identity, and sodA analysis, which had up to 100% interspecies identity (Table 1). The highest interspecies nucleotide identity using gyrB was 95.8%, between S. salivarius and S. vestibularis.

When phylogenetic analysis was performed, we found that the gyrB nucleotide sequence successfully delineated all of the 102 VGS strains into individual species branches that were consistently in accordance with the MLSA assignment (Fig. 1C). Importantly, the gyrB sequences successfully resolved all S. oralis and S. mitis strains into their respective species. The bootstrap values for most of the branches dividing different species were ≥84, validating the confidence of the branching and therefore the species identifications achieved with the gyrB sequence analysis.

However, the considerable intraspecies variations of the gyrB sequences observed (Table 1) indicated that a simple BLAST alignment, instead of a detailed phylogenetic analysis, could not readily assign VGS strains to a particular species. Thus, we sought to determine whether the GyrB amino acid sequence could be used in this regard. We discovered that each VGS species showed a distinct amino acid signature sequence for the portion of GyrB analyzed (Table 2). In some instances, a distinct amino acid at one particular position separated a particular species from all other VGS species. For example, all S. sanguinis isolates had an arginine at position 378, whereas the remaining VGS species had a lysine at this position. Similarly, all S. vestibularis isolates had a serine at position 489 in contrast to an alanine in other VGS species (Table 2). In other circumstances, it was the combination of amino acids that constituted their signature GyrB sequence. For example, S. mitis isolates were distinguished from S. oralis by a serine at position 494 versus a threonine for S. oralis (Table 2). S. parasanguinis was the only other species that contained a threonine at position 494; however, S. parasanguinis was differentiated from S. oralis by a leucine at position 371, a glutamine at 425, and an isoleucine at 503 (Table 2). Thus, for our invasive VGS isolates, the analysis of the partial GyrB sequence correctly identified species. Furthermore, although we had previously shown that 79/102 of our strains were fluoroquinolone nonsusceptible, none of the strains had determinative polymorphisms in the GyrB QRDR (32). Therefore, in our cohort of invasive VGS strains, fluoroquinolone resistance does not influence the ability of the GyrB sequence to distinguish VGS species (32).

TABLE 2.

Numbers of isolates from MD Anderson Cancer Center and NCBI and GyrB signature sequences of VGS species

Streptococcus species	MDA (n)a	NBCI (n)b	GyrB amino acid at position:
			371	376	378	379	423	424	425	426	427	489	494	499	503
S. mitis	61	9	Mc	I	K	R	N	P	A	E	T	A	S	Q	L
S. oralis	16	6	M	I	K	R	N	P	A	E	T	A	T	Q	L
S. parasanguinis	11	6	L	I	K	K	N	P	Q	E	T	A	T	Q	I
S. infantis	4	2	L	I	K	R	N	P	A	E	T	A	S	Q	I
S. sanguinis	4	15	L/I/F	V	R	K	D	P	Q	E	T	A	S	Q	I
S. salivarius	2	5	L	I	K	K	N	A	E	M	N	A	S	H	I
S. vestibularis	2	2	L	I	K	K	N	A	E	M	N	S	S	H	I
S. australis	1	1	L	I	K	K	N	P	A	E	T	A	S	Q	I
S. anginosus	1	4	L	V	K	K	D	P	H	E	T	A	S	Q	I
S. constellatus	0	4	L	V	K	K	D	P	H	E	T	A	N	Q	I

^aThe number of invasive VGS isolates collected at MD Anderson Cancer Center (MDA) per species.

^bThe number of VGS isolates available at NCBI per species.

^cBold type denotes the amino acids used to differentiate each VGS species; see also Fig. 3.

Validation of utilizing the gyrB nucleotide and amino acid sequences for VGS species-level identification using NBCI isolates.

Our VGS strains were derived from a single institution over a short period of time, and thus our findings required validation. To do this, we performed the same phylogenetic analyses using the sequences of 54 VGS isolates available from the NCBI. Similar to our invasive VGS data, two NCBI S. oralis strains grouped with S. mitis strains when the full-length 16S gene sequence was analyzed (Fig. 2A). Additionally, S. salivarius and S. vestibularis strains were not resolved into their appropriate species by the full-length 16S gene sequence either. Phylogenetic analyses using the sodA sequence differentiated most of the 54 NCBI isolates in accordance with their MLSA species designation, although the S. infantis and S. australis strains were located on the same branch (Fig. 2B). When phylogenetic analyses were performed using the gyrB nucleotide sequence for the NCBI strains, we found that again, gyrB successfully segregated each species in accordance with their MLSA designation, with most major branches having bootstrap values of >90 (Fig. 2C). Validation of the GyrB amino acid sequences also confirmed that all 54 NCBI strains displayed the same amino acid sequence for each species, as was seen in the 102 clinical VGS isolates (Table 2). Additionally, three S. constellatus strains were available from the NCBI (a species not represented in our 102 clinical VGS isolates), and we found that S. constellatus isolates had an asparagine at position 494, whereas all other VGS species had either a serine or threonine at that position (Table 2). Thus, our partial GyrB amino acid sequence scheme performs equally well with VGS type strains and other strains available from the NCBI.

FIG 2.

Phylogenetic analyses using the full-length 16S rRNA, sodA fragment, and gyrB fragment sequences from 54 diverse NCBI isolates. Phylogenetic trees were created for each gene, as described in Materials and Methods. The robustness of the nodes was evaluated via bootstrapping (1,000 replicates). (A) Phylogenetic analysis using the 16S rRNA sequence from nucleotides 50 to 1457. (B) Phylogenetic analysis using the sodA sequence fragments from nucleotides 46 to 423. (C) Phylogenetic analysis using the gyrB sequence from nucleotides 1113 to 1512. (A to C) Strains were assigned to particular VGS species based on MLSA and given a particular color, as noted in the key. The scale bar indicates genetic distance. If strains were genetically identical at a locus, they appear as one circle (two overlapping circles). The numbers represent the bootstrap values of >80 at each major branch point.

DISCUSSION

In this work, we discovered that partial sequencing of the gyrB gene and subsequent determination of the corresponding amino acid sequence is a simple and effective means for discriminating VGS species, including those strains of the mitis and salivarius groups that are difficult to differentiate using other means (7–10, 19). The recent development of accurate VGS species assignment using the seven-gene MLSA has already augmented the understanding of clinical correlations with VGS species assignment (5, 6, 39). Thus, having a practical and effective means for both research and clinical microbiology laboratories to assign VGS strains to particular species is becoming increasingly imperative.

The gyrB nucleotide sequence accurately assigned species for both our invasive VGS isolates and the VGS strains present in the NCBI database. Still, the clinical applicability of gyrB sequence analysis is limited, as it requires phylogenetic tree analysis using a large cohort of strains. Circumventing this obstacle by converting the gyrB DNA sequence to its corresponding amino acid sequence allows for a single VGS strain to be readily assigned to a species using only a small number of amino acids. In Fig. 3, we offer a workflow for VGS species identification that can be used in either a research or clinical setting. Importantly, although gyrB encodes part of DNA gyrase, which is a target of fluoroquinolones, polymorphisms in GyrB are not a major mechanism by which VGS strains develop fluoroquinolone resistance (32, 36). Therefore, fluoroquinolone resistance does not influence the ability of GyrB to assign VGS strains to a particular species.

FIG 3.

Proposed workflow for using the partial GyrB amino acid sequence to differentiate invasive viridans group streptococcal species. The diagram shows the order of events that should be used in order to determine the species of an invasive viridans group streptococcal isolate incorporating the novel molecular technique using the GyrB fragment amino acid sequence.

A major concern with any single-gene approach is that recombination might lead to inaccurate VGS species assignment, as has been observed for sodA and other single-gene analyses (3, 8, 9). However, out of 156 strains, the partial GyrB amino acid sequence was 100% successful for species identification, indicating that gyrB is perhaps located in a genomic area that does not experience frequent homologous recombination. In support of this hypothesis, we found that the genes on both sides of gyrB are present in the same locations in all NCBI VGS strains used in this study, multiple strains of S. pneumoniae, and even in the beta-hemolytic streptococcal species S. pyogenes and S. agalactiae (data not shown).

Our findings have two limitations worth discussing. First, although the gyrB nucleotide and amino acid sequence proved effective for discriminating all VGS species studied here, gyrB does not differentiate S. pneumoniae or S. pseudopneumoniae from S. mitis (data not shown). Typically, however, S. pneumoniae can be distinguished from nonpneumococcal VGS in the clinical setting by the use of optochin susceptibility and bile solubility tests (4). It is worth noting that some strains of S. mitis and, to a greater degree, S. pseudopneumoniae, can be optochin susceptible and bile soluble (29). However, Scholz, Poulsen, and Kilian (35) presented a novel molecular methodology in which S. pneumoniae can be differentiated from VGS species within the mitis group due to a conserved nucleotide difference of a cytosine at position 203 of the 16S rRNA sequence, compared to an adenine in other mitis group streptococci. Second, for relatively rare VGS species, such as S. infantis, S. constellatus, and S. australis, the low number of strains available for analysis means that additional strains will need to be tested to confirm our findings. Of note, although mutans group isolates are classically considered viridans group streptococci, they are not common causes of invasive disease and thus were not considered in these analyses (3, 5, 6).

In conclusion, we show here that the partial amino acid sequence of GyrB is sufficient to accurately assign invasive VGS strains to a specific species. These data provide an economical and reliable method for VGS species identification that can be employed in both clinical microbiology and research settings.