Relationships among streptococci from the mitis group, misidentified as Streptococcus pneumoniae

Sixty-three misidentified streptococcal isolates (‘misID’ streptococci) from 22 centres in 20 cities were included in the study (Table 1). These isolates originated from the NRCBM collection including approximately 3100 respiratory tract and approximately 3600 invasive isolates received as S. pneumoniae from 1997 until the end of 2015 as well as from the archival collection of the laboratory of approximately 500 isolates from the early 1990s. Thus, misID streptococci accounted for approximately 1% of all pneumococcal isolates and approximately 2% of isolates from the respiratory tract, which constituted the principal source of misID isolates. In particular, among 63 misID streptococci, 29 were obtained from bronchoalveolar aspirates/lavages (BAL), 26 from sputum and six from throat swabs, and single isolates were derived from a central catheter and blood. Patients were aged from 13 to 92 years; information on age was not provided for five patients. Forty-one (65%) and 15 (24%) patients were male and female, respectively; in seven cases, sex was not reported. The male to female ratio was similar to the one observed for 5328 patients with S. pneumoniae infections in Poland in 2006–2015 (62% and 37%, respectively; p = 0.1).

All misID isolates were retested with standard procedures used for S. pneumoniae identification, i.e. for bile solubility and optochin susceptibility (https://​www.​cdc.​gov/​meningitis/​lab-manual/​chpt08-id-characterization​-streppneumo.​pdf; 8th August 2019, date last accessed). Both these tests were performed at least twice. Antimicrobial susceptibility testing was performed as recommended by the European Committee on Antimicrobial Susceptibility Testing (EUCAST) using the viridans group streptococci breakpoints for penicillin, amoxicillin, ceftriaxone, clindamycin and vancomycin; and S. pneumoniae breakpoints for erythromycin, telithromycin, linezolid, chloramphenicol and rifampicin [32]; for ciprofloxacin, isolates with the MIC values > 4 mg/L were considered nonsusceptible as previously for S. pneumoniae [33].

Total DNA was purified using the Genomic DNA Prep Plus kit (A&A Biotechnology, Gdynia, Poland) following the manufacturer’s instructions. The 3′ terminal part of the lytA gene was amplified as described [24] and analysed by sequencing of the amplified product. The PCR-RFLP with BsiHKAI restriction endonuclease (New England BioLabs, Hertfordshire, UK) was used to detect the A203C polymorphism in the 16S rRNA genes [25]. The R6 strain DNA was used as a positive control for PCR-RFLP.

Genomic sequencing was performed with MiSeq (Illumina, San Diego, CA) as an external service (GENOMED, Warsaw, Poland) with sequencing depth of at least 40×. Runs were assembled into contigs using the CLC software v9.0.1 (QIAGEN, Aarhus, Denmark). The presence of acquired antimicrobial resistance genes was verified using the ResFinder 3.1 database (https://​cge.​cbs.​dtu.​dk/​services/​ResFinder/​; 8th August 2019, date last accessed). The online tool https://​pubmlst.​org/​rmlst/​ (8th August 2019, date last accessed) was used for the identification of rMLST-types (rSTs). Novel alleles of ribosomal protein gene loci and novel rMLST profiles found among studied isolates were submitted to the rMLST database. In silico MLST [34] was performed using the S. pneumoniae MLST database [35] (https://​pubmlst.​org/​spneumoniae/​; 8th August 2019, date last accessed).

To compile the reference set for the study, the following data were downloaded from GenBank (http://​www.​ncbi.​nlm.​nih.​gov/​genbank/​, accessed the 17th December 2018): (1) complete genomes and whole-genome sequences of S. pseudopneumoniae and S. mitis; (2) whole-genome sequence of the 596553 strain, a potentially novel representative of the Mitis group [36]; (3) complete genomes of S. pneumoniae; (5) complete genomes of other representatives of Mitis group, including S. oralis, Streptococcus cristatus, Streptococcus peroris, Streptococcus australis, Streptococcus gordonii, Streptococcus infantis, Streptococcus parasanguinis and Streptococcus sanguinis. rSTs were verified using the rMLST database [35] (https://​pubmlst.​org/​rmlst/​; 8th August 2019, date last accessed), and strains with unique complete rMLST profiles were used in further analyses (26 S. pseudopneumoniae, 61 S. mitis, the 596553 strain, 40 S. pneumoniae and 8 other Mitis streptococci; Supplementary Table 1). The genomic sequences of studied isolates and reference strains were uploaded to a private instance of BIGSdb and used for MLSA, rMLST and gene presence/absence analyses with the BLAST tool using default parameters [37]. For MLSA, the scheme including seven loci as described elsewhere [30] was set up in BIGSdb. The rMLST-based phylogenetic analysis was performed using the MUSCLE algorithm [38] and the scheme available in the BIGSdb, excluding the BACT00062 locus (rpmG), which has paralogous genes in streptococcal genomes [31]; i.e., the scheme encompassed 52 loci. The concatenated alignments resulting from MLSA and rMLST analyses were used to construct neighbor-nets using the SplitsTree v.4 [39] and maximum likehood (ML) trees in MEGA-X [40] with 500 bootstrap replicas. Genome annotations were performed using Prokka [41], and the core genome was established using Roary [42]. Core-genome alignments were used to construct approximately-maximum-likelihood (AML) trees in FastTree [43, 44], which were visualized in FigTree (available at http://​tree.​bio.​ed.​ac.​uk/​software/​figtree/​; 18th February 2019, date last accessed).

The differences in distributions were evaluated with the chi-squared test, with the p value < 0.05 considered significant. The adjusted Wallace coefficient (AW) with confidence intervals (CI) as a measure of congruence between identification methods was calculated using the online tool Comparing Partitions at http://​www.​comparingpartiti​ons.​info/​?​link=​Home (8th August 2019, date last accessed). An in silico DNA-DNA hybridization (dDDH) was carried out using the Genome-to-Genome Distance Calculator (GGDC) [45].

This Whole Genome Shotgun project has been deposited at DDBJ/ENA/GenBank under the accession VMKU00000000-VMNE00000000 in the BioProject PRJNA556140. The version described in this paper is version VMKU01000000-VMNE01000000.

All 63 misID isolates presented at least one phenotypic feature typical for pneumococci, such as bile solubility (44 isolates) and optochin susceptibility (62 isolates, of which five repeatedly showed a borderline zone of inhibition of 14–15 mm) (Table 1). Antimicrobial susceptibility results are summarized in Table 2 for 62 isolates (a single isolate repeatedly did not grow under the test conditions, i.e. in cation-adjusted Mueller Hinton broth supplemented with 5% lysed horse blood). Thirty-one isolates (50%) demonstrated a multi-drug-resistant (MDR) phenotype; i.e., they were not susceptible to antimicrobials belonging to three or more classes tested. Analysis of genomic sequences (see below) using ResFinder revealed that 26 and nine isolates carried the erm(B) and mef(A) genes, respectively, including two isolates containing concomitantly both genes; two isolates had the cat gene and 32 isolates were positive for the tet(M) gene. Antimicrobial susceptibility phenotypes to erythromycin, clindamycin, chloramphenicol and tetracycline were generally in a good agreement with the distribution of acquired antimicrobial resistance genes (Table 1) with the exception of three erythromycin-susceptible isolates and a single tetracycline-susceptible isolate, harbouring presumably silent copies of the erm(B) and tet(M) genes. The analysis of the translated amino acid sequences of quinolone-determining regions (QRDR) in ParC revealed single isolates with S₇₉R and S₇₉I changes (Table 1); no mutations typically associated with quinolone nonsusceptibility were observed in the QRDR of GyrA.

The 6-bp deletion in the terminal part of the lytA gene, specific for Mitis streptococci other than S. pneumoniae [24], was detected in all 63 isolates. Patterns observed in the PCR-RFLP of 16S rRNA genes were consistent with the presence of nucleotides other than the signature cytosine at the position 203, specific for S. pneumoniae [25] in all but two isolates, which presented mixed patterns. A subsequent detailed examination of sequencing genomic reads (see below) revealed the presence of both adenosine and cytosine at the position 203, suggesting the heterogeneity of 16S rRNA copies.

Genomic sequencing and assembly yielded genomes ranging approximately from 1.85 to 2.28 Mb in size for investigated misID streptococcal isolates (Table 1). When in silico MLST was performed on obtained genomic sequences, alleles specific for S. pneumoniae were observed in 47 (74.6%) isolates, ranging from a single locus up to all seven loci (two isolates representing ST11884; Table 1). This ST was reported for the NTPn 138 isolate, which in our core-genome analysis clustered with S. pseudopneumoniae and not S. pneumoniae (see below). It is worth noting, however, that particular alleles in the profile of this ST (139-371-345-325-389-677-656) can be found in the association with several other STs of S. pneumoniae in the database. In the rMLST analysis, all 54 observed profiles were novel. Five rSTs were observed for more than a single isolate, including rSTs 106395 (5 isolates) and 23930, 106387, 106395 and 106415 (two isolates each). rMLST allele characteristics for misID streptococci were compared with the set of allelic profiles specific for 7270 rST of 27217 isolates of S. pneumoniae available at https://​pubmlst.​org/​rmlst/​ (as of the 12th May 2019). Alleles common for both misID streptococci and S. pneumoniae were found for all the rMLST loci (Supplementary Table 2). BLAST performed with sequences of genes proposed as specific for S. pneumoniae revealed the counterparts of psaA and ply in 100% and 81.0% of isolates, respectively, and the presence of the spn9802 and spn9828 loci in 49.2% and 38.1% of isolates, respectively (Table 1).

Concatenated and aligned genes from the MLSA and rMLST schemes were used to construct phylogenetic trees and networks that included misID streptococci and reference strains representing unique rSTs. In neighbor-nets based on both MLSA and rMLST, all S. pneumoniae strains formed a distinct cluster with relatively short branches. In particular, this group did not show clustering with any of the misID streptococci (Fig. 1a, b). This cluster was supported by 99–100% bootstrap values on the corresponding ML trees (Supplementary Fig. 1AB). The second cluster comprised 15 reference S. pseudopneumoniae strains and 28 misID streptococci on the MLSA-based neighbor-net and 15 reference S. pseudopneumoniae strains and 30 misID streptococci on the rMLST-based neighbor-net (Fig. 1a, b); the incongruence was observed for four misID streptococci and four reference strains (Fig. 1; Table 1 and Supplementary Table 1). The single misID isolate 12_5905 showed the closest clustering with the S. oralis Uo5 type strain [46]. However, considering the depth of these branches and their support, this isolate could not be unambiguously classified as S. oralis. The Uo5 strain and the 12_5905 isolate as well as S. cristatus, S. peroris, S. australis, S. gordonii, S. infantis, S. parasanguinis and S. sanguinis were clearly separated from other analysed strains and isolates, forming long branches in neighbor-nets and usually well-supported branches in ML trees. The remaining misID streptococci and S. mitis and S. pseudopneumoniae reference strains formed several branches, and the position of these was in several cases incongruent among MLSA and rMLST neighbor-nets. The 596553 strain proposed as a potentially novel species was present in this particular group. Generally, the AW coefficient between MLSA and rMLST for species identification of misID streptococci as S. pseudopneumoniae or S. mitis was 0.768 (CI 0.570–0.966).

Core-genome analysis was performed on misID streptococci (except for a single S. oralis-like isolate 12_5905); reference strains of S. pseudopneumoniae, S. mitis and S. pneumoniae; and the 596553 strain. After construction of an initial AML tree (data not shown), six strains belonging to major observed branches of S. pneumoniae were chosen. Finally, core-genome analysis on the group of 62 misID streptococci and 94 reference strains identified 523 common genes (Supplementary Table 3) and produced an alignment 561 519 bp in length, i.e. covering approximately 25–27% of genomes of the studied misID streptococci. The AML tree included the well-separated S. pneumoniae branch, the S. pseudopneumoniae cluster, comprising 31 misID streptococci and 16 reference strains, and several deep branches associated with S. mitis (Fig. 2). Ten reference strains, reported to the GenBank as S. pseudopneumoniae, were not included within this species (Supplementary Table 1), similarly to other reports [47]. The AW coefficient between the results of core-genome analysis with MLSA and rMLST for distinguishing S. pseudopneumoniae and S. mitis among the misID streptococci was 0.815 (CI 0.639–0.991) and 0.934 (CI 0.814–1.000), respectively. Among S. mitis, 25 misID streptococci formed a separate cluster that contained also the 596553 strain. Such clustering was not clearly apparent in earlier MLSA- and rMLST-based networks and trees (Fig. 1AB, Supplementary Fig. 1AB). The remaining six misID isolates were distributed among other S. mitis-like strains in the AML tree (Fig. 2). All isolates belonging to the S. pseudopneumoniae cluster harboured MLST alleles typical for S. pneumoniae while only 16 (51.6%) isolates of 596553-like and other S. mitis-like isolates showed the presence of such alleles (p = 0; Table 1). All misID representing S. pseudopneumoniae carried counterparts of ply and the 9808 locus while the presence of the 9828 locus was variable. In contrast, the presence of the ply counterparts was characteristic for 19 isolates (61.3%) of the 596553 cluster isolates and remaining S. mitis-like misID isolates, and all these lacked the 9808 and 9828 loci. Nonsusceptibility to penicillin was significantly more common among the 596553 cluster and other S. mitis-like isolates compared with S. pseudopneumoniae (p = 0.005). Such differences were not observed for other antimicrobial classes.