Characteristics of the BSAC and UKHSA IE BSI S. mitis isolates
To characterise the genetic diversity, AMR gene profiles, and identify virulence genes associated with invasive S. mitis infection, we obtained and performed WGS of 217 presumed S. mitis isolates from patients with BSI and clinically diagnosed IE from the UK and Ireland between 2001 and 2016 (Fig. 1, Supplementary Fig. 1, and Supplementary Data 1 and 2). These isolates were collected as part of BSAC’s Resistance Surveillance Project (n = 172), and UKHSA’s voluntary identification service (n = 45). While the retrospective nature of our analysis has preluded the rigorous application of the modified Duke/European Society of Cardiology (ESC) 2023 diagnostic criteria for IE33, the isolates were all from patients where IE had been clinically diagnosed (and given the BSI, likely fulfilling the “definite” or “possible” category), and had referred to the reference laboratory for species confirmation and further antibiotic sensitivity testing. Two of 172 BSAC and three of 45 UKHSA isolates were not viable after bacterial culture of transport swabs and were therefore not sequenced. One UKHSA isolate subsequently failed WGS quality control due to low DNA concentration. Accurate species determination among the VGS has been a challenge using conventional and molecular approaches34, therefore, we used a bioinformatic approach for species confirmation among the presumed S. mitis isolates obtained from BSAC and UKHSA. Overall, S. mitis was confirmed by WGS in 106 of 170 (62.4%) and 23 of 41 (56.1%) of the viable BSAC and UKHSA isolates respectively (Fig. 1). S. mitis was most frequently misidentified as S. oralis and S. infantis for 29 of 211 (13.7%) and 27 of 211 (12.8%) sequenced isolates, respectively (Supplementary Table 1), highlighting the importance of WGS-based approaches for species confirmation among the VGS34. The S. mitis isolates from patients with IE were from patients of all ages (0 to 99 years) with the age range 50–59 years having the highest frequency of S. mitis isolates (20.2%; 26/129) (Supplementary Table 2). Of the 129 confirmed S. mitis isolates, 70 (54.3%) were collected from men and 58 (45.0%) from women (Supplementary Table 2).
Population genetic diversity of IE-associated S. mitis
S. mitis strains are known to be highly diverse16,25,32,35–38, but accurately assessing S. mitis population structure has been a challenge due to limited S. mitis genome sequences and species specific molecular genotyping tools. Because of this, it remains unclear whether specific lineages predominantly cause BSIs associated with S. mitis in localised populations and over time. To address this, we therefore quantified the genetic diversity of the IE-associated S. mitis isolates collected from the UK and Ireland, and we determined if population-level genetic diversity changed over time during the 16-year surveillance period. We calculated the number of single nucleotide polymorphisms (SNPs) between all the pairs of isolates to quantify the genetic diversity (Supplementary Data 3). The number of SNPs between the pairs of isolates ranged from approximately 45,000 to 55,000 bp (Supplementary Fig. 2) and varied significantly over time (Kruskal-Wallis test p < 0.001) across the 16-year period. The observed high number of SNPs distinguishing pairs of isolates sampled in the same year suggested a high genetic diversity of the IE-associated S. mitis strains during the surveillance period, reflecting the presence of multiple different S. mitis lineages.
Next, we confirmed the high genetic diversity of the IE-associated S. mitis isolates by the identification of long internal and terminal branches separating the isolates in the constructed maximum likelihood phylogenetic tree (Fig. 2a). The long phylogenetic branches indicated the existence of several distinct lineages. To confirm the presence of multiple distinct lineages, we then used our recently developed S. mitis MLST scheme available on the PubMLST website (https://pubmlst.org/smitis) and a complementary whole-genome-based sequence clustering approach using the PopPUNK framework (https://www.bacpop.org/poppunk/).30 This analysis identified 127 (98.4%) unique sequence types (STs) based on the S. mitis MLST scheme and an equal number of the PopPUNK lineages, that we defined as Global Sequence Clusters (GSCs), among the 129 IE-associated S. mitis isolates (Supplementary Data 4). Two isolates isolated in 2014 and 2016, which belonged to sequence type ST30, belonged to a single lineage, namely, GSC27, and differed from each other by 3,024 SNPs. However, the isolate pair had the same penicillin susceptibility profile. Similarly, two ST36 isolates collected in 2007 and 2015 belonged to lineage GSC28 and differed from each other by 6,413 SNPs and by their penicillin susceptibility profiles (Fig. 2a). Since we did not have access to patient-identifiable data from BSAC or UKHSA, we could not exclude the possibility that S. mitis isolates of the same STs represented recurrent infection of the same patient. Nonetheless, considering that 127 out of 129 isolates (98.4%) belonged to different STs and lineages, and the large pairwise distances between isolates of the same ST, this appears less likely. Therefore, using multiple approaches, these observations suggest that S. mitis BSI-associated IE in the UK and Ireland is not predominantly caused by a select few dominant lineages.
Temporal trends in AMR among the IE-associated S. mitis isolates
Due to the limited focus, misdiagnosis, and low incidence of S. mitis BSIs, temporal AMR trends for S. mitis associated with IE have not been well described. We therefore assessed the susceptibility of the IE-associated S. mitis isolates against commonly used antibiotics to treat suspected S. mitis IE (penicillin, amoxicillin, gentamicin, and vancomycin)39. We found the distribution of penicillin and amoxicillin non-susceptible isolates across the entire phylogeny, not restricted to only specific phylogenetic branches containing closely related isolates (Fig. 2a). Among isolates with phenotypic MIC data, 23.3% (30/129) and 6.2% (6/97) of the isolates were non-susceptible to penicillin and amoxicillin, respectively (Supplementary Data 1 and 2, antibiotic abbreviations used by BSAC and UKHSA are explained in Supplementary Table 3). We show that all 30 penicillin non-susceptible isolates belonged to different STs and GSCs (Supplementary Data 5). We observed non-susceptibility to penicillin among isolates across the surveillance period (Fig. 2b), which would impact the use of penicillin as a first-line antibiotic for Streptococcal IE39. All the isolates had low-level gentamicin resistance (MIC ≤ 128 µg/mL), which would not impact its use as a synergistic antibiotic in IE management39. Additionally, all isolates showed full susceptibility to vancomycin. Our AMR findings support the use of the current antibiotic treatment regimens for the management of IE, however, our data empasises that continued surveillance remains critical for monitoring AMR trends particularly for penicillin.
To further assess the temporal changes in antimicrobial susceptibility, we aggregated phenotypic MICs for several antibiotics into four 3-year intervals, namely 2001–2004, 2005–2008, 2009–2012, and 2013–2016. Due to the small number of phenotyped isolates per year, these 3-year intervals ensured derivation of more robust estimates for the phenotypic MIC trends based on a sufficient number of isolates. Using this approach, we found no statistically significant differences in the median MICs across all four-time intervals for seven out of eight antibiotics (Table 1 and Supplementary Fig. 3). Conversely, the median MICs for gentamicin showed a significant overall decrease over time (Kruskal-Wallis test p = 0.001). Despite the variability of the MIC changes over time due to the limited number of phenotyped isolates, our findings provide baseline data for the genomic surveillance of AMR in S. mitis associated BSIs, including IE, in the UK and Ireland, regionally and globally.
Table 1
Antibiotic median MIC for Streptococcus mitis isolates from patients with IE across four 3-year intervals
|
Year
|
Number of isolates
|
Median MIC (µg/mL)
|
Number of non-susceptible isolates and proportion (%)
|
P value
|
Amoxicillin
|
2001–2004
|
34
|
0.03
|
2 (5.9)
|
|
n = 97
|
2005–2008
|
21
|
0.06
|
4 (19.0)
|
|
|
2009–2012
|
25
|
0.03
|
2 (8.0)
|
|
|
2013–2016
|
17
|
0.06
|
2 (11.8)
|
0.06
|
Cefotaxime
|
2001–2004
|
38
|
0.06
|
2 (5.3)
|
|
n = 118
|
2005–2008
|
23
|
0.06
|
1 (4.3)
|
|
|
2009–2012
|
29
|
0.06
|
3 (10.3)
|
|
|
2013–2016
|
28
|
0.125
|
0 (0)
|
0.403
|
Gentamicin
|
2001–2004
|
38
|
6
|
0 (0)
|
|
n = 118
|
2005–2008
|
23
|
4
|
0 (0)
|
|
|
2009–2012
|
29
|
4
|
0 (0)
|
|
|
2013–2016
|
28
|
2
|
0 (0)
|
0.001
|
Penicillin
|
2001–2004
|
41
|
0.03
|
4 (9.8)
|
|
n = 129
|
2005–2008
|
28
|
0.064
|
10 (35.7)
|
|
|
2009–2012
|
29
|
0.03
|
6 (20.7)
|
|
|
2013–2016
|
31
|
0.06
|
10 (32.3)
|
0.111
|
Vancomycin
|
2001–2004
|
38
|
0.5
|
0 (0)
|
|
n = 118
|
2005–2008
|
23
|
0.5
|
0 (0)
|
|
|
2009–2012
|
29
|
0.5
|
0 (0)
|
|
|
2013–2016
|
28
|
0.5
|
0 (0)
|
0.230
|
Other Streptococcus species, including S. pneumoniae, S. agalactiae or Group B Streptococcus (GBS), and S. pygenes or Group A Streptococcus (GAS), are characterised by lineages that are more likely to cause invasive diseases in humans40–43. The identification and tracking of lineages has facilitated genomic surveillance and guided clinical interventions against these species44, an approach that could be equally valuable for monitoring the epidemiology of invasive S. mitis strains. We therefore analysed the 129 S. mitis isolates from patients with clinically diagnosed IE in the context of globally sampled strains to better understand the global genetic diversity and the distribution of AMR and virulence genes amongst invasive S. mitis. We compiled a total of 322 confirmed whole-genome sequenced S. mitis isolates, from the present study and publicly available genomic sequence repositories, representing 258 PopPUNK lineages and 259 STs (Fig. 3a, Fig. 3b, and Supplementary Data 4). Analysis of the metadata for our sequenced isolates and the contextual publicly available sequences revealed that 158 out of 322 isolates (49.1%) were from carriage, 152 (47.20%) were from invasive disease, and 12 (3.7%) were from unknown sources (Supplementary Data 4). Of the invasive isolates, 138 (42.9%) were from patients with infective endocarditis (129 of these were from this study), 13 (4.0%) were from bacteraemia, and 1 (0.3%) was from pneumonia. Overall, there were no shared lineages between the S. mitis isolates obtained from patients in the UK and Ireland with clinically diagnosed IE and other global strains from carriage or invasive disease (Fig. 3c). However, the global strains that shared STs and GSCs were part of a previous carriage study that sampled the same individuals, such that the same strain was sampled multiple times32. Therefore, S. mitis isolates from asymptomatic carriage and invasive disease are distributed across the entire phylogeny of the global isolates, indicating the potential for all, rather than a select few lineages to cause IE. Furthermore, we found no clustering of isolates based on the AMR genes (Supplementary Fig. 4), or major virulence genes associated with pneumococcal pathogenicity (Fig. 3c). Two of the latter genes, encoding the pneumococcal capsular polysaccharide synthesis (cps) region and pneumolysin (ply) gene, were found in distinct positions on the global S. mitis phylogeny (Fig. 3c). Pneumococcal adherence and virulence protein A (pavA) and pneumococcal surface adhesin A (psaA) genes were present among all 322 isolates (100%) (Supplementary Data 6). Together, these findings demonstrate that even when viewed from the global context, invasive S. mitis strains are not predominantly associated with a single or limited number of lineages.
Distribution of known pneumococcal virulence genes in IE-associated S. mitis isolates
Due to the close genetic relatedness between S. mitis and the more pathogenic S. pneumoniae16,25, both species have been suggested to have evolved from a most recent common ancestor26. However, it has been suggested that S. mitis has evolved through genome reduction, which has resulted in the loss of several virulence genes, possibly explaining its typical commensal and opportunistic lifestyle26. As a more diverse species and a known donor of diverse AMR and virulence genes to the pneumococcus27,45,46, it may therefore be possible that IE-associated S. mitis isolates may have retained a majority of genes associated with virulence in the pneumococcus. To investigate this hypothesis, we investigated the distribution of pneumococcal virulence genes among IE S. mitis isolates in the context of other closely related Streptococcus species (Fig. 4a). S. mitis isolates formed multiple highly diverse lineages associated with long internal and terminal phylogenetic branches from the other species, which further confirmed that the IE isolates were not misclassified as other VGS. We observed a variable presence of the virulence genes among the IE-associated S. mitis isolates (Fig. 4a), such as those found in the cps locus (23.3%; 30/129), autolysin A (lytA; 18.6% [24/129]), autolysin C (lytC; 58.1% [75/129]) and ply genes (11.6%; 15/129). However, the IE-associated isolates grouped together into more than four sub-populations defined by distinct virulence gene profiles, in which one or more virulence genes were dominant (Fig. 4a). For example, lytC was found in 91.5% (54/59) of isolates of a clade grouping 37 IE-associated S. mitis isolates, 20 from asymptomatic carriage, one from bacteraemia, and one from an unknown isolation condition. We also observed that 34.1% (44/129) of IE-associated S. mitis isolates were genetically more closely related to the pneumococcus, and these formed two main groups that were either cps, lytA, and lytC dominant, or cps, lytA, lytC, and ply dominant. There was also no predominance of pneumococcal virulence genes among IE-associated compared to carriage isolates (Fig. 4b). We therefore conclude that all the IE-associated S. mitis caused IE regardless of pneumococcal virulence gene presence or closer genetic relatedness to the pneumococcus, further emphasising the opportunistic nature of S. mitis infections.
Differential abundance of putative pathogenicity enhancing genes among S. mitis isolates from IE-associated BSI and carriage
Previous studies have suggested that horizontal gene transfer (HGT) between S. mitis and other more virulent members of the Streptococcus genus, specifically S. pneumoniae, drives the spread of AMR and virulence between these species25,27,47. Additionally, S. mitis is known to harbour pneumococcal virulence genes, including those involved in the biosynthesis of serotypes 1 and 5 capsules21,22 and homologs of other pathogenicity-associated genes, including Zinc metalloproteases (zmpC, zmpC, and zmpD)48, neuraminidases (nanA and nanB), pneumolysin (ply), immunoglobulin A protease (iga), and autolysins (lytA-C)18 and glucan binding protein B (gbpB or pcsB)49. However, no systematic analyses to assess the abundance of all the functionally characterised (or known) and hypothetical genes in the S. mitis pan-genome between the invasive and carriage isolates have been conducted to date. Since S. mitis is widely regarded as a source of virulence factors which enhance the pathogenicity of pneumococcal strains25, we therefore hypothesised that the S. mitis strains associated with BSIs may show a higher abundance of virulence genes compared to the isolates sampled from the asymptomatic carriage.
To address this, we employed a genome-wide association study (GWAS) approach (Fig. 4c), increasingly used to identify genomic loci associated with bacterial phenotypes50–54. Our null hypothesis was that no gene influenced the pathogenicity of S. mitis. Therefore, we expected the distribution of any gene would be similar among the BSI and carriage isolates due to the inclusion of phylogenetically similar but phenotypically distinct pairs of isolates. However, we found fifteen orthologous gene clusters which were differentially overrepresented among either IE-associated or carriage isolates (Table 2, Supplementary Fig. 5, and Supplementary Data 7). Among these genes were TP-binding cassette (ABC) transporters, competence-specific and Hca operon transcription regulators, phage-associated proteins, autolysin (a known pneumococcal virulence factor30,55–57), and several uncharacterised hypothetical proteins. Twelve of the genes were more overrepresented in the carriage isolates than in the invasive disease isolates, while three genes showed the opposite association. As the pneumococcus is a close relative of S. mitis, we screened 493 randomly selected invasive pneumococcal isolates for the presence of these 15 overrepresented genes in S. mitis (Fig. 5). The pneumococcal isolates were obtained from the Global Pneumococcal Sequencing Project and were part of the Centers for Disease Control and Prevention’s (CDC) active bacterial core surveillance (Supplementary Data 8). We identified the presence of 6 out of 15 overrepresented orthologous S. mitis gene clusters that were also prevalent among the pneumococcal isolates. Among these gene clusters, the pneumococci had a high prevalence of 2 out of 3 genes that were more overrepresented in the invasive S. mitis than carriage isolates, a transcriptional regulator and hypothetical gene. Together, these findings suggest that these identified genes may potentially modulate the pathogenicity of S. mitis, facilitating potential rare transition from a typical commensal to a pathogenic lifestyle58,59.
Table 2
Summary of the S. mitis genes, which were differentially abundant among a subset of phylogenetically paired invasive disease and carriage isolates.
Gene cluster ID*
|
Gene name
|
Gene presence in the paired invasive disease and carriage isolates
|
P-value**
|
Gene product/description
|
None
|
Carriage isolates only
|
Disease isolates only
|
Both
|
SCLS1
|
btuD
|
17
|
12
|
1
|
14
|
0.0055
|
ABC transporter, ATP-binding protein
|
SCLS2
|
|
17
|
12
|
1
|
14
|
0.0055
|
ABC-2 family transporter protein
|
SCLS3
|
|
31
|
1
|
9
|
3
|
0.0269
|
Transcriptional regulator ComX2
|
SCLS4
|
|
29
|
9
|
1
|
5
|
0.0269
|
DNA-binding phage protein
|
SCLS5
|
|
9
|
6
|
17
|
12
|
0.0371
|
Hypothetical protein
|
SCLS6
|
|
19
|
10
|
2
|
13
|
0.0433
|
ComC/BlpC family leader-containing pheromone/bacteriocin***
|
SCLS7
|
|
20
|
2
|
10
|
12
|
0.0433
|
Hypothetical protein
|
SCLS8
|
lytA
|
31
|
10
|
2
|
1
|
0.0433
|
Autolysin
|
SCLS9
|
|
34
|
8
|
1
|
1
|
0.0455
|
SPFH domain-containing protein***
|
SCLS10
|
|
35
|
8
|
1
|
0
|
0.0455
|
Major Facilitator Superfamily (MFS) transporter***
|
SCLS11
|
|
34
|
8
|
1
|
1
|
0.0455
|
Phage protein
|
SCLS12
|
hcaR
|
35
|
8
|
1
|
0
|
0.0455
|
Hca operon transcriptional activator HcaR
|
SCLS13
|
|
35
|
8
|
1
|
0
|
0.0455
|
YbhB/YbcL family Raf kinase inhibitor-like protein***
|
SCLS14
|
|
34
|
8
|
1
|
1
|
0.0455
|
Hypothetical protein
|
SCLS15
|
|
34
|
8
|
1
|
1
|
0.0455
|
Phage transcriptional regulator, Cro/CI family protein
|
*Specific nucleotide sequences of the representative genes in each orthologous gene cluster inferred from pan-genome clustering analysis using Panaroo are available in Supplementary Data 5.
**The P-value was calculated using McNemar’s exact test based on phylogenetically paired invasive disease and carriage isolates.
***Gene description determined through the online NCBI BLAST tool, it’s databases, and using default parameters
|