Here we report the core resistome of oral streptococci by leveraging thousands of curated whole-genome sequences. We show that while some of these ARGs are anchored in the chromosome and seem to show little recombination in the population of sequences tested, others are almost exclusively present in ICEs, indicating their high chance of recombination frequency. In addition, we show that the rate of ARG and ICE per genome varies significantly across streptococcal species, with the mutans group exhibiting low counts for both whereas the anginosus group seems to be prolific both in ARG diversity as well as in the presence of uncharacterized ICEs. Furthermore, we investigated diversity and recombination in the pbps of these species, with the finding that a high diversity is present particularly in the mitis group. Intra-species and inter-species recent recombination events in this group were observed at a high frequency, and particularly S. mitis and S. oralis are prolific donors and receivers of genetic fragments from other species.

We report high diversity in the pbp genes particularly in the non-pneumococcal species of the mitis group. In other groups of streptococci, these genes were more conserved, often clustering together. In line with our findings, previous studies focused on the mitis group have also reported large heterogeneity in commensal species compared to the pneumococcus (Jensen et al. 2015). We observed that highly resistant strains clustered mostly together in pbp1a. Further, cross-species recombination was more often observed for pbp2b and pbp2x. These findings go in line with previous reports indicating that the pbp1a governs high-level resistance in S. pneumoniae (D’Aeth et al. 2021; Smith and Klugman 1998). Fewer recombinations were detected in pbp2b compared to pbp1a and pbp2x. We hypothesized that this could have happened due to sequence variations across species resulting in poor alignment. To answer this, we attempted to align only S. pneumoniae, S. mitis, and S. oralis and observed little differences to the previous data, suggesting that the lack of recombination was not due to poor alignment. It has been reported previously that recombinations in pbp2b carry a fitness cost (Albarracin Orio et al. 2011), which could be explanation as to the lower frequency of events observed in our findings. In addition, we corroborate previous findings (Kalizang’oma et al. 2021; van der Linden et al. 2017) and expand on the finding that high variance in pbp2x with promiscuous recombination by showing that nearly all resistant strains showed recombination, in comparison less than 10% of the susceptible isolates.

Prolific recombination events were observed in the mitis group compared to other groups, which indicates the mitis group as particularly prone to transformation in vivo. As oral streptococci are genetically closely related to S. pneumoniae, we have previously shown that tight cell-to-cell contact in biofilms can facilitate genetic exchange of large DNA fragments, likely taking place in the upper airways (Cowley et al. 2018). S. mitis and S. oralis encompassed less than 5% of pneumococcal genomes and, regardless, they were overrepresented as the source of genetic material. The inclusion of new sequences and public access to a higher quantity of genomes is likely to allow for further research to be developed on the topic. However, it should be highlighted that detection of recombination events is highly dependent on the diversity of genomes included, coupled with the quality of alignment. For instance, in the findings presented here, the mutans group was excluded from recombination analyses due to high divergencies in sequence structure, particularly for pbp2b. As such, the data presented here shows bias towards the mitis group of streptococci. We have also explored recombination events in individual groups of streptococci (data not shown), i.e. anginosus group alone, and we detect more recombination events likely due to better alignment as a product of sequence similarity. While interspecies recombination was our focus due to the potential of transfer to S. pneumoniae, it should be noted that future research exploring recombination should be performed with diverse strategies to understand the dynamics of genetic exchange in microbes. Finally, the involvement of other genes in streptococcal antibiotic resistance such as the murMN operon has been reported to play a role (Smith and Klugman 2001) and other genes such as murE have also been implicated (Todorova et al. 2015).

With 50% of all oral streptococci genomes analyzed containing at least one complete ICE, these elements likely play a significant role in the transfer of ARGs in the oral microbiome. The distribution of such ICEs was uneven across species: members of the mutans group contained considerably fewer ICEs per genome, often less than one, whereas species in the anginosus group exhibited a much higher frequency, averaging 1.75 ICEs per genome, with very few genomes lacking any complete ICE or ICE-like elements. In line with our findings, a recent study revealed a diverse array of chromosomally integrated mobile genetic elements (cciMGEs) in selected oral streptococcal species (Lee et al. 2024). Moreover, their findings included the identification of novel relaxases, highlighting the dynamic nature of these elements in oral streptococci. Overall, ARGs responsible for tetracycline and macrolide resistance were the most observed association with high frequency found in tet(M) and erm(B) across most species, which goes in line with previous studies assessing presence of ICEs in bacteria from the oral cavity (Wang et al. 2024). As streptococci are the most abundant genus of the oral cavity, these findings can also correlate to the frequent identification of tetracycline and macrolide resistance in saliva samples (Brooks et al. 2022; Vanhout et al. 2024). Noteworthy were the challenges in identifying mef(A) and msr(D) resistance genes associated with elements via ICEscreen, which have previously been reported to be present in mobile genetic elements (Ambroset et al. 2015; Mingoia et al. 2014). Initially, we hypothesized the issue to be related to genome fragmentation, but upon inspection of the subset of completely assembled genomes, the issue persisted as only two were found to harbor an ICE associated with mef(A) and msr(D), both of which being Tn5252 elements. Employing homology searches with ORF3 and ORF4, which known conserved elements of the macrolide efflux genetic assembly (mega) element harboring mef(A) and msr(D) in streptococci, 36 and 35 genomes contained fragments of high identity (E≤0.001) out of 37 completely assembled genomes with mef(A) and msr(D). Across all genomes irrespective of assembly level, the same ORFs were found in 300 and 299 genomes out of 533 genomes with mef(A) and msr(D). The mega element is ∼5.5kb and lacks enzymes required for DNA transposition (Gay and Stephens 2001), as such being non-conjugative on its own. Further, in addition to its presence in composite ICEs as mentioned above, previous reports also indicated its possibility of transfer via natural transformation (Del Grosso et al. 2006; Schroeder and Stephens 2016). Identifying ICEs can be challenging as the backbones can be atypical, the gene order can be scrambled due to recombination, and there is divergence in the core genes. In addition, detection of concatenation across these genetic entities can be difficult in WGS at the contig level. We utilized a fairly strict approach to characterize these elements, similar to previous studies (Johansson et al. 2025; Lao et al. 2022); however, it should be noted that the results here are likely an underestimation of the total of ICEs present given the strictness of our approach. Finally, other MGE components such as integrative and mobilizable elements, phages, and integrons likely contribute significantly to antibiotic resistance transfer and should be further studied.

Despite increases in the last decade, the scarcity of genomes creates challenges for extrapolation and interpretation of the findings at a larger scale. We recently have been able to characterize over 75,000 genomic sequences of S. pneumoniae, employing this data to understand temporal and geographical patterns of AMR distribution (Shawrob et al. 2025). While there are over 10,000 curated pneumococcal sequences available at NCBI, for all oral streptococci together there are only 2,000. Surveilling and gathering data on the resistome of commensal species is important to the development of strategies to combat AMR (Anderson etal. 2023; Goytia and Wadsworth 2022; Hendriksen et al. 2019; Nji et al. 2021; Vihta et al. 2021). Altogether, the data presented in this study reinforce the clinical relevance of understanding gene flow between commensals and pathogens, especially in habitats such as upper respiratory tract, where close spatial interactions facilitate natural transformation, which reflects current evidence that interspecies recombination drives the evolution of β-lactam resistance in pneumococci (D’Aeth et al. 2021). In addition, as penicillin is often the first line of treatment for streptococcal infections - primarily from the pathogenic species S. pneumoniae, Streptococcus pyogenes, and Streptococcus agalactiae - understanding the dynamicity of pbp fragments is paramount for diagnostics and surveillance efforts. Nevertheless, it should be noted that while in silico prediction of resistance has been available for S. pneumoniae (Li et al. 2016), there are no current reliable methods in oral streptococci (Eriksen et al. 2023). Recently, we reported an average MDR score of 0.9 antibiotic resistant determinants per genome in S. pneumoniae (Shawrob et al. 2025) and here we show a slightly higher MDR score of 1.2 ARGs in oral streptococci. While the previous analysis did not include intrinsic genes, e.g. pmrA, patAB, and RlmA(II), it did include a prediction of pbp divergence suggesting resistance. Our study shows the high burden and accumulation of resistance genes/alleles in oral streptococci and their relevance to resistance in major human pathogens. It is important to note, however, that oral streptococcal genomes potentially present a bias risk of resistance overrepresentation compared to more frequently sampled species such as S. pneumoniae. Often grouped together as oral or the viridans group of streptococci, our findings call attention for the importance of investigating each group of species for a better understanding of their physiology and ability to act as a reservoir of antibiotic resistance.