Our results suggest that the most probable PTA-causing pathogens are Streptococcus pyogenes and two species of fusobacteria, Fusobacterium nucleatum and Fusobacterium necrophorum. Some other streptococci (including S. anginosus) and some other Gram negative anaerobic rods (Prevotella, Porphyromonas) may contribute to the infection as well. S. pyogenes and F. necrophorum tend to cause the infection rather individually than in combination while in case of other bacteria, polymicrobial infection is more likely. It is important to point that these bacteria are members of the normal nasopharyngeal microbiota [9]. Therefore, in a large number of cases, they may remain undetected in routine diagnostics or it may not be clear which of the isolated microbes is the cause of the infection, thus, it is important to address the methodological challenges [17] as well as pathogenic potential of the suspected bacteria. Our analysis was based on the results of NGS, the new golden standard of microbiome studies, and our results may be important in the context of optimizing the handling of the PTA patients.

Streptococcus pyogenes is an acknowledged causative agent of tonsillitis and also PTA. In previous studies, S. pyogenes has been isolated in about 20–40% of PTA patients [3, 10, 11, 18]. As the previous studies have applied cultures or PCR-based methods, the results were quite incomparable with our data – in case of S. pyogenes we showed an incidence of 66 and 70% as opposed to the previous highest finding of 42% [19]. However, in nearly one third of our samples we did not detect S. pyogenes, strongly suggesting the existence of other causative agents of PTA in a significant proportion of patients. According to the current diagnostics practices, these other possible agents, especially anaerobic bacteria may remain frequently undetected.

Fusobacterium necrophorum has been suggested as another possible causative agent of PTA [13, 20]. Studies applying NGS have detected F. necrophorum seldom in the healthy tonsils while both the prevalence and abundance of this species was significantly increased in case of tonsillitis [21]. In our study, we identified F. necrophorum in both sample types for over 75% of PTA patients with high abundancies in half of non-S. pyogenes cases, indicating a clear link between PTA development and F. necrophorum. Thus, our data are in line with other studies suggesting that F. necrophorum might be a substantial causative agent of PTA [3, 8, 20]. A synergistic effect of F. necrophorum and S. pyogenes in the formation of PTA is unlikely provided the observed negative correlation between the two species. Streptococcus pyogenes may have an inhibitory effect on F. necrophorum as described earlier [22]. Considering these complex relationships between the microbes, it is important to consider F. necrophorum as the causative agent of PTA in the absence of S. pyogenes and vice versa.

Fusobacterium nucleatum ranked third in terms of mean relative abundance in both sample types and was the third most frequently predominating species in individual samples as well. It predominated in 15% of the communities, suggesting that it could belong to the main causative agents in those cases. This species has not previously been considered a potential cause of PTA, except in a patient with Lemierre syndrome [23]. At the same time, it has been isolated from PTA patients in several studies [24,25,26], including in our previous culture-based study [13]. In addition, significant increase in antibodies titre against F. nucleatum has been found previously in PTA patients [27]. Our results showed a significant increase in the abundance of F. nucleatum in the absence of F. necrophorum and a negative correlation between the two. Again, an antagonistic relationship is a plausible possibilty.

In addition to the above-mentioned three bacteria, or data suggest that some PTA cases may be associated with other opportunistic bacteria like some streptococcal species (S. anginosus, S. mitis) or other Gram negative anaerobic rods (Porphyromonas, Prevotella). Although S. pyogenes and F. necrophorum tended to cause the infection rather individually than in combination, this disease might be polymicrobial in case of other microbial combinations [10]. To clarify the clinical importance of other bacteria, additional studies are needed.

Correlations between different bacteria and clinical findings were mild in our study. We found S. pyogenes to be associated with increased duration of symptoms before hospitalization possibly indicating a slower development of severe symptoms as compared to fusobacteria.

The sampling sites for microbiological analyses need to be chosen wisely [17]. In our study we compared biopsy material from tonsils and pus. The tonsils displayed significantly higher levels of less pathogenic and environmental bacteria in terms of both incidence and relative abundances, such as Chryseobacterium hominis and Sphingomonas faeni, and also the poorly defined Allorhizobium-Neorhizobium-Pararhizobium-Rhizobium group. Although C. hominis is considered to be of human origin [28] and S. faeni an environmental microbe [29], a positive correlation was shown between them. Their role in the disease should be evaluated with caution, since contamination cannot be ruled out. Additionally, tonsils form an important part of localised immune protection and they are exposed to various microbes both inhaled and ingested [14]. This aspect may also explain the higher content of S. faeni, C. hominis and Allorhizobium group in the tonsil tissue as compared to pus. Specimen collection and handling took place in the same environment for pus and biopsies. As concerns putative causative agents of PTA, there was no significant difference between their prevalence in pus and tonsils. In clinical practice, pus has several advantages over tonsil biopsies: it is easier both to collect and to manage in the laboratory and it was found to be less contaminated by non-pathogenic bacteria complication interpretation of results. Therefore, we recommended aspiration of pus for routine diagnostic bacterial diagnostics. However, it is important to note that in case of cultures, microbes in pus may not always be viable.

Treatment approaches for PTA patients vary remarkably, depending on physicians’ preference and also regional aspects considering antibiotic susceptibility. The first choice of antibiotic for PTA is frequently penicillin alone or in combination with metronidazole. Other common antibiotic treatment options include amoxicillin in combination with metronidazole, cephalosporins or clindamycin [11, 30]. Penicillin is effective against S. pyogenes and in most cases targets fusobacteria as well. Recent evidence reports a continuous susceptible state and even a decrease in MIC values of common antibiotics used for fusobacterial infections, such as penicillin, clindamycin and others [31, 32]. Thus, penicillin remains an effective treatment option in majority of PTA cases. However, the possibility of other anaerobes cannot be ruled out, therefore, the addition of specific anaerobe coverage with metronidazole or clindamycin in poor response to penicillin alone should be considered. In addition, the polyols erythritol and xylitol inhibit growth of S. pyogenes and thus may have potential in preventing PTA in patients with frequent episodes of tonsillitis [33].

Limitations. Our study cohort was predominated by male patients. However, some studies have identified male gender as a risk factor of PTA [34], which could help to explain this phenomenon. Another limitation was antibiotics usage prior hospitalization that might influence microbial community composition by reducing the microbial diversity [17]. Antibiotics are often useful in order to treat the tonsillitis and to prevent several complications [3]. At the same time, culture independent methods are less prone to bias caused by antimicrobial treatment in particular in sample types with slow clearance of microbial DNA like undrained abscesses. To our knowledge, this is the first report using NGS, the new golden standard of microbiome studies, for identifying the microbial communities in tonsils and pus of PTA patients. Though being useful in revealing the wide spectrum of causative agents, it does not allow to determine their antibiotic susceptibility if needed.

In conclusion, the most probable causative agents of PTA include Streptococcus pyogenes, Fusobacterium necrophorum and Fusobacterium nucleatum. Some other streptococci (e.g. S. anginosus) and anaerobes (e.g. Prevotella and Porphyromonas) might have role in PTA development, too. Hence, it is important to consider anaerobic bacteria both in the laboratory work and the treatment of PTA. Pus from the peritonsillar abscess appears to be a more representative specimen for microbiological examination than the tonsillar tissue. Our results are important in the context of optimizing the handling of the PTA patients.