Several studies yielding conflicting results have attempted to compare the clonal composition of GAS populations causing invasive and non-invasive infections in order to identify particularly virulent clones or properties that may be used as epidemiological markers of invasiveness [7, 8, 11, 12, 16]. However, many of those studies were limited in the size and diversity of the GAS collections studied or in the typing methodologies used, with most of them relying essentially on emm typing, which has been shown to be insufficient for the complete identification of GAS clones . In this work, we used several different typing methods to compare a collection of genetically diverse GAS isolates recovered from normally sterile sites during a period of six years in Portugal  with isolates recovered from pharyngeal exudates of patients presenting with tonsillo-pharyngitis, during the same time period and in the same geographical region. The nasopharyngeal mucosa has been suggested to be the main reservoir for GAS isolates associated with invasive infections [19, 20].
The differences among the GAS clones associated with invasive infections and pharyngitis were reflected in antimicrobial resistance, with the invasive group of isolates presenting lower macrolide resistance and higher tetracycline resistance, when compared with the pharyngitis group. Among isolates belonging to the same emm types, namely emm1 and emm4, only the macrolide-susceptible clones were associated with either invasive infections or pharyngitis. The macrolide-resistant clones of these emm types are reflected in invasive infections according to their prevalence in pharyngitis, suggesting that these are translating more the antibiotic selective pressure than the invasive capacity of the clones. Tetracycline is not currently used in the treatment of GAS infections but resistance to this antimicrobial in S. pyogenes is usually acquired by horizontal transfer, since the resistance genes are frequently encoded in mobile genetic elements with a wide host range . These elements often carry macrolide resistance genes as well, and in S. pyogenes a significant association between the presence of the genes erm(B) and tet(M) has been reported and it has been suggested that tetracycline use could contribute to the selection of macrolide-resistant GAS isolates [21, 22]. In our study, the association between the presence of the genes erm(B) and tet(M) was observed globally, but not among the invasive isolates, suggesting that the genetic elements carrying tetracycline resistance conferring genes may be different between the two bacterial populations.
Bacitracin susceptibility is routinely used for the presumptive identification of GAS, although resistant clones have been identified in several countries [23–25]. In our GAS collection, all the bacitracin-resistant isolates (5%), regardless of the type of infection, were clustered in the same PFGE clone (H26) and belonged to ST52, although one was emm22-T12 while the others were all emm28-T28. Isolates with such characteristics had been previously reported in Portugal associated with tonsillo-pharyngitis, skin infections and asymptomatic carriage [26, 27], but not with invasive infections. Bacitracin resistance among invasive isolates has been previously reported only among isolates recovered in France and in San Francisco [24, 25].
Although 74% of the invasive isolates in our collection belonged to clones which were equally frequent among pharyngitis, suggesting that a significant part of the invasive GAS population mirrors the clonal structure of the circulating GAS isolates, the remaining isolates represented clones that had an enhanced capacity to cause invasive disease. We also found significant associations between individual properties or pairwise combinations of properties and disease presentation. Since in most cases these were also characteristics of the more invasive clones, we cannot exclude that the associations of individual properties or pairwise combinations of properties can reflect, at least partially, the distribution of genetic lineages in the two GAS populations analyzed.
Individually, emm types 1 and 64 were associated with invasive infections. Isolates belonging to these emm types presented the only two SAg profiles significantly associated with invasive infections. Surprisingly, only one of these SAg profiles includes a phage-encoded SAg gene (speA). In agreement with our observation, a previous study found that within the same PFGE-emm group, the SAg profiles significantly associated with invasive infections had a smaller number of SAg genes than the dominant profiles in pharyngitis . These results suggest that although some SAg genes may significantly contribute to the virulence of S. pyogenes, the rise and success of highly virulent GAS clones may not hinge upon the acquisition of phage-encoded SAg genes. Still, in our study, the SAg genes speA and speJ were both significantly more prevalent among invasive isolates. This association was not substantially affected by emm type, PFGE clone, nor by the presence of other SAg genes, suggesting that speA and speJ can be regarded by themselves as markers for invasiveness. Although such association has not been previously noted for speJ, the speA gene has been frequently associated with invasive infections [6, 8, 16] and the production of SpeA by GAS isolates has been linked to streptococcal toxic shock syndrome .
On the other hand, we identified an association of pharyngitis isolates with emm types 4 and 75, and with the SAg genes speC, ssa, and speL/M. The association of speC with non-invasive infections has been previously reported [6, 16, 30], but in our collection this association could be explained simply by a high frequency of co-occurrence of this gene with ssa which was strongly associated with pharyngitis, as was also noted in a recent study . The presence of the genes speL and speM was not previously associated with non-invasive infections.
Since there is a strong correlation of the SAg profile with emm type and of both these properties with PFGE type, some of these individual factors frequently co-occurred in the same clones. Therefore, combinations of these characteristics were also significantly associated with disease presentation. However, we could not detect any synergistic or antagonistic interactions between most of these characteristics, meaning that their co-occurrence in a particular isolate does not make it more invasive than isolates sharing only one of these characteristics.
Two PFGE clusters were significantly more prevalent among isolates associated with invasive disease than among those causing tonsillo-pharyngitis. One of these was a cluster of macrolide-susceptible isolates characterized as emm1-T1-ST28 and by the presence of the SAg genes speA, speG, speJ, and smeZ (B49), which accounted for 18% of the invasive isolates. M1T1 isolates have been frequently associated with severe invasive GAS disease, and the acquisition of prophage-encoded virulence genes, as well as horizontal gene transfer events by homologous recombination were implicated in the increased virulence of these isolates [31, 32]. However, many epidemiological studies found a similar prevalence of this clone among invasive and non-invasive isolates [10–12], questioning its enhanced invasive capacity. In contrast to these findings, but similarly to those of others [6–9], we found an association between this clone and invasive GAS disease in Portugal, although it can also frequently cause milder infections such as pharyngitis (it accounted for 6% of the pharyngitis isolates analyzed in this study).
The other cluster significantly associated with invasive infections in Portugal was J16, which was dominated by isolates belonging to emm64-ST164 and carrying the SAg genes speG and smeZ. A clone with these characteristics has not been previously associated with invasive disease and emm64 has been infrequently reported among invasive GAS isolates [4, 33, 34]. The higher invasive capacity of this clone cannot be attributed to its SAg repertoire, since these isolates do not harbor any of the SAg genes associated with invasive infection. Other, still unidentified, characteristics may be responsible for the properties of this clone.
In contrast to these PFGE clones, the F29 clone of macrolide-susceptible isolates characterized by emm4-T4-ST39 and harboring the genes speC, ssa and smeZ was associated with pharyngitis, suggesting that this clone may have a reduced ability to cause invasive disease, in agreement with the negative association between emm4 and invasive infection that has been suggested elsewhere . The association of emm75 with pharyngitis has not been previously reported and was not translated into particular PFGE clusters due to the high diversity of emm75 isolates.
Our data confirms that the widely dispersed M1T1 clone has enhanced invasiveness but we also identified clones with increased or decreased invasive capacity that may have emerged locally and that have a more limited temporal and geographical spread. The emm alleles and the SAg genes characteristic of these clones were associated with particular disease presentations. Other individual emm alleles and SAg genes were also associated with a higher propensity to cause invasive infections or pharyngitis indicating the importance of these characteristics in determining an isolate’s invasive capacity.
Other factors that were not evaluated in this study may contribute to a different distribution of GAS clones in less severe and more severe infections. These include bacterial factors, such as the occurrence of mutations in transcriptional regulators controlling the expression of virulence factors, which seems to play an important role in the pathogenesis of some GAS isolates . For other clones, the ability to cause invasive infections may be more dependent on exploiting host factors, like the HLA class II haplotype , which may vary in frequency in different human populations.