Molecular Epidemiology of Staphylococcus aureus Bacteremia: Association of Molecular Factors With the Source of Infection

Staphylococcus aureus bacteremia (SAB) is associated with high morbidity and mortality, which varies depending on the source of infection. Nevertheless, the global molecular epidemiology of SAB and its possible association with specific virulence factors remains unclear. Using DNA microarrays, a total of 833 S. aureus strains (785 SAB and 48 colonizing strains) collected in Spain over a period of 15 years (2002–2017) were characterized to determine clonal complex (CC), agr type and repertoire of resistance and virulence genes in order to provide an epidemiological overview of CCs causing bloodstream infection, and to analyze possible associations between virulence genes and the most common sources of bacteremia. The results were also analyzed by acquisition (healthcare-associated [HA] and community-acquired [CA]), methicillin-resistant (MRSA) and methicillin-susceptible (MSSA) strains, and patient age (adults vs. children). Our results revealed high clonal diversity among SAB strains with up to 28 different CCs. The most prevalent CCs were CC5 (30.8%), CC30 (20.3%), CC45 (8.3%), CC8 (8.4%), CC15 (7.5%), and CC22 (5.9%), which together accounted for 80% of all cases. A higher proportion of CC5 was found among HA strains than CA strains (35.6 vs. 20.2%, p < 0.001). CC5 was associated with methicillin resistance (14.7 vs. 79.4%, p < 0.001), whereas CC30, CC45, and CC15 were correlated with MSSA strains (p < 0.001). Pathogen-related molecular markers significantly associated with a specific source of bacteremia included the presence of sea, undisrupted hlb and isaB genes with catheter-related bacteremia; sed, splE, and fib genes with endocarditis; undisrupted hlb with skin and soft tissue infections; and finally, CC5, msrA resistance gene and hla gene with osteoarticular source. Our study suggests an association between S. aureus genotype and place of acquisition, methicillin resistance and sources of bloodstream infection, and provides a valuable starting point for further research insights into intrinsic pathogenic mechanisms involved in the development of SAB.


INTRODUCTION
Staphylococcus aureus is an opportunistic pathogen that can potentially cause a wide range of infections. It is a leading cause of bacteremia and represents a significant global health problem (Weiner et al., 2016). S. aureus bacteremia (SAB) is often associated with severe metastatic infections, such as infective endocarditis, septic arthritis and osteomyelitis and complications, such as sepsis and septic shock, which lead to adverse outcomes that are challenging to manage (Shorr et al., 2006;Wyllie, 2006).
The incidence of SAB is difficult to determine and there are major geographical differences that reflect discrepancies in health care systems and infection control practices. In developed countries, the estimated incidence 80-190 cases per 100,000 inhabitants per year (Laupland, 2013;Le Moing et al., 2015). Despite the improvements in SAB management, including greater understanding of this infection and mandatory surveillance implemented in several countries over recent decades, SAB still causes significant morbidity and mortality, with an associated early mortality that appears to have plateaued at approximately 20-30% (van Hal et al., 2012). Certainly, little is known about global SAB epidemiology in terms of the circulating clones causing SAB in different patient subgroups, such as adults and children, or those most commonly found in the community or hospital settings. Because it is becoming progressively more difficult to differentiate between healthcareassociated and community-acquired infections due to changes in the complexity of present health care systems, it is important to identify the specific clones that are traditionally associated with the community but may be entering hospitals and replacing common nosocomial clones, and vice versa. Moreover, it would be especially interesting to study clonality taking into account that bacterial phenotype and genotype have been shown to have a possible influence on infection outcome, since different clones can adopt different strategies to overcome host responses and cause severe pathology (Recker et al., 2017). The overall mortality rate from SAB varies depending on the primary focus of infection (the highest rates occur in patients with infective endocarditis and pulmonary infections, and the lowest in patients with catheter-related infections) and on the complications deriving from SAB. This association makes it necessary to regard SAB not as a single entity, but as a heterogeneous group of infections that can evolve differently and therefore require source-specific management (van Hal et al., 2012). However, the characteristics of the most common clones causing SAB according to source of infection remain unknown. Furthermore, determining the role of particular genetic backgrounds (clonality and virulence) in bloodstream infections caused by S. aureus has become a real challenge due to the diversity, redundancy and host specificity of the virulence factors.
The aim of the present study was to explore the molecular characteristics of S. aureus strains causing bacteremia in order to provide an epidemiological overview of the circulating clones causing bloodstream infection and to analyze the possible association between virulence and the most common sources of bacteremia.

Data Collection
A total of 785 strains causing bacteremia with different source and 48 colonization strains collected over a period of 15 years (2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017) were analyzed. These strains were obtained from different sources in hospitals geographically distant from each other spread across the territory of Spain (Table 1). Specifically, these strains were identified in 10 different collections: six were single-center studies developed at the Hospital 12 de Octubre in Madrid, and the remaining four corresponded to multi-center studies developed at various Spanish hospitals (Muñoz-Gallego et al., 2017;San-Juan et al., 2017;Fernández-Hidalgo et al., 2018). The main focus and objective of the studies for which these strains were collected was source of staphylococcal bacteremia, mainly endocarditis (N = 214), catheter-related bacteremia (CRB) (N = 212), skin and soft tissue infections (SSTI) (N = 66), and bone and joint infections (N = 100). Eight of these collections corresponded to SAB infections in adults, and two in children (<15 years of age). The percentage of MRSA strains included in each collection varied. The studies were approved by the ethics committee of the University Hospital 12 de Octubre (Madrid, Spain). It was not considered necessary to obtain written informed consent because the participants were anonymized (IRH-ANT-2013-01).
Cases were classified according to acquisition: healthcareassociated (HA) or community-acquired (CA). HA included both nosocomial cases with a positive blood culture obtained from patients who had been hospitalized for 48 h or longer (Garner et al., 1988) and healthcare-associated cases following Friedman et al.'s criteria (Friedman et al., 2002). CA cases were those with a positive blood culture obtained at the time of hospital admission or within 48 h after hospital admission.
Methicillin resistance was defined on the basis of results of microdilution techniques, cefoxitin susceptibility testing and/or the presence of the mecA gene.

Molecular Studies
Blood cultures were processed with an automated blood culture system (BACTEC 9240, Becton Dickinson Microbiological System, USA). Automatic microdilution techniques were used for identification and susceptibility testing of isolates. Bacterial DNA was extracted using commercial extraction kits (Qiagen, Germany) according to the manufacturer's recommendations. DNA microarrays (Alere, Germany; Monecke et al., 2008) covering 334 target sequences and approximately 187 different genes that included species-specific markers, antimicrobial resistance genes, exotoxins, genes encoding microbial surface components recognizing adhesive matrix molecules (MCSCRAMMs), capsule genes, clonal complexes (CC) and agr group typing markers were run on the whole collection of strains. Those cases with ambiguous array results were considered as missing values for further analysis.
Only genes found with a frequency of between 5 and 95% in the whole collection were considered for statistical analysis.

Statistical Analysis
Categorical variables were compared using the chi-squared or Fisher's exact test, as appropriate. Significance levels of DNA microarray results were corrected using the Bonferroni correction for multiple tests. Pairwise comparisons of the main CCs, agr types and virulence genes were performed with source of bacteremia. Potential associations were investigated by univariate and multivariate logistic regression, in which CCs, agr types and virulence factors were considered as independent dichotomous variables, and source of bacteremia as the dependent variable. For multivariate analysis, variables with a p-value <0.1 in the univariate analysis were included in a backward stepwise algorithm. All statistical tests were two-tailed and a p-value of <0.05 was considered statistically significant. Analyses were performed using the SPSS statistical package, version 21.0 (SPSS Inc., Chicago, IL).

Distribution of S. aureus Strains by CC and agr Type According to Acquisition, Methicillin Resistance and Age of Population
Healthcare-Associated vs. Community-Acquired In our collection, there was a higher proportion of HA strains (68.4%) compared to CA strains, which accounted for 31.6%. Healthcare-associated strains were assigned to 26 different CCs, with CC5 (35.6%) being the most common, followed by CC30 (18.6%). The CC diversity was slightly lower among CA strains, which were assigned to 23 CCs, with CC30 (23.4%) and CC5 (20.2%) being the ones most commonly found ( Table 2). Although most clones circulated in the healthcare and community settings, a higher proportion of CC5 was found among HA than among CA strains (35.6 vs. 20.2%, p < 0.001; Table 2).
The distribution of agr types is presented in Table 3. Note that while agrI was the main agr type among CA strains (31.0% vs. 39.2%, p: 0.030), agrII was associated with HA acquisition, (46.6 vs. 31.8%, p < 0.001).

Adult vs. Child Population
While 28 CCs detected in this study were represented among strains isolated from adults, lower CC diversity was detected in children, with only 19 CCs identified. A comparison of clonality in the adult and children populations revealed that while CC5 was associated with strains from adults (34.2 vs. 20.0%, p < 0.001), CC30 was significantly related to strains from the child population (18.2 vs. 27.2%, p < 0.008; Table 2).
The distribution of agr types by patient age is shown in Table 3. This analysis showed a significant association between agrII and strains from adults (44.7 vs. 32.7%, p < 0.001), while agrIII was associated with the child population (19.9 vs. 29.1%, p < 0.008).

Clonal Complex Diversity, agr Type and Virulence Genes Among S. aureus Strains From Different Sources of Bacteremia
The main objective was to explore the distribution of CCs and virulence genes according to source of bacteremia. A collection of S. aureus strains from healthy carriers was also added to the analysis in order to evaluate potential differences between colonizing and bacteremic strains.
Remarkably, CC5 and agrII predominated in SAB from osteoarticular infections (Tables 4A,B). When we focused on antibiotic resistance genes, significant differences were identified in strains from different sources of infection for the mecA, msrA, aadD, aphA3, and sat genes (Tables 4A,B). Our results seemed to indicate a higher proportion of these resistance genes among osteoarticular infections compared with other bacteremia sources. In general, significant differences for source of bacteremia and colonization were also detected in virulence genes, such as sea, sed, hla, undisrupted hlb, splE, cna, fib, and isaB among others.
Various regression models were performed in order to measure the role of pathogen-related molecular markers (CC, agr type and virulence genes) adjusted for different sources of bacteremia and colonization ( Table 5). All adjusted models in multivariate analysis showed that these variables were the presence of agrIV type and sed gene for colonizing strains; the presence of sea, undisrupted hlb and isaB genes for CRB; sed, splE and fib genes for an endocarditis source; undisrupted hlb for the SSTI group; and finally, CC5, msrA resistance gene and hla gene with respect to bacteremia from an osteoarticular source (Table 5).

DISCUSSION
The present study describes and gives a global epidemiological overview of the molecular epidemiology of a large collection of S. aureus strains focused on bloodstream infections in Spain over a 15-years period. In this scenario, our study provides important findings regarding the distribution of clonality and virulence genes and their association with specific sources of SAB.
Our results revealed high clonal diversity among SAB strains, although the most prevalent CCs were CC5, CC30, CC45, CC8, CC15, and CC22, which together represented 80% of all cases. Additionally, substantial differences were found between strains causing MRSA and MSSA bacteremia, which indicated that MSSA strains were much more genetically diverse than their MRSA counterparts, which is consistent with studies developed in Europe (Aamot et al., 2012;Grundmann et al., 2014) and the USA (Miko et al., 2013;Park et al., 2017). The most common clone among MSSA strains was CC30, followed by CC45 and CC15, whereas among MRSA strains, there was a significant representation of CC5 in more than 75% of strains. Similar results have been reported in Latin America (Arias et al., 2017) and Germany (Schaumburg et al., 2012), where the CC5-MRSA clone was the most prevalent in the setting of bloodstream infections. Furthermore, in our collection, CC5 was found to be significantly associated with HA acquisition and the adult population, a finding which lends support to the interest aimed at investigating the pathogenic and molecular characteristics of CC5 and those factors that enhance its spread.
Several studies have suggested that while the agrI type is the most common one among clinical isolates (van Leeuwen et al., 2000;Moore and Lindsay, 2001), others (Sakoulas et al., 2003) have determined that more than half of clinical MRSA bloodstream isolates belong to agr group II. In our collection, agrII was also associated with MRSA, which may explain the higher percentage of agrII found in the nosocomial setting and among adults, in whom the prevalence of MRSA was higher. By contrast, agrI was related to MSSA strains and CA acquisition. Interestingly, a statistically significant association was also found between agrII and agrIII and adult and child populations, respectively. These associations are probably due to the correlation between agr type and CC, since CC5 (agrII) was the majority clone among adults and CC30 (agrIII) among children.
To date, different studies have explored the association between bacterial genotype, especially S. aureus virulence genes, and various clinical syndromes (Gillet et al., 2002;Jarraud et al.,   2002; Peacock et al., 2002). This study focuses specifically on bacteremia. Our collection included S. aureus strains from the most common primary clinical sources of infection: CRB, SSTI, osteoarticular infection and endocarditis, as well as nasal carriage strains. We found no major differences between colonizing and bacteremia-producing strains of S. aureus, which supports the fact that most strains of S. aureus are capable of causing bacteremia. Nevertheless, and in accordance with other studies (Fowler et al., 2007;Giulieri et al., 2016), we identified specific clonal backgrounds and various molecular markers that have been associated with bloodstream infections and certain sources of bacteremia in particular. In this regard, our findings showed that CC5 in addition to hla and msrA genes were more frequently present in strains causing osteoarticular bacteremia. The association of the hla gene, present in most S. aureus strains, with different types of infection has already been reported (Stulik et al., 2014;Sharma-Kuinkel et al., 2015). Further studies are needed to elucidate the role of this important virulence factor in the pathogenesis of bacteremia. With respect to the adhesin genes (MSCRAMMs), which play an essential role in the pathogenesis of intravascular, osteoarticular and device-associated S. aureus infections (Foster et al., 2014), our study revealed an association between the fib and isaB genes and endocarditis and CRB sources, respectively. Other adhesins like clfA/B, fnbA/B, and cna and their linkage with bacteremia, endocarditis and CRB, have also been reported (Giulieri et al., 2016;San-Juan et al., 2017). Another finding of note in this study was the presence of the undisrupted β-hemolysin (undisrupted hlb) which was significantly related to sources, such as CRB and SSTI. Different studies have demonstrated its contribution to SSTI (Hedström and Malmqvist, 1982;Lebughe et al., 2017) and biofilm-related infections (Salgado-Pabón et al., 2014). Although β-toxin is encoded in S. aureus, most strains are reported not to secrete β-toxin because the bacteriophage (φSa3) inserts into the hlb gene (Winkler et al., 1965;Coleman et al., 1991), inactivating it in the majority of S. aureus strains recovered from humans. Moreover, the φSa3 bacteriophage encodes the immune evasion cluster (IEC) sak-chip-scn (Coleman et al., 1989;de Haas et al., 2004). Coinciding with other studies (Pantucek et al., 2004;Van Wamel et al., 2006), these genes were relatively abundant in our collection, ranging between 73% (sak) and 87% (scn). Interestingly, the absence of the intact hlb gene (or which amounts to the same thing, the presence of hlb truncated by the IEC-carrying φSa3 phage) was significantly associated with an osteoarticular source. This intriguing association should be investigated further since other studies have reported the association between these phageintegrated genes and less severe staphylococcal infections (Jin et al., 2003).
This study presents several limitations that should be mentioned. First, the heterogeneity and non-continuity of the SAB collection (geographical origin, time points and hosts) precluded us from adjusting for these variables in multivariate analysis. Moreover, the proportion of colonization strains was small in comparison with the number of SAB strains. The results therefore should be interpreted with caution. At the same time, our study includes a large number of S. aureus strains causing bacteremia, with relevant information on place of acquisition, methicillin resistance and source of infection. Second, the lack of clinical data regarding the outcome of the bacteremic episodes makes it impossible to make inferences about the prognostic importance of the molecular factors. Other studies evaluating associations between bacterial genotype and virulence have led to conflicting results (Day et al., 2001(Day et al., , 2002Feil et al., 2003;Melles et al., 2004), due in part to the heterogeneous nature of the S. aureus infections included, as well as the absence of a large, well-characterized collection of isolates. Third, our study methodology was based on DNA microarrays, which should be noted in the case of the hla gene. Despite the fact that hla is present in virtually all S. aureus strains, some studies, such as Sharma-kuinkel et al. (JCM) have reported up to 12 different variants of hla. In our study, the hla gene was detected in 92.2% of strains. We think that the low frequency of this gene observed in our collection may have been due to the DNA microarray technology, which may underestimate the presence of certain minority hla variants due to lack of sensitivity. Whole genome sequencing may be a more effective genotypic characterization approach for detecting different genetic variants that may not be detected by hybridization procedures, although previous studies have shown good agreement between the genotypic results obtained using a DNA array-based methodology and those using high-throughput sequencing (Strauß et al., 2016). Finally, we did not perform gene expression studies, which would be key to determining whether a particular gene or set of genes was responsible for the specific pathogenic behavior observed in SAB from particular clinical sources. Nevertheless, our findings offer a valuable starting point for further research insights into intrinsic pathogenic mechanisms involved in the development of SAB.
In conclusion, the current study suggests a potential association between S. aureus genotype and acquisition, Various multivariate models were explored that included different numbers of variables according to the number of events by bacteremia source.
-: variables included in the initial model of multivariate analysis then discarded in a backward stepwise process. Only variables consistently retained in exploratory models are shown. CRB, catheter-related bacteremia; aOR, adjusted Odds Ratio; 95% CI, 95% confidence intervals; α undisrupted hlb.
methicillin resistance and bloodstream infection sources. The results of this study reinforce the view that SAB continues to represent a major clinical challenge. Thus, a better understanding of S. aureus epidemiology and pathogenesis is crucial to the detection of prognostic biomarkers as well as to the development of potential therapeutic targets aimed at improving patient outcomes.