Staphylococcus aureus is an important cause of human infections ranging from skin abscesses to life threatening conditions such as bacteremia and pneumonia. This pathogen is able to cause hospital as well as community acquired infections (Chen and Huang, 2014). These infections are often associated with high rates of antibiotic resistance representing a serious challenge for patient management. Of particular importance, methicillin resistant S. aureus (MRSA) has emerged as a notorious etiologic agent for a wide range of infections worldwide (Shittu et al., 2011).

The ability of S. aureus to cause multiple infections has been associated with the expression of myriads of different toxins, virulence factors, cell wall adhesion proteins such as MSCRAMMs (Microbial Surface Components Recognizing Adhesive Matrix Molecules) and other proteins involved in immune evasion (Shambat et al., 2012). MSCRAMMs are among the factors of interest, as they are known to have the capacity to bind extracellular matrix proteins such as collagen, fibrinogen and fibronectin, all potentially important for the ability of S. aureus invasion. In addition, different S. aureus may have different constellations of MSCRAMMs and so may be predisposed to cause certain types of infections (Gordon and Lowy, 2008). Toxins are another group of critical virulence factors, among them the Panton–Valentine Leukocidin (PVL) is of particular importance. PVL is a bi-component (lukS-PV and lukF-PV) pore-forming cytotoxin that has been shown to target polymorphonuclear cells, monocytes and macrophages in humans and rabbits (Boyle-Vavra and Daum, 2007) and has been associated with a highly aggressive and often fatal form of community acquired infections (Stryjewski and Chambers, 2008). Furthermore, capsular polysaccharides and regulators such as the accessory gene regulatory (agr) may also play a role on S. aureus pathogenicity. Capsular polysaccharide or capsule is a cell wall bacterial component which protects bacterium from phagocytic and enhances microbial virulence (Verdier et al., 2007) while the agr locus is a quorum sensing system, essential for the global regulation of S. aureus virulence factors and other accessory gene functions (Chong et al., 2013). Due to polymorphisms in the agr locus, the S. aureus isolates are assigned to agr groups I to IV (Rasmussen et al., 2013).

While S. aureus infections from developed countries have been extensively studied concerning their virulence patterns and clonal relatedness, the corresponding data from Africa is limited. Available data have shown differences on clonal structure and virulence patterns of African isolates (Schaumburg et al., 2011), often associated with high mortality compared to those from industrialized countries (Shittu et al., 2011).

In Manhiça, southern Mozambique, data from the ongoing surveillance of invasive bacterial infections showed that S. aureus is the leading cause of community-acquired neonatal bacteremia/sepsis, accounting for 39% (Sigaúque et al., 2009); however the molecular epidemiology and pathogenicity determinants remain unknown. Therefore, in the present study we aimed to determine genetic diversity and virulence factors of a collection of S. aureus causing community-acquired bacteremia in children less than 5 years admitted to Manhiça District Hospital, a rural area in southern Mozambique.

This is one of the few studies in the sub-Saharan Africa and probable globally assessing the molecular profile of virulence markers of S. aureus causing bacteremia in children. Our data suggest the presence of high diversity of S. aureus causing bacteremia in Mozambican children, with a limited number of potential lethal clones particularly in young children. Although the sample size was small, the high case fatality rate among children infected with CC45 and CC8 compared to other CCs, may suggest possible differences on the degree of pathogenicity among CCs. The high prevalence of CC8 and CC15 in our study is consistent with findings from the African German multi-centric study conducted from 2010 to 2013 in Africa and Germany on infection biology and epidemiology of S. aureus (http://www.african-german-staph.net) (Ruffing et al., 2017). Similarly, these CCs have been consistently reported in patients with bacteremia in Germany (Rieg et al., 2013), Sweden (Rasmussen et al., 2013) and from different infection sites in Nigeria, Northern Africa (Shittu et al., 2012). Indeed, CC8 is among the major CCs which includes most of MRSA causing both nosocomial and hospital infections worldwide (Argudín et al., 2011). However, our isolates are from community-acquired infection as blood cultures were collected upon admission and no further blood culture was collected during the hospital stay of the patients; in addition none of the patients had previous hospitalization in the preceding 3 months.

Although PVL has been primarily associated with skin and soft tissue infections such as furunculosis and skin abscesses (Chiu et al., 2012), the prevalence found here is a matter of concern as we recently described a case of CA-MSSA necrotizing pneumonia complicated with multifocal osteomyelitis, pericardial effusion and endocarditis in a 6-year-old boy PVL positive hospitalized in our community with poor outcome. This prevalence (30%) is higher compared to 2.4% in developed countries (Aamot et al., 2012) and up to 25% in other African countries (Kechrid et al., 2011; Orth et al., 2013) among patients with community-acquired bacteremia. In fact, the distribution of PVL may vary from different geographical regions (Correa-Jiménez et al., 2016) and it has been observed that African S. aureus carries the highest rates of PVL compared to developed countries, although the reason for this scenario still is a matter of debate (Schaumburg et al., 2011). Additionally, all identified PVL encoding genes were among MSSA isolates although this toxin has been shown to be present in both MRSA and MSSA strains (Muttaiyah et al., 2010). The fact that all MRSA isolates belonged to the USA500 epidemic clone (data not shown) may explain the lack of PVL in our MRSA cohort, as USA500 strains also lacks the PVL encoding genes and other mobile genetic elements contributing to virulence and transmissibility (Roberts, 2014).

The high prevalence of pyrogenic toxin super-antigens support the hypothesis that classical pyrogenic toxin genes are common in S. aureus and as many as 73% of S. aureus carry at least one of the genes encoding a classical pyrogenic toxin (Shukla et al., 2010). The high proportion of these toxins was also reported in similar studies of S. aureus blood isolates from German (Becker et al., 2003; Rieg et al., 2013), Sweden (Rasmussen et al., 2013) and Gabon (Central Africa) (Schaumburg et al., 2011). In contrast, the exfoliative toxins, etA and etD were much more frequent in our study compared to others (0–8%) (Becker et al., 2003; Rasmussen et al., 2013; Rieg et al., 2013).

Most of the isolates possessed the accessory gene regulator I (agrI) which supports the observation that agr group I strains comprise a significant majority of clinical isolates (Sakoulas et al., 2002). In addition, as expected all isolates carried either capsule type 5 or 8 (Riordan and Lee, 2004; Fischer et al., 2014) being cap8 the most frequent with some geographical variations (Fischer et al., 2014). The high prevalence of adhesion molecules in our strain collection may correlate with the fact that MSCRAMMs are crucial for establishment of S. aureus infections (Gordon and Lowy, 2008). Several MSCRAMMs shown to be important for the invasiveness are highly conserved into the staphylococcal genome (Rasmussen et al., 2013). The ClfA is the major fibrinogen binding protein of S. aureus and has been suggested to bind to the C-terminal region of fibrinogen γ-chain resulting in platelet aggregation or clumping of bacteria in plasma. The contribution of ClfA into S. aureus pathogenesis has been demonstrated in several animal models of infection including endocarditis, arthritis and sepsis while ClfB has been found to enhance the attachment of S. aureus to the anterior nares during colonization. Fibronectin-binding-proteins (FnBP) A and B enable S. aureus to adhere and invade a range of cell types including the epithelia, endothelia, fibroblasts and osteoblasts (Lacey et al., 2016).

Additionally, the high prevalence of proteases may help to explain the potential invasiveness of the studied isolates. Proteases are essential for cell invasion, destruction of host tissues and creation of metastasis to other sites (Gordon and Lowy, 2008). As expected, the hemolysin-α gene (hla) was present in almost all isolates, with some occasional exceptions probably due to mishybridization reactions while the gene encoding for δ-hemolysin was detected in all isolates. Similarly, hlb was also present at high frequency. Hemolysins are known to cause membrane damage of red blood cells (Wardenburg and Peptides, 2012).

Although S. aureus is recognized to be highly resistant to the most common used antibiotics, the rate of resistance to clindamycin in our study was very surprising. In contrast to erythromycin, which is commonly used, clindamycin is almost unavailable in the study area. Therefore, possible explanation for this scenario could be erythromycin-induced resistance. Indeed all clindamycin resistant isolates showed induction of phenotypic resistance which was confirmed by the presence of ermC gene coding for inducible resistance to macrolide-lincosamide-streptogramin antibiotics (Ii and Jorgensen, 2005; Prabhu et al., 2011). On the other hand, despite the fact that MRSA was detected in less than 10% of isolates, this finding is a matter of concern as MRSA are often associated with increased health care costs by either increasing the duration of patient hospitalization or need for often unavailable second line treatment (Rybak and LaPlante, 2005).

In summary, this is the first report of molecular characterization of S. aureus bacteremia in Mozambican children. These data provide a snapshot on the genetic diversity and pathogenicity markers of S. aureus causing pediatric bacteremia, which may explain its potential role as leading cause of neonatal bacteremia in our community. This emphasizes the urgent need of its recognition for prompt treatment. Further studies involving a large number of isolates are needed to explore in detail the potential impact of clonal complexes with patient outcome.

The strains characterized here were isolated from the ongoing invasive bacterial surveillance system which included several study protocols reviewed and approved by the Mozambican National Bioethics Committee for Health and institutional review boards of Hospital Clinic of Barcelona, Spain; the US Centers for Disease Control and Prevention and the University of Maryland School of Medicine.

IM, BS, PA have been contributing in design of the study. MH, UR, LV have been contributing in experiment design of microarray. DV, MG, SA have been contributing in study implementation, seeing patients (SA) and laboratory experiments. All authors have contributing in writing and revision of the manuscript.

This study has been supported by the grant of the Deutsche Forschungsgemeinschaft HE 1850/11-1 to MH and IM. CISM receives core funds from Spanish Agency for International Cooperation and Development (AECID).