As a commensal microorganism, S. epidermidis retains the ability to specifically adhere to host proteins in the skin. In a surgical wound, the bacterium utilizes these adhesion mechanisms in order to adhere to the deeper tissues and to the implanted device, or more specifically, the conditioning layer of host proteins deposited upon the device. Initial adhesion of bacteria to implant surfaces is mediated by non-specific interactions such as hydrophobic interactions (Gristina, 1987), and then as shown schematically in Figure 1, by specific adhesins such as autolysin (AtlE) (Heilmann et al., 1997), extracellular DNA (eDNA) (Qin et al., 2007; Izano et al., 2008), and staphylococcal surface protein 1 and 2 (SSP-1, SSP-2) (Veenstra et al., 1996). AtlE, SSP-1, and SSP-2 have been primarily associated with adhesion to native surfaces (Veenstra et al., 1996; Heilmann et al., 1997), whilst eDNA is generated in S. epidermidis through an AtlE-mediated lysis of a subpopulation of the bacteria, promoting biofilm formation within the remaining population (Qin et al., 2007). In the context of medical devices, the surface of the device becomes coated with host-derived plasma proteins, extracellular matrix (ECM) proteins and coagulation products (platelets and thrombin) immediately following implantation (Baier et al., 1984). Cell-wall-anchored (CWA) proteins/adhesins, such as the microbial surface components recognizing adhesive matrix molecules (MSCRAMMs) (Foster and Hook, 1998) bind bacteria like S. epidermidis directly to these molecules (Figure 2). In S. epidermidis, adhesins for fibrinogen [as serine-aspartate repeat protein G (SdrG/Fbe) (Hartford et al., 2001; Brennan et al., 2009)], fibronectin [extracellular matrix-binding protein (Embp) (Arciola et al., 2003)], collagen [SdrF/GehD (Bowden et al., 2002; Arrecubieta et al., 2007)], vitronectin [AtlE or autolysin/adhesin (Aae) (Heilmann et al., 2003)] and elastin [elastin-binding protein (EbpS)] have all been identified. Peptidoglycan-bound wall teichoic acids (WTA) are an essential part of the S. epidermidis cell wall and also play an important role in bacterial adhesion. WTA enhances the initial adhesion of S. epidermidis to medical devices by binding to adsorbed fibronectin (Hussain et al., 2001) and fibrin clots (Chugh et al., 1990).

The ability to adhere to a surface represents the first step in biofilm formation, commonly believed to be the most important virulence factor possessed by S. epidermidis (Figure 1). Biofilm development facilitates resistance against host defense mechanisms (Myrvik et al., 1989; Kristian et al., 2008; Cerca et al., 2011; Schommer et al., 2011) and confers antibiotic resistance (Cerca et al., 2006; Mack et al., 2006). Biofilm formation also complicates medical and surgical treatment protocols because implant removal is often required to remove the biofilm.

Biofilms are defined as complex communities of adherent bacteria encased in a matrix of self-produced extracellular polymeric substances (EPS) (Costerton et al., 1995) (Figure 1). The accumulation and maturation of the S. epidermidis biofilm occurs via a number of mechanisms. Polysaccharide intercellular adhesin [PIA, or poly-N-acetyl-glucosamine (PNAG)], synthesized by icaADBC encoded proteins (Heilmann et al., 1996; Mack et al., 1996a) is responsible for biofilm formation in the majority of S. epidermidis isolates (Mack et al., 1996b) and was believed to be the most common molecule associated with biofilm formation (Heilmann et al., 1996; Mack et al., 1996a; Figure 2). This was endorsed by the observation that the ica operon was absent in most commensal S. epidermidis strains (Zhang et al., 2003; Chokr et al., 2007). However, not all S. epidermidis have the icaADBC genes (Heilmann et al., 1996; Harris et al., 2016) and these isolates mediate biofilm formation by proteinaceous factors, such as the accumulation associated protein (Aap) (Rohde et al., 2005) that contributes to biofilm formation upon cleavage by extracellular or host proteases (Figure 2). The aap gene has been observed in both pathogenic and commensal isolates, more frequently than the ica operon (Gill et al., 2005; Los et al., 2010; Harris et al., 2016). Other PIA-independent mechanisms include biofilm associated homolog protein (Bhp) (Bowden et al., 2005; Tormo et al., 2005), Embp (Williams et al., 2002; Christner et al., 2010), and S. epidermidis surface protein (Ses)C (Shahrooei et al., 2009), and SesE (Harris et al., 2016). Interestingly, Rohde et al. suggested that PIA-dependent biofilms are more robust than those formed by proteinaceous factors (Rohde et al., 2007), and another study found they result in a different morphotype or biofilm substructure (Harris et al., 2016). WTA have also been linked with S. epidermidis biofilm formation. TagO encodes the first enzymatic step in WTA biosynthesis and a tagO mutant has been shown to have a biofilm negative phenotype. This is partly attributed to an increase in cell surface hydrophobicity, impairing its initial adhesion to the surface, and a decreased production of PIA by activating the icaADBC repressor, icaR (Holland et al., 2011).

Both CWA proteins and biofilm formation mechanisms are regulated by several global regulators, such as the accessory gene regulator (agr), staphylococcal accessory homologous sar genes, sigma factor B (σ^B), and luxS (Vuong et al., 2003; Knobloch et al., 2004; Xu et al., 2006; Christner et al., 2012). Further information on regulation of biofilm in S. epidermidis can be obtained in other review articles (Kong et al., 2006; Mack et al., 2007; Le and Otto, 2015; Paharik and Horswill, 2016).

As already mentioned, biofilms play a role in immune evasion, primarily by providing a barrier to immune cells. PIA may contribute to innate immune system evasion by promoting generation of complement C5a fragment (Satorius et al., 2013; Al-Ishaq et al., 2015), inhibiting phagocytes and neutrophil killing (Vuong et al., 2004b,c), and reducing the activity of AMPs (Vuong et al., 2004b; Otto, 2006). Recently, other studies have reported slightly opposite findings, with PIA-producing bacteria inducing greater inflammatory responses and enhanced phagocytosis (Spiliopoulou et al., 2012; Ferreirinha et al., 2016), although Spiliopoulou et al. did observe reduced killing in PIA-producing strains as discussed elsewhere recently (Nguyen et al., 2017). S. epidermidis also produces a second exopolymer, the poly-γ-glutamic acid (PGA), although at comparatively lower levels. Synthesized by the gene products of the cap locus, PGA is important in mediating S. epidermidis resistance to neutrophil phagocytosis and AMPs, and promoting growth at high salt concentrations (PGA is induced under such conditions) (Kocianova et al., 2005).

It has yet to be elucidated if WTA has a direct role in S. epidermidis immune system evasion. However, like S. aureus, S. epidermidis contains the genes for D-alanylation of WTA, a modification known to protect the bacteria from the activity of AMPs (Peschel et al., 1999).

Although the majority of S. epidermidis strains remain susceptible to the newer antibiotics such as daptomycin, tigecycline, linezolid and dalbavancin (Hellmark et al., 2009; Pinheiro et al., 2016), high endemic antimicrobial resistance within this species represents a significant challenge in the treatment of S. epidermidis infections, especially DRI (Diekema et al., 2001). Methicillin resistance in S. epidermidis (MRSE) is an important characteristic of infecting isolates as it is often associated with additional antibiotic resistance mechanisms. Resistance to other antibiotics, such as erythromycin (encoded by erm genes), ciprofloxacin, clindamycin, aminoglycosides (encoded in aacA/aphD gene) or trimethoprim-sulfamethoxazole, are also often observed, especially in MRSE (Cherifi et al., 2013). Methicillin resistance is encoded by mecA, an alternative penicillin binding protein with decreased affinity to β-lactam based antibiotics such as penicillin, methicillin and oxacillin (Chambers et al., 1985). It is carried on the mobile genetic element, staphylococcal cassette chromosome mec (SCCmec), of which several types have been identified for S. epidermidis (Miragaia et al., 2005). MRSE have been found to be common in infection-causing isolates (70–87% of all S. epidermidis isolates) (Cherifi et al., 2013; Farina et al., 2016; Morgenstern et al., 2016c; Salgueiro et al., 2017), and even higher (90%) in specific patient cohorts (Morgenstern et al., 2016b). MRSE prevalence in healthy individuals is low (3–18% of S. epidermidis commensal isolates) (Rolo et al., 2012; Cherifi et al., 2013; Farina et al., 2016), although prevalence is increased for individuals exposed to the healthcare system, as observed in hospitalized patients or in healthcare workers (Rohde et al., 2004; Morgenstern et al., 2016a; Widerstrom et al., 2016). The specific causes of the increased prevalence of resistant isolates in the hospital environment is unknown, although is likely associated with high antibiotic exposure and direct or indirect interpersonal transmission.

It remains unclear whether infection with resistant organisms results in a worse clinical outcome in comparison with susceptible counterparts. In a recent study of patients with S. epidermidis ODRIs, methicillin resistance status did not influence the clinical course and outcome of treatment (Morgenstern et al., 2016c), although further studies are required to confirm this finding. In any case, clear therapeutic guidelines are available for the treatment of both MRSE and MSSE, with a high likelihood of treatment success in both cases when guidelines are followed closely.

Until relatively recently it was thought that S. epidermidis did not produce toxins. However, the identification and characterization of the PSMs have now changed that concept (Mehlin et al., 1999). The PSMs are a family of genome-encoded peptides, and like the CWA proteins/adhesins, are under the strict regulation of the agr quorum sensing system (Figure 2; Mehlin et al., 1999; Vuong et al., 2004a; Yao et al., 2005). In S. epidermidis, the PSM family consists of PSM-α, PSM-β1, PSM-β2, PSM-δ, PSM-ε, and PSM-γ/δ-toxin (Mehlin et al., 1999; Vuong et al., 2004a; Yao et al., 2005). PSMβ peptides are the primary PSMs produced by S. epidermidis, are expressed at high levels during biofilm formation, and have been shown to have a role in the structuring and dispersal of the biofilm (Yao et al., 2005; Wang et al., 2011). They are specifically associated with the formation of channels observed between the biofilm layers, which are considered important for nutrient uptake (Wang et al., 2011). S. epidermidis-derived PSMδ is strongly cytolytic against neutrophils, similar to S. aureus. However, S. epidermidis culture filtrates were observed to have a very low cytolytic potential in vitro (Cheung et al., 2010). As growing conditions are likely to have an influence on PSM production, the role of S. epidermidis PSMδ in vivo needs to be further addressed.

Finally, certain S. epidermidis strains have been shown to produce PSM-mec, a PSM encoded in the mobile genetic element SCCmec, in contrast to the other PSMs that are chromosomal encoded (Qin et al., 2016). PSM-mec has cytolytic potential against neutrophils in vitro and its presence has been associated with decreased bacterial clearance and higher mortality rates in a murine model of sepsis (Qin et al., 2017).

Small colony variants (SCVs), a colony phenotype characterized by small size, slow growth and downregulation of virulence genes, are recognized as a pathogenic mechanism for several bacterial species, including S. epidermidis, and are often associated with chronic infections (Johns et al., 2015). SCVs seem to be less susceptible to antibiotics and to the immune system, potentially by being able to survive intracellularly and inducing a more anti-inflammatory environment due to increased secretion of IL-10 (Magrys et al., 2015). The topic has been extensively reviewed recently (Kahl et al., 2016).

Finally, internalization and intracellular persistence in non-professional phagocytes (e.g., osteoblasts) is a described evasion mechanism for S. aureus (Mempel et al., 2002; Hamza and Li, 2014). A few internalization mechanisms have been described for S. epidermidis, involving AtlE (Hirschhausen et al., 2010) and SdrG (Claro et al., 2015). This represents a potentially new pathogenic mechanism for S. epidermidis and a location where bacteria could survive to cause persistent/relapsing infections; however its relevance in vivo has not yet been proven.

Innate immune responses are triggered by the detection of microbial structures through pattern-recognition receptors (PRRs) on immune and tissue cells. The most studied PRRs are toll-like receptors (TLRs), which recognize a broad range of bacterial derived macromolecules (Akira and Hemmi, 2003). S. epidermidis triggers immune responses partly via TLR-2 (which often forms heterodimers with TLR-1 and TLR-6; Fournier, 2012), similar to S. aureus (Yoshimura et al., 1999; Morath et al., 2002). TLR-2 can recognize different bacterial cell wall molecules including lipoproteins, LTA and peptidoglycan (PDG) (Figure 3; Akira et al., 2006; Fournier, 2012), although some of its ligands are still controversial (van Bergenhenegouwen et al., 2013). Secreted components can also be recognized and activate the immune system, as it was shown for S. epidermidis PSM, which is recognized by TLR-2/TLR-6 heterodimers (Hajjar et al., 2001).

Recognition of S. epidermidis via TLR-2 has been shown in keratinocytes (Wanke et al., 2011; Ommori et al., 2013), endothelial cells (Robertson et al., 2010), or human fibroblasts (Hatakeyama et al., 2003), and has also been demonstrated in TLR-2 transfected human embryonic kidney (HEK)293 cell line (Strunk et al., 2010). Furthermore, in preclinical models of S. epidermidis bacteremia or subcutaneous/soft tissue foreign-body infection, an up-regulation of TLR-2 and the adaptor molecule MyD88 has been observed upon infection (Kronforst et al., 2012; Svensson et al., 2015, 2017). The use of TLR-2 knock-out (KO) in bacteremia models with neonatal and adult mice resulted in delayed clearance, especially at early time-points after infection (Strunk et al., 2010; Bi et al., 2015; Cole et al., 2016). These data suggest that TLR-2 is involved in the early responses to S. epidermidis infections although is not essential for clearance of the infection (Cole et al., 2016).

Responses toward S. epidermidis can also occur independently of TLR-2, as it was shown in the models using TLR-2 KO mice (Bi et al., 2015). Other PRRs that may potentially be involved in S. epidermidis sensing are NOD-like receptors, as they recognize S. epidermidis-derived PDG (Natsuka et al., 2008). CD14, expressed mostly in monocytes and macrophages, is a TLR-2 co-receptor which may contribute to S. epidermidis recognition in some cell subsets (Hatakeyama et al., 2003). PSMs produced by S. epidermidis can be sensed by formyl peptide receptor 2 (FPR2/ALX) (Kretschmer et al., 2012, 2015), expressed in neutrophils and involved in their recruitment to the infection site (Rautenberg et al., 2011). To date, the contribution of these receptors in vivo has not been addressed.

Human AMPs are a heterogeneous group of amphipathic peptides, which may be subdivided depending on their structure and function. AMPs functions include rapid, direct killing of microbes and activation/modulation of immune responses, such as cell recruitment or chemokine production. One of the most effective early responses of the host to pathogenic insults is mediated through human β-defensins (hBD). In vitro experiments with keratinocytes or skin explants have shown that S. epidermidis or its culture supernatants can elicit high levels of hBD-2 and hBD-3 but not hBD-1 (Lai et al., 2010; Li et al., 2013; Ommori et al., 2013; Percoco et al., 2013; Park et al., 2014), and RNase7 and cathelicidin LL-37 in epithelial cells (Burgey et al., 2016). This AMP induction may be beneficial under healthy conditions to counteract more pathogenic species (Lai et al., 2010; Li et al., 2013) but can be also expected to contribute to defense in S. epidermidis superficial or ocular infections. Of relevance, some of them (hBD-2, hBD-3, LL-37 and human alpha defensin (HNP)-1) have been proven, to different extents, to be effective against S. epidermidis in vitro (Turner et al., 1998; Gordon et al., 2005; Huang et al., 2007; Dapunt et al., 2016b), although no data is available from in vivo studies. Nevertheless, the studies mentioned above showed some discrepancies in terms of AMP killing capacity, which could be explained by differences in strains used, as some of them may possess mechanisms against AMP. More relevant in the context of S. epidermidis DRI, other cell types including neutrophils and monocytes can produce AMPs. These AMPs will often be located in the phagolysosomes, where they can contribute to bacteria killing. Of interest, hBD-3, LL-37 and hepcidin 20, a liver-derived AMP, have been shown to reduce S. epidermidis attachment and/or biofilm formation in vitro (Hell et al., 2010; Zhu et al., 2013; Brancatisano et al., 2014). The mechanisms of action is currently unknown, although for hBD-3 a decrease in icaA and icaD expression and increase of icaR were associated with the observations (Zhu et al., 2013).

Phagocytosis by neutrophils is one of the most important mechanisms for elimination of contaminating or infecting bacteria. Neutrophils migrate to the site of infection, following host signals (e.g., chemokines, AMPs) or sensing bacterial components as mentioned above. At the infection site, neutrophils will internalize opsonized bacteria forming a phagosome and, finally, bacteria will be destroyed in the phagolysosome by the action of reactive oxygen species (ROS), proteases and AMPs. An additional mechanism to kill bacteria has been described for neutrophils: the generation of neutrophil extracellular traps (NETs) or NETosis. Nuclear and mitochondrial DNA is released to the extracellular space to form NETs, which contain high local concentrations of intracellular antimicrobial proteins. Although literature is still limited, S. epidermidis biofilms have been shown to induce DNA release and NETosis in vitro (Meyle et al., 2012; Dapunt et al., 2016a). Macrophages are also able to phagocytose and destroy S. epidermidis (Riool et al., 2014) with similar mechanisms, and further present antigens to T cells. Phagocytosis of S. epidermidis by macrophages is enhanced following stimulation with IFN-γ in vitro (Magrys et al., 2015) and in vivo (Boelens et al., 2000a).

Phagocytes will also act against biofilms. It has been shown that neutrophils can bind to opsonized but also non-opsonized biofilms, partly by recognizing EPS (Meyle et al., 2012). Nevertheless, it is generally accepted that the biofilm mode of growth will protect bacteria from phagocytosis, despite some discrepancies in the literature that have been discussed elsewhere (Nguyen et al., 2017). Furthermore, biofilm mode of growth, most often studied in PIA-producing strains, has been shown to decrease killing efficiency in macrophages and neutrophils (Vuong et al., 2004c; Cerca et al., 2006; Kristian et al., 2008; Spiliopoulou et al., 2012).

Interesting observations were made when comparing the phagocytosis of S. epidermidis and S. aureus biofilms, with the latter being more likely infiltrated and engulfed (Guenther et al., 2009). However, although S. aureus was more likely phagocytosed, this does not always correlate with the capacity of neutrophils to kill the bacteria. In fact S. aureus has several mechanisms to avoid lysis by neutrophils and to persist intracellularly (Foster, 2005). S. epidermidis does not appear to possess similar mechanisms. However, some strains are killed less efficiently, potentially by having a low capacity to prime the oxidative response of neutrophils (Nilsdotter-Augustinsson et al., 2004), or as described before by their biofilm mode of growth. These observations, together with lower induction of neutrophil apoptosis, may lead to intracellular survival and could partially explain the low inflammatory nature and chronicity often associated with S. epidermidis infections.

Cytokines are a broad group of secreted proteins that play a role in intercellular communication, with a broad range of functions within the immune system as cell recruitment, differentiation and activation. Interleukins and other factors play an essential role in leukocyte communication and differentiation, while chemokines are mainly involved in cell recruitment. In vitro stimulation of peripheral blood mononuclear cells with different staphylococcal species showed a rapid release of pro-inflammatory cytokines such as IL-1β, IL-6, IL-12p70, or IFN-α (Megyeri et al., 2002). Of note, S. epidermidis induced lower levels of pro-inflammatory cytokines compared to S. aureus (Megyeri et al., 2002). Studies with monocytes/macrophages have also observed IL-6, tumor necrosis factor (TNF)-α and IL-1β release after S. epidermidis stimulation (Wilsson et al., 2008; Strunk et al., 2012). Laborel-Préneron et al. reported that stimulation of moDC with commensal S. epidermidis induced a more anti-inflammatory profile in contrast to stimulation with commensal strains of S. aureus, with high levels of IL-10 being a key differentiator. Nevertheless, pro-inflammatory cytokines such as IL-6 and TNF-α were also detected (Laborel-Preneron et al., 2015). Similar observations have been made from in vivo studies: IL-6, TNF-α, and IL-1β are typically observed in serum in the first hours post-challenge with live or inactivated S. epidermidis (Wakabayashi et al., 1991; Simojoki et al., 2011; Bi et al., 2015; Ferreirinha et al., 2016; Qin et al., 2017), or in tissue exudates/homogenates from experimental DRI models (Boelens et al., 2000b; Svensson et al., 2015). The regulatory cytokine IL-10 is also present in vivo (Ferreirinha et al., 2016) and it has been shown that S. epidermidis inoculation result in higher IL-10 levels compared to P. aeruginosa in an intradermal infection model (Bialecka et al., 2005). In a S. epidermidis DRI mouse model it was shown that IL-10 was involved in reducing infection-associated morbidity, with higher levels of pro-inflammatory cytokines and greater weight loss in IL-10 KO animals. Interestingly, bacterial counts were the same in both wild-type and KO strains, suggesting that IL-10 does not impact bacterial clearance (Gutierrez-Murgas et al., 2016). Overall, despite differences due to different S. epidermidis strains and its effect in different tissues, it can be hypothesized that lower induction of pro-inflammatory cytokines together with high IL-10 production, can contribute to the sub-acute nature of S. epidermidis infections.

Multiple chemokines are also released upon S. epidermidis infection. Secretion of IL-8, important for neutrophil recruitment, has been described in vitro and in the first hours post-infection in in vivo studies (Wakabayashi et al., 1991; Boelens et al., 2000b; Simojoki et al., 2011; Svensson et al., 2015). CXCL-1 and CXCL-2, mostly produced by macrophages (via TLR-2 recognition but also by other mechanisms), have also been observed in bacteraemia and peritonitis models (Strunk et al., 2010; Bi et al., 2015; Ferreirinha et al., 2016; Qin et al., 2017). Additionally, a murine peritonitis model revealed increasing levels of numerous chemokines upon challenge with S. epidermidis supernatants (Perks et al., 2016).

The aggregation and activation of platelets in the presence of bacteria was first described over 25 years ago (Usui et al., 1991) and yet the nature of this interaction has only recently been elucidated. Platelets and bacteria can interact in three ways: the indirect binding of bacteria to a plasma protein (which is a ligand of a platelet receptor), the direct recognition of bacteria by platelet receptors and the binding of secreted bacterial products to platelets (Hamzeh-Cognasse et al., 2015). Only the first type has been described for S. epidermidis, where the SdrG has been described to bind platelets in a fibrinogen and Ig-dependent manner; an interaction that leads to platelet aggregation (Brennan et al., 2009). S. aureus or Streptococcus have been shown to interact with platelets in other ways, which can lead to sepsis or thrombosis but also can play a role in internalization of bacteria by platelets or release of antimicrobial components and immunomodulatory factors (Hamzeh-Cognasse et al., 2015). Future studies will be required to elucidate if S. epidermidis-platelets interaction is limited to SdrG or if, like other bacteria, possess multiple mechanisms.