OUP user menu

Virulence strategies of the dominant USA300 lineage of community-associated methicillin-resistant Staphylococcus aureus (CA-MRSA)

Lance R. Thurlow, Gauri S. Joshi, Anthony R. Richardson
DOI: http://dx.doi.org/10.1111/j.1574-695X.2012.00937.x 5-22 First published online: 1 June 2012


Methicillin-resistant Staphylococcus aureus (MRSA) poses a serious threat to worldwide health. Historically, MRSA clones have strictly been associated with hospital settings, and most hospital-associated MRSA (HA-MRSA) disease resulted from a limited number of virulent clones. Recently, MRSA has spread into the community causing disease in otherwise healthy people with no discernible contact with healthcare environments. These community-associated MRSA clones (CA-MRSA) are phylogenetically distinct from traditional HA-MRSA clones, and CA-MRSA strains seem to exhibit hypervirulence and more efficient host : host transmission. Consequently, CA-MRSA clones belonging to the USA300 lineage have become dominant sources of MRSA infections in North America. The rise of this successful USA300 lineage represents an important step in the evolution of emerging pathogens and a great deal of effort has been exerted to understand how these clones evolved. Here, we review much of the recent literature aimed at illuminating the source of USA300 success and broadly categorize these findings into three main categories: newly acquired virulence genes, altered expression of common virulence determinants and alterations in protein sequence that increase fitness. We argue that none of these evolutionary events alone account for the success of USA300, but rather their combination may be responsible for the rise and spread of CA-MRSA.

  • Staphylococcus aureus
  • ACME
  • PVL
  • Agr
  • SaeRS

Multidrug resistance in Staphylococcus aureus: the rise of methicillin-resistant S. aureus (MRSA)

The Gram-positive pathogen Staphylococcus aureus remains one of the most problematic and costly sources of bacterial infection worldwide (Diekema et al., 2001). Disease typically presents as mild skin/soft tissue infections but can also be the source of more serious bacteremia, endocarditis, osteomyelitis and necrotizing pneumonia (Lowy, 1998). Staphylococcus aureus asymptomatically colonizes the skin and, more commonly, the anterior nasal passages of healthy people (Foster, 2009). Nasal colonization is the most significant predictor of invasive disease; however, in some studies, nearly half of patients carrying S. aureus are strictly colonized extranasally (Schechter-Perkins et al., 2011). Thus, estimates of S. aureus carriage at ∼ 25% of the human population may be an underestimate of true colonization levels. Given the near ubiquity of S. aureus among the human population combined with its virulence potential, it is no wonder this organism has been recognized as a significant healthcare burden for over a century. Staphylococcus aureus was first described by Alexander Ogston in 1881 as the sole microorganism within the fluid drained from a severe knee abscess (Ogston, 1881). Then, he noted that ‘once established the micrococci are hard to kill…’ underscoring the recalcitrant nature of S. aureus toward antiseptic treatment (Newsom, 2008). During this time, Joseph Lister's influence on surgical procedures through the implementation of carbolic acid (phenol) to sterilize wounds and instruments had greatly reduced the occurrence of post-operative infections (Lister, 1867). However, it was subsequently shown that S. aureus was inherently resistant to phenol explaining its association with surgical infections despite good ‘sterile technique’ (Reddish, 1925). Thus, S. aureus was recognized as an important hospital-associated pathogen over 130 years ago in the pre-antibiotic era and little has changed to this day.

Perhaps because of its intimate association with hospitals and patients, S. aureus has always been among the first bacterial species reported to develop resistance to new antimicrobials, from sulfonamide resistance in the early 1940s (Landy et al., 1943) to the identification of penicillinase in 1944 (Kirby, 1944) just months after US penicillin production reached full scale. Interestingly, these progenitor β-lactamase positive S. aureus clones were isolated from patients that had not even been treated with penicillin. Nonetheless, penicillin-resistant S. aureus was here to stay and became pandemic in hospitals during the late 1950s and early 1960s (Rountree & Freeman, 1955). Subsequently, a penicillinase-resistant β-lactam derivative, methicillin (Celbenin; Beecham Pharmaceuticals), was approved for use in the US in 1959. Less than 2 years later, the first report of MRSA was published documenting the isolation of MRSA clones from a patient and hospital staff in the UK, again none of which were treated with methicillin (Jevons, 1961). It was immediately recognized that methicillin resistance was mechanistically different than penicillin resistance in that the MRSA phenotype did not involve direct inactivation of the drug. Rather, resistance was mediated through the acquisition of an alternative penicillin-binding protein (PBP2a) with lowered affinity for β-lactam antibiotics. Within 20 years after the first discovery of MRSA, it became a leading cause of hospital-acquired infections (Archer & Mayhall, 1983). Currently, it can still be responsible for nearly 60% of skin/soft tissue infections presenting to US emergency rooms (Moran et al., 2006).

The methicillin resistance determining PBP2a is encoded by mecA harbored on a mobile genetic element (MGE), staphylococcal cassette chromsome (SCCmec). A nearly identical homologue, now thought to be the ancestral mecA, was recently discovered in Staphylococcus fleuretti, an animal colonizing staphylococcal species (Tsubakishita et al., 2010). Unlike a previously identified mecA homologue in Staphylococcus sciuri that does not confer methicillin resistance (Couto et al., 1996), S. fleuretti is fully resistant to β-lactam antibiotics. Interestingly, the S. fleuretti mecA homologue is not found on a mobile SCC, but rather in the core chromosome between the mevalonate biosynthetic and xylose utilization operons, explaining the presence of mva and xyl gene fragments in some S. aureus SCCmec elements (Tsubakishita et al., 2010). These mobile islands have diversified considerably over the 50-year history of MRSA such that there are currently eight distinct SCCmec types circulating among S. aureus as well as some species of coagulase negative staphylococci (Center for Disease Control & Prevention, 2009a). SCCmec elements can vary greatly in size and composition with the largest (SCCmec type II) spanning 52 kb and additionally encoding erythromycin, spectinomycin and tobramycin resistance determinants (Katayama et al., 2000). Depending on the particular SCCmec type, these mobile islands peppered with insertion sequence (IS) elements, transposons and integrated plasmids, can confer multidrug resistance determinants that significantly diminish treatment options in a clinical setting. Thus, in addition to methicillin resistance, MRSA isolates have evolved multidrug resistance leading to what the popular press refers to as an emerging superbug (McKenna, 2010).

Paradigm shift: the rise of CA-MRSA

After 1961, MRSA spread worldwide causing significant morbidity and mortality almost entirely as hospital-acquired infections. Advances in molecular epidemiology allowed for in-depth analyses of MRSA spread and expansion at the evolutionary level. For instance, spa-typing (polymorphisms in Protein A coding sequence) and SCCmec-typing discriminated unrelated clones and identified clusters of related MRSA lineages responsible for disease (Shopsin et al., 1999; Okuma et al., 2002). Multi-locus sequence typing involves the sequencing of fragments from seven ‘housekeeping’ genes (arcC, aroE, glpF, gmk, pta, tpi and yqiL) yielding unique sequence types (STs) (Enright et al., 2000). STs sharing identity at the majority of these loci are grouped into clonal complexes (CCs) encompassing related lineages of MRSA (Enright et al., 2002). Another highly discriminatory approach that can identify genomic rearrangements and insertions/deletions is pulsed-field gel electrophoresis (PFGE) whereby SmaI digested chromosomal DNA is separated and similarities in banding patterns reflect relatedness among lineages (Bannerman et al., 1995; McDougal et al., 2003). This allows for the classification of S. aureus strains into the now familiar PFGE types USA100-1200. Employing these epidemiological approaches, researchers appreciated that most MRSA disease worldwide (nearly 70% of reported infections) was caused by five major CCs: CC5, CC8, CC22, CC30, and CC45 (McDougal et al., 2003; Robinson & Enright, 2003) (Fig. 1). CC5 includes clones belonging to the USA100 PFGE type (e.g. SCCmec-II New York/Japan clone), the most common source of US hospital-acquired MRSA as well as USA800 (SCCmec-IV Pediatric clone). CC8 includes the archaic, or original MRSA clones as well as the related Iberian clone, the SCCmec-III Brazilian/Hungarian clone, and the SCCmec-IV USA500 clones. CC22 includes the EMRSA-15 clones that dominated hospital infections in the UK during the 1990s along with strains from CC30 encompassing EMRSA-16 as well as the USA200 PFGE type. Finally, CC45 consists of clones belonging to USA600 PFGE type (e.g. Berlin clone) that caused widespread MRSA hospital infections in northern Europe. In essence, after 30 years of investigation, the scientific community began to understand the population structure of the MRSA clones responsible for the majority of hospital-acquired disease. The source of high virulence potential inherent to these five CCs was never fully appreciated before everything we knew about MRSA epidemiology changed at the turn of the century.

Figure 1

Schematic representation of the evolution of MRSA. STs belonging to established CCs are colored as follows: CC1, purple; CC5, green; CC8, red; CC22, orange; CC30, blue; CC45, black. ST59 has not been assigned to a CC. Roman numerals reflect acquired SCCmec type. Commonly used Staphylococcus aureus strains are depicted around their relevant ST symbol.

Initially reported in 1993, patients without any contact with healthcare settings contracted invasive MRSA infections in Kimberly Australia, a region in the northern part of Western Australia (Udo et al., 1993). It was later discovered that simultaneously, strains related to these ‘community-acquired’ MRSA (CA-MRSA) clones were causing serious and fatal respiratory infections in Chicago, again in patients without direct contact with hospital environments (Center for Disease Control & Prevention, 1999). Prior to these reports, MRSA infections were exclusively associated with healthcare settings. These new clones belong to CC1 (USA400 PFGE type), a CC unrelated to the five traditional hospital-associated MRSA (HA-MRSA) complexes (Center for Disease Control & Prevention, 1999). CC1 clones spread quickly through Australia, the mid- and northwestern United States as well as Canada and Alaska where they still cause significant CA-MRSA disease (Center for Disease Control & Prevention, 1999; Coombs et al., 2004; Mulvey et al., 2005; David et al., 2008; Van De Griend et al., 2009). Recent studies show that USA400 can account for over 98% of MRSA infections in northern Canada (Golding et al., 2011) and has been implicated in isolated MRSA disease in southern Europe (Vignaroli, 2009; Neocleous et al., 2010). However, about 10 years ago, a new source of CA-MRSA arose from one of the ‘traditional’ virulent CCs, CC8. Descending from a USA500 clone through acquisition of various MGEs (Robinson & Enright, 2003; Li et al., 2009), USA300 became the dominant CA-MRSA clone in US (Moran et al., 2006; Hulten et al., 2010; Talan et al., 2011), effectively replacing USA400 clones in most regions (Como-Sabetti et al., 2009; Simor et al., 2010), and has also been isolated from patients in Canada and Mexico (Nichol et al., 2011; Velazquez-Meza et al., 2011). The explosion of USA300 CA-MRSA across North America resulted from a very recent clonal expansion of a successful CA-MRSA clone as demonstrated by very low sequence divergence among geographically distinct USA300 isolates (Kennedy et al., 2008).

Given the occurrence of multiple CA-MRSA clones in the population, a formal definition was put forth by the Center for Disease Control and Prevention for CA-MRSA disease as that which is contracted within 48 h of hospital admission by patients not having recently undergone surgery, hemodialysis, prolonged hospitalization, catheterization, or MRSA colonization (Morrison et al., 2006). Currently in the US, MRSA disease fitting these criteria is almost always caused by USA300 clones, followed by USA400 and occasionally USA1000 and USA1100 (Talan et al., 2011). To complicate matters further, USA300 clones have recently been implicated in causing significant HA-MRSA disease (Popovich et al., 2008; Jenkins et al., 2009; Moore et al., 2009; Hulten et al., 2010), blurring the lines between the two disease onset environments (Popovich et al., 2008; Jenkins et al., 2009; Moore et al., 2009; Hulten et al., 2010). In some studies, USA300 accounted for at least half of hospital-acquired MRSA infections (Popovich et al., 2008; Hulten et al., 2010). Thus, USA300 represents a highly successful S. aureus clone that emerged in the community and quickly spread throughout the North American continent to become the leading cause of MRSA infection even in healthcare settings. For now, USA300 seems to be primarily limited to North America, while in Europe, South America and Asia, CA-MRSA disease is dominated by divergent clones unrelated to CC8 (e.g. ST30, ST80 and ST59) (Deleo et al., 2010). Given the rapid and efficient transmissibility of USA300 in North America (Pan et al., 2005), it remains to be seen whether these clones will become the dominant source of MRSA disease worldwide.

USA300 virulence

Animal models of S. aureus infection have repeatedly demonstrated the hypervirulence associated with USA300 compared with other MRSA strains (Montgomery et al., 2008; Li et al., 2009, 2010; Cheung et al., 2011). USA300 strains exhibited enhanced production of dermonecrotic lesions in skin abscess models when compared to HA-MRSA clones (Li et al., 2009, 2010; Cheung et al., 2011), and USA300 was more lethal in a rat model of pneumonia compared with a USA400 isolate (Montgomery et al., 2008). Furthermore, USA300 strains were more lethal in septic infections compared with archaic and Iberian clones as well as ST239 clones (Brazilian clones) (Li et al., 2009). When compared with other CA-MRSA clones, USA300 isolates generally exhibit increased virulence with the exception of ST80 and USA1000, which also possess enhanced virulence (Li et al., 2010). In contrast, nearly every clone of HA-MRSA tested was significantly less virulent than USA300 with the only exception being USA500 HA-MRSA (Li et al., 2009, 2010). This is of particular interest in that USA300 clones descended from USA500 via the acquisition of a prophage containing panton-valentine leukotoxin (PVL), a mobile arginine catabolic mobile element (ACME) and enterotoxins K and Q (see below) (Li et al., 2009). Thus, the source of USA300 hypervirulence may have originally evolved in the HA-MRSA isolates belonging to USA500. However, for unknown reasons, despite exhibiting hypervirulence in animal infection models, USA500 clones remain relegated to healthcare settings and do not cause significant CA-MRSA disease. Whether CA-MRSA USA300 clones exhibit hypervirulence in human disease has been difficult to directly discern, however, recent population-based clinical data are beginning to corroborate conclusions drawn from laboratory animal model experiments.

In humans, USA300 S. aureus primarily causes skin infections of which, it can account for up to 98% of all MRSA presenting as skin/soft tissue infections to US emergency rooms (Talan et al., 2011). In addition, USA300 can also cause more invasive disease such as bacteremia (Seybold et al., 2006), endocarditis (Haque et al., 2007), and necrotizing fasciitis (Miller et al., 2005), a condition almost never associated with S. aureus. In particular, pulmonary infections caused by USA300 S. aureus can lead to aggressive and often fatal necrotizing pneumonia (Francis et al., 2005; Hageman et al., 2006; Klevens et al., 2007). The populations most at risk for contracting USA300 CA-MRSA are military personnel (Ellis et al., 2009), athletes (Center for Disease Control & Prevention, 2003b, c, 2009b), prisoners (Center for Disease Control & Prevention, 2001, 2003a; Maree et al., 2010), African Americans (Klevens et al., 2007; Kempker et al., 2010), daycare attendees (Buckingham et al., 2004; Kaplan et al., 2005), and men who have sex with men (Sztramko et al., 2007). Patients contracting CA-MRSA are, on average, younger than those with HA-MRSA and otherwise generally healthy (Nair et al., 2011; Whitby et al., 2011). Furthermore, CA-MRSA is often associated with worse clinical outcomes. For instance, USA300 infections were associated with increased in-hospital mortality and a higher occurrence of severe sepsis than HA-MRSA infections (Kempker et al., 2010; Kreisel et al., 2011). USA300-related strains were also more prone to spread from the initial infection site and caused more severe infections than HA-MRSA in patients suffering from pneumonia with pulmonary emboli (Ganga et al., 2009; Hota et al., 2011). However, other reports describe better clinical outcomes associated with USA300 infections (Lalani et al., 2008; Moore et al., 2009). Although some studies that reported more positive clinical outcomes with CA-MRSA also describe hypervirulent CA-MRSA trends that merely lack full statistical significance, such as increased risk of being admitted into intensive care (OR = 1.8, P = 0.09) (Popovich et al., 2008). Additionally, effective treatment, which is easier to achieve when treating CA-MRSA infections given their inherent susceptibility to clindamycin, tetracyclines, rifampicin and trimethoprim/sulfonamide, can reduce the severity of CA-MRSA disease outcomes in population-based studies (Bassetti et al., 2011). Unfortunately, this trend of increased antibiotic susceptibility may be diminishing as new reports show increased antibiotic resistance among USA300 isolates, possibly through direct acquisition of resistance determinants from multidrug-resistant HA-MRSA strains (McDougal et al., 2010). Thus, the future clinical outlook appears grim with respect to USA300 infections given their increased prevalence in both hospital- and community-acquired infections, their propensity to acquire new antibiotic resistance determinants, and the steady decline in positive clinical outcomes associated with USA300 infections.

Genetic determinants contributing to USA300 success

Given the recent impact of USA300 on human health, significant research effort has been exerted to elucidate the source of USA300 success. Here, we review these findings and broadly categorize them into three main classes: (1) newly acquired genes that promote virulence and/or fitness, (2) altered regulation of core genes resulting in elevated virulence and/or fitness, and (3) nonsynonymous mutations in core genes that enhance virulence and/or fitness.

Newly acquired genes

Many different lineages of CA-MRSA (USA400, USA1000, and USA1100) cause outbreaks and invasive infections, but in North America, none are as prevalent as epidemic USA300. These clones have acquired many genes in the form of MGEs that may confer a selective advantage over other CA-MRSA strains. Several groups have investigated many of these MGEs with the goal of elucidating factors (if any) that have contributed to the overwhelming success of USA300.

Enterotoxins K and Q

USA300 CA-MRSA isolates contain genes encoding enterotoxins K and Q (sek2 and seq2) in a unique pathogenicity island SaPI5 (Diep et al., 2006a). Sek2 and Seq2 are thought to contribute to pathogenesis by stimulating T-cells through binding of the Vβ chain of αβ T-cell receptors. Sek2 and Seq2 share 98% amino acid homology with enterotoxins (Sek and Seq) found on SaPI3 in S. aureus COL an archaic HA-MRSA clone belonging to ST250 that is less virulent than CA-MRSA isolates (Yarwood et al., 2002). USA400 isolates (e.g. MW2) harbor νSA3, a pathogenicity island that shares similarity to SaPI3 of COL and SaPI5 of USA300, however, νSA3 does not contain the genes for Sek or Seq (Diep et al., 2006a). Thus, the acquisition of these toxins by USA300 and not US400 may potentially explain the differences in pathogenicity although direct demonstration of this has not been reported.


The mecA gene encodes a penicillin-binding protein and is located on a MGE known as the Staphylococcal Cassette Chromosome mec (SCCmec). There are currently eight recognized SCCmec types (I–VIII). SCCmec types I, II, and III contain additional drug resistance determinants, whereas types IV, V, VI, and VII cause resistance only to β-lactams (Carvalho et al., 2010). Initial sequence comparisons show that both USA400 and USA300 strains contain a nearly identical SSCmecIVa (Baba et al., 2002; Diep et al., 2006a). As it turns out, SCCmecIV is the most common form of SCCmec found across divergent S. aureus lineages in addition to ST8 (USA300) including ST1 (USA400), ST80, ST72 (USA700) and ST8 (USA500) (Daum et al., 2002; Goering et al., 2007). It has been shown that SSCmecIV does not impose a fitness cost in vitro or in vivo, whereas acquisition of the SSCmec types I, II, and III resulted in decreased in vitro growth rates (Ender et al., 2004; Lee et al., 2007; Diep et al., 2008a). Thus, it is thought that harboring SSCmecIV as opposed to other SCCmec types imparts CA-MRSA with an advantage in its ability to cause infection in healthy individuals. However, although SSCmecIV may provide a selective advantage to CA-MRSA over other SCCmec types, the fact that nearly all CA-MRSA isolates contain SSCmecIVa suggests that it is not a major contributing factor to the dominance of USA300 among CA-MRSA isolates.

Panton-Valentine leukocidin (PVL)

The PVL is a bicomponent pore-forming toxin that induces necrosis and apoptosis in leukocytes (Coulter et al., 1998). PVL is encoded by the genes lukS-PV and lukF-PV located on the prophage φSA2pvl (Diep et al., 2006a). This phage is highly associated with CA-MRSA clones in that nearly all USA300, USA400, and USA1100 clinical isolates are positive for PVL as are many USA1000 strains (Diep et al., 2006b; Coombs et al., 2010). Furthermore, epidemiological and clinical reports indicate a strong correlation between PVL production and severe skin/soft tissue infections, as well as necrotizing pneumonia and fasciitis, suggesting PVL may be a major contributor to the virulence of CA-MRSA (Cribier et al., 1992; Lina et al., 1999; Gillet et al., 2002). Moreover, PVL can be directly detected in human skin abscesses at levels known to result in rapid neutrophil lysis (Badiou et al., 2008, 2010). Thus, PVL is significantly correlated with invasive CA-MRSA disease; however, recent clinical studies demonstrate that CA-MRSA strains lacking PVL can still cause disease outbreaks (Diep et al., 2008b; Otter & French, 2008; Zhang et al., 2008).

Until recently, demonstrating a direct role for PVL in model disease has proven difficult. This likely stems from the host specificity of PVL in that it is rapidly leukocidal for rabbit and human neutrophils, but much less active against murine, rat, or simian neutrophils (Loffler et al., 2010). Consequently, a virulence effect of PVL in murine or rat pneumonia, sepsis, and skin infection models has never been reproducibly defined (Voyich et al., 2006; Bubeck Wardenburg et al., 2007a, 2008; Labandeira-Rey et al., 2007; Brown et al., 2009; Villaruz et al., 2009). Moreover, there was no demonstrable role for PVL in a pneumonia model involving nonhuman primates (Olsen et al., 2010). In contrast, using PVL susceptible rabbit models, isogenic USA300 strains lacking PVL were less virulent in pneumonia, osteomyelitis, and skin abscess models (Cremieux et al., 2009; Diep et al., 2010; Kobayashi et al., 2011; Lipinska et al., 2011). However, the attenuation of mutants lacking PVL in rabbit skin lesions was not nearly as striking as a mutant lacking α-hemolysin or phenol-soluble modulin (PSM) production underscoring the contributory nature of PVL toward S. aureus pathogenesis (Hongo et al., 2009; Kobayashi et al., 2011). Furthermore, the nearly ubiquitous presence of PVL among CA-MRSA isolates clearly suggests that this toxin cannot explain the particular success of the USA300 lineage.

Arginine catabolic mobile element

Of all the genetic elements acquired by CA-MRSA isolates, only the ACME is completely unique to USA300 (Diep et al., 2006a). The type 1.02 ACME carried by USA300 is juxtaposed to the SCCmecIV island and was acquired from Staphylococcus epidermidis through horizontal gene transfer via a mechanism likely involving the SCCmec-related CcrAB recombinases (Diep et al., 2006a, 2008a; Miragaia et al., 2009). The physical linkage of ACME with SCCmecIVa is mirrored by an epidemiological linkage in that nearly all USA300 strains harboring SCCmecIVa also carry ACME, while USA300 clones with other SCCmec islands, with rare exceptions, do not (Goering et al., 2007; Shore et al., 2011). The ACME of USA300 contains a complete arginine deaminase (arc) system that converts l-arginine to l-ornithine for both ATP and ammonia production. The island also encodes a putative oligopeptide permease, a zinc-containing alcohol dehydrogenase, and a spermine/spermidine acetlytransferase (SpeG) as well as several hypothetical proteins (Diep et al., 2006a). While a role for ACME in USA300 virulence was demonstrated in a rabbit sepsis model (Diep et al., 2008a), no effect of ACME was observed in murine pneumonia or skin abscess models (Montgomery et al., 2009). Thus, it has been proposed that ACME aids primarily in USA300 colonization, in part, through the Arc-mediated ammonification of the acidic skin environment; though, this has never been experimentally verified (Diep et al., 2008a; Otto, 2010).

We have additionally observed a peculiar phenotype in S. aureus suggestive of a selective advantage afforded by the ACME cassette. Polyamines, including spermine, spermidine, and putrescine, are a group of polycationic compounds reportedly synthesized from l-arginine by all living organisms. Not only does S. aureus lack the ability to synthesize polyamines de novo, but spermine and spermidine are bactericidal to this organism at levels found within mammalian tissue (Baze et al., 1985; Joshi et al., 2011). Polyamine-sensitivity was apparent in all tested strains except those belonging to USA300, and in these isolates, polyamine resistance was dependent on speG encoding a spermine/spermidine aceytltrasferase harbored on ACME. Could speG provide USA300 with a selective advantage by nullifying the staphylocidal effects of host polyamines? While no direct measure of host polyamine levels during S. aureus infections have been reported, several indirect lines of evidence may suggest that polyamines do affect the outcome of staphylococcal disease and/or colonization.

Upon wounding, the host response in the skin is proinflammatory and dominated by cytokines such as IL-1, INF-γ, and TNF-α (Mahdavian Delavary et al., 2011). The resulting inflammation is mediated, among other effectors, by the production of reactive oxygen and nitrogen species, the latter of which, nitric oxide (NO·) is synthesized from l-arginine by the inducible NO-synthase (iNOS, Fig. 2). This enzyme competes for available l-arginine with host enzymes such as Arginase-1 (Fig. 2) as well as with arginine-auxotrophic S. aureus (Emmett & Kloos, 1979). Once tissue damage signals resulting from the primary inflammation outweigh pathogen-associated signals, the host response shifts away from proinflammatory mediators and initiates the profibrotic response (Mahdavian Delavary et al., 2011). This phase is dependent on the production of TH2-like anti-inflammatory cytokines such as IL-4, IL-10, IL-13, and TGFβ and results in induction of host fibrotic response involving Arginase-1 expression. At this stage, the l-ornithine produced by Arginase-1 can be converted to staphylocidal polyamines that will additionally promote fibroblast proliferation, collagen deposition, and inhibition of inflammation (e.g. blocking iNOS translation) (Maeno et al., 1990). It therefore may be during this TH2-dominant fibrotic phase that host polyamines exert their effects on invading S. aureus thereby selecting for ACME-encoded SpeG. Indeed, inhibiting IL-4 signaling in mice increased organism burdens during S. aureus sepsis, while INF-γ−/− mice (lacking robust inflammatory wound response) survived better than WT mice (Sasaki et al., 2000). Thus, TH2-dependent signaling, as opposed to an inflammatory TH1 response, proved critical to the host's ability to control S. aureus infections. Recently, protection against chronic implant infections was also highly dependent on an effective TH2/Treg response (Prabhakara et al., 2011). Furthermore, polymorphisms in the human IL-4 gene associated with reduced IL-4 production are significantly linked with increased S. aureus colonization (Emonts et al., 2008). These data are consistent with the TH2 anti-inflammatory fibrotic response as being critical for controlling S. aureus infection. Whether this is directly because of the induction of polyamine synthesis has yet to be reported, but the acquisition of speG-encoding ACME would counter increased spermine levels in fibrotic tissue perhaps explaining the association of USA300 CA-MRSA with severe skin/soft tissue infections.

Figure 2

Association between arc gene cluster and speG in ACME. Top: ACME type I, found in USA300 Staphylococcus aureus and also found in many Staphylococcus epidermidis isolates, and ACME type II, found primarily in S. epidermidis, both harbor arc gene clusters as well as speG. ACME type III (not shown) lacks an identifiable arc gene cluster but does contain an opp-3 locus. Bottom: Fate of host arginine depends on competition between iNOS and arginase-1 enzyme activities. The net production of ornithine by Arc-expressing S. aureus may skew the fate of host arginine down the polyamine synthesis pathway thereby necessitating speG.

How do we reconcile a significant role for SpeG in S. aureus pathogenesis with the lack of a strong ACME phenotype in most model infections (Diep et al., 2008a; Montgomery et al., 2009)? One explanation could be that the observed increase in α-hemolysin and Protein A expression upon ACME inactivation in USA300 could overcompensate for the resulting polyamine sensitivity (Diep et al., 2008a). Another possibility is that the Arc operon on ACME actually drives excess polyamine production necessitating SpeG-mediated spermine detoxification. The Arc operon consists of genes that convert l-arginine to l-ornithine and CO2 while producing ATP and ammonia. The resulting l-ornithine is exchanged for extracellular l-arginine by the l-arginine/l-ornithine antiporter ArcD effectively converting extracellular l-arginine to l-ornithine. Thus, the Arc operon could skew the flux of host l-arginine away from iNOS toward polyamine synthesis rendering speG essential (Fig. 2). Deleting all of ACME might allow the host to partition available l-arginine toward NO-production, an immune effector that S. aureus is known to effectively resist (Richardson et al., 2006, 2008; Hochgrafe et al., 2008). This is consistent with the presence of speG on ACME islands that harbor the auxiliary arc gene cluster (Fig. 2). While this hypothesis could explain the modularity of ACME that results in ∆speG attenuation, it has several aspects that require experimental attention. First, all strains of S. aureus already encode an Arc operon on the core chromosome that could also result in excess host polyamine synthesis, yet SpeG is only associated with ACME-positive USA300 S. aureus. This could be explained by the fact that the chromosomal Arc operon is only expressed under conditions of low oxygen and low glucose and little is known about ACME Arc expression in S. aureus (Makhlin et al., 2007). Second, a dominant MRSA clone of ST22 lineage in Irish hospitals harbors an ACME island with an arc gene cluster but appears to lack a speG homologue (Shore et al., 2011). Another issue is that significant CA-MRSA disease in Latin America is caused by USA300 clones that lack ACME (Reyes et al., 2009). Thus, ACME may contribute to colonization and virulence, but it cannot fully explain the predominance of USA300 in CA-MRSA disease in North America.

Enhanced virulence gene expression

Staphylococcus aureus elaborates a wide variety of toxins and proteases that have proven critical for efficient dissemination, inflammation, and disease progression (Dubin, 2002; Arvidson, 2006; Bohach, 2006). For instance, α-toxin or α-hemolysin (Hla) is a potent heptameric pore-forming toxin known to be critical for virulence in nearly every tested disease model from skin lesions and endocarditis to murine mastitis (Jonsson et al., 1985; O'Reilly et al., 1986; Bayer et al., 1997). Upon interacting with susceptible cells, which include leukocytes, keratinocytes, platelets, and endothelial cells, it forms a 100 Å deep pore in the plasma membrane resulting in rapid cell lysis (Song et al., 1996; Gouaux, 1998). Recently, a number of reports have shown that Hla expression is highly elevated in USA300 clones compared with other S. aureus isolates (Montgomery et al., 2008; Li et al., 2009, 2010; Cheung et al., 2011). Moreover, deletion of hla abrogates USA300 virulence in murine and rabbit skin lesion models as well as pneumonia (Bubeck Wardenburg et al., 2007a; Kennedy et al., 2008, 2010). However, it should be noted that hla mutants in almost any S. aureus background are attenuated (O'Reilly et al., 1986; Patel et al., 1987; Bramley et al., 1989; McElroy et al., 1999; Bubeck Wardenburg et al., 2007b); thus, the loss of virulence in USA300 ∆hla mutants is consistent with α-toxin in general being a critical pathogenicity factor to S. aureus. δ-toxin (encoded by hld) and related α-type PSMs (αPSMs) are amphipathic α-helical peptides with potent leukocidal and chemotactic properties (Wang et al., 2007). They have been shown to be overproduced by CA-MRSA clones with respect to most HA-MRSA isolates (Wang et al., 2007; Li et al., 2009, 2010). Their abundant production is essential for full virulence in murine and rabbit skin models of infection as well as murine sepsis (Wang et al., 2007; Kobayashi et al., 2011). Interestingly, they have recently been shown to exert potent antimicrobial activity against multiple Gram-positive bacterial species (Joo et al., 2011). This property may prove critical for efficient colonization of nonsterile sites such as skin and nasal passages, thereby providing CA-MRSA with a selective advantage during transmission. Finally, S. aureus expresses a number of secreted proteases that, while antagonistic to in vitro biofilm formation, likely mediate the breakdown of host fibrotic tissue synthesized to confine S. aureus-containing lesions thereby promoting bacterial dissemination and disease progression. As with α-toxin and αPSMs, USA300 clones are also known to excrete proteases in excess, potentially limiting the host's ability to control minor skin and soft tissue infections (Lauderdale et al., 2009). Thus, several groups have consistently reported the robust expression of numerous virulence determinants in USA300 compared with other clinical isolates. It has therefore been hypothesized that this over-production of toxins/proteases confers the selective advantage that explains the overwhelming success of USA300 clones. If true, the regulatory mechanisms explaining these virulence trait expression phenomena are poorly defined.

Agr quorum sensing system

Staphylococcus aureus expresses a peptide-based quorum sensing system known as Agr for Accessory Gene Regulator (Bohach, 2006; Thoendel et al., 2011). Signaling is mediated through a peptide form of AgrD [processed by the combined activity of the AgrB endopeptidase and a type I signal peptidase, SpsB (Kavanaugh et al., 2007)] that stimulates the two-component system sensor kinase, AgrC. The resulting activation of the response regulator AgrA leads to induction of the agrBDCA operon as well as the divergently transcribed RNAIII. While RNAIII encodes δ-toxin, the RNA molecule itself mediates a significant proportion of Agr regulation by affecting the expression of α-toxin (Novick et al., 1993), protein A (Vandenesch et al., 1991), repressor of toxins (Rot) (Geisinger et al., 2006), and others (Vanderpool et al., 2011). Active AgrA is also known to directly control the expression of other virulence determinants including the PSMs (Queck et al., 2008). Thus, the reported overproduction of Hla, Hld, and PSMs in USA300 clones may be explained by a hyperactive Agr system in these clones. Indeed, the RNAIII molecule was shown to be expressed to a higher level in USA300 clones than in other S. aureus isolates explaining the overabundance of δ-hemolysin production (Montgomery et al., 2008; Li et al., 2010). Additionally, the overactive USA300 Agr system was the source of excess PSM and protease production associated with these clones and was partially responsible for excessive Hla expression (Cheung et al., 2011). Consistent with these data, ∆agr mutants in USA300 are highly attenuated in murine sepsis, pneumonia, and skin abscess models (Montgomery et al., 2010; Cheung et al., 2011; Kobayashi et al., 2011). Though, given the importance of Agr in virulence gene regulation, it is not surprising that mutants exhibit such attenuation. Moreover, overproduction of PSMs was reported for USA400 CA-MRSA clones implying that the greater success of USA300 cannot be fully attributed to overactive Agr (Wang et al., 2007; Li et al., 2010). In fact, USA500 clones, thought to be ancestral to USA300, also exhibit phenotypes with hyperactive Agr as well as being highly virulent in murine model infections (Li et al., 2009, 2010). Thus, the high virulence potential of USA300, including high Agr activity, likely evolved in the HA-MRSA clones belonging to USA500. Still, ∆agr mutants of USA300 are highly attenuated and exhibit no increased virulence relative to non-USA300 agr mutants underscoring its importance in the evolution of USA300 (Cheung et al., 2011).

SaeR two-component system

The S. aureus exoprotein expression (Sae) locus contains four genes, saePQRS the latter of which comprise a two-component regulatory system (Giraudo et al., 1994, 1999; Adhikari & Novick, 2008). The response regulator/sensor kinase genes (saeRS) are preceded by genes encoding a membrane protein (SaeQ) and a lipoprotein (SaeP) of unknown function. All four genes are cotranscribed from a promoter that is strongly induced by active SaeR (Geiger et al., 1994). A second promoter drives the expression of saeRS alone and is modestly repressed by these regulatory gene products (Geiger et al., 1994). Activation of the Sae system seems to involve sensing changes in the overall integrity of the cell envelope and is highly stimulated by hydrogen peroxide and cationic peptides including α-defensins (Geiger et al., 1994; Novick & Jiang, 2003). Active SaeR promotes the induction of a number of virulence genes in S. aureus through binding of a consensus sequence found upstream of promoters for hla, sbi, efb, lukS-PVL, splA, and saeP (Nygaard et al., 2010). Additionally, expression of β-hemolysin, fibrinogen-binding proteins, lactose catabolizing enzymes, and the chromosomal arginine deiminase operon are all highly affected by Sae (Voyich et al., 2009). It has been shown that SaeRS expression is higher in USA300 than in USA400 clones (Geiger et al., 1994; Montgomery et al., 2008), which may be a result of overactive Agr system (see above) because RNAIII is known to positively regulate Sae expression (Novick & Jiang, 2003). Deletion of saeRS resulted in almost complete loss of Hla expression and a significant drop in PVL levels as well (Montgomery et al., 2010; Nygaard et al., 2010). Moreover, ∆sae USA300 was attenuated in murine sepsis, peritonitis, dermonecrosis, and pneumonia models (Voyich et al., 2009; Montgomery et al., 2010; Nygaard et al., 2010; Watkins et al., 2011). This was surprising given that in USA400, Sae was only essential for sepsis and peritonitis and not for survival within skin abscesses (Voyich et al., 2009; Watkins et al., 2011). However, USA400 clones do not induce the same level of dermonecrosis and do not express high levels of Hla as in USA300 infections (Montgomery et al., 2008; Li et al., 2010). Thus, it appears as though some of the hypervirulence attributed to USA300 clones in skin/soft tissue infections is likely due to Sae-mediated Hla overproduction. However, HA-MRSA USA500 clones also exhibit severe dermonecrosis during skin infections and overproduce Hla and PSMs yet have not disseminated as widely as USA300.

Source of overactive Agr

While it has not been directly tested, it is tempting to hypothesize that the overactive Agr system inherent to USA300 results in excessive PSMs and Sae expression, the latter of which leads to high Hla expression. However, the mechanism driving high Agr activity in USA300 is not defined. Agr activity can be modulated through the actions of a number of trans-acting regulators including SarA (Cheung & Projan, 1994), Stk1 (Tamber et al., 2010), MgrA (Ingavale et al., 2005), SigB (Lauderdale et al., 2009), CodY (Majerczyk et al., 2008), CcpA (Seidl et al., 2006), Sar-family proteins other than SarA (Schmidt et al., 2001; Manna & Cheung, 2003, 2006; Tamber & Cheung, 2009), ArlRS (Liang et al., 2005), Rsr (Tamber et al., 2010), and SrrAB (Yarwood et al., 2001). Many of these regulators are presumed to affect Agr expression indirectly; however, some [CodY (Majerczyk et al., 2010), SrrA (Pragman et al., 2004) and SarA (Heinrichs et al., 1996)] have been shown to directly bind to the Agr locus. It is intriguing that many of these regulators are involved in modulating metabolic adaptation to various environments (CodY, CcpA, Rsr, and SrrAB) given the apparent increase in fitness associated with USA300 (Herbert et al., 2010) (see below). Though, any one of these or other unknown regulatory systems may be responsible for enhanced Agr activity in USA300; therefore, investigations into strain-specific differences in activity among these regulators may prove enlightening. For instance, SarA positively affects Agr expression (Cheung & Projan, 1994; Reyes et al., 2011), and deletion of sarA in USA300 leads to drastic reductions in Hla and PSM levels (Weiss et al., 2009; Zielinska et al., 2011). However, recently, it was demonstrated that the loss of cytolytic expoprotein expression in the ∆sarA mutant was attributed to the resulting overproduction of extracellular proteases and not because of altered exoprotein gene transcription (Zielinska et al., 2011).

While trans-acting regulators may prove to be major influences on USA300 Agr activity, cis-acting polymorphisms may also be involved. RNAIII transcripts among sequenced ST8 isolates are 100% conserved, but there is a single nucleotide polymorphism (SNP) 3 bp upstream of a known AgrA binding site within the RNAIII promoter that is only found among USA300 isolates. While this is the only SNP among ST8 and ST1 clones specific to USA300, other sites of variation exist when compared to USA100 and USA200 promoter sequences. SNPs in the Hla promoter were recently shown to drive its overexpression in bovine isolates by modulating SarZ binding (Liang et al., 2011). It remains to be determined whether SNPs in the RNAIII promoter region of USA300 isolates affect expression leading to high Agr activity. Regardless of the mechanism behind hyperactive toxin production in USA300, it is important to remember that similar high-level expression is observed in the HA-MRSA progenitor clone, USA500. Thus, while the high virulence potentials of USA300 and USA500 may result from overproduction of exoproteins, this phenomenon alone cannot fully explain the enormous success of USA300 in human disease.

Nonsynonymous mutations in core genes

The evolutionary forces that drive diversification in S. aureus have been recently examined, in part, because of the availability of more than 15 published S. aureus genome sequences. While a significant level of divergence is achieved through acquisition of MGEs, variability within the S. aureus core genome (∼ 2000 orthologous genes shared among most S. aureus strains) is primarily generated through mutation (Feil et al., 2003; Kuhn et al., 2006). The most common forms of mutation are SNPs or short insertion/deletions (indels) that have been estimated to be ∼ 15-fold more attributable to de novo mutation than to recombination (Feil et al., 2003). However, recent reports contend that the contribution of homologous recombination to core diversity in S. aureus may be underestimated (Chan et al., 2011). Nevertheless, mutation is a significant driving force in S. aureus diversification allowing for evolutionary classification of strains into ST types (see above) (Enright et al., 2000). Most SNPs are within coding regions reflecting the fact that ∼ 80% of the core genome encodes protein (Highlander et al., 2007). Synonymous SNPs, those that do not result in amino acid changes, by far outweigh amino acid substituting nonsynonymous SNPs in S. aureus (Herron et al., 2002; Gill et al., 2005; Herron-Olson et al., 2007; Sivaraman & Cole, 2009). This is likely because nonsynonymous mutations are more often detrimental and are therefore subject to evolutionary loss via purifying selection. Consequently, the relative ratio of nonsynonymous to synonymous substitution rate (dN/dS) among staphylococci is generally less than 1. In contrast, a recent report comparing the complete genome sequences of 10 newly isolated USA300 clones with the published FPRF3757 USA300 sequence revealed an unusually high ratio of nonsynonymous : synonymous SNPs (as high as 2.6 : 1, much higher than reported in comparisons of non-USA300 S. aureus lineages) (Kennedy et al., 2008). This discrepancy can be rationalized by assuming a recent clonal expansion of the USA300 lineage such that new isolates still harbor nonsynonymous SNPs that have not yet undergone purifying selection (Holden et al., 2004). To be sure, the unusually high dN/dS ratio of USA300 clones is inconsistent with evolutionary convergence among distantly related clones, an event that would only be consistent with normal to low dN/dS ratios if the converging progenitors were of sufficiently diverse origins (Kennedy et al., 2008).

It is important to note that overall low dN/dS ratios are not necessarily constant across all functional gene families. For instance, while housekeeping and metabolic genes generally exhibit low dN/dS ratios, genes encoding surface associated or secreted proteins can often have elevated dN/dS ratios (Jordan et al., 2002; Rocha & Danchin, 2004). This is indicative of forward selective pressures driving variability in these genes either to promote functional differences (e.g. an adhesin adapting to a host receptor molecule) or immune avoidance through changes in antigenicity. Indeed, comparisons among divergent S. aureus clones reveal higher dN/dS ratios for genes encoding components of the cell envelope and secreted proteins than genes encoding housekeeping or metabolic enzymes (Herron et al., 2002; Herron-Olson et al., 2007; Highlander et al., 2007). USA300 clones, however, seem to be an exception to this rule. A recent comparison of genome sequences from USA200, USA300, and a distantly related S. aureus strain revealed high dN/dS ratios indicative of forward selection in a large number of USA300 metabolic genes (Holt et al., 2011). The largest subset of USA300 genes predicted to be under positive selection (45%) were involved with metabolism, whereas only 7% encoded components of the cell envelope. This phenomenon cannot be explained by the fact that metabolic genes make up a large proportion of the core genome because this same study showed that in USA200, the most prominent class of genes undergoing positive selection were those encoding cell envelope components (a third of all genes with elevated dN/dS) (Sivaraman & Cole, 2009; Holt et al., 2011). An independent study verified that all of the metabolic genes in USA300 exhibiting forward selection were completely conserved among 10 sequenced USA300 genomes (Kennedy et al., 2008). Moreover, data from this same study showed that, while relatively few SNPs were found among 10 different USA300 genomes, genes encoding cell envelope proteins more commonly exhibited high dN/dS ratios (57% of all genes with multiple nonsynonymous substitutions) (Kennedy et al., 2008). Thus, the peculiar overrepresentation of S. aureus metabolic genes among those undergoing positive selection is only evident when comparing USA300 with non-USA300 genomes implying that USA300 clones in general seem to be adapting to disproportionately high selective pressures at the metabolic level.

It is possible that the resulting adaptive mutations in the overall metabolism of USA300 directly contribute to the evolutionary success of this clone. For instance, it has been observed that USA300 clones simply grow faster than any other tested S. aureus isolate (Herbert et al., 2010). Taken together, it would appear that USA300 is more metabolically fit and/or adaptable than other S. aureus lineages. This may provide an advantage when competing for limiting nutrients with endogenous microbial communities as well as contribute to severe disease given a rapid growth rate within sterile sites of the body. Further inspection in our laboratory revealed that USA300 clones have growth advantages when metabolizing many different carbon sources (Table 1). In general, USA300 clones exhibited higher growth rates than other clones when cultivated on nutrients that are abundant in human sweat and skin (Harvey et al., 2010), consistent with the high prevalence of skin/soft tissue infections associated with USA300 clones. But can a relatively small set of amino acid changes in metabolic genes really account for such drastic growth differences? Laboratory adaptation of Escherichia coli to growth on lactate resulted in strains that exhibited nearly twice the growth rate on lactate alone (Hua et al., 2007). These adapted strains exhibited major alterations in metabolic flux capacity through gluconeogenic and pyruvate catabolic pathways, yet none of these changes were because of altered gene expression. This would be consistent with subtle changes in protein sequence (nonsynonymous SNPs) that alter enzyme activity or response to allosteric regulation. Furthermore, a laboratory-adapted clone of Caulobacter crescentus exhibited a ∼ 20% greater growth rate than its progenitor strain and this entire phenotype was explained by a single SNP altering the expression of glucose-6-phosphate dehydrogenase (zwf) (Marks et al., 2010). This enzyme controls the primary flux between energy generating glycolysis and the precursor generating pentose-phosphate pathway (PPP). It was shown that lower flux through PPP with concomitant increased glycolytic activity lead to higher growth rates in laboratory-adapted C. crescentus (Marks et al., 2010). Interestingly, one of the very genes exhibiting signs of positive selection in USA300 was zwf along with two glycolytic genes (pgm and pfkA) potentially linked to the USA300 growth advantage on numerous carbon sources (Holt et al., 2011). Whether or not SNPs within these metabolic genes account for enhanced USA300 growth rates and whether that contributes to the success of this clone remain to be proven; however, the unusual SNP distribution among metabolic genes in USA300 combined with its enhanced growth rate suggest there may be more to USA300 virulence than newly acquired or overexpression of virulence genes.

View this table:
Table 1

Maximal growth rates of Staphylococcus aureus strains on various carbon sources

Glucose1.0% lactate1.0% pyruvate1.0% glycerol1.0% Cas A.A.1.0% tryptone
SF8300 (USA300)0.92 ± 0.020.60 ± 0.010.60 ± 0.010.90 ± 0.030.74 ± 0.020.65 ± 0.02
LAC (USA300)0.89 ± 0.030.57 ± 0.020.57 ± 0.010.89 ± 0.030.72 ± 0.010.63 ± 0.02
Newman (ST8 MSSA) 0.71 ± 0.01 0.43 ± 0.02 0.46 ± 0.02 0.73 ± 0.02 0.43 ± 0.02 0.36 ± 0.01
MW2 (USA400)0.85 ± 0.04 0.41 ± 0.02 0.52 ± 0.02 0.76 ± 0.04 0.75 ± 0.02 0.51 ± 0.05
UAMS-1 (USA200) 0.73 ± 0.02 0.48 ± 0.01 0.52 ± 0.01 0.79 ± 0.01 0.57 ± 0.01 0.68 ± 0.02
  • Rates (µ, h−1) were calculated from at least four independent curves and mean ± SD are reported. Chemically defined medium (Pattee, 1976) was used varying only the primary carbon source. Underlined values indicates rates significantly lower than those of USA300 strain SF8300 (P ≤ 0.05, two-tailed Student's t-test).


The overwhelming success of USA300 in North America as the dominant source of CA-MRSA infections represents a fascinating example of a pathogenic variant emerging as a new threat to human health. The adaptations acquired by USA300 clones in the form of novel genetic components, altered gene regulation, and sequence polymorphisms likely act in concert to provide these strains with a selective advantage. It appears as though USA300 hypervirulence, as assayed in animal models of infection, correlates with increases in virulence gene expression and is apparent in HA-MRSA progenitors as well as other unrelated CA-MRSA lineages. Whether this is because of hyperactive Agr resulting in elevated PSM production and Sae expression (which in turn could lead to excess Hla and other exoprotein excretion) remains to be proven. In contrast to overt virulence, traits that affect transmission and colonization efficiency are inherently difficult to model in the laboratory. It may prove, however, that this aspect of USA300 biology is as critical to its success as is high virulence potential. It remains to be determined whether newly acquired genetic components (e.g. ACME) and/or sequence polymorphisms contribute to the rapid transmission and success of USA300 in the community. In the end, we may appreciate that none of the three evolutionary events (gene acquisitions, altered gene regulation, protein sequence divergence) outlined here can alone explain the success of USA 300. Rather, the amalgamation of all these events created the highly successful pathogen that we must contend with today.


This work was supported by funding from the NIH (AI088158 to A.R.R.)


  • Editor: Nicholas Carbonetti


View Abstract