Global Distribution and Evolution of Panton- Valentine Leukocidin-Positive Methicillin-Susceptible Staphylococcus aureus, 1981–2007

Abstract

Background. Panton-Valentine leukocidin (PVL)-positive methicillin-susceptible Staphylococcus aureus and methicillin-resistant S. aureus (MSSA and MRSA, respectively) are both associated with severe infections, such as necrotizing pneumonia. The epidemiological profile of PVL-positive community-acquired (CA) MRSA has been extensively studied, but few corresponding data on PVL-positive MSSA are available.

Objectives. The objectives of the study were to investigate the global population structure of PVL-positive MSSA, to compare it with that reported for CA-MRSA, and thus to examine the phylogenetic relationship between these pathogens.

Methods. We determined the agr types, multilocus sequence types, and toxin gene profiles of 211 PVL-positive MSSA clinical isolates collected in 19 countries throughout the world between 1981 and 2007.

Results. The predominant lineages of PVL-positive MSSA were agr3/ST30, agr4/ST121, agr3/ST1, agr2/ST5, and agr3/ST80. Except for agr4/ST121, these lineages are also reported to be prevalent among CA-MRSA. PVL-positive MSSA lineages that are genetically related to CA-MRSA have gradually replaced other lineages (especially agr4/ST121) over the past 2 decades. Within a given sequence type, the toxin gene content of PVL-positive MSSA strains was very similar to that of PVL-positive CA-MRSA.

Conclusions. The molecular epidemiological profiles of PVL-positive MSSA and CA-MRSA are dynamically interrelated, with the former appearing to constitute a reservoir for the latter.

Staphylococcus aureus is a major human pathogen that causes diseases ranging from minor skin and soft-tissue infections (SSTIs) to life-threatening pneumonia and toxin-mediated diseases, such as toxic shock syndrome [1]. S. aureus is the most frequently occurring pathogen in hospitals and the second most common pathogen in outpatient settings [2]. S. aureus SSTIs and severe deep-seated infections, such as necrotizing pneumonia, are frequently associated with strains harboring the lukPV locus encoding the Panton-Valentine leukocidin (PVL). PVL, a pore-forming protein exotoxin, can recruit polymorphonuclear cells and monocytes and trigger their apoptosis and lysis [3, 4]. Both methicillinsusceptible S. aureus (MSSA) and community-acquired methicillin-resistant S. aureus (CA-MRSA) can express PVL [5]. It is unclear why lukPV is frequently associated with the methicillin resistance locus in community isolates, especially because these transferable elements integrate distinct parts of the genome [4].

Epidemiological studies of PVL-positive S. aureus (mainly MRSA) have revealed a highly clonal population structure, and the worrisome recent upsurge in PVL-positive CA-MRSA is due to a limited number of pandemic lineages. The most prevalent genotypes, designated either by their multilocus sequence type or by their pulsotype, are ST80 in Europe, ST8-USA300 in the United States, and ST30 (Southwest Pacific clone), which is ubiquitous [6–8]. Less-prevalent CA-MRSA lineages include ST1-USA400, ST5-USA100, and ST59-USA1000 [6, 9, 10].

PVL-positive MSSA strains appear to belong to more-diverse genetic backgrounds than their methicillin-resistant counterparts [11]. For example, Monecke et al [12], using DNA microarray analysis, found that 30 PVL-positive MSSA strains belonged to 8 different sequence types (STs). Interestingly, some PVL-positive MSSA lineages are closely related to epidemicassociated CA-MRSA lineages. This association raises the possibility that PVL-positive CA-MRSA arises from preexisting MSSA lineages by methicillin resistance gene transfer, although one cannot exclude the transfer of PVL genes in some healthcare- associated MRSA genetic backgrounds. Thus, the success of CA-MRSA as a pathogen could be due not only to its antibiotic resistance but, also, to other factors related to its genetic background. However, epidemiological data on PVL-positive MSSA are sparse, and most relevant studies have involved limited geographic areas, such as a single town or healthcare center, and/or short time frames [12, 13].

The aim of the present study was to investigate the population structure and distribution of PVL-positive MSSA, as well as their evolution over time. The study period covered the emergence and spread of CA-MRSA. A total of 211 PVL-positive MSSA isolates were collected from 19 countries throughout the world over 26 years (from 1981 through 2007). The isolates were characterized by means of multilocus sequence typing (MLST) and spa typing, and they were screened for 19 virulence genes. Population diversity was assessed using the Simpson diversity index. The population structure of PVL-positive MSSA was then compared with that reported for CA-MRSA, to assess the phylogenetic relationship between these closely related pathogens.

Materials And Methods

Bacterial isolates. A total of 211 PVL-positive MSSA clinical isolates were selected from the strain collection of the French National Reference Center for Staphylococci. These isolates were voluntarily sent to the center, for toxin gene profiling, by microbiologists throughout the world. The selected strains were collected in 19 countries spanning 5 continents (Table 1). The study lasted from 1981 through 2007. The strains were associated with a wide variety of diseases, comprising SSTIs (n = 85 [40.3%]), necrotizing pneumonia (n = 71 [33.6%]), bone and joint infections (n = 12 [5.7%]), other infections (n = 12 [5.7%]), and unknown infections (n = 31 [14.7%]). Each strain was isolated from a different patient.

Table 1

Geographic Origins and Isolation Periods of 211 Panton-Valentine Leukocidin-Positive Methicillin-Susceptible Staphylococcus aureus Isolates

DNA extraction, mecA testing, and identification of agr alleles. Genomic DNA was extracted using a standard procedure [14]. gyrA amplification was used to confirm the quality of each DNA extract and the absence of polymerase chain reaction (PCR) inhibitors. agr alleles and the absence of the mecA methicillin resistance gene were identified by PCR, as described elsewhere [14, 15].

Detection of toxin genes. The isolates were screened by PCR for 19 specific staphylococcal virulence genes, as described elsewhere [14], comprising sequences specific for enterotoxins (sea-d, seh, sek, sel, and sep-r), the enterotoxin gene cluster (egc), toxic shock syndrome toxin-1 (tst), exfoliative toxins (eta, etb, and etd), PVL (lukS-PV and lukF-PV), class F lukM leukocidin (lukM), epidermal cell differentiation inhibitor (edinA/ B/C), and β-hemolysin (hlb). This panel of genes was selected because it represents well-characterized virulence factors that are not uniformly distributed in the S. aureus population [14].

MLST. All the isolates were studied by MLST, as described elsewhere [16]. The allelic profiles were determined by sequencing 500-bp internal fragments of 7 housekeeping genes (arcC, aroE, glpF, gmk, pta, tpi, and yqiL). Nucleotide sequences were entered on the MLST home page [17], where 7 numbers depicting the allelic profile were assigned to define the STs. STs that differed by only one locus were grouped together into clonal complexes (CCs) by use of eBURST software (version 3) [18, 19]. To examine possible changes in ST distribution over time, we studied subsets of isolates collected in France during 4 different periods: 1981–1990 (n = 16 isolates), 1991–2000 (n=26), 2001–2005 (n=45), and 2006–2007 (n = 18). This analysis was restricted to French isolates, to limit the bias resulting from variations over time in the inclusion of samples from other countries.

spa typing. All the isolates were also spa-typed, as described elsewhere [20]. The repeat region of the protein A gene (spa) was amplified by PCR and sequenced. spa types were determined using Ridom Staph Type software (version 1.5; Ridom), which automatically detects spa repeats and assigns a spa type according to Harmsen et al [21, 22]. spa types were clustered into CCs (ie, spaCCs) by use of the integrated BURP (Based Upon Repeat Patterns) algorithm [23, 24]. User-definable parameters were set as follows: “cluster spa types into spaCC if cost distances are less than or equal to 4” and “exclude spa types shorter than 5 repeats”. This parameter combination ensures optimal concordance (95.3%) between BURP and e- BURST [25].

Antimicrobial susceptibility testing and detection of remnant SCCmec elements in ST80 isolates. Most ST80 CAMRSA isolates exhibit a particular resistance profile, with resistance to β-lactams, kanamycin, tetracycline, and fusidic acid [6]. We determined the resistance profiles of all ST80 PVLpositive MSSA isolates and searched for similarities between their profile and that of their CA-MRSA counterparts. Susceptibility to penicillin, oxacillin, cefoxitin, kanamycin, tobramycin, gentamicin, erythromycin, clindamycin, tetracycline, pristinamycin, ofloxacin, fusidic acid, vancomycin, teicoplanin, fosfomycin, trimethoprim/sulfamethoxazole, rifampin, mupirocin, quinupristin/dalfopristin, and linezolid was determined with the standard agar diffusion technique, as recommended by the French Society for Microbiology [26].

Because of the similarities observed between PVL-positive ST80 MSSA and CA-MRSA with respect to their susceptibility patterns, we raised the hypothesis that certain ST80 MSSA isolates were descended from ST80 CA-MRSA after SCCmec excision. The Infectio Diagnostic Incorporated (IDI)-MRSA PCR assay (BD Geneohm) was applied to all ST80 isolates, as described elsewhere [27, 28]. This assay detects SCCmec elements remaining in MSSA after partial excision of SCCmec. It uses 5 primers that are specific for the different SCCmec sequences located downstream of the mecA gene, plus a primer specific for the S. aureus chromosomal gene orfX. This conserved open-reading frame flanks the SCCmec integration site attB. SCCmec-specific amplicons were detected with 3 molecular beacon probes.

To identify the SCCmec type of the remnant SCCmec elements, all ST80 PVL-positive MSSA isolates were screened for SCCmec markers—namely, ccrA, ccrB, dcs region, ΔmecR, mecR, mecI, and ugpQ—by use of DNA array (CLONDIAG Chip Technologies), as described elsewhere [29].

Population diversity assessment. Genetic diversity was estimated using the modified Simpson index λ = 1 - ∑p²₁ where p_i is the proportion of isolates belonging to the ith genotype, as defined by MLST [30, 31]. The λ value can be described as the probability that 2 isolates chosen at random will have different genotypes, and it thus can be expressed as a percentage [32]. λ increases with the number of genotypes and, also, with more-balanced genotype distributions. Approximate 95% confidence intervals (CIs) were calculated as described elsewhere [32]. The λ value was calculated for the PVL-positive MSSA population, by comparison with (1) a set of CA-MRSA strains characterized by the French National Reference Center for Staphylococci [6] and selected for their similar clinical and geographic origins, and (2) the global S. aureus population, based on the MLST database [17] available online on 22 January 2009 (which included 2314 isolates belonging to 919 different STs). To optimize the comparison of PVL-positive MSSA and CA-MRSA, diversity was calculated for a subset of 118 PVLpositive MSSA isolates collected during the same period as the CA-MRSA panel (1999 through 2005).

Results

agr types, STs, and CCs. The 211 PVL-positive MSSA isolates were agr1 (n = 61 [28.9%]), agr2 (n = 18 [8.5%]), agr3 (n = 89 [42.2%]), or agr4 (n = 43 [20.4%]) (Table 2). They belonged to 28 STs. eBURST analysis identified 3 CCs—namely, CC1 (ST1, ST567, and ST1278), CC22 (ST22 and ST217), and CC30 (ST30 and ST37). The remaining 21 STs were singletons. The most frequent STs were ST30 (n = 42 [19.9%]), ST121 (n = 42 [19.9%]), ST1 (n = 20 [9.5%]), ST25 (n = 15 [7.1%]), ST5 (n = 14 [6.6%]), and ST80 (n = 14 [6.6%]). The STs were strictly specific for agr types.

Table 2

Distribution of Panton-Valentine Leukocidin-Positive Methicillin-Susceptible Staphylococcus aureus Isolates, According to agr Type, Sequence Type (ST), and spa Type

Twelve (42.9%) of the 28 STs—namely, ST1 (CC1), ST5, ST8, ST22 (CC22), ST25, ST30 (CC30), ST37 (CC30), ST59, ST80, ST88, ST152, and ST398—had been attributed to CAMRSA in other studies, whereas 4 other STs—namely, ST567 (CC1), ST1278 (CC1), ST217 (CC22), and ST776 (CC30)—belonged to CA-MRSA CCs [6, 7, 9, 33–43]. Thus, 16 STs (57.1%) were shared by PVL-positive MSSA and CA-MRSA. To our knowledge, no CA-MRSA belonging to the other 12 STs (42.9%; ie, ST15, ST45, ST121, ST188, ST291, ST446, ST669, ST918, ST946, ST1279, ST1280, and ST1290) have been identified to date. These latter 12 STs accounted for 67 (31.8%) of the 211 PVL-positive MSSA isolates, and nearly two-thirds of these isolates (43 [64.2%]) had the agr4 (ST121 and ST946) genetic background. The proportion of PVL-positive MSSA isolates that were unrelated to CA-MRSA decreased from 56.3% before 1990 to 27.6% after 2005 (Figure 1).

Figure 1

Distribution of community-acquired methicillin-resistant Staphylococcus aureus (CA-MRSA)-related and -unrelated lineages among Panton- Valentine leukocidin-positive methicillin-susceptible S. aureus (MSSA) (n = 211), as well as changes over time. CA-MRSA-related lineages were accessory gene regulator (agr)1/clonal complex (CC) 22, sequence type (ST) 25, ST152, ST8, ST59, and ST398; agr2/ST5; and agr3/CC30, CC01, ST80, ST88, and ST776. CA-MRSA-unrelated lineages were mostly agr4/CC121, an agr group in which no MRSA strains have been described to date. The other CA-MRSA-unrelated lineages were agr1/ST188, ST291, ST45, ST669, ST1280, and 1290; agr2/ST15 and ST1279; and agr3/ST446 and ST918

The ST distribution of the French PVL-positive MSSA isolates changed drastically from 1981 through 2007 (Figure 2). ST80 strains, none of which were found before 2000, accounted for 15.6% of isolates collected during 2001–2005, and ST80 became the most frequent ST (27.8%) identified during 2006–2007. Conversely, ST121 accounted for 56.3% of isolates collected before 1990 but for only 30.8% of isolates in 1991–2000, 13.3% in 2001–2005, and 0% after 2005.

Figure 2

Changes in the distribution of Panton-Valentine leukocidin-positive methicillin-susceptible Staphylococcus aureus (MSSA) in France, 1981–2007. Sequence types (STs) designated as “others” were, in order of decreasing frequency, ST5, ST217, ST15, ST188, ST291, ST8, ST22, ST37, ST59, ST398, ST669, ST918, ST1278, and ST1279.

spa types and spaCCs. The 211 isolates belonged to 96 different spa types, corresponding to 14 spaCCs and 13 singletons (14% of spa types) (Table 2). Seven spa types (8% of spa types; 3.8% of all isolates) were excluded from BURP analysis because they consisted of <5 repeats. The most frequent spaCCs were spaCC21 (MLST CC30) (n = 44 [20.9%]), spaCC159 (mostly ST121) (n = 29 [13.7%]), spaCC127 (MLST CC01) (n = 21 [10.0%]), and spaCC44 (mostly ST80) (n = 19 [9.0%]). Of 15 STs comprising >1 isolate, 14 encompassed >1 spa type. spa types and spaCCs were mostly specific for agr types and STs. Four spa types (t044, t131, t645, and t1710) were found in >1 ST. t645 (agr4/ST121 and agr4/ST946) and t1710 (agr3/ST446 and agr3/ST918) were present in STs that were not totally independent (double-locus variants). In contrast, t044 (agr3/ST80 and agr1/ST152) and t131 (agr3/ST80 and agr1/ST1290) were present in totally unrelated lineages.

Geographic distribution. Most of the dominant lineages were widely distributed (Table 2). ST30, ST121, and ST1 were pandemic. The most prevalent lineages were, in decreasing frequency, ST30, ST121, and ST80 in Europe and North Africa; ST121, ST5, and ST152 in sub-Saharan Africa; ST121, ST30, and ST25 in North America; and ST30, ST5, ST25, and ST1 in South America.

Toxin genes. Most isolates (n = 193 [91.5%]) carried ⩾1 virulence factor other than pvl (Table 3). ST188, ST398, ST1279, and ST1280 were the only lineages that did not carry any of the toxin genes sought, apart from pvl. The most prevalent toxin locus in PVL-positive MSSA, egc, was carried by 61.1% of isolates. egc was specific for certain lineages: all isolates belonging to CC30, CC121, ST5, ST25, and CC22 carried egc, whereas no isolates belonging to CC1, ST80, ST152, ST188, and ST8 carried egc. All ST80 isolates carried etd and edin and no other toxin genes, except for 1 isolate that also carried ser. Apart from egc, which is highly prevalent in the general S. aureus population [44], the prevalence of the toxin genes ranged from 0.5% (eta and etb) to 24.6% (edin).

Table 3

Toxin Gene Content of Dominant Lineages of Panton-Valentine Leukocidin-Positive Methicillin-Susceptible Staphylococcus aureus

Antimicrobial resistance and SCCmec remnants in ST80 MSSA isolates. Of the 14 ST80 MSSA isolates, 6 (42.9%) were resistant to penicillin, 1 (7.1%) was resistant to kanamycin, 2 (14.3%) were resistant to erythromycin, 8 (57.1%) were resistant to tetracycline, and 6 (42.9%) were resistant to fusidic acid. The most common phenotype (4 strains [28.6%]) included resistance to penicillin, tetracycline, and fusidic acid. This phenotype resembles that of most ST80 CA-MRSA strains, which are resistant to β-lactams, kanamycin, tetracycline, and fusidic acid [6].

Two ST80 PVL-positive MSSA isolates (14.3%) were positive in the IDI-MRSA assay and carried remnant SCCmec elements. DNA array-based characterization of these strains showed that 1 strain carried the dcs region, ccrA-2, and ccrB-2, whereas the other strain carried only the dcs region. The dcs region is found in SCCmec types I, II, and IV. ccrA-2 and ccrB-2 are found in SCCmec types II and IV [29]. It has been postulated that such strains carrying SCCmec remnants may have arisen fromMRSA having undergone partial SCCmec excision [45]. Each strain was resistant to only one antibiotic class (erythromycin and kanamycin, respectively).

Population diversity. The Simpson diversity index was high for PVL-positive MSSA, with λ = 88.9%(95% CI, 86.8%–90.9%) for the 1981–2007 collection and λ = 89.2% (95% CI, 86.8%–91.7%) for the 1999–2005 collection. On the basis of published data, the index was much lower (λ = 41.0% [95% CI, 35.4%–46.7%]) for the 1999–2005 CA-MRSA collection [6]. The index calculated for the entire S. aureus MLST database was λ =98.0% (95% CI, 97.8%–98.2%).

Discussion

Several interesting findings emerge from this study of 211 PVL-positive MSSA strains isolated on 5 continents over nearly 3 decades. First, PVL-positive MSSA was more diverse than CAMRSA: the modified Simpson diversity index λ value for PVLpositive MSSA (89.2%) was 2.2 times higher than that calculated for CA-MRSA (41.0%). This finding indicates that PVLencoding phages are able to integrate a wide range of unrelated genetic backgrounds. Second, with the exception of ST8, the predominant CA-MRSA lineages reported elsewhere (ST80, ST30, ST8, ST1, and ST5 [6, 7]) were all represented among the predominant lineages of PVL-positive MSSA observed here. Third, among the PVL-positive MSSA lineages, those that were genetically related to CA-MRSA have gradually replaced those unrelated to CA-MRSA (especially agr4/ST121) over the past 2 decades. These findings suggest that the molecular epidemiologies of PVL-positive MSSA and CA-MRSA are dynamically interrelated, with the former acting as a reservoir for the latter.

Several predominant genetic backgrounds were found in our collection of PVL-positive MSSA. agr3/ST30 and agr4/ST121 were found to be pandemic, and each accounted for 19.9% of the PVL-positive MSSA isolates (Table 2q). Among the other prevalent lineages, ST1, ST5, ST25, and ST152 were also pandemic, whereas ST80 was found only in Europe and Africa, and ST188 was found only in France, New Caledonia, and Polynesia. However, the observed distribution of PVL-positive MSSA is not fully representative of the current global epidemiology, because of the long inclusion period and the sources of the isolates received by the French National Reference Center for Staphylococci. The low prevalence of ST8 strains in our study might be related to the limited number of strains collected from the United States.

A global analysis of the whole S. aureus MLST database showed that MRSA strains were only represented among 215 (23.4%) of the 919 STs. This finding is consistent with the report by Enright et al [46], who found that MRSA strains were represented in only 38 (23.5%) of the 162 S. aureus genetic backgrounds described in 2002. This suggests that only a few S. aureus lineages are capable of acquiring SCCmec [46]. In contrast, most of the PVL-positive MSSA lineages studied here were genetically related to CA-MRSA (16 [57.1%] of 28 STs). We can thus speculate that PVL phages integrate preferentially into SCCmec-capable lineages. Further investigations are needed to address this question.

Most of the PVL-positive MSSA strains studied here that were not genetically related to CA-MRSA had the agr4 background. As far as we know, this background has never been reported in MRSA and thus seems to be incompatible with stable integration of SCCmec elements.

ST80 is the main PVL-positive CA-MRSA lineage in Europe, accounting for 76.1% of CA-MRSA isolates sent to the French National Reference Center for Staphylococci from 1999 through 2005 [6]. It started to emerge in the early 2000s. We found that ST80 PVL-positive MSSA emerged and spread during the same period: ST80 PVL-positive MSSA started to emerge in France between 2001 and 2005, and it was the most frequent PVL-positive MSSA lineage after 2005. The spa types, non-β-lactam resistance profiles, and toxin gene content of ST80 PVL-positive MSSA were highly analogous to those of ST80 PVL-positive MRSA [6]. Moreover, 14.3% of ST80 MSSA isolates carried SCCmec remnants, indicating that they may correspond to former CA-MRSA strains that have undergone partial SCCmec excision. Together, these findings suggest that ST80 PVL-positive S. aureus is a successful lineage in which SCCmec acquisition and excision have occurred several times.

spa typing with BURP clustering can be used to group S. aureus isolates into clonal complexes and to infer their phylogenetic relationships. The inferred phylogeny is consistent with that obtained by MLST and eBURST clustering, with the concordance between these two methods being 95% [25]. However, the same spa sequence is sometimes found among phylogenetically unrelated lineages [47], possibly because of repeated convergent evolution (homoplasy) of spa sequences rather than because of recombination events. The spa gene product, protein A, is a secreted virulence factor involved in host immune system evasion [48]. Because homoplasy results from evolutionary convergence toward the best-adapted phenotype in response to a particular selection pressure, homoplastic spa sequences are likely to confer an evolutionary benefit on their host. In the present study, 2 spa types (t044 and t131) appeared to be homoplastic, because they were each found in 2 independent lineages (t044 in agr3/ST80 and agr1/ST152, and t131 in agr3/ST80 and agr1/ST1290). Interestingly, these closely related spa types are the most prevalent spa types in ST80 PVL-positive MSSA and CA-MRSA (Table 2) [6]. The putative evolutionary benefit associated with such spa loci might contribute to the success of this lineage. Additional studies are needed to determine whether homoplastic spa loci are associated with increased fitness or infectivity.

In conclusion, we found that most predominant genetic backgrounds of PVL-positive MSSA are pandemic and are phylogenetically related to CA-MRSA. In France, the historical ST121 PVL-positive MSSA genetic background has been gradually replaced over the past decade by MSSA lineages related to CA-MRSA (mostly ST80 PVL-positive MSSA). Because PVLpositive MSSA is still the most common cause of severe PVLassociated infections in many countries and is also the most plausible reservoir of CA-MRSA, epidemiological surveillance and infection control efforts must be reinforced.

Acknowledgments

We thank the bacteriologists throughout the world who sent us Panton- Valentine leukocidin (PVL)-positive methicillin-susceptible Staphylococcus aureus isolates. We also thank C. Courtier, C. Gardon, C. Bouveyron, and C. Spinelli for technical help; O. Dumitrescu and O. Dauwalder for helpful advice; B. Pangon for performing the IDI-MRSA assay; and David D. Young for editorial assistance. This publication made use of the Multi Locus Sequence Typing Web site [17] at Imperial College London, developed by David Aanensen and funded by the Wellcome Trust.

Potential conflicts of interest: none reported.

Financial support: French Ministry of Health; French Ministry of Education; Institut National de la Santé et de la Recherche Médicale.