Epidemiological typing uses highly discriminatory genetic markers that characterize human pathogens allowing the identification of isolates that are closely related due to recent common ancestry. The spa locus of S. aureus codes for protein A, a species-specific gene product known for its IgG binding capacity. This locus is highly polymorphic due to an internal variable region of short tandem repeats, which vary not only in numbers but also because of nucleotide substitutions within individual repeat units [11]. DNA sequences of the spa gene therefore provide portable, unambiguous, and biologically meaningful molecular typing data, which have demonstrated their utility for epidemiological purposes such as transmission and outbreak investigations at various geographical levels [12],[13].

During three workshops organised for technical personnel from 28 European SRLs, participants received hands-on training in spa sequence typing and spa sequence analysis according to a standard protocol using a purpose-designed software tool, StaphType, developed by Ridom GmbH [13]. Proficiency testing was carried out by mailing each SRL five well-characterized S. aureus isolates and five sequence chromatograms (trace files) of known spa types as described previously [14],[15]. All laboratories participating in the structured survey described here fulfilled quantifiable quality criteria, which consisted of an unambiguous base-calling of all sequenced nucleotides for both forward and reverse sequencing runs of the test panel.

A protocol was drawn up and circulated for comment to all SRLs and agreed upon in April 2006. Following this protocol, European SRLs were asked to identify approximately 20 laboratories that serve hospitals and that are geographically and demographically representative of their country, and secure their participation. These laboratories were chosen from those that regularly participate in the European Antimicrobial Resistance Surveillance System (EARSS). For 6 mo, from September 2006 until February 2007, these participants were asked to submit the first five successive MSSA isolates and the first five successive MRSA isolates from individual patients with invasive infection. An invasive infection was defined as a localised or systemic inflammatory response to the presence of S. aureus at otherwise sterile anatomical sites. Isolates were dispatched by the participating laboratories to the SRLs accompanied by additional information, including the EARSS laboratory identifier, the sample number, the date of isolation, origin of clinical specimen, demographic detail (such as age and gender), epidemiological context (hospital-acquired when symptoms developed more than 48 h after admission or as community-onset otherwise), isoxazolylpenicillin- (i.e., oxacillin) or cefoxitin-resistance, and all-cause mortality 14 d after isolation of the first invasive isolate. SRLs confirmed MRSA by mecA PCR or determination of minimum inhibitory concentration for oxacillin together with PBP2a agglutination. Additional data were uploaded to the database and Web application if available. These consisted of staphylococcal cassette chromosome mec (SCCmec) typing, and identification of luk-PV genes, indicative of Panton-Valentine Leukocidin (PVL). All SRLs preserved the isolates in strain collections and performed spa sequence typing according to the standard protocol, uploaded the sequence information, and made this available by synchronisation with the central Ridom SpaServer (www.spaserver.ridom.de) curated by SeqNet.org at the Institute of Hygiene, University Hospital Münster, Germany [15],[16]. Submitted sequences were quality controlled by comparison with accompanying chromatograms (trace files) and excluded if stringent quality criteria for excellent sequencing data were not fulfilled. spa types were grouped into spa complexes if a single genetic event such as single insertions, single deletions, or single nucleotide polymorphism could account for the observed sequence divergence. In the following the designation of spa types indicated by lower case “t” follow the nomenclature used by the spa server and multilocus sequence types are reported as sequence type (ST) according to convention [17]. Finally, the SCCmec type is also added to a string consisting of spa type/ST/SCCmec all in parenthesis, e.g., (t032/ST22/SCCmecIV).

Epidemiological and typing data were communicated in parallel to a central SQL database at the National Institute for Public Health and the Environment (RIVM) of the Netherlands. For each participating laboratory, SRLs also provided the postal address and indicated their administrative region (such as region, province, state, department, or NUTS-2 level) if the location of laboratories were to be aggregated on a higher administrative geographical level for display on the interactive mapping tool (which was done only for Austria, Belgium, Czech Republic, and Poland). All data were anonymous and collected in accordance with the European Parliament and Council decision for the epidemiological surveillance and control of communicable disease in the European community [18],[19]. Ethical approval and informed consent were thus not required.

All data were inspected for inconsistencies and analysed on a country-by-country basis and returned to SRLs for feedback, clarification of inconsistencies and final approval in February 2008. After final approval, data were analysed using Stata version 10 and SAS version 9.1 (SAS Institute Inc.) using chi-square test for proportions and Student's t-test for continuous variables. The index of diversity is an unbiased measure of the probability of drawing two different spa types given the distribution of spa types in the sample. The 95% confidence intervals (CIs) were calculated as described previously [20]. Postal address and location of all sampling laboratories were converted into decimal cartesian coordinates using the geocoding facility at www.spatialepidemiology.net [21]. Pairwise distances of laboratories that isolated MSSA and MRSA with identical spa types were calculated and distance matrices were summarised and compared by nonparametric tests. The Web application SRL-Maps (http://www.spatialepidemiology.net/srl-maps) was developed to interrogate the data on the basis of mapping of laboratory locations and on centroids of administrative regions (when laboratory results were aggregated at the level of administrative region).

A spatial scan statistic was used to identify the geographic distribution of specific spa types at two levels: (i) country-specific frequencies that take into account all spa types within national boundaries and (ii) regional clusters of varying size independent of national boundaries. To identify spa types with higher than expected occurrence in any of the participating countries, the observed number of all spa types isolated within each country was compared with the number expected under the assumption of a European-wide random distribution. For the identification of regional clusters, circular windows around all laboratory locations were projected. For each location, the radius of the window was varied from 0 to 1,000 km. In this way, a finite number of distinct windows was created. For each window, the observed number of isolates with a specific spa type was compared with the expected number under the assumption of a random distribution. 10,000 random distributions were obtained by varying the composition of local spa types at all laboratory locations consistent with cumulative spa-specific frequencies using Monte Carlo Simulations. The test statistic was calculated by log-likelihood ratio test, whereby countries or windows where the observed spa type frequencies differed from those obtained by simulation were considered to contain significant clusters with an alpha error of p<0.0001. The scan-statistic was executed with SaTScan software using SAS macros [22],[23].

26 SRLs from 26 countries satisfactorily fulfilled the proficiency criteria for spa sequence typing and contributed data for final analysis. Between September 2006 and February 2007, 357 laboratories serving 450 hospitals (Figure 1) collected 2,890 successive MSSA and MRSA isolates from patients with invasive S. aureus infection (2,603 from blood cultures, 90.1%; 17 from cerebrospinal fluid, 0.6%; and 270 from puncture fluids of other normally sterile anatomical sites, 9.3%). Final inspection of data revealed missing information for gender (28 isolates, 1%), age (54 isolates, 1.9%), sampling dates (74 isolates, 2.6%), epidemiological context (community onset or hospital acquisition, 568 isolates 19.7%), all-cause mortality 14 d after S. aureus isolation (1052 isolates, 36.4%), and spa type (40 isolates, 1.4%). Table 1 gives a summary overview of the number of participating laboratories, isolates, and spa types submitted by country. Many laboratories were unable to collect five invasive MRSA isolates within the sampling period because of a low MRSA incidence in the hospitals they serve. Therefore, the combined collection consisted of two-thirds MSSA (1,923; 66.5%) and one-third MRSA (967; 33.5%, Table 2). Patients with invasive MRSA infections were older (Figure 2) with a median age of 69 y compared to a median age of 63 y in MSSA patients (p<0.001). Moreover, MRSA patients had higher all-cause mortality (20.8%) 14 d after the first isolation of S. aureus than MSSA patients (13.2%, p<0.0001). More males (1,765/2,862; 61.7%) than females had invasive S. aureus infections. The proportion of MRSA compared to MSSA did not differ between the sexes (p = 0.2). Of the 231 MRSA that were reported as community-onset (CO-MRSA), 226 (97.8%) were tested for the presence of PVL-associated genes luk-PV and ten (4.4%) were positive. Of the 585 hospital-acquired MRSA (HA-MRSA), 551 (94.2%) were tested for PVL and six (1.1%) were positive. The difference was significant (p = 0.009).

A total of 660 spa types were reported (Table 2). Among all spa types, 565 and 155 were assigned to MSSA and MRSA, respectively, of which, 505 were exclusively identified as MSSA and 95 for MRSA alone. 60 spa types contained both MSSA and MRSA. 27 of the MSSA (1.4%) and 13 of the MRSA (1.3%) isolates were nontypeable. Table 3 shows the rank order of the most frequent spa types isolated during the survey and Table 4 the three most frequent spa types by country. MRSA was less diverse than MSSA. Only five spa types accounted for almost half (48.1%) of all MRSA isolates, whereas the same proportion of MSSA isolates comprised the 26 most frequent MSSA spa types. Estimates of the genetic diversity differed significantly with an index of diversity for MSSA of 0.985 (95% CI 0.983–0.987) and 0.940 (95% CI 0.933–0.947) for MRSA. While MSSA diversity ranged between 0.934 in Latvia and 1.0 in Iceland (unpublished data), MRSA diversity between countries was more heterogeneous ranging from 0.62 in Romania to 0.91 in Austria (Figure 3), indicating the presence of few dominant MRSA spa types in several countries. Accordingly, MRSA showed a higher degree of geographic clustering as the average distance between laboratories that isolated the same spa type was significantly smaller than for MSSA (Table 2). No correlation between genetic diversity of MRSA and overall proportion of MRSA among S. aureus blood stream infections at country level as reported to the EARSS database for 2007 was found (r = −0.09, p = 0.75) [24], indicating that single successful spa types cannot explain the variance in the proportion of MRSA causing S. aureus blood stream infections observed in Europe.

In 17 countries, 22 spa types were found with frequencies that were unexpected when applying the hypothesis of a random distribution, indicative of local epidemics (Table 5). Most (86.9%) of these were MRSA. In ten countries two spa types coexisted with unexpected frequencies. In four of them, these two spa types showed a close genetic relationship and belonged to the same spa complex whereby a single genetic event could account for the sequence divergence between the types. There was also a frequent regional coincidence with neighbouring countries sharing identical epidemic spa types. The Czech Republic and Germany shared spa type t003 (t003/ST225/SCCmecII), Bulgaria and Romania shared t030 (t030/ST239/SCCmecIII), the UK and Ireland t032 (t032/ST22/SCCmecIV), Italy and Croatia shared t041 (t041/ST228/SCCmecI), and strain t067 (t067/ST5125/SCCmecIV) whose dominance in Spain was first identified through this initiative [25], was also found in southern France. The notion of regional spread was supported by the cluster statistic that projected windows beyond national boundaries for this dataset (Table 6). The degree of unexpectedness, which is an indication of the significance of each cluster, is expressed by the log likelihood ratio (LLR). The majority of regional clusters extended beyond national boundaries and 74% of all isolates that occurred in these clusters were MRSA. The most significant cluster was identified in Spain and consisted of spa type t067 (t067/ST5&125/SCCmecIV). A northern Balkan/Adriatic cluster consisting of spa type t041 (t041/ST228/SCCmecI) was found in Austria, Hungary, Slovenia, Croatia, and northern and central Italy. In Britain and Ireland, t032 (t032/ST22/SCCmecIV), known as epidemic MRSA 15 (EMRSA-15), was the dominant type and represented the third most significant cluster. An additional cluster of spa type t032, albeit less significant and much smaller, was located in the Brandenburg area of Germany. Central Germany, the Czech Republic, and western Poland were included in a large regional cluster of spa type t003 (t003/ST225/SCCmecII), which was geographically centred in Saxony and had a radius of approximately 400 km corresponding to the German hospitals participating in the study. Figure 4 provides a geographical illustration of these clusters. The largest cluster in size as well as in numbers (radius 930 km, 119 isolates) consisted of spa type t008 and was centred in southern France. This cluster consisted of ST8 and contained different subclones as it included both MSSA and MRSA, and MRSA isolates exhibited two different SCCmec elements (SCCmecI and IV). A smaller cluster, ranking in sixth position in terms of significance, was located in Flanders on the Belgian-Dutch border and consisted of spa type t740 (t740/ST45/SCCmecIV). Interestingly, regional spa clusters with overlapping geographical range were frequently made up of spa types that belonged to the same spa complex, a clear indication that local spread is accompanied by local evolution of the rapidly evolving spa locus. Clusters with the smallest size (0 km) included those submitted by single laboratories most likely reflecting single hospital outbreaks. Three regional clusters consisted of MSSA alone. They were located in Latvia (t435/ST425), Poland (t127/ST1), and Denmark (t230/ST45), indicating that regional spread of S. aureus is not limited to MRSA alone.

The Web application SRL-Maps (http://www.spatialepidemiology.net/srl-maps) provides a community tool for the interrogation of the geographic distribution of different spa types. All laboratory and regional locations across Europe are represented as placemarks on a Google map. Clicking on a placemark displays, below the map, all spa types identified at that location (and their frequency) along with the number of isolates (and number of geographic locations) found elsewhere (if any) for each of these spa types (Figure 5). The European distribution of any spa type can then be displayed and placemarks on the map are colour-coded on the basis of whether the isolates at each location are all MSSA (green), all MRSA (red), or are a mix of MSSA and MRSA (yellow), with the number of isolates inside the placemark. Graphical charts are displayed showing spa type-specific proportion of MRSA/MSSA, all-cause mortality at 14 d, and the age distribution among cases (Figure 5). This method allows the identification and mapping of strains with particular public health importance and further exploration by the scientific community is encouraged.

For both MSSA and MRSA isolates, there was no association between specific spa types and all-cause mortality after 14 d, indicating that no spa type was associated with hyper-virulence. Of the ten CO-MRSA isolates that were found to be PVL-positive, three were assigned to spa type t044 (t044/ST80/SCCmecIV), the so-called European community-acquired (CA)–MRSA, and another three had spa type t008 (t008/ST8/SCCmecIV) and are indistinguishable from USA300 CA-MRSA. Of the four other PVL-positive CO-MRSA, two belonged to t622 (spa complex 8/ST8/SCCmecIV), one to t529 (ST1043/SCCmecV), and one to t437 (not further characterised). MRSA belonging to ST398 have recently emerged in several European countries and are regarded as being livestock-associated (LA-MRSA). Of all spa types typically associated with this clone, spa types t011, t034, t571, t1255, and t2383 were identified on 12 occasions (1.3%) in eight countries during this survey. None of these isolates, however, displayed an MRSA phenotype or harboured the mecA gene.