A preliminary guideline for the assignment of methicillin-resistant Staphylococcus aureus to a Canadian pulsed-field gel electrophoresis epidemic type using spa typing

ethicillin-resistant Staphylococcus aureus (MRSA) has had a dramatic impact on the health care systems of many countries worldwide. With rates of MRSA infections steadily increasing each year, it has become important to quickly distinguish strains of MRSA for the purpose of epidemiological investigation. MRSA isolates are generally characterized by pulsed-field gel electrophoresis (PFGE), which is a proven and very powerful technique for determining the degree of relatedness among isolates (1,2). The Canadian PFGE typing scheme and identification of Canadian ‘epidemic’ strains has allowed comparison with international MRSA clones (1,3). In Canada, a 13-year national surveillance study through the Canadian Nosocomial Infection Surveillance Program (CNISP) has revealed 10 Canadian MRSA (CMRSA1-10) epidemic PFGE fingerprint clusters (1,3). In the early stages of surveillance,

CMRSA1 (resembling USA600) was the most prevalent (1), but it is quickly being replaced by CMRSA2 (resembling USA100/USA800/New York), which accounted for approximately 55% of all strains in 2004 (3).CMRSA3 has virtually disappeared, being replaced by the closely related CMRSA6 (1,3).CMRSA8 is genetically similar to EMRSA15, a common European epidemic strain type (4).PFGE patterns of CMRSA7 and CMRSA10 were indistinguishable from USA400/MW2 and USA300, respectively, which are strain types linked with community-associated MRSA (CA-MRSA) outbreaks in Canada and the United States (5,6).The occurrence of CMRSA7 and CMRSA10 has increased in Canada over the past few years; however, CMRSA10 has become the more prevalent of the two (3).This has also been observed in the United States, where USA300 is more prevalent than USA400 (7).
Despite the excellent discriminatory power of PFGE, there are several disadvantages of using this technique, such as long wait times to obtain final results, a high degree of labour intensity, lack of standardization between laboratories and the inability to easily develop an MRSA typing database due to the subjective interpretation of results associated with gel-based systems (8,9).The desire to quickly and easily type MRSA led to the development of spa typing, a DNA sequencing technique that examines the polymorphic X or short-sequence repeat (SSR) region of the protein A gene (spa) (8).The SSR region consists of a variable number of short-sequence repeats, usually 24 base pairs long, and is located immediately upstream of the region encoding the C-terminal cell wall attachment sequence.Changes in the SSR region arise due to deletions, duplications and point mutations, which results in a diverse collection of 'spa types', where each spa type consists of a specific combination of SSRs (10).Advantages of DNA sequencing as a method for strain typing include speed, ease of use, unambiguous data interpretation and the ability to easily develop an MRSA typing database.
The present study examines the validity and feasibility of using spa typing for national surveillance by comparing PFGE and spa types for 1488 CMRSA isolates obtained from across Canada over a 13-year time period.Results from the present study were used to create a preliminary guideline for Canadian laboratories interested in using spa typing as a technically less demanding method to identify CMRSA isolates.

Bacterial isolates
A total of 1488 CMRSA isolates were selected from a collection of over 14,000 MRSA isolates obtained through the CNISP (1) and through routine services performed at the National Microbiology Laboratory (Winnipeg, Manitoba).Isolates were chosen on the basis of unique PFGE patterns that were associated with one of the 10 CMRSA strains (3).When possible, multiple isolates of the same epidemic PFGE pattern were chosen by varying geographical distribution and/or time of isolation.The samples were collected over a 13-year time span, between 1995 and 2007, and from across Canada (245 from British Columbia, 191 from Alberta, 83 from Saskatchewan, 119 from Manitoba, 471 from Ontario, 189 from Quebec, 78 from Newfoundland, 48 from New Brunswick, 50 from Nova Scotia and 14 from Prince Edward Island).spa sequencing DNA was prepared using a glass bead DNA extraction method.Briefly, a 1 μL loop was used to collect colonies from an overnight culture grown on Tryptic Soy agar plates containing 5% sheep blood, and then emulsified in a microfuge tube containing 600 μL neutralization buffer (500 mL stock solution [15 mL 1 M Tris-HCl pH 8.4, 2 mL 0.5 M EDTA pH 9 and 483 mL deionized distilled water]) and 50 μL of cell disruption media (0.1 mM) (Scientific Industries Inc, USA).The samples were heated for approximately 3 min at 95°C to 100°C in a heating block, followed by disruption of cells for 3 min using a Genie vortex adaptor (Ambion, USA).The samples were then centrifuged at 3000 rpm for 1 min.Five microlitres of the supernatant was used as the template for the polymerase chain reactions (PCRs).PCR amplification of the spa repeat region was performed as previously described using primers, spa-1113f (5'-TAA AGA CGA TCC TTC GGT GAG C-3') and spa-1514r (5'-CAG CAG TAG TGC CGT TTG CTT-3') (10).For individual PCRs, amplicons were purified using YM-100 Microcon filtres (Millipore, USA).For high throughput, Agencourt AMPure Reagent (Agencourt Bioscience Corporation, USA) was used to purify amplicons in a 96-well format.Purified amplicons were sequenced in-house by the DNA core facility (National Microbiology Laboratory).The DNA sequences of the spa repeat region in both directions were imported as ABI files and analyzed using the spa typing program provided with BioNumerics version 4.6 (Applied Maths Inc, USA).DNA sequences were compared using the spa typing Web sites <http://tools.egenomics.com/public/login.aspx> and <http://www.spaserver.ridom.de>, the latter of which was developed by Ridom GmbH (Germany) and curated by SeqNet.org<http://www.SeqNet.org/>(10).
Cluster analysis was performed using the spa typing module of BioNumerics version 4.6.The default cost matrix was used to correct for the evolutionary distances between the repeats.Cluster analysis settings were set to 400% gap creation cost, 70% gap extension cost, 25% duplicate creation cost, 25% duplicate extension and a maximum duplication length of three repeats.The minimal spanning tree was calculated from the distance matrix using the default distance bin size of 100%.A total of 36 isolates had repeat successions (successions less than five), and were excluded from the cluster analysis as previously discussed (11).PFGE PFGE was performed as previously described (12).CMRSA1-10 were determined using guidelines as previously described (1,3).

spa typing
In total, 1488 CMRSA isolates representing 802 unique PFGE patterns (Figure 1) were chosen for spa typing in the present study.Multiple isolates (range two to 119) from 256 indistinguishable PFGE patterns were chosen by varying geographical distribution and/or time of isolation, when possible (refer to Table S1 provided online <www.pulsus.com>).
For the 1488 CMRSA isolates tested, spa typing revealed a total of 114 unique spa types averaging eight to nine repeats (Table 1).Comparison of spa types obtained from multiple isolates of 256 indistinguishable PFGE patterns revealed that 167 were associated with only one spa type.The remaining isolates from 89 indistinguishable PFGE patterns contained anywhere between two and five different spa types (Table S1).
Cluster analysis, based on the degree of similarity between spa repeat successions, was performed for 1452 isolates using the spa minimal spanning tree application of BioNumerics version 4.6.This analysis revealed six spa clonal complexes (spaCC), which were arbitrarily named spaCC1-6 (Figure 2).spaCC1 contained only isolates with PFGE epidemic fingerprint types CMRSA1, spaCC2 contained only isolates with PFGE epidemic fingerprint types CMRSA2, spaCC3 contained only isolates with PFGE epidemic fingerprint types CMRSA7 and spaCC4 contained only isolates with PFGE epidemic fingerprint types CMRSA8.The remaining two clusters consisted of a mixture of PFGE types, which included isolates of PFGE epidemic fingerprint types CMRSA3, 4, and 6 in spaCC5 and CMRSA5, 9 and 10 in spaCC6 (Figure 1).
Despite the inability to clearly differentiate all PFGE epidemic types using spa clustering (Figure 2), the majority of individual spa types within these clusters (110 of 114) were only associated with a single PFGE epidemic type, which was even noted for spaCC4 (Table 1).On the basis of these results, a table is provided to be used as a preliminary guideline for spa typing of CMRSA epidemic isolates (Table 1).

Validation of the spa guideline
The first 300 MRSA isolates received in 2007 through the CNISP were used to validate the spa typing guideline.Resolved spa types and PFGE banding patterns were independently assigned CMRSA epidemic types and were then compared (refer to Table S2 provided online <www.pulsus.com>).Of the 300 isolates, 285 were assigned Canadian epidemic types (CMRSA1-10) using spa typing and the guideline alone.Of the 15 isolates that were not assigned epidemic types, eight contained spa types that were not listed in the guideline, but based on similarities of repeat successions; Panton-Valentine leukocidin (PVL) results were assigned to their corresponding PFGE epidemic types (CMRSA10 [t068, t1767, t2067, t3135 and t3154], CMRSA2 [t688] and CMRSA7 [t2593 and t555]) (Tables 1 and S2).Six isolates did not cluster with any of the 10 Canadian epidemic spa types, which included four isolates with spa types t216, t186 and t901.The other two were spa type t437, which is associated with a ST59 pandemic CA-MRSA strain (also referred to as USA1000/WA MRSA15), and one with t044, which is associated with a ST80 pandemic CA-MRSA strain (also referred to as the 'European CA-MRSA clone') (Table S2).The last nonassigned epidemic isolate was spa type t008 and was PVL-negative and, therefore, required PFGE for designation.
Of the same 300 isolates, 283 were assigned Canadian epidemic types (CMRSA1-10) using PFGE.PFGE was consistent for the six non-CMRSA epidemic isolates predicted using spa typing (Table S2).However, the PFGE banding pattern for the remaining 11 isolates clustered within a PFGE epidemic type, but they were not assigned to that epidemic cluster because they differed by seven or more bands from the epidemic type strain.These 11 nonepidemic-assigned isolates clustered within the same PFGE epidemic fingerprint types that were predicted using spa typing (Table S2).

Cost and time comparison of spa and PFGE
To compare the costs and required time for either spa typing or PFGE, a number of assumptions had to be made.For PFGE, the availability of two PFGE units for MRSA was assumed and for spa typing, in-house capability of DNA sequencing was assumed.Using these assumptions, the cost and time required to process 48 MRSA isolates were compared.For spa typing, the time required for DNA extraction, PCR, amplicon cleanup, DNA sequencing and the analysis of 48 isolates was estimated to be approximately 13 h.Whereas for PFGE, the time required to prepare and digest plugs, load and run gels, stain and destain gels, and the analysis of 48 isolates was estimated to be approximately 58 h.Estimated 'hands-on' technical times were approximately 6 h for spa typing and 12 h for PFGE.The estimated cost for spa typing of a single isolate was approximately between $9.50 and $12.00.Nonaccess to in-house DNA sequencing would obviously add to the cost and turnaround time of spa typing, but could be worth the increased objectivity of the analysis.The estimated cost of PFGE for a single isolate was approximately between $9.50 and $10.00.The following estimates will vary depending on laboratory capabilities and availability of equipment, but were used as baselines to compare the two methods.

Continued on 278
revealed that spa types were not confined to a specific PFGE banding pattern, and that indistinguishable PFGE patterns could contain anywhere from one to at least five different spa types.As a result, spa typing could not reliably be used as an indicator of a specific PFGE fingerprint type.However, closer examination of the repeat successions of the spa region for these isolates were similar (Table 1) and, therefore, a less discriminatory approach involving the clustering of related spa and PFGE epidemic types was examined.Analysis of the clustering of related spa types revealed that two of the spa clonal clusters contained multiple PFGE epidemic types and, therefore, could not be used for the specific classification of PFGE epidemic types.The clustering of PFGE epidemic types CMRSA5, 9 and 10 in spaCC6 (Figure 2) was not surprising because a dendrogram generated from the PFGE fingerprint types of these strains indicated that they were closely related (Figure 1).Secondly, all three PFGE epidemic types, CMRSA5, 9 and 10, have the same multilocus sequence typing (MLST) designation (ST8) (Table 2).However, the clustering of PFGE epidemic types CMRSA4 with CMRSA3 and CMRSA6 in spaCC5 was initially surprising (Figure 2).
*Refer to discussion section for assignment of spa types t008, t030, t037 and t275 to a CMRSA epidemic type

Number of
Figure 2) Minimal spanning tree analysis for the spa types of 1452 Canadian methicillin-resistant Staphylococcus aureus (CMRSA) isolates (types 1 to 10).Clusters of related spa types were arbitrarily assigned clonal complexes (spaCC1-6) and are indicated in bold.Pulsedfield gel electrophoresis epidemic types (CMRSA1-10) associated with each spaCC is also indicated CMRSA3 and CMRSA6 are genetically similar by both MLST (Table 2) and PFGE (Figure 1), and would therefore be expected to be closely related.However, CMRSA4 is distantly related to both CMRSA3 and CMRSA6 when comparing these epidemic types in the PFGE dendrogram (Figure 1).In addition, CMRSA4 belongs to a different MLST CC in comparison with CMRSA3 and CMRSA6 (Table 2).Previously observed 'violations' of MLST ST assignment by spa typing have been attributed to intergenomic recombination involving the small region of the chromosome that spa typing focuses on (13).This was previously shown for lineages belonging to MLST CC8 and CC30 (14), and likely explains the clustering of spa types obtained from CMRSA4 with CMRSA3 and CMRSA6.As a result, isolates identified as unrelated by MLST, could be seen as related by spa typing because they would both share the same repeats in the polymorphic X region of the protein A gene.It is also interesting to note that CMRSA3 and CMRSA6 do not cluster with any of the MLST CC8 and CC30 lineages, thereby suggesting that PFGE is also unable to correctly classify recombinants from multiple genetic backgrounds.These violations cannot be resolved by spa typing alone, but are observed when both spa typing and MLST analyses are used.Despite the inability to assign epidemic types on the basis of spa clustering analysis, small genetic differences within the spa region still enabled us to differentiate a CMRSA4 isolate from a CMRSA3 or a CMRSA6 isolate, which could be the result of micro-and/or macroevolution of this small chromosomal region over time following the recombination event (15).

Assignment of Canadian epidemic types using spa typing
In the present study, and many others (13,15,16), spa typing displayed excellent concordance with PFGE fingerprint clusters.For the four spa types, t008, t030, t037 and t275, where multiple PFGE epidemic types were found, additional molecular typing would be required for classification.For instance, spa type t008 was associated with three different PFGE epidemic types of CMRSA (CMRSA5, CMRSA9 and CMRSA10).However, of these three epidemic types, only CMRSA10 is known to carry PVL.Therefore, for isolates with spa type t008, an additional PCR reaction is proposed to detect the presence or absence of PVL.A PVL-positive result for a spa type t008 isolate would likely differentiate CMRSA10 from CMRSA5 and CMRSA9 isolates and could, therefore, be reported as CMRSA10.For PVL-negative strains, a PFGE is proposed for identification of CMRSA5, CMRSA9 or a PVL-negative CMRSA10 isolate.To date, CMRSA5 and CMRSA9 strains containing PVL have not been reported, and PVL-negative CMRSA10 isolates in Canada appear to be rare and should, therefore, not result in a large number of samples requiring PFGE.
Although not examined in the present study, additional PCR-based methods for delineating different strains of the same spa type could also include the use of staphylococcal cassette chromosome mec typing (17) or the arginine catabolic mobile element (18), alone or in combination with PVL, as an indicator of CMRSA10 strains.
spa type t037 was associated with three different PFGE epidemic types of CMRSA (CMRSA3, CMRSA4 and CMR-SA6); however, the occurrence of spa type t037 in CMRSA4 was only seen once and might represent a rare recombination event.Both CMRSA3 and CMRSA6 are staphylococcal cassette chromosome mec type III (3), but MLST could be useful in differentiating spa type t037 isolates.However, MLST requires DNA sequencing of seven loci, which is an expensive and time consuming method.PFGE is, therefore, recommended to differentiate MRSA isolates of spa type t037.It should be noted that CMRSA3 has dramatically declined in Canada, and it is, therefore, expected that the majority of t037 isolates would be CMRSA6 (3).PFGE would also be recommended for spa types t030 and t275, which were both seen in two different epidemic PFGE types.To date, these spa types appear to be rare in Canada, having only been seen in four of the 1488 isolates tested, and would, therefore, not be expected to result in large numbers of samples requiring PFGE for identification.
The assignment of CMRSA epidemic types using the provided spa guideline (Table 1) was highly successful for the molecular classification of the first 300 MRSA isolates received in 2007 through the CNISP.The high discriminatory ability of spa typing has been attributed to the small size of repeats, which are more prone to duplication and deletion via slip-strand mispairing, and a high number of synonymous substitutions per synonymous site (dS=0.72)(15).These slow point mutations and fast-occurring changes in the number of repeats enables spa  typing to be effectively used in both long-and short-term epidemiologic studies (15).It should be noted that the use of spa typing in an outbreak setting was not assessed in the present study, but has previously been shown to be successful in outbreak investigations (8,10) and early warning systems for the detection of outbreaks (19).It is, therefore, suggested that the combination of PFGE and spa typing would be useful for obtaining additional comparative data in such settings and will be important for the continued national surveillance and reporting of MRSA in Canada.For instance, spa typing of the first 300 CNISP MRSA isolates received in 2007 revealed that performing PFGE alone potentially misclassified 11 isolates (3.6%) as nonepidemic (Table S2).This has identified a potential limitation of the past definition of epidemic CMRSA isolates in Canada, which is the assignment of epidemic types based primarily on comparisons of PFGE banding patterns with one representative epidemic type strain.spa typing of additional nonepidemic-assigned isolates is, therefore, warranted to examine possible alternative definitions of epidemic MRSA types in Canada.
International comparison of spa types spa types are freely comparable using international databases such as Ridom (http://www.spaserver.ridom.de)(10), which currently contains over 3300 spa types that were obtained from the submission of over 47,000 isolates from 45 different countries.Briefly, spa types of MRSA strains defined as epidemic in Canada were also prevalent around the world (Table 2).Comparison of spa types obtained from the present study to the Ridom database also revealed two spa types, t044 and t437, associated with two pandemic CA-MRSA strains (ST80 'European' and ST59 clones) that have not previously been reported in Canada.The potential emergence of these two CA-MRSA clones in Canada should continue to be monitored.

SUMMARY
The high discriminatory power of PFGE provides an excellent tool for studying short-term outbreaks, and was supported in the present study and others as a useful method for long-term epidemiological surveillance studies.However, PFGE is labourintensive and technically demanding.The dependence on strict adherence to standardization protocols between laboratories and the subjective interpretation of PFGE results has led to an inability to easily develop MRSA typing databases on a large scale.In comparison with PFGE, spa typing was capable of yielding results faster, enabled the processing of more isolates, did not require any type of subjective interpretation and was comparable in cost with PFGE.The spa data were also easily exportable, and could be uploaded rapidly into databases for 'real-time' surveillance of MRSA on a national or international scale.The provided guideline for the assignment of Canadian epidemic PFGE types using spa typing is preliminary and, therefore, a periodically updated table, including newly defined spa types, will be made accessible on-line <www.nml.ca> or available on request.
spa typing of Canadian epidemic MRSA Can J Infect Dis Med Microbiol Vol 19 No 4 July/August 2008 279

TABLE 1
Preliminary guideline for the assignment of Canadian epidemic types using spa typing (see also TableS1<www.pulsus.com>).spa types indicated in bold were identified as Canadian methicillin-resistant Staphylococcus aureus (CMRSA) pulsed-field gel electrophoresis (PFGE) epidemic types from the 300 isolates used for validation from the 2007 Canadian Nosocomial Infection Surveillance Program and were subsequently added to the table (see also Table S2 <www.pulsus.com>)

TABLE 1 -
CONTINUEDPreliminary guideline for the assignment of Canadian epidemic types using spa typing (see also TableS1<www.pulsus.com>).spa types indicated in bold were identified as Canadian methicillin-resistant Staphylococcus aureus (CMRSA) pulsed-field gel electrophoresis (PFGE) epidemic types from the 300 isolates used for validation from the 2007 Canadian Nosocomial Infection Surveillance Program and were subsequently added to the table (see also Table S2 <www.pulsus.com>)

TABLE 1 -
CONTINUEDPreliminary guideline for the assignment of Canadian epidemic types using spa typing (see also TableS1<www.pulsus.com>).spa types indicated in bold were identified as Canadian methicillin-resistant Staphylococcus aureus (CMRSA) pulsed-field gel electrophoresis (PFGE) epidemic types from the 300 isolates used for validation from the 2007 Canadian Nosocomial Infection Surveillance Program and were subsequently added to the table (see also Table S2 <www.pulsus.com>)