|Home | About | Journals | Submit | Contact Us | Français|
This study evaluates the multiple-locus variable-number tandem-repeat assay (MLVA) and pulsed-field gel electrophoresis (PFGE) when using restriction enzymes BstZI, SacII, and ApaI to fingerprint a diverse collection of methicillin (meticillin)-resistant Staphylococcus aureus (MRSA) sequence type 398 (ST398) isolates. These isolates had been characterized previously by multilocus sequence typing, spa typing, and staphylococcal cassette chromosome mec (SCCmec) typing. Typeability and discriminatory power were analyzed, and the concordance between the various methods was determined. All MRSA ST398 isolates were typeable by the MLVA and PFGE using BstZI, SacII, and ApaI. With each method, the MRSA ST398 isolates formed a separate group from the two non-ST398 MRSA strains. PFGE, performed with any of the three restriction enzymes, had the most discriminatory power, followed by MLVA, spa typing, and SCCmec typing. The MLVA showed the highest concordance with PFGE using ApaI and spa typing. As further expressed by the Wallace coefficient, the MLVA type was poorly predicted by spa typing, whereas the spa type was well predicted by MLVA. PFGE, using a combination of all three restriction enzymes, had the highest concordance with the MLVA but had a low probability of being predicted by MLVA. PFGE, using a combination of all three restriction enzymes, was able to predict SCCmec type and MLVA type completely and had a high probability of predicting spa type. Both the MLVA and PFGE could be used to discriminate among the MRSA ST398 isolates. Although the MLVA is a faster technique, PFGE had more discriminatory power than the MLVA, especially when a combination of restriction enzymes was used.
Infections caused by methicillin (meticillin)-resistant Staphylococcus aureus (MRSA) have been a problem in hospitals and nursing homes for many decades. These MRSA isolates are therefore called health care-associated MRSA (HA-MRSA) isolates (1). Since the early 1990s, MRSA has emerged in healthy persons without risk factors for MRSA infections. These isolates are the so-called community-associated MRSA (CA-MRSA) isolates (20). In the last few years, MRSA has been isolated from livestock animals (pigs in particular) and pig farmers (5, 6, 30). These MRSA strains are called animal-associated MRSA (AA-MRSA) strains. It seems that (livestock) animals form a new, separate reservoir. These AA-MRSA strains all appear to belong to the new clonal complex 398 (CC398), with sequence type 398 (ST398) as the basic type, as determined by multilocus sequence typing (MLST) (29). MRSA ST398 has already been isolated in Europe, Asia, and North America (32). Considering the worldwide spread of MRSA, epidemiological questions arise about its transmission within farms, among farms, and from farms to the population. Fast and inexpensive typing methods with good discriminatory power are necessary to conduct large-scale epidemiological studies.
Traditionally, human MRSA isolates have been typed by pulsed-field gel electrophoresis (PFGE), using SmaI as the restriction enzyme (19). The advantages of using PFGE are good discriminatory power and good reproducibility at the interlaboratory level when standardized protocols are used. However, AA-MRSA is not typeable by this method, as the activity of SmaI is blocked due to methylation of the restriction site (2).
More recently, methods based on DNA sequencing, such as MLST and spa typing, are increasingly being used to discriminate among different MRSA strains. Given their excellent interlaboratory reproducibility, online databases have been made to collate and harmonize data from various geographic regions. The drawback of MLST, which measures sequence variation at seven housekeeping loci, is its limited use with epidemiological studies due to its weak discriminatory power, time-consuming protocols, and high costs. spa typing, based on the variation in repeats present in the X-region of staphylococcal protein A, has a discriminatory power that lies between those of PFGE and MLST. Within ST398, several spa types have been distinguished, although the number of spa types seems rather limited in most countries.
One promising method is the multiple-locus variable-number tandem-repeat assay (MLVA), a PCR-based method, based on the analysis of the number of repeats in the variable-number tandem-repeat regions of various individual genes. This method has proven to be useful for typing both Staphylococcus aureus and clinical MRSA isolates with good reproducibility and good discriminatory power. Because the MLVA is also simple, inexpensive, and easy to interpret, it is useful as a typing method for large-scale epidemiological studies (10, 11, 12, 15, 16, 17, 23, 24, 27).
This study aimed to investigate various methods for typing MRSA ST398 isolates. An MLVA, consisting of a selection of primers from three existing MLVA systems, was tested with a collection of MRSA ST398 isolates. In addition, PFGE with restriction enzymes other than SmaI was performed with this set of isolates. These isolates had been previously characterized by MLST, spa typing, and staphylococcal cassette chromosome mec (SCCmec) typing. Typeability and discriminatory power were analyzed for all methods, and the concordance among the different methods was determined.
This study used 34 MRSA strains (Table (Table1).1). MLST typing revealed that 32 isolates belong to MRSA ST398 (8; http://saureus.mlst.net). The isolates had various spa types (www.ridom.de/staphtype), SCCmec types, and origins (18, 21, 33). The strains were isolated from humans who were in contact with pigs, chickens, cows, and the farm environment. Twenty-one strains were isolated in Belgium during studies performed in 2007 and 2008 (7, 22, 31). Another 11 strains originated in The Netherlands (5, 28). The strains ATCC 33592 and ATCC 43300 were included as non-ST398 MRSA reference strains.
Chromosomal DNA was prepared by suspending a few colonies from a tryptic soy agar plate in water containing 10% lysostaphin (1 mg/ml). After incubation for 10 min at 37°C and centrifugation for 2 min at 14,000 × g, DNA was prepared further according the protocol of Flamm et al. (9).
First, the primer pairs listed in Table Table22 were used separately in order to evaluate the typeability and discriminating power of each primer pair on the collection of 34 MRSA isolates. Each PCR mixture contained 1× PCR buffer, 1.3 mM MgCl2, 2.5 U AmpliTaq DNA polymerase (Applied Biosystems, Foster City, CA), 200 μM of each deoxynucleoside triphosphate, 1.5 μM of a primer pair, and 2 μl of template DNA (25 ng/μl) in a final volume of 25 μl. PCR was performed with predenaturation at 94°C for 5 min, followed by 25 cycles of 30 s at 94°C, 30 s at 58°C, and 60 s at 72°C, with a final extension at 72°C for 5 min on a GeneAmp PCR system 9700 (Applied Biosystems, Foster City, CA). Ten microliters of PCR product was used for electrophoresis in 2% SeaKem LE agarose with 1× Tris-borate-EDTA (TBE; pH 8) for 60 min at 100 V.
Only primers that discriminated among the 32 MRSA ST398 isolates were included in the final MRSA ST398 MLVA. This multiplex PCR was performed under the same conditions described above. Eighteen microliters of PCR product was electrophoresed in 1.5% SeaKem LE agarose with 1× TBE (pH 8) for 240 min at 140 V. The DNA size standard was a 100-bp DNA ladder (Invitrogen, Carlsbad, CA). The gels were stained in ethidium bromide, visualized on an UV transilluminator, and digitally photographed. Gel patterns were analyzed with BioNumerics version 5.10 (Applied Maths, Sint-Martens-Latem, Belgium) and clustered using Dice's coefficient (tolerance, 1%; optimization, 0.1% to 1%) and the unweighted-pair group method using average linkages (UPGMA).
The plugs were prepared according the protocol of Struelens et al. (26), with modifications. Briefly, an overnight-grown culture was suspended in EET buffer (100 mM EDTA, 10 mM EGTA, 10 mM Tris-HCl [pH 8]) and adjusted to an optical density at 600 nm value of 0.9. The suspension was mixed with equal volumes of a 2% solution of low-melting-temperature agarose (InCert agarose; Cambrex Bio Science Rockland, Rockland, MA) in EET buffer. After solidification, the agarose plugs were incubated for 4 h at 37°C in EET buffer containing 1 mg lysozyme and 50 μg lysostaphin per ml. The plugs were transferred into EET buffer containing 1% sodium dodecyl sulfate and 1 mg proteinase K per ml of buffer and incubated overnight at 50°C. Afterwards, the plugs were washed thoroughly with TE buffer (10 mM Tris-HCl, 1 mM EDTA [pH 8]).
The restriction enzyme chosen was isoschizomer XmaI, because it is a methylation-insensitive version of SmaI. XmaI cuts the same recognition sequence as SmaI, but at a different position. The other restriction enzymes were chosen based on the restriction digest and PFGE program provided at http://insilico.ehu.es/ (3). DNA was digested overnight with 40 U SmaI (Invitrogen, Carlsbad, CA), SacII, XmaI (New England Biolabs, Ipswich, MA), ApaI, or BstZI (Promega, Madison, WI) at the recommended assay temperatures. DNA fragments were separated on a CHEF Mapper system (Bio-Rad Laboratories, Hemel Hempstead, United Kingdom) in a 1% SeaKem gold agarose gel (Lonza, Rockland, MA). The running conditions were 6 V/cm in 0.5× TBE (45 mM Tris, 45 mM boric acid, 1 mM EDTA [pH 8]) at 14°C. For separation of SmaI and XmaI DNA fragments, runs lasted 18 h, with switch times from 2.63 s to 63.8 s. ApaI fragments were separated during runs of 16 h, with switch times of 0.22 s to 17.33 s. DNA fragments digested with BstZI and SacII were separated during a run of 20 h, with an initial switch time of 0.46 s and a final switch time of 35.38 s for BstZI and a switch time of 44.69 s for SacII. Gels were stained with ethidium bromide, destained in water, and digitally captured under UV light. Gel images were visually analyzed with BioNumerics version 5.10 (Applied Maths, Sint-Martens-Latem, Belgium) using Salmonella Braenderup H9812 digested with XbaI as a normalization reference. The similarity between the fingerprints was calculated using the Dice coefficient (with an optimization of 1% and a position tolerance of 0.5% to 1%), and they were grouped together according to their similarities using UPGMA.
An MLVA type was assigned on the basis of a difference of at least one band. As this method relies on the number of the tandem repeats in the variable-number tandem-repeat regions of different individual genes, even a small shift in the position of one band was considered to show another MLVA type. MLVA types are indicated by capital letters, as shown below.
For each restriction enzyme, a corresponding pulsotype was assigned, which was based on the difference in the presence or absence of at least one band. This was indicated by the name of the restriction enzyme followed by a number (e.g., BstZI-1).
The discriminatory power (D) was calculated according to the formula described by Hunter and Gaston (13, 14). The concordance of different typing techniques was calculated with the software described by Carriço et al. (4). Rand's index is an index that shows the proportion of agreement, whereas the adjusted Rand's index (R) shows the proportion of agreement corrected for the presence of chance agreement. Wallace's coefficient (W) indicates the probability that two strains classified as the same type by one method will also be classified as the same one when using the other method.
For these calculations, isolates originating from the same farm and having the same spa type, SCCmec type, MLVA type, and pulsotype were considered to be the same strain and were excluded from further analysis.
All primer pairs generated one band for at least one of the two reference strains, except the primers for sdr, which generated multiple bands. Data for the number of types for the 32 MRSA ST398 isolates and the discriminatory power of each primer pair are given in Table Table2.2. The primer pairs SIRU5 and SIRU16 generated no amplicons for the MRSA ST398 isolates. Primer pairs sspA, fnbAB, cna, SIRU1, SIRU7, SIRU13, and SIRU15 did not discriminate among the MRSA ST398 isolates. The ClfA and ClfB primer pairs generated three and five different amplicons, respectively, although the difference in size for clfA was minimal. Both primer pairs spa and SIRU21 had good discriminating power: both found six types among the tested MRSA ST398 isolates. For each isolate, primer pair SIRU21 consistently generated a band that was approximately 50 bp bigger than the bands generated by the spa primers. The locus of SIRU21 has also been identified as spa (12). Last to be evaluated were the two primer pairs sdrCDE and sdrCDE2, used to detect the polymorphism of three individual genes, namely sdrC, sdrD, and sdrE. Staphylococcus aureus strains do not always possess all three of these genes (25). The 32 MRSA ST398 strains possess only the sdrC and sdrE genes, as tested in this study according to the protocol of Sabat et al. (25) (data not shown). Given this information, only two amplicons for each isolate can be expected when using these two primer pairs. Only the sdrCDE2 primer pair generated two bands for each MRSA ST398 isolate, and five different types were obtained.
The following primer pairs were chosen for their discriminatory power and included in the ST398 MLVA: SIRU21, ClfA, ClfB, and sdrCDE2. The best results were obtained by running the sdrCDE2 primer set separately from the other three primer sets and mixing 10 μl of the sdrCDE2 PCR product with 8 μl of the multiplex PCR product before electrophoresis.
All 34 isolates were typeable using the MLVA described above. The MRSA ST398 isolates clustered together with a similarity of 33%. As some differences were so small, it was not possible to determine a cutoff value in order to automatically categorize the types with the BioNumerics software. For example, MB4393 and RIVM10 were 100% related according to the BioNumerics software, but it was possible to visually distinguish a small shift due to a smaller clfB band. Using visual categorization, 10 different MLVA types were obtained within the ST398 isolates (Fig. (Fig.1).1). Six MLVA types consisted of only one isolate. Two isolates belonged to MLVA type A, but these two isolates originated from the same farm. MLVA type H consisted of five isolates, all isolated in The Netherlands, whereas MLVA types B and C consisted of 13 and 6 isolates, respectively. These two last types were very similar; the only difference was a small shift of the clfA band.
None of the AA-MRSA isolates was typeable by SmaI digestion, which confirmed this characteristic as being typical for MRSA CC398. However, they were all typeable using the restriction enzymes BstZI, SacII, and ApaI. XmaI, an alternative to SmaI, as mentioned above, was also evaluated. However, the fingerprints obtained yielded several weak bands, even after optimization of the restriction conditions. Because of this, XmaI was excluded from the study.
A cutoff value of 97% for delineation of the different pulsotypes was determined for each restriction enzyme, according to the criteria of the delineation of pulsotypes as described above in Materials and Methods.
The fingerprints of the MRSA ST398 isolates obtained from PFGE using restriction enzyme BstZI clustered together with a similarity of 79%, whereas the two reference non-ST398 MRSA isolates had a similarity with the MRSA ST398 isolates of less than 57%. Based on the cutoff value of 97%, 13 BstZI pulsotypes were determined within the MRSA ST398 isolates (Fig. (Fig.22).
Using SacII, the AA-MRSA isolates clustered together with a similarity of 76%, whereas the two reference strains were an outgroup, with a similarity of 61% with the MRSA ST398 isolates (Fig. (Fig.3).3). Thirteen SacII pulsotypes were distinguished among the AA-MRSA isolates, with a cutoff value of 97%. The ST398 cluster could be further subdivided into two major groups (cutoff value of 80%), one group of isolates with SCCmec type V, and another group of isolates with SCCmec type IVa. The same isolates grouped together with the restriction enzyme SacII as those that grouped with restriction enzyme BstZI, with the exception of three pulsotypes. Pulsotype Sac-9 contained pulsotypes BstZI-8 and BstZI-11; thus, in this case, BstZI was more discriminating. In contrast, restriction enzyme SacII was more discriminating for pulsotypes Sac-6 and Sac-7, which shared the same fingerprints when using restriction enzyme BstZI.
The AA-MRSA isolates clustered together with a similarity of 71% when using restriction enzyme ApaI (Fig. (Fig.4).4). Again, the two reference strains had a similarity of 62% or less with the MRSA ST398 cluster. As with restriction enzyme SacII, the ST398 cluster could be further subdivided into two groups (cutoff value of 80%), separating isolates carrying SCCmec type IVa and those carrying SCCmec type V. With the cutoff value of 97%, 13 different Apa pulsotypes were distinguished. However, there was only little correlation between the pulsotypes obtained by BstZI or SacII and the pulsotypes obtained by ApaI. For example, pulsotype Apa-2 contained four different BstZI/SacII pulsotypes. On the other hand, isolates belonging to pulsotype BstZI-8 or Sac-9 were further divided into three pulsotypes using ApaI.
The discriminatory indices (DI) of the methods, and those of the combination of the methods, are shown in Table Table3.3. The discriminatory power of SCCmec typing, spa typing, and the MLVA for ST398 were quite low, as follows: 0.51, 0.74, and 0.81, respectively. The combination of spa typing and the MLVA resulted in more discriminatory power (D = 0.87) than that of the MLVA used alone but was still lower than that of PFGE when using a single restriction enzyme (D = 0.88 to 0.92). For this data set, the DI of PFGE with the restriction enzymes BstZI and ApaI was as high as the DI of PFGE using all three restriction enzymes together or using those enzymes in combination with the MLVA (D = 0.97). The combination of PFGE (when using all restriction enzymes or BstZI and ApaI) and spa typing had the most discriminatory power (D = 0.98).
The following isolates were indistinguishable by all means applied: MB4392 and RIVM38; MB4362 and MB4366; RIVM29 and RIVM30; and RIVM21, RIVM33, and RIVM34.
The concordance among different typing methods is shown in Tables Tables44 and and5.5. For the MLVA, the highest concordance was with PFGE using ApaI and with spa typing. As further expressed by the Wallace coefficient, the MLVA type was completely predicted by PFGE using a combination of all the restriction enzymes (W = 1). The MLVA type was poorly predicted only by spa typing (W = 0.48), whereas the spa type was well predicted by MLVA (W = 0.68). As already indicated above, the concordance between PFGE using restriction enzyme BstZI and PFGE using SacII was very high (Rand's index = 0.98; R = 0.89). PFGE, using a combination of all three restriction enzymes, was able to predict the SCCmec type and MLVA type completely (W = 1.00) and had a high probability of predicting the spa type (W = 0.64). SCCmec type was a poor predictor for all typing methods (W = 0.06 to 0.38), although it was well predicted by MLVA (W = 0.98) and completely predicted using all PFGE methods (W = 1.00).
AA-MRSA strains of CC398 are not typeable using PFGE with SmaI digestion, the gold standard for MRSA typing. For this reason, until now, sequence-based typing methods, such as MLST and spa typing, have been used for typing AA-MRSA in epidemiological studies. Although these methods have the advantage of reliability and excellent interlaboratory reproducibility, their discriminatory power is weak and, consequently, less useful for short-term epidemiological studies such as an examination of contamination routes on farms. To our knowledge, this is the first study exploring alternative fingerprinting techniques for subtyping MRSA ST398.
In this study, PFGE using the restriction enzymes XmaI (an isoschizomer of SmaI), BstZI, SacII, and ApaI was tested to obtain the enzymes’ discriminatory power with a set of 32 MRSA ST398 isolates. In addition, three recently developed MLVA techniques, successfully applied to Staphylococcus aureus and more specifically to HA-MRSA and CA-MRSA, were evaluated and optimized for use with ST398 strains. Both the initial Polish system (15, 16, 17, 23, 24) and the Swiss system (10) are based on tandem repeats in genes, whereas the British system is based on staphylococcal interspersed repeat units, called SIRUs (11, 12). In this study, only primers with discriminatory power were included in this optimized MLVA specific for AA-MRSA.
All isolates were typeable using both the MLVA and PFGE. However, fingerprints obtained by PFGE using XmaI had weak patterns and were thus excluded from the study. The discriminatory power of PFGE, when using any of the remaining three restriction enzymes, was higher than that of SCCmec typing, spa typing, the MLVA, or spa typing in combination with the MLVA. Within PFGE, the combination of all three restriction enzymes had higher discrimination than the use of only one restriction enzyme. The discriminatory power of PFGE using only BstZI and ApaI was equally as good as using all three restriction enzymes. The discriminatory power of the MLVA, the second best typing method, was only a bit higher than that of spa typing, although spa typing was also able to discriminate among identical MLVA types in a few cases and within a pulsotype on one occasion (using all three restriction enzymes or BstZI and ApaI). In this set of AA-MRSA isolates, the MLVA was not able to distinguish among isolates with the same PFGE profile. In agreement with our results, previously published reports of MLVA typing of clinical MRSA isolates had more discriminatory power than those of spa typing and MLST (12, 15, 16, 17). In contrast, most studies comparing the MLVA with PFGE have shown that the discriminatory power of PFGE was comparable to that of the MLVA or only slightly better. Luczak-Kadlubowska et al. (16) and Sabat et al. (23) demonstrated that their MLVA system had a comparable discriminatory power to that of PFGE. The same group had reported previously that PFGE was superior to the MLVA. These findings were based on the setup of the study and the selection of the MRSA isolates (17). The results of the MLVA based on staphylococcal interspersed repeat units (11) depended on the set of isolates: for some isolates, the MLVA was able to distinguish within the same PFGE profile, but for others, the MLVA was less discriminatory than PFGE.
It should be kept in mind that in the publications mentioned above, the MLVA consisted of a different combination of primer sets than the MLVA described here. Furthermore, in those publications, PFGE was performed using only the restriction enzyme SmaI. The MLVA and PFGE were also used on a wide range of MLST and spa types belonging to HA- and CA-MRSA but not to MRSA ST398. Those studies aimed to investigate whether the faster MLVA technique was able to predict PFGE typing, MLST, or spa typing. In order to do so, relatively low similarity cutoff values between 70% and 80% were used for the MLVA and PFGE (17, 27). This contrasts with the present study, during which the MLVA and PFGE techniques were evaluated for their power to distinguish among MRSA ST398 isolates. To this end, isolates that differed even slightly were considered to belong to different types. For the MLVA, some differences were small. It is thus advisable to use capillary electrophoresis for better resolution between bands.
Even when combining all methods, some isolates from different origins were indistinguishable from each other: pig isolates MB4392 and RIVM38 (one isolated in Belgium, the other in The Netherlands), MB4362 and MB4366 (the first isolated from a cow, the second from a pig), RIVM29 and RIVM30 (both isolated from Dutch pigs), and finally, RIVM21, RIVM33, and RIVM34 (also found in Dutch pig herds). It may be possible to distinguish between these isolates using other techniques. Another possibility, given the large number of animals transported within and between Belgium and The Netherlands, is that those isolates are the same strain circulating in different countries and in different animal species.
The concordance among the different methods was calculated according to Rand's and adjusted Rand's indices and Wallace coefficients. As this is calculated for only a limited set of data, it should be interpreted carefully. According to the obtained results, the spa type was best predicted by MLVA, whereas the MLVA type was poorly predicted only by spa typing. The MLVA type was completely predicted by PFGE using a combination of all restriction enzymes (W = 1) but not vice versa. Tenover et al. (27) also concluded that the MLVA cannot be used to predict PFGE types.
In conclusion, both the MLVA and PFGE can be used to discriminate among isolates belonging to the newly emerging MRSA ST398 type. All isolates were typeable using both of the fingerprinting techniques and had better discriminatory power than the commonly used techniques. Although the MLVA is a faster technique, PFGE had more discriminatory power than the MLVA, especially when two restriction enzymes (i.e., BstZI and ApaI) were used. Both techniques can be used for epidemiological tracing of the MRSA ST398 isolates.
We are grateful to M. Struelens from the Laboratoire de Référence MRSA—Staphylocoques, Department of Microbiology, Université Libre de Bruxelles, Hôpital Erasme, Brussels, Belgium, for providing strains MB4387, MB4388, and MB4389. We thank Miriam Levenson for critical reading of the manuscript.
Published ahead of print on 26 August 2009.