Cross-species spread of SCCmec IV subtypes in staphylococci

Abstract

Staphylococcal chromosomal cassette mec (SCCmec) is a mobile genetic element that carries resistance genes for beta-lactam antibiotics. Coagulase-negative staphylococci, such as S. epidermidis, are thought to be a reservoir of diverse SCCmec elements that can spread to the most virulent staphylococcal species, S. aureus, but very little is known about the extent of cross-species spread of these elements in natural populations or their dynamics in different species. We addressed these questions by using a sample of 86 S. aureus and S. epidermidis isolates with SCCmec type IV that were collected from a single hospital over a period of six months. To subtype SCCmec IV, we used multiplex PCR to detect structural variations and we used sequences from a fragment of the ccrB gene and from the dru repeats to detect additional variations. Multiplex PCR had significantly lower typeability than ccrB:dru sequencing, due to more nontypeable isolates among S. epidermidis. No statistically significant differences in diversity were detected by subtyping method or species. Interestingly, while only 4 of 24 subtypes (17%) were shared between species, these so-called shared subtypes represented 58 of 86 isolates (67%). The shared subtypes differed significantly between species in their frequencies. The shared subtypes were also significantly more concordant with genetic backgrounds in S. aureus than in S. epidermidis. Moreover, the shared subtypes had significantly higher minimum inhibitory concentrations to oxacillin in S. aureus than in S. epidermidis. This study has identified particular SCCmec IV subtypes with an important role in spreading beta-lactam resistance between species, and has further revealed some species differences in their abundance, linkage to genetic background, and antibiotic resistance level.

1. Introduction

Methicillin-resistant Staphylococcus aureus (MRSA) infections are a global health problem (Grundmann et al., 2006). Resistance to beta-lactam antibiotics is mediated by the mecA gene, which is carried on a mobile genetic element known as the staphylococcal chromosomal cassette mec (SCCmec) (Katayama et al., 2000). In addition, resistance genes for aminoglycosides, macrolides, tetracyclines, and heavy metals such as cadmium and mercury, can accrue on SCCmec elements. Eight different SCCmec types have been defined, all of which are chromosomally integrated close to the origin of replication at an open reading frame of unknown function called orfX (IWG-SCC, 2009). SCCmec type IV is among the shortest types at 20.9 to 24.2 kb and is thought to be relatively more mobile and to have a relatively lower cost on strain fitness in comparison to some other types (Lee et al., 2007; Okuma et al., 2002; Robinson and Enright, 2003). SCCmec IV occurs frequently in both hospital-acquired and community-acquired MRSA. For example, it currently resides in the most prevalent MRSA clone in the United States, ST8-MRSA-IV (USA300) (Tenover et al., 2006). Eight subtypes of SCCmec IV have been defined (Milheiriço et al., 2007). These subtypes differ from each other structurally in the length of the 3′ end of the element known as the J1 region. Additionally, SCCmec IVa carries pUB110 that confers bleomycin resistance and SCCmec IVc carries Tn4001 that confers broad aminoglycoside resistance.

S. epidermidis is a coagulase-negative staphylococci that has been identified previously as a potential source of conjugative plasmids and other mobile genetic elements that can spread to S. aureus (Archer et al., 1994; McDonnell et al., 1983, Mongkolrattanothai et al., 2004). Several observations are consistent with the hypothesis that S. epidermidis is a reservoir of diverse SCCmec elements: i) it is ubiquitous on human skin, ii) >50% of its isolates are consistently resistant to antistaphylococcal penicillins regardless of geographic locale, and iii) >10% of its isolates yield nontypeable SCCmec elements (Diekema et al., 2001; Garza-González et al., 2010; Hanssen et al., 2007; Ibrahem et al., 2009; Miragaia et al., 2005, 2007; Ruppé et al., 2009). In particular, S. epidermidis may be a reservoir of SCCmec IV. Although the origin of SCCmec IV has not been identified, it has been found among S. epidermidis isolates that were collected in the 1970s, whereas some of the earliest reported S. aureus isolates with SCCmec IV are from the 1980s (Wisplinghoff et al., 2003). Among more recent isolates, SCCmec IVa is common in both S. aureus and S. epidermidis in Japan and France (Barbier et al., 2010; Jamaluddin et al., 2008). Wisplinghoff et al. (2003) indicated that these two species could carry SCCmec IV elements with 98–99% nucleotide identity to each other. Furthermore, Bloemendaal et al. (2010) confirmed an in vivo transfer event of SCCmec IVa from S. epidermidis to S. aureus in which the elements differed by only a single bp (Bloemendaal et al., 2010; Wielders et al., 2001).

The occurrence of very similar SCCmec elements in different staphylococcal species is more parsimoniously explained by their relatively recent horizontal genetic transfer between species rather than their convergent or parallel evolution within species. While these observations provide evidence that cross-species spread of SCCmec can occur in nature, they do not provide an indication of its contribution to the pool of SCCmec elements found in circulating MRSA. In some geographic locales, such as in Norway, the prevalence of MRSA infections is low (<1%) and cross-species spread of SCCmec may be an important mechanism for generating sporadic MRSA (Hanssen et al., 2004). However, in other locales, such as in the United States, the prevalence of MRSA infections is high (>50%) and both coagulase-negative staphylococci as well as other MRSA might be donors of SCCmec. In this study, we characterized a local population of staphylococci from the United States to determine how often isolates carry SCCmec elements that occur in both S. aureus and coagulase-negative staphylococci. We found that a small proportion of the total number of SCCmec IV subtypes were common to both S. aureus and S. epidermidis, but these so-called shared subtypes represented a majority of isolates in both species. We also found that these shared subtypes differed in several characteristics between the two species.

2. Materials and methods

2.1. Bacterial isolates

The Microbiology Laboratory of Westchester Medical Center in Valhalla, NY provided weekly samples of both S. aureus and coagulase-negative staphylococci from clinical specimens from January through June 2007. The isolates were sampled from multiple patients and without regard to their clinical relevance. Upon receipt, the isolates were purified and examined for beta hemolysis and mannitol fermentation phenotypes on appropriate agar plates. Routine bacterial growth was done overnight on tryptic soy agar plates at 37°C. Long-term storage was at −80°C in a solution of tryptic soy broth and 15% glycerol. Bacterial genomic DNA was isolated with a DNeasy kit (Qiagen), according to the manufacturer’s instructions. Species identification was done by PCR amplification and sequencing of both strands of a fragment of the tuf gene, as described previously (Heikens et al., 2005).

tuf gene sequencing identified 12 different species in the isolate collection (181 S. aureus, 3 S. auricularis, 7 S. capitis, 1 S. cohnii, 160 S. epidermidis, 6 S. haemolyticus, 40 S. hominis, 2 S. lugdunensis, 1 S. saprophyticus, 1 S. sciuri, 3 S. simulans, 5 S. warneri). Subsequent SCCmec typing revealed that only SCCmec IV was present in sufficient numbers in S. aureus and coagulase-negative staphylococci, S. epidermidis in this sample, to allow comparisons. For example, there were 50 S. aureus isolates with SCCmec II but only single isolates of S. capitis, S. cohnii, and S. hominis with this type.

Multiple isolates were available from some patients because either more than one colony was picked from an agar plate or specimens were available on more than one date. To avoid bias in frequency-sensitive calculations, we included a single isolate where multiple, genetically indistinguishable isolates were recovered from the same patient. In total, the study sample included 26 S. aureus and 60 S. epidermidis isolates with SCCmec IV. For these isolates, oxacillin minimum inhibitory concentrations (MICs) were determined by Etest (AB Biodisk), according to the manufacturer’s instructions. Characteristics of these isolates are listed in Table S1.

2.2. Characterization of genetic backgrounds and SCCmec elements

Multilocus sequence typing (MLST) was used to identify the genetic backgrounds of the isolates. MLST was done according to published methods for S. aureus (Enright et al., 2000) and S. epidermidis (Thomas et al., 2007). Briefly, fragments of seven housekeeping genes were amplified by PCR and sequenced on both strands. Alleles and sequence types (STs) were determined using the S. aureus and S. epidermidis MLST databases (http://www.mlst.net).

SCCmec IV was identified using PCR methods that score the mec class and ccr allelic group, as done previously (Robinson and Enright, 2003). These results were confirmed using multiplex PCR reactions 1 and 2 of Kondo et al. (2007). SCCmec IV was subtyped using the multiplex PCR method of Milheirço et al. (2007), which was supplemented with a separate PCR reaction to detect the dcs region (Oliveira and de Lencastre, 2002), as recommended. Control strains used to represent the multiplex PCR subtypes included JCSC4744 (SCCmec IVa), CDC OR-9 (SCCmec IVb), JCSC4788 (SCCmec IVc), and JCSC4469 (SCCmec IVd). SCCmec IV was also subtyped by PCR amplification and sequencing on both strands of a fragment of the ccrB gene and the direct repeat units (dru repeats), as done previously (Smyth et al., 2010). The combined subtyping data of multiplex PCR and ccrB:dru sequencing identified indistinguishable SCCmec IV elements that occurred in both species; throughout this study, these elements were referred to as shared subtypes.

2.3. Statistical analyses

Typeability was defined as the proportion of typeable isolates, p. Assuming a binomial sampling distribution, as is commonly done with allele frequency data, the variance of p was calculated as p(1−p)/n, where n is the number of isolates tested. 95% confidence intervals (CI) were constructed as p +/− 2√Var(p).

Diversity was measured with Simpson’s index (Hunter and Gaston, 1988) as well as the k_e3 estimator of the effective number of types (Nielsen et al., 2003). 95% CI for these two measures of diversity were calculated as described previously (Grundmann et al., 2001; Smyth et al., 2010). Simpson’s index provides an estimate of the probability that two isolates picked at random from the population belong to different types, and the effective number of types provides an estimate of the number of equally frequent types that will produce the observed diversity.

Differentiation in type frequencies between species was measured with Jost’s D (Jost, 2008). We used the SPADE computer program of Chao and Shen (2003) to calculate the D estimator in equation 13 of Jost (2008) and to calculate 95% CI based on bootstrap resampling with 1000 replicates. Jost’s D will be 0 when subpopulations are completely undifferentiated (i.e. identical) in type frequencies and 1 when subpopulations are completely differentiated. When used simply as a measure of differentiation, D does not require knowledge of the processes that underlie variation in the markers and therefore may be applicable to a variety of typing datasets.

The distributions of MICs were represented with box-and-whisker plots. Differences in median oxacillin MICs were tested with a two-tailed Mann-Whitney U test, using InStat v3.1 (GraphPad Software).

2.4. Nucleotide sequence accession numbers

The sequences of each different ccrB allele has been deposited in the GenBank database with accession numbers HQ236719-21. Unique dru repeats and dru types were deposited in a publicly available dru typing database (http://www.dru-typing.org).

3. Calculation

To measure the concordance between genetic backgrounds and SCCmec IV subtypes, we cross-classified all pairs of isolates according to whether they matched or mismatched for MLST-defined STs and SCCmec IV subtypes. For n isolates, there are N = (n²−n)/2 such pairwise comparisons. When comparing two markers, X and Y, these pairwise comparisons can be summarized in a 2×2 table accordingly: a, number of isolate pairs that match for both markers; b, number of isolate pairs that match for marker X but mismatch for marker Y; c, number of isolate pairs that mismatch for marker X but match for marker Y; d, number of isolate pairs that mismatch for both markers. The Rand index (Rand, 1971) represents the proportion of concordant pairs and is calculated as

$$RI=\frac{a+d}{a+b+c+d}=\frac{a+d}{N}$$

Carriço et al. (2006) noted that this measure does not take into account the concordance due to chance and may therefore overestimate the concordance between markers. The adjusted Rand index (Hubert and Arabie, 1985) was recommended for typing datasets to correct for chance concordance and can be calculated with a general formula (Albatineh et al., 2006) as

$$\mathit{ARI}=\frac{\text{observed}RI-\text{expected}RI}{\text{maximum}RI-\text{expected}RI}$$

The maximum RI is always 1. The expected RI due to chance concordance can be calculated from the marginal frequencies of the 2×2 table as

$$\text{expected}a=\frac{(a+b)(a+c)}{N}$$

$$\text{expected}d=\frac{(b+d)(c+d)}{N}$$

Substituting these into the numerator of the observed RI, we have

$$\text{expected}RI=\frac{[(a+b)(a+c)+(b+d)(c+d)]}{N^2}$$

With the observed, maximum, and expected RI, we can then calculate the ARI as indicated above. We point out that the Wallace index, which is another measure of concordance recommended for typing datasets (Carriço et al., 2006), is also influenced by chance concordance and should be correctable with an analogous procedure as outlined above. In fact, we find that the expected Wallace index due to chance concordance is very similar, if not identical, to the W_i of Pinto et al. (2008). Since the Wallace index has been used to assess the value of including additional typing methods in studies of both S. aureus and S. epidermidis (Faria et al., 2008; Miragaia et al., 2008; Shore et al., 2010), an adjusted version of this index should be examined with some priority; however, such a development is beyond the scope of this work.

In order to make statistical comparisons of the ARI for different markers, we developed and partially validated a jackknife resampling procedure. We initially investigated several bootstrap resampling procedures, but found that they performed poorly with our datasets. We note that the ARI, like Simpson’s index (Grundmann et al., 2001), uses a sample to make an inference about a population parameter. In this case, the population parameter is the unknown concordance between markers in the larger population. Thus, the ARI is subject to sampling variation and our jackknife procedure attempts to account for that source of variation. It remains to be determined whether the ARI is the preferred population parameter to be estimated and whether these sorts of cross-classification indices are most appropriate for comparing all classes of markers; for example, linkage disequilibrium indices may be more appropriate when the markers consist of discrete genetic loci. Nonetheless, our jackknife procedure is based on a general approach (Crowley, 1992). To begin, each of the n isolates are removed in turn from the dataset and the ARI for each of these n datasets are calculated. Next, n pseudovalues are calculated as

$${\mathit{ARI}}_i={\mathit{nARI}}_o-(n-1){\mathit{ARI}}_{-i}$$

where ARI_o is the observed ARI with all isolates included, and ARI_−i is the ARI with isolate i removed (i=1, 2, 3, …n). The jackknife estimator of the ARI is the average of the pseudovalues

$$\overline{\mathit{ARI}}=\frac{\sum {\mathit{ARI}}_i}{n}$$

Its variance and 95% CI are calculated as

$$\text{Var}(\overline{\mathit{ARI}})=\frac{\sum {({\mathit{ARI}}_i-\overline{\mathit{ARI}})}^2}{n(n-1)}$$

$$\text{CI}=\left[\overline{\mathit{ARI}}-2\sqrt{\text{Var}(\overline{\mathit{ARI}})},\overline{\mathit{ARI}}+2\sqrt{\text{Var}(\overline{\mathit{ARI}})}\right]$$

To examine the validity of this approach, we first simulated two artificial populations of 1000 isolates each, where each isolate had three markers and each marker had ten types in the population. Marker A was made with 820 isolates of one type followed by 20 isolates each of the remaining nine types. Marker B was made with 460 isolates of one type followed by 60 isolates each of the remaining nine types. Marker C was made with 100 isolates for each of the ten types. Simpson’s index for these three markers was 0.324, 0.757, and 0.901, respectively. In the second population, the types for each marker were randomized among the isolates 100 times to simulate recombination and lower the concordance. Separate random samples of 25, 50, and 100 isolates were taken from each population, and this sampling was repeated 100 times. We then counted the number of samples where the jackknife confidence intervals included the population ARI. With one exception, which achieved 93.3% accuracy, the confidence intervals included the population ARI in 95% or more of the samples.

4. Results

4.1. Typeability and diversity of SCCmec IV

Comparisons of typeability revealed an overall statistically significant difference between the multiplex PCR and ccrB:dru sequencing methods (Table 1); that is, the 95% CI for the proportion of typeable isolates for these two methods did not overlap. The typeability of the multiplex PCR method was relatively low because of the many nontypeable S. epidermidis isolates. For the isolates of both species that were nontypeable by multiplex PCR, only the mecA positive control gene was amplified on repeated attempts. This finding indicated that more uncharacterized structural variations existed in the SCCmec IV elements of S. epidermidis than S. aureus. On the other hand, the ccrB:dru sequencing method yielded a single isolate of S. epidermidis that was nontypeable for its ccrB gene, and the dru repeats were typeable for all 86 staphylococcal isolates (Table 1).

Table 1.

Typeability of SCCmec IV subtyping methods.

Species	No. of isolates tested	Proportion of typeable isolates by method (95% CI)
		multiplex PCR	ccrB:dru
S. aureus	26	0.885 (0.759, 1.000)	1.000 (1.000, 1.000)
S. epidermidis	60	0.683 (0.563, 0.803)	0.983 (0.950, 1.000)
Total	86	0.744 (0.650, 0.838)	0.988 (0.965, 1.000)

The diversity of genetic backgrounds, as reflected by MLST, did not differ significantly between the two species (Table 2) even though this sample of SCCmec IV-carrying isolates included more than twice as many S. epidermidis as S. aureus. Moreover, there were no significant differences in the diversity of SCCmec IV based on subtyping method or species. These observations were confirmed using both measures of diversity (Table 2).

Table 2.

Diversity of genetic backgrounds and SCCmec IV subtypes.

Species	Method	No. of typesa	Simpson’s index of diversity (95% CI)	Effective no. of types (95% CI)
S. aureus	MLST	6	0.511 (0.294, 0.728)	2.04 (0.60, 3.48)
	multiplex PCR	4	0.403 (0.178, 0.628)	1.67 (0.92, 2.42)
	ccrB:dru	7	0.563 (0.351, 0.775)	2.28 (0.45, 4.11)
S. epidermidis	MLST	13	0.794 (0.707, 0.881)	4.84 (2.73, 6.95)
	multiplex PCR	6	0.636 (0.547, 0.725)	2.75 (2.11, 3.38)
	ccrB:dru	11	0.561 (0.418, 0.704)	2.28 (0.64, 3.91)

^aS. aureus STs (no. isolates): ST1 (1), ST5 (4), ST8 (18), ST59 (1), ST72 (1), ST1183 (1)

S. epidermidis STs (no. isolates): ST2 (6), ST5 (25), ST6 (1), ST16 (7), ST23 (1), ST59 (1), ST69 (5), ST83 (6), ST89 (3), ST175 (1), ST177 (1), ST179 (2), ST180 (1)

multiplex PCR and ccrB:dru subtypes are given in Table 3

4.2. Species distributions of SCCmec IV subtypes

Multiplex PCR subtype IVa was the most prevalent subtype in both species (Table 3). This observation extends the reports that IVa is highly prevalent in both species (Barbier et al., 2010; Jamaluddin et al., 2008). Jost’s D revealed that, with these sample sizes, S. aureus and S. epidermidis were not differentiated from each other based on the frequencies of their multiplex PCR subtypes (Table 3). Although there was a large difference between the two species in the number of nontypeable isolates, this difference was not large enough to genetically differentiate between the two species.

Table 3.

Differentiation between S. aureus and S. epidermidis based on SCCmec IV subtypes.

	multiplex PCR subtypes	No. of isolates		ccrB:dru subtypes	No. of isolates		shared subtypes	No. of isolates
		S.a.	S.e.		S.a.	S.e.		S.a.	S.e.
	IVa	20	31	3:10a	4	39	IVa:3:10a	1	18
	IVb	0	2	3:10m	0	1	IVa:3:9g	17	3
	IVc	0	3	3:11b	0	7	IVg:3:10a	1	3
	IVd	2	0	3:11q	0	1	IVnt:3:10a	1	14
	IVg	1	3	3:12f	0	1	Total	20	38
	IVh	0	2	3:7f	0	1
	IVnt	3	19	3:7n	0	1
	Total	26	60	3:9g	17	6
				3:9q	0	1
				7:10a	0	1
				nt:8a	0	1
				3:10e	1	0
				3:10y	1	0
				3:8q	1	0
				3:9d	1	0
				6:7j	1	0
				Total	26	60
Jost’s D (95% CI):	0.093 (0.000, 0.244)			0.622 (0.367, 0.878)			0.789 (0.591, 0.987)

Note: S.a. is S. aureus, S.e. is S. epidermidis, nt is nontypeable

The ccrB:dru sequencing revealed two prevalent subtypes, 3:10a and 3:9g (Table 3). In contrast to the multiplex PCR method, the ccrB:dru sequencing method was able to differentiate between S. aureus and S. epidermidis, as revealed by a significant non-zero Jost’s D. This result was partly due to the two prevalent ccrB:dru subtypes having different frequencies in the two species. Note that both species also had unique ccrB:dru subtypes, but these were represented by single isolates in all except one subtype (3:11b) (Table 3).

When both the multiplex PCR and ccrB:dru sequencing data was combined, a total of 24 subtypes were identified. While only four of these subtypes occurred in both species, these so-called shared subtypes represented two-thirds of all isolates (Table 3). Furthermore, these shared subtypes represented more than half of each species’ isolates: 20/26 (77%) in S. aureus and 38/60 (63%) in S. epidermidis. By Jost’s D, the shared subtypes alone were able to differentiate between the two species (Table 3). These findings indicated that the shared subtypes were among the most successful elements overall and differed in frequency between species.

4.3. Linkage of SCCmec IV subtypes to genetic backgrounds

The concordance between genetic backgrounds and SCCmec IV subtypes was compared using a new jackknife estimator of the ARI and its confidence interval (see Calculation section). Within species, there were no significant differences in concordance between the different markers (Table 4). The concordance between genetic backgrounds and multiplex PCR subtypes was similar for both species. However, genetic backgrounds were significantly more concordant with ccrB:dru subtypes and with the shared subtypes in S. aureus than in S. epidermidis (Table 4). The multiplex PCR subtypes were also significantly more concordant with ccrB:dru subtypes in S. aureus than in S. epidermidis. These findings revealed that marker concordance was much higher in S. aureus than in S. epidermidis, on average 7.5x higher.

Table 4.

Concordance between genetic backgrounds and SCCmec IV subtypes.

Species	Comparisona	Observed Rand index	Expected Rand index	Adjusted Rand index	Jackknife estimator (95% CI)
S. aureus	MLST, multiplex PCR	0.677	0.498	0.357	0.370 (−0.009, 0.749)
	MLST, ccrB:dru	0.818	0.501	0.636	0.652 (0.323, 0.982)
	multiplex PCR, ccrB:dru	0.803	0.488	0.616	0.628 (0.281, 0.975)
	MLST, shared subtypes	0.911	0.559	0.797	0.821 (0.416, 1.000)
S. epidermidis	MLST, multiplex PCR	0.630	0.580	0.119	0.119 (0.021, 0.217)
	MLST, ccrB:dru	0.569	0.536	0.072	0.072 (−0.065, 0.208)
	multiplex PCR, ccrB:dru	0.501	0.517	−0.032	−0.034 (−0.139, 0.071)
	MLST, shared subtypes	0.642	0.569	0.169	0.171 (−0.003, 0.346)

^aMarker A, Marker B

We note that the crude ARI and the jackknife estimator produced results that were very similar to each other (Table 4). Although sufficient to distinguish between species, the confidence intervals for the S. aureus data were very wide, which might be due to the relatively small sample sizes available for that species (n=20 for the shared subtypes, n=26 for other data). This observation would argue against the use of thresholds to interpret the ARI as “moderate”, “good”, and so on, because the uncertainty in the estimate may span several thresholds.

4.4. Oxacillin susceptibilities of SCCmec IV subtypes

MICs to oxacillin were determined for all isolates with the goal of assessing whether different SCCmec IV subtypes had different levels of resistance. The distributions of MICs for these two species showed mostly high MICs for S. aureus and a wide range of MICs for S. epidermidis (Figure 1). The median MIC for S. aureus isolates was >256 μg/mL compared to 14 μg/mL for S. epidermidis isolates, which is a statistically significant difference (Mann-Whitney U=269.5, P<0.0001). These same patterns extended to the shared subtypes, where S. aureus isolates had a median MIC of >256 μg/mL compared to 12 μg/mL for S. epidermidis isolates (Mann-Whitney U=93.5, P<0.0001).

Figure 1.

Distribution of oxacillin MICs for S. aureus and S. epidermidis isolates with SCCmec IV. Top and bottom lines indicate maximum and minimum values, top and bottom of the boxes indicate the 75^th and 25^th percentiles, and the dark horizontal lines represents the medians.

The most prevalent shared subtype in S. aureus, identified as IVa:3:9g (Table 3), had a median MIC of >256 μg/mL. Its genetic background included ST8 and a single locus variant of ST8 that was newly identified from this population, ST1183 (Wong et al., 2010). These isolates are closely related to the USA300 MRSA clone that is currently prevalent in the United States (Tenover et al., 2006). In S. epidermidis, this same shared subtype occurred in two STs that differed from each other at three MLST loci and its median MIC was 8 μg/mL. Likewise, the most prevalent shared subtype in S. epidermidis, identified as IVa:3:10a (Table 3), was spread among six STs that differed from each other at an average of 2.6 MLST loci and had a median MIC of 12 μg/mL. In S. aureus, this same shared subtype occurred in a single isolate of ST59 and had an MIC of >256 μg/mL. These results demonstrated that the dynamics of SCCmec IV subtypes that have spread between species can differ in different species; in S. aureus, the shared subtypes have spread among a few genetic backgrounds with high oxacillin MICs whereas, in S. epidermidis, they have spread to more genetic backgrounds with lower MICs.

5. Discussion

The estimated number of acquisitions of SCCmec elements by S. aureus has been upwardly revised with advancements in typing technology (Branger and Goullet, 1987; Enright et al., 2002, Musser and Kapur, 1992; Nübel et al., 2008; Robinson and Enright, 2003). These estimates may differ substantially between clonal groups, from zero acquisitions for much of the current genetic diversity found in S. aureus, to a single acquisition in the ST239 clonal group (Smyth et al., 2010), to a couple of dozen acquisitions in the ST5 clonal group (Nübel et al., 2008). However, with some notable exceptions (Blomendaal et al., 2010; Hanssen et al., 2004), we have very little insight into where the acquired elements come from in individual cases.

Here, we provide new information about the extent of cross-species spread of SCCmec type IV and some dynamics of this element in S. aureus and S. epidermidis. In summary, only a small proportion of the total number of SCCmec IV subtypes circulating in the population was found in both species but, importantly, these shared subtypes were the most prevalent subtypes in the population. Since these shared subtypes are indistinguishable in the two species based on several genetic markers, we infer that they have spread between species. We conclude that these particular SCCmec IV subtypes have an important role in spreading beta-lactam resistance between species and in the population at large. We further determined that these shared subtypes differed significantly between the two species in several characteristics, including abundance, linkage to genetic background, and antibiotic resistance level. While the species differences in abundance might be used to infer the direction of spread (e.g. high abundance indicating the donor and low abundance indicating the recipient), we avoid such inferences in this study because there is the potential of selection for the different MICs to impact abundances. The species differences in the linkage of shared subtypes to strain genetic backgrounds might be explained by differences in the relative clonality of these species, but this result was not entirely expected since the dynamics of mobile genetic elements can be independent of their hosts.

Almost one-third of the S. epidermidis isolates were nontypeable by the multiplex PCR method. These data indicated that uncharacterized variations existed in the S. epidermidis elements, which is consistent with the hypothesis that this species is a reservoir of diverse SCCmec IV elements. However, there was no excess of nontypeable elements by the ccrB:dru sequencing method and the overall diversities of the elements in the two species did not differ significantly. Moreover, 14/19 (74%) of the S. epidermidis isolates with nontypeable elements by multiplex PCR were of the same ccrB:dru subtype (IVnt:3:10a; Table 3), indicating that they probably do not represent an extremely diverse collection of SCCmec IV elements but rather a few uncharacterized structural variants. This observation also justifies our use of the nontypeable isolates as a discrete type during the analyses.

Variation at the ccrB locus is part of the definition of SCCmec types, and its sequence variation appears to agree qualitatively with PCR-detected variation (Lina et al., 2006; Oliveira et al., 2006). The dru sequencing method has been used previously to improve discrimination among MRSA isolates with SCCmec IV from Scottish and Irish hospitals (Goering et al., 2008; Larsen et al., 2009; Shore et al., 2010). We previously compared a multiplex PCR method with the ccrB:dru sequencing method among global strains of the ST239-MRSA-III clonal group, which all carry SCCmec type III (Smyth et al., 2010). Within that clonal group, variations in the ccrB:dru sequences reflected the phylogenetic history of the strains. In contrast, the multiplex PCR method reflected homoplasious (i.e. non-treelike) variations representing independent losses of accessory regions of the element in different strains. In the current study, we observed a difference in the ability of the multiplex PCR and ccrB:dru sequencing methods to differentiate between staphylococcal species. This ability might reflect different mutation rates for these markers. Length variation in the J1 region of SCCmec IV elements is the primary basis for the different multiplex PCR subtypes. Variation at the dru locus, which arises by slipped-strand mispairing and nucleotide substitutions, might accumulate more rapidly than does variation among the multiplex PCR targets or at the ccrB locus. Our results demonstrate that sequence-based subtyping of SCCmec IV has advantages in typeability and differentiation ability over multiplex PCR subtyping in a population-based sample of staphylococci.

We developed a new estimator of the ARI and its confidence interval using a jackknife procedure. This innovation allowed us to make statistical comparisons of the concordance between genetic backgrounds and SCCmec IV subtypes. Based on previous MLST analyses (Feil et al., 2003, Kozitskaya 2005; Miragaia et al., 2007; Ruimy et al., 2008), S. aureus appears to undergo effectively less strain-to-strain recombinations in core genes than does S. epidermidis, resulting in S. aureus having a relatively more clonal population structure than S. epidermidis. Although evidence for multiple acquisitions of SCCmec within diverse samples of both species has been presented (Miragaia et al., 2007; Robinson and Enright, 2003), it has been unknown whether SCCmec was more tightly linked to genetic background in S. aureus or in S. epidermidis. Since mobile genetic elements are often acquired through nonhomologous recombinations mediated by recA-independent transposases or recombinases, it is not obvious whether a correlation should be expected between homologous background recombinations and nonhomologous acquisitions of mobile elements. Our results suggest such a correlation may exist in staphylococci because the concordance between genetic backgrounds and SCCmec IV subtypes was much higher in S. aureus than in S. epidermidis. We find this result surprising given that there were no significant differences in the diversity of these species in terms of their genetic backgrounds or SCCmec IV subtypes.

While the occurrence of indistinguishable subtypes in different species is consistent with a history of cross-species spread of SCCmec IV, the shared subtypes defined here on the basis of all available subtyping data probably represent the most recent subtypes to be spread between these species. Consider that subtypes that both differ slightly from each other and that occur in different species might have accumulated their differences after having been spread between species. Phylogenetic analyses of complete SCCmec IV sequences might be able to identify related elements that have crossed species boundaries, but the markers used here are not ideal for conducting such analyses. On the other hand, previous studies have indicated that variations in the excision of SCCmec IV from the chromosome may influence the propensity of different elements to spread between strains (Higgins et al., 2009; Jansen et al., 2006; Noto and Archer, 2006). Accordingly, the shared subtypes could represent the more transferrable subtypes. Further work is necessary to address these issues.

Questions also remain about where and how frequently SCCmec IV subtypes are spread between species; our data address the impact of cross-species spread on standing diversity of SCCmec IV but not these other questions. The actual horizontal genetic transfer events that resulted in the shared subtypes need not have taken place in the population that we sampled. In fact, the most prevalent shared subtypes from both species occur in MRSA isolated outside our local population (unpublished data). In addition, even though the shared subtypes are the most prevalent in our local population, the frequency of transfer in nature could be very low due to, for example, barriers to horizontal genetic transfer (Corvaglia et al., 2010; Marraffini and Sontheimer, 2008; Waldron and Lindsay, 2006).

Finally, it is clear that in the presence of beta-lactam antibiotics, SCCmec IV could provide a selective advantage to staphylococci. Some data suggest that the cost of carrying type IV is lower than the cost of carrying some other SCCmec types (Lee et al., 2007). Other data indicate that the oxacillin resistance levels conferred by the element have the major effect on fitness, with an inverse relationship between MIC and growth rate (Ender et al., 2004). We observed that the shared subtypes differed significantly in oxacillin MICs between the species. For example, subtype IVa:3:10a was rare in S. aureus where it had a high MIC, whereas it was common in S. epidermidis where it had a low MIC. The same type of pattern, with S. aureus having the higher MIC, was observed for subtype IVa:3:9g even though its frequencies in the two species were reversed. It has been reported that S. epidermidis tends to display a broader range of oxacillin MICs than does S. aureus (McDonald et al., 1995); however, to our knowledge, the results presented here are the first to show that this pattern extends to isolates that share indistinguishable SCCmec subtypes. It is possible that this pattern reflects species differences in the numerous non-SCCmec loci that are known to impact oxacillin MICs (Berger-Bächi et al., 2009). However, it is also possible that this pattern reflects species differences in the intensity of selection, which could be driven by antibiotic treatment of the more aggressive infections that S. aureus causes. These explanations are not necessarily mutually exclusive, since a favorable MIC could be reached by selection on multiple loci.