A number of factors drive the emergence and spread of antibiotic resistance, including antibiotic usage, infection control practices and the organism's genetics . Previous studies carried out in South Africa have reported large proportions of rifampicin-resistant MRSA isolates [2–5], and this study is no exception with the prevalence of rifampicin-resistance among MRSA isolates ranging from 39.7% to 46.4% (Figure 1). It is likely that the frequent use of rifampicin to treat tuberculosis in South Africa has driven the high prevalence of rifampicin-resistance among local MRSA. Support for this suggestion comes from the work of Sekiguchi et al.  who reported a significantly higher prevalence of rifampicin-resistant MRSA in tuberculosis wards compared to non-tuberculosis wards in two hospitals in Japan.
A previous study showed that ST612-MRSA-IV was the dominant clone circulating in public hospitals in Cape Town. The 44 isolates corresponding to this clonal type were uniformly resistant to rifampicin. Only one other isolate of the 100 MRSA investigated was resistant to this antibiotic and corresponded to ST5-MRSA-I . Analysis of the RRDR of 14 rifampicin-resistant MRSA (rifampicin MICs ≥ 256 mg/L), including the ST5-MRSA-I isolate, nine representatives of Cape Town ST612-MRSA-IV isolates and four previously described ST612-MRSA-IV isolates, identified three rpoB genotypes; no amino acid substitutions were detected in the two rifampicin-susceptible isolates (rifampicin MICs ≤ 0.016 mg/L) (Table 2).
The high-level rifampicin-resistant ST5-MRSA-I isolate carried a single mutational change within RpoB, H481Y. This substitution, previously associated with high-level resistance, is one of the most common rifampicin resistance genotypes and has been reported previously in several laboratory mutants and clinical isolates [11–13, 16, 17]. Molecular modelling has demonstrated that the H481Y substitution disrupts an H bond between rifampicin and RNA polymerase, and also reduces hydrophobic interactions within the binding cavity, thereby decreasing the affinity of the drug for its target .
A relatively uncommon genotype, H481N, I527M, previously reported in two clinical rifampicin-resistant MRSA from Italy  and a single vancomycin intermediate S. aureus (VISA) isolate from Brazil , accounted for 12 of the 13 high-level rifampicin-resistant ST612-MRSA-IV isolates, including N83, N84 and 04-17052. These results differ from the findings of Mick et al.  who detected four markedly different rifampicin resistance genotypes among 32 ST228-MRSA-IV isolates, expressing various levels of resistance, which were collected from a single hospital over three years.
The third rpoB genotype, H481N, I527M, K579R, was present in 09-15534, the remaining Australian ST612-MRSA-IV isolate. To the best of our knowledge, K579R, which occurs outside the RRDR, has not been reported previously, hence H481N, I527M, K579R represents a novel rpoB genotype. Whether the latter substitution impacts rifampicin resistance is unknown because the RRDR of this isolate contains two other mutations associated with resistance to this antibiotic. It is possible that this novel K579R substitution represents the latest mutational change in ST612-MRSA-IV as isolate 09-15534 was isolated in 2009, whereas the other MRSA strains included in this study were collected between 2004 and 2008.
A number of silent SNPs were detected in the 16 isolates when using the nucleotide sequence of RN4220 as a reference (Table 2). One SNP at amino acid position 498 (GCG → GCT) was common to all 16 isolates, which belonged to four different S. aureus clonal complexes (CCs) (Table 2). This SNP has also been reported in ST247-MRSA-I control strains ATCCBAA44 and PER88 (CC8), and in ST228-MRSA-I (CC5) isolates from Spain . Codon usage tables derived from genome sequences of six S. aureus control strains (NCTC8325, COL, Newman, USA300, N315 and Mu50), indicated that the codon GCT is twice as prevalent as GCG . It is possible that the SNP arose on separate occasions in multiple S. aureus lineages and that its prevalence is related to codon bias in this organism. However, it seems more likely that RN4220 contains the SNP (GCT → GCG), which arose once in this strain. This can only be confirmed when more rpoB sequences of S. aureus isolates from a variety of genetic backgrounds become available.
Of greater interest is the only other conserved silent SNP found in the codon for arginine at amino acid position 512 (CGT → CGC) that was observed in all ST612-MRSA-IV isolates (Table 2). This mutation was notable for two reasons: firstly, AT-rich organisms such as S. aureus more commonly favour AT-rich codons with either adenine or thymine bases, rather than cytosine, at the third position [21, 22]; secondly, codon usage tables indicated that CGT is more common than CGC for arginine . Thus, it is possible to suggest that the SNP (CGT → CGC) has not arisen on multiple occasions in ST612-MRSA-IV, but instead was inherited from a common ancestor and has been conserved within the lineage.
Interestingly, ST612-MRSA-IV has also recently been reported as the predominant clone in a population of horses in Australia . All of the equine ST612-MRSA-IV isolates that were tested were rifampicin-resistant, making it tempting to speculate that they may be related to those described in this study; however, the equine strains carried SCCmec type IVa , while the ST612-MRSA-IV isolates from Cape Town and Australia carried SCCmec type lished data), which suggests at least two separate SCCmec acquisitions in this genetic background.
Although mutations associated with resistance frequently evince an initial fitness cost to the organism, it has been shown that rifampicin-resistant E. coli do not revert to wild-type susceptibility in the absence of this antibiotic. Rather, they persist because of their capacity to develop compensatory mutations, which restore bacterial fitness . Other studies have also suggested that the reduction of antibiotic pressure may not necessarily result in reversion to susceptibility , which is worrying in our setting given that ST612-MRSA-IV is multidrug-resistant .
Vancomycin remains the drug of choice for the treatment of multidrug-resistant MRSA infections; however, the emergence of vancomycin-resistant S. aureus poses a new challenge. Watanabe et al.  have suggested that certain mutational changes in rpoB, including H481Y, may be linked to reduced vancomycin susceptibility in S. aureus. In light of these facts, the vancomycin MICs of isolates selected for rpoB genotyping in the current study were determined by E-test. Interestingly, the ST5-MRSA-I isolate, with rpoB genotype H481Y, was susceptible to vancomycin (MIC of 2 mg/L). Of interest is the observation that isolates with MICs of 2 mg/L have been associated with a poor clinical response to vancomycin . All ST612-MRSA-IV were susceptible to vancomycin (MICs of ≤ 1 mg/L), suggesting that the mutational changes present in rpoB in these isolates are not associated with resistance to vancomycin.