The present study reports several important findings regarding the antimicrobial susceptibility, resistance mechanisms, sequence type distributions, and clonality of MRSA strains obtained from ICU respiratory infections in Spain and Italy. At the participating ICUs, S. aureus was not found to be resistant to first-line antibiotics such as linezolid and vancomycin. Although the prevalence of MRSA in the participating centres was low, the mechanisms of resistance described may also be representative for sites with high MRSA prevalence because the MRSA collected during this study corresponded to highly disseminated clonal complexes (CC22 and CC5). In addition, these strains harbored a wide range of antimicrobial mechanisms to second-line antibiotics, including ciprofloxacin, erythromycin, gentamicin, tetracycline, clindamycin, and fusidic acid.
Although our strains did not show resistance to linezolid, the detection of subpopulations resistant to this antimicrobial is a finding that merits comment. Hetero-resistance (HR) is an unstable phenomenon with a high incidence in several bacterial strains, according to recent reports. It is considered unstable because subpopulations defined as hetero-resistant in one susceptibility test may no longer appear as such if the test is repeated. The recent finding of plasmid-associated HR mechanisms emphasizes the problem, since these HR mechanisms may spread horizontally between pathogens. The lack of routine determination by many laboratories and the decrease in antimicrobial activity may have clinical implications. Another important feature is that the decrease in antimicrobial activity caused by this phenomenon is not reflected in the MIC value. Previous research also suggests that HR can indeed be responsible for treatment failure in S. aureus infections. However, this phenomenon is not detected by established procedures and new methods are needed for rapid identification of HR in pathogenic bacteria .
Few reports on MRSA resistance mechanisms contain an exhaustive evaluation of antibiotics. In our study we often identified a combination of several resistance mechanisms for MRSA strains, such as spontaneous mutations that decrease bacterial replication, transferable and chromosomal efflux pumps, and antimicrobial or target-modifying enzymes. In terms of antimicrobial susceptibility, similar results have been reported in other studies. For instance, in a series of MRSA strains isolated from different samples, Kitti et al.  found high levels of resistance to ciprofloxacin (72.1%), erythromycin (86.9%), gentamicin (72.1%), and clindamycin (86.9%).
In agreement with Sierra et al. , we found mutations in the quinolone resistance-determining regions (gyrA, grlA, gyrB, grlB) of ciprofloxacin-resistant S. aureus strains. Several reports have indicated that topoisomerase IV is the primary target for quinolone resistance in Gram-positive microorganisms, including S. aureus with DNA-gyrase acting as secondary target, with specific point mutations at GrlA (subunit A of the topoisomerase IV) and GyrA (subunit A of the DNA-gyrase) as the most relevant . However, our results differ because the mutation S144P in grlA gene may be a polymorphism, given that it is found in both susceptible and resistant strains. This means that the primary target in our strains is the DNA-gyrase. Studies in Japan have not shown mutations in the gyrB and grlB genes , which in any case tend to be infrequent in MRSA strains. Nevertheless, some of our strains showed more than three mutation points in each gene. Therefore, further studies are needed to confirm whether these mutations determine resistance or genetic polymorphisms. In addition, in gyrA and grlA, some rare mutations were described (Table 2). Because antibiotic combinations are used during nosocomial pneumonia treatment, these strains are exposed to strong antibiotic selection pressure which may contribute to the high number of mutations found here compared with prior studies. For instance, it has been demonstrated that hospital-acquired MRSA harbors higher levels of antimicrobial resistance than community-acquired MRSA .
Analyzing 206 strains of S. aureus from different centers in Canada, China, and France, Martineau et al. attributed erythromycin resistance to ermA (98%), ermB (21%), ermC (2.4%), and msrA (1%) genes. In our study, erythromycin resistance in S. aureus was mediated by the ermA (15%), ermC (62%) and msrA (23%) genes. In an Algerian study, erythromycin resistance was much lower (37.8%) than in our study (65%). However, the authors of the Algerian study included S. aureus from food, nosocomial, and community-acquired infections and identified only the ermC gene [6, 19]; the fact that we only isolated strains from nosocomial pneumonia, whereas the other studies used different sources, could explain the differences observed.
Similarly, Yilmaz et al. also included S. aureus from different clinical samples and found the ermC and ermA genes in strains with resistance to clindamycin . They reported a lower prevalence (6%) of resistance to clindamycin compared to ours (20%), and detected the lnuA gene instead of the erm genes [20, 21]. Some S. aureus strains have shown inducible resistance to clindamycin after exposure to subinhibitory concentrations of erythromycin . In our study, the D-test  revealed two MRSA isolates with inducible resistance to clindamycin. This finding is important because clindamycin is used in the treatment not only of pneumonia but also of muscle, bone, skin, and soft tissue infection.
In our study, gentamicin resistance was related to the aac(6′)/aph(2′′). Choi et al. detected higher proportions of the aac(6′)/aph(2′′) gene in MRSA isolates from blood, sputum, urine, and pus samples (65%) than we did in ETT specimens (35%), but they also found prevalences of ant(4′)-Ia and aph(3′)-IIIa of 41 and 9% respectively . Yilmaz et al. found that four of six MRSA isolates carried the same aac(6′)/aph(2′′) gene. Nevertheless, our findings are consistent with those of Martineau et al., who observed a higher number of S. aureus isolates, among which all those with gentamicin resistance had the aac(6′)/aph(2′′) gene .
Tetracycline resistance in S. aureus at our ICU was mediated only by the tetK gene. By contrast, other studies have found different proportions of involvement of the tetK or tetM genes alone or in combination; for instance, Strommenger et al. identified ten strains of S. aureus carrying tetK, tetM, or both genes . Yilmaz et al. identified nine strains of S. aureus with the tetM gene and ten with the tetK gene . Finally, Achek et al. detected both the tetK and tetM genes in ten S. aureus isolates from clinical samples .
Although fusB was initially thought to be the only gene to encode a protein capable of protecting EF-G, a whole family of related fusB-like proteins has since been described. Thus, mutations in two more genes (fusC and fusD) can lead to staphylococcal resistance to fusidic acid . Several studies have also reported an increase in resistance to fusidic acid. We suggest a chromosomal location of the fusB gene because the primers we used were developed in previous studies by O’Neil et al. in which the fusB gene was detected in total DNA preparations but not in plasmid DNA preparations, indicating a chromosomal location for this resistance determinant (different fusB genes have been discovered on plasmid pUB101 and plasmid pUB102). Some previous data suggest that chromosomal fusB was associated with epidemic strains of S. aureus . Interestingly, in another study fusB-type resistance (fusB and fusC) was found in 87% of MRSA isolates , with an association between fusB and clonal complexes CC45 and CC97. By contrast, we found only one strain with fusB, and this was the singleton ST1535. Fusidic acid is a topical drug that is used for the treatment of staphylococcal skin infections, but its increased use appears to have led to the emergence and dissemination of resistant staphylococci .
The molecular epidemiology of MRSA in bloodstream infections has been described previously, but less frequently in respiratory infections contracted in the ICU [27, 28]. CC22 is one of the largest circulating clonal complexes associated with hospital-acquired MRSA in Europe (UK) and Asia (Kuala Lumpur, China) [11, 29], while studies of nosocomial pneumonia indicate that CC5 is associated with MRSA strains originating mostly from the US, Europe (Portugal), Asia (China), Africa (Algeria) and Latin America (Argentina and Chile) [18, 19, 30, 31]. Despite the marked heterogeneity of the sequence types in this study, CC22 and CC5 were the main clonal complexes detected. Specific resistance mechanisms can be associated with clonality, since a higher number of these mechanisms were found in the widely expanded CC5 and CC22 clones than in the CC45 and CC59 clones . Consistent with our results, previous studies (in the US, Portugal and Japan) have found CC5 and CC59 to be associated with extensive multi-drug resistance, but not CC45 . Another important point to stress is that these CCs had previously been associated with virulent S. aureus strains.
The heterogeneity of MRSA sequence types at each hospital suggests that ICU cross-transmission has decreased, probably due to the introduction of VAP prevention bundles, isolation measures, and hospital hygiene measures over the last 10 years. Thus, our study indicates that other sources of MRSA transmission such as nasal carriage constitutes risk factors for ICU and nosocomial pneumonia. Although we did not assess nasal MRSA carriage in our study, it has been shown to be an independent risk factor for ICU pneumonia in previous work .
This study has some limitations. First, the number of strains is relatively low because S. aureus and MRSA are infrequent causes of nosocomial pneumonia in Spain. However, we also included strains from Italy and found a high heterogeneity of sequence types, which may be representative of the current clones circulating as causes of hospital-acquired MRSA in Europe. Second, although we did not assess the virulence of our MRSA strains, some of the clonal complexes identified, such as the CC59 and CC45, have been shown to be closely related to virulent strains. Immune evasion cluster (IEC) genes have been associated with CC59 (IEC-hemolysin genes) and CC45 (IEC-enterotoxin-hemolysin genes) .
Despite the limitations mentioned, we think that this study is important for establishing the epidemiology of S. aureus. Little recent information is available on the resistance mechanisms of action of S. aureus strains obtained from mechanically ventilated patients, and it is unclear whether or not these mechanisms are associated with particular circulating S. aureus clones. We also observed the presence of linezolid hetero-resistance and high resistance to second-line antibiotics in MRSA strains isolated from endotracheal tubes in humans mechanically ventilated for long periods in the ICU. These findings show that MRSA infection is still relevant in southern Europe, with a high capacity of resistance to different antimicrobials, an extensive battery of resistance mechanisms, and a wide clonal variability.