We collected MRSA isolates from consecutive unrelated patients in intensive care unit with MRSA BSI at 10 tertiary hospitals from 2007 to 2009. MRSA BSI was considered to be present if one or more blood cultures had positive results, and if clinical signs and course were consistent with MRSA infection. Further case definitions were defined, as previously described [11].

Identification and antimicrobial susceptibility testing were done using Vitek II (bioMérieux, Craponne, France) or MicroScan Pos Combo Panel Type 6 (Siemens, Munich, Germany) and confirmed by PCR for the presence of mecA gene. We used the following S. aureus strains: ATCC25923, ATCC29213, ATCC43300 (mecA, positive control) from the American Type Culture Collection (Manassas, VA, USA). In addition, antimicrobial susceptibility testing were performed with the disc diffusion method, if needed, according to the Clinical and Laboratory Standards Institute guidelines. S. aureus ATCC 29213 were used as quality control strains for MICs.

The biofilm formation assay in microtiter wells was performed as previously described [12]. First, overnight cultures were diluted 1:100 in brain heart infusion broth (Becton Dickinson and Co., Franklin Lakes, NJ, USA) supplemented with 1% glucose. Cell suspensions (200 μL) were transferred to individual wells of a flat-bottom 96-well polystyrene microtiter plate (Nunclon; Nunc, Roskilde, Denmark). After incubation at 35 °C for 48 hours, detached cells were gently rinsed three times with sterile water, and the bacteria that attached to the surface were stained with crystal violet, rewashed, and destained with 1 mL of ethanol-acetone (95:5, vol/vol). A total of 200 μL of the mixed solution was transferred to a 96-well microtiter plate for spectrophotometric analysis at optical density (OD) 570 nm. The absorbance was recorded by Micro-ELISA autoreader (Titertek Multiscan Plus; Labsystems, Helsinki, Finland). Each assay was performed in triplicate and the mean OD₅₇₀ value of tested wells was applied to biofilm-forming ability. Uninoculated medium was used to determine background. The biofilm formation were divided into three categories in this study, the strains with OD₅₇₀ < 0.2, 0.2 ≤ OD₅₇₀ < 1.0, and OD₅₇₀ ≥ 1.0 were defined as biofilm non-formers, biofilm formers of week level, and strong level, based on the ODs in brain heart infusion broth with and without two supplements. S. aureus SA113 (ATCC35556) and Staphylococcus epidermidis RP62A (ATCC35984), well-characterized biofilm-forming strains, were purchased from American Type Culture Collection for use as positive controls.

SCCmec types were determined by using a multiplex PCR strategy according to the method described by Oliveira and de Lancastre [13]. The agr type (1–4) was assigned by PCR as previously described [14]. Multilocus sequence typing was carried out according to protocol previously described [15,16]. Data were analyzed by comparing the database at the multilocus sequence typing website (http://saureus.mlst.net), and the sequence type (ST) for each strain was determined.

All 126 strains were analyzed by PCR assay as previously described [12,17]. These included nine staphyloccocal enterotoxin genes (sea, seb, sec, sed, see, seg, seh, sei, sej), three exfoliative toxin genes (eta, etb, etd), six adhesin genes (icaA, icaD, cna, atl, fnbA, fnbB), two surface-associated genes (cap5HK, cap8HK), and two staphylococcal regulators (sarA, arlRS).

Medical, laboratory, and pharmacy records were reviewed. Data from patients infected with biofilm-forming and non-forming MRSA isolates were compared. The data collected included age, sex, primary site, and onset of infection. In addition, clinical outcome was assessed for all assessable cases according to biofilm-forming capacity. The outcome measures used were crude mortality, MRSA-related death, and eradication of MRSA. Healthcare-associated risk factors included the following: presence of invasive devices (central venous catheter, urinary catheter, and other indwelling devices), prior use of antibiotics, residence in long-term care facility, prior hospitalization, prior surgery, receipt of hemodialysis, and prior MRSA colonization.

Data analysis was performed using Statistical Package for the Social Sciences (SPSS) Software version 10.0 (SPSS Inc., Chicago, IL, USA) and MedCalc Statistical Software version 10.0.1.0 (MedCalc Software Inc., Mariakerke, Belgium). Statistical significance was assessed via the Pearson χ² test or the Fisher exact test for categorical variables and the Student t test or the Mann–Whitney U test for continuous variables. Logistic regression analysis was used for multivariate analysis. Variables that achieved a probability of <0.1 in univariate analyses were considered for inclusion in logistic regression models.

A total of 126 MRSA isolates from blood cultures were analyzed; 86 strains (68.3%) had biofilm-forming ability with five strong and 81 weak level and the other strains (40/126, 31.7%) did not form biofilms (Figure 1). All isolates were resistant to oxacillin and penicillin. Biofilm-forming isolates were less frequently resistant to oxacillin and penicillin (14.0% vs. 42.5%; p = 0.005), and they were more likely to be multidrug resistant (three or more non-β lactam antibiotics) than biofilm nonforming ones (85.0% vs. 57.5%; p < 0.001; Table 1).

The genetic characteristics of 126 MRSA strains were differed between biofilm-forming and nonforming isolates (Table 2). Most biofilm-forming (55 isolates; 64.0%) and biofilm-nonforming (20 isolates; 50.0%) isolates were of agr group II, with no difference in the distribution of the agr group. The most common SCCmec type was II in both biofilm-forming and nonforming isolates, and the next most frequent type was SCCmec type IV (p = 0.055). Biofilm-forming isolates were more likely to contain ST5 (69.8% vs. 52.5%; p = 0.060), ST239 (8.1% vs. 2.5%; p = 0.434) and significantly less likely to contain ST72 (18.6% vs. 42.5%; p = 0.005) than biofilm nonforming isolates.

Table 3 presents the distribution of virulence-associated genes (adhesin-encoding, toxin-encoding, surface-associated, and gene regulators). With a range over 90%, most of the isolates had similar distribution of adhesion genes (icaA, icaD, cna, atl, and fnbA), toxin genes (SEs, hla, and hlb), and staphylococcal regulators (sarA and arlRS) between biofilm-forming and nonforming isolates and both isolates showed low frequency in detection of eta, etb, etd, PVL, and cap8HK. The biofilm-forming isolates were significantly more detected in distribution of fnbB than biofilm-nonforming isolates (74.4% vs. 45.0%; p = 0.001; odds ratio 3.556; 95% confidence interval 1.615–7.827).

Epidemiological and clinical data were available for 126 patients and were included in the analysis of factors associated with biofilm formation. Patient characteristics are shown in Table 4. Biofilm-forming isolates were more frequently detected in nosocomial infection (77.9%) than biofilm-nonforming isolates (60%). Among 86 nosocomial infection, biofilm-forming isolates were involved in 67 cases (77.9%), which was higher than community onset cases (22.1%). There were no significant differences in sites of infection. CRI and primary infection were two most common origins of MRSA infection regardless of biofilm-forming ability. As shown in Table 4, there were no significant differences between clinical outcomes of biofilm-forming and nonforming cases for crude mortality and MRSA-related death, in univariate analysis shown in Table 5, significant risk factors associated with healthcare-associated biofilm-forming isolates were the presence of invasive devices (central venous catheter, urinary catheter, and other indwelling devices) and prior hospitalization (p < 0.05). Odds ratios (95% confidence interval) for the variable selected in multivariate analysis were: presence of invasive devices, 3.87 (1.60–8.96); and prior hospitalization, 0.40 (0.18–0.88).