Moxifloxacin HCl (microbiological standard; potency: 90.9%) was provided by Bayer HealthCare (Leverkusen, Germany). Resazurin sodium salt, crystal violet solution, calcofluor white (CFW), proteinase K, and cation-adjusted Muller Hinton broth 2 (CA-MHB) were obtained from Sigma-Aldrich (St-Louis, MO); Tryptic Soy Broth (TSB), Tryptic Soy Agar (TSA), and multiwell plates, from VWR (Radnor, PA); bovine serum albumin, from Thermo Fisher Scientific Rockford, IL); 5-cyano-2, 3-ditolyl tetrazolium chloride (CTC) (RedoxSensor vitality kit), from Invitrogen (Carlsbad, CA); and Quick Start™ Bradford Protein Assay, from BioRad (Hercules, CA).

The methicillin-susceptible S. aureus (MSSA) ATCC 25923 and the methicillin-resistant S. aureus (MRSA) ATCC 33591 were used as references. Two strains expressing GFP under the control of a tetracycline-inducible promoter (RN4220-palc and 1214-palc, Nguyen et al., 2020b), were used in specific experiments. In addition, 18 clinical, non-isogenic, isolates of S. aureus collected from patients with persistent or recurrent infections at the Bach Mai hospital, Hanoi, Vietnam, were included in the study. The patients were hospitalized in urology or rheumatology, and the collected samples included mainly infected catheters, pus, tophi, abscesses, or blood. These isolates were previously characterized (Supplementary Table 1) by molecular spa and agr typing; their MRSA character was established by measuring cefoxitine MIC and detecting mecA and mecC by PCR (Nguyen et al., 2020a). The MIC of moxifloxacin was previously determined (Nguyen et al., 2020b) by broth microdilution following CLSI recommendations (Clinical and Laboratory Standards Institute, 2020) and susceptibility breakpoints were those established by EUCAST;¹ the relative persister fraction was previously established (Nguyen et al., 2020b) by measuring the number of CFU after 5 h incubation of a stationary phase culture with moxifloxacin at 100 times its MIC, according to a published procedure. In this method, the persister character of each isolate was evaluated by its persister fraction relative to that of a reference (De Groote et al., 2009), here ATCC 25923, to allow for a more accurate comparison of persister fractions determined in independent experiments. The relative persister fraction of ATCC 33591 is 1 (same as that of ATCC 25923), so that this strain was used as a reference rather than ATCC 25923 in specific experiments because it forms a biofilm more comparable to clinical isolates in terms of CFUs counts (see Supplementary Figure 1). Isolates were categorized in three groups according to their persister and resistant character (Nguyen et al., 2020b), respectively, referred to as susceptible with low relative persister fraction [S-LP; MICs ≤ 0.25 mg/L; persister fraction relative to ATCC 25923 ≤ 10 (6 isolates)], as susceptible with high relative persister fraction [S-HP; MICs ≤ 0.25 mg/L; persister fraction relative to ATCC 25923 > 10 (3 isolates)], and resistant with high relative persister fraction [R-HP; MICs > 0.25 mg/L; persister fraction relative to ATCC 25923 > 10 (9 isolates)]. Metabolic activity was assessed by following the capacity of each isolate to metabolize the non-fluorescent substrate resazurin in fluorescent resorufin (λ_exc 560 nm/λ_em 590 nm) (Bauer et al., 2013). Planktonic cultures at different OD_{620 nm} were incubated with resazurin (10 mg/L in PBS) for up to 60 min and fluorescence was measured over time in a M3 SPECTRAmax microplate spectrofluorometer (Molecular Devices LLC, Sunnyvale, CA).

Biofilms were obtained using as a starting inoculum bacteria transferred from frozen stocks onto TSA and incubated overnight at 37°C, after which 10 colonies were inoculated in TGN (TSB supplemented with 2% NaCl and 1% glucose), and the bacterial density of the starting inoculum was adjusted to an OD_{620 nm} of 0.005 (approx. 10⁷ CFU/mL). Biofilms were grown in 96-well plates at 37°C for 24 h (Bauer et al., 2013), after which the medium was renewed (controls) or replaced by a medium containing moxifloxacin for 24 h.

These analyses were performed on control biofilms after 2 successive cycles of 24 h of growth with mid-period renewal of the medium as described above. Biofilm biomass was evaluated by measuring the absorbance of crystal violet at 570 nm using a M3 SPECTRAmax microplate spectrofluorimeter as previously described (Bauer et al., 2013), except that we used 200 μL of crystal violet (Sigma-Aldrich) at 10% (V/V, final concentration 2.3 g/L) in water to stain the dry biofilms for 15 min. Bacterial viability was determined by CFU counting (Nguyen et al., 2020a). Biofilms were gently washed twice with sterile phosphate buffer saline (PBS), collected in sterile water, and sonicated 5 min using a sonication bath to insure the liberation of bacteria from the matrix. After appropriate dilution in PBS, aliquots were spread on TSA and CFU were counted after overnight culture.

To obtain a gross estimation of their polysaccharide content, biofilms were grown in 96-well plates, the medium was removed and wells were washed twice with PBS. Biofilms were fixed by heat at 60°C for about 2 h (Bauer et al., 2013). Polysaccharides were then stained by Calcofluor White (CFW), a fluorophore that preferentially binds to β-1,3 and β-1,4 polysaccharides (Maeda and Ishida, 1967), using a procedure based on that of Stiefel et al. (2016). In brief, 200 μL of CFW (0.25 mg/mL in H₂O) was added in each well and incubated during 15 min in the dark, after which the staining solution was removed, and the plates were washed twice with PBS. Two hundreds μL of absolute EtOH were added per well and incubated for 5 min in the dark to resolubilize the dye, after which fluorescence was read (λ_exc 360 nm/λ_em 460 nm) in a M3 SPECTRAmax microplate spectrofluorometer.

To quantify proteins, biofilms formed in 96-well plates were washed thrice with sterile PBS and detached with 200 μL cold sterile distilled water by pipetting. Proteins were quantified using the Quick Start Bradford Protein Assay according to the manufacturer’s instructions, except that 50 μL of samples were mixed with 200 μL of reagent. After 15 min incubation in the dark under agitation at 100 rpm, absorbance was read at 595 nm, using a SPECTRAmax microplate spectrofluorometer and protein content was calculated using a standard curve obtained with bovine serum albumin.

To quantify extracellular DNA (eDNA), biofilms were grown in 4 mL TGN in 6-well plates, then washed thrice with PBS, detached with 1 mL cold sterile distilled water by rapid pipetting. eDNA was purified by a method adapted from Kreth et al. (2009). Briefly, samples were incubated at 37°C for 1 h in the presence of 5 μg/mL proteinase K and centrifuged at 16,000 g for 2 min. The supernatant was collected and eDNA extracted with phenol:chloroform:isoamyl alcohol (25:24:1). Samples were centrifuged at 16,000 g for 5 min, and 800 μL of aqueous phase were collected. eDNA was precipitated by the addition of 500 μL isopropanol and 80 μL 3 M sodium acetate, pelleted by centrifugation at 16,000 g for 10 min, air-dried, and re-suspended in 50 μL nuclease-free H₂O. The purity of DNA in each sample was checked on agarose gels and its concentration determined using a NanoDrop™ spectrophotometer (Thermo Fisher Scientific; Waltham, MA).

Moxifloxacin activity was determined against 24-h biofilms, as previously described (Bauer et al., 2013). In brief, the culture medium was removed and replaced by a control medium or a medium containing moxifloxacin at increasing concentrations (0.001–1,000 mg/L). Biofilms were reincubated for 24 h at 37°C. Remaining biomass was quantified by crystal violet staining, and viability, by CFU counting using the same procedures as for control biofilms. Data were plotted as concentration-response curves and used to fit Hill equations, which allowed us to calculate pharmacodynamic parameters, like the maximal efficacy (E_max; i.e., maximal reduction in CFU or crystal violet absorbance as extrapolated for an infinitively large antibiotic concentration), and relative potency, estimated by the calculation of C_–1log (i.e., concentration needed to reduce of 1 log₁₀ CFU the bacterial counts in biofilms) or of C_20% (i.e., concentration needed to reduce of 20% crystal violet absorbance).

The presence of the ica locus in clinical isolates was checked by PCR. Genomic DNA was extracted from overnight cultures in CA-MHB using the DNeasy Blood and Tissue Kit (Qiagen, Hilden, Germany) following the manufacturer’s instructions. The purity of DNA in each sample was checked on agarose gels. icaA was amplified by PCR using 1 μL of DNA, 2.5 μL of 10x DreamTaq Buffer, 0.8 μL of 10 mM dNTPs, 1 μL of 10 mM of each primer (icaA-F :CGAGAAAAAGAATATGGCTG; icaA-R: ACCATGTTGCGTAACCACCT), 0.125 μL DreamTaq DNA polymerase (Thermo Fisher Scientific, Waltham, MA), and 18.58 μL sterile nuclease-free water (Thermo Fisher Scientific) and the following PCR program: denaturation at 98°C for 1 min, 30 amplification cycles (98°C for 10 s, 60°C for 30 s, and 72°C for 60 s), and a final cycle of 72°C for 10 min. PCR products were visualized on agarose gel. As icaA was detected in all isolates, the expression of the gene was then quantified by quantitative real-time PCR (q-RT-PCR). To this effect, RNAs were isolated from 3 h-old biofilms grown in 6-well polystyrene plates (4 mL of culture at OD_{620 nm} = 0.005). Biofilms were collected by pipetting, pelleted by centrifugation at 4,000 rpm for 7 min, washed twice in sterile phosphate buffered saline (PBS) and pelleted again. Cell pellets were resuspended in 100 μL of solution containing 270 μg lysozyme and 10 μg lysostaphin (Sigma) and incubated 30 min at room temperature to achieve bacterial lysis. Total RNA was isolated using The InviTrap^® Spin Universal RNA Mini Kit (Stractec, Berlin, Germany). Samples were cleaned from DNA using the TURBO DNA-free™ Kit (Thermo Fisher Scientific). RNA quality and quantity were checked by measuring A_{260 nm} and A_{280 nm} using a NanoDrop spectrophotometer. Purified RNA was converted to cDNA using the transcription first strand cDNA synthesis kit (Roche Applied Science) with random hexamer primers according to the manufacturer’s instructions, and then diluted 10 times. Quantitative PCR reactions were performed in triplicates in 96-well plates (Greiner) using 5 μL of cDNA, 12.5 μL of SYBR Green Master Mix (Bio-Rad), 2 μL of 5 mM of each primer, and 3.5 μL of sterile RNase-free water (Ambion). gmk gene (encoding a guanylate kinase) was used as an housekeeping gene and amplified with following primers: gmk-F: TCAGGACCATCTGGAGTAGGTAAAG; gmk-R: TTCACGCATTTGACGTGTTG. For icaA, we used the same primers as those described above for PCR amplification of gene and the following program: denaturation at 95°C for 3 min, 40 amplification cycles at 95°C for 15 s and 60°C for 60 s. A melting curve was run at the end of the PCR cycles to check for the presence of a unique PCR reaction product. Relative expression levels of icaA were calculated using the ΔΔCt method, with gmk expression levels as a control and compared with the expression level measured in the reference strain ATCC 33591.

According to a previously described procedure (Siala et al., 2014), 24 h biofilms were grown on coverslips, incubated during 1 h with 20 mg/L moxifloxacin, washed twice with 1 mL PBS and stained for 30 min in the dark with 1 mL 0.5 mM CTC, a colorless, non-fluorescent and membrane-permeable compound, which is reduced by viable bacteria to fluorescent, insoluble CTC-formazan (Kim et al., 2008). Stained biofilms were then washed with 1 mL PBS and then examined in a Cell Observer^® SD confocal fluorescent microscope (Carl Zeiss AG, Oberkochen, Germany) using spinning disc technology (Yokogawa Electric Corporation, Tokyo, Japan) and controlled by the AxioVision software (AxioVs40 V 4.8.2.0; Zeiss). Excitation/emission wavelengths were set at 415 nm/500–550 nm for moxifloxacin and 488 nm/570–620 nm for CTC-formazan. Moxifloxacin concentrations within biofilms were then calculated using calibration curves built using moxifloxacin solutions at concentrations ranging from 5 to 50 mg/L and examined in the microscope using the same settings as for samples.

We used two clinical isolates with susceptible low-persister (RN4220) and susceptible high-persister (1,214) phenotype, respectively, which were transformed by the pALC2084 to express Green fluorescent protein (GFP) under a xyl/tetO inducible promoter (Nguyen et al., 2020b). They were used to form biofilm as described above except that they were maintained in the presence of 125 ng/mL tetracycline (to induce GFP production) and 10 μg/mL chloramphenicol (as a selecting agent) during the 24 h of biofilm growth. They were then incubated with moxifloxacin in the absence of tetracycline and chloramphenicol during 24 h, after which bacteria were isolated from biofilms as described above and used for CFUs counting and flow cytometry analysis as previously described (Nguyen et al., 2020b) using a FACSVerse cytometer (BD Biosciences) and the FlowJo 10.5.2 software (TreeStar Inc., Ashland, OR) for data analysis.

In a first step we compared 18 clinical isolates (9 susceptible and 9 resistant to moxifloxacin) and the reference strain ATCC 25923 for biofilm formation after 48 h of incubation with renewal of the medium at 24 h. Bacterial survival was first assessed by measuring their metabolic activity toward the non-fluorescent dye resazurin (reduction into fluorescent resorufin). However, since a large proportion of the clinical isolates showed a high relative persister fraction, which, by definition, implies a decreased metabolic activity, we checked in preliminary experiments whether this method could be used to correctly quantify the amount of living bacteria in biofilms. As shown in Supplementary Figure 2, the fluorescence signal generated by the clinical isolates was not only often lower than that of ATCC 25923, but was also highly variable among isolates, irrespective of the proportion of persisters they could generate. Thus, the method was found unusable for quantitative determination of viable bacteria in our experimental setting, as already reported by others in similar experiments (Sandberg et al., 2009; Alonso et al., 2017). Bacterial viability was therefore evaluated using CFU counting, while biomass was measured using crystal violet staining. Supplementary Figure 1 shows that all clinical isolates were high biofilm producers, with crystal violet absorbance values being globally 2 or 3 times higher than for the two reference strains (ATCC 33591 and ATCC 25923) (Supplementary Figure 1A), and CFU counts reaching values 10-fold (1 log₁₀) higher than for ATCC 25923, but similar to those recovered from ATCC 33591 biofilms (Supplementary Figure 1B). An in-depth analysis of biofilm formation by clinical isolates depending on their persister or resistant phenotype is presented in Figure 1, with reference strains added as internal controls. A significant correlation was observed between biomass and CFU counts (Figure 1A). We then examined the correlation between biomass or CFU counts in the biofilms and the relative persister fraction (RPF) of each isolate (Nguyen et al., 2020b) or the MIC of moxifloxacin. A highly significant correlation was noticed between biomass and RPF (Figure 1B), and to a lower extent, between biomass and MIC (Figure 1C). In contrast, the correlations between CFU and the two same characteristics of the isolates did not reach significance (Figures 1D,E).

In order to better characterize the differences in biomass production among isolates, we analyzed their main matrix constituents in the same conditions. A strong correlation was found between crystal violet absorbance and fluorescence of calcofluor-white (CFW), which binds to β-1,3 and β-1,4 polysaccharides, including PNAG (Figure 2A), but not with the protein or eDNA contents in the biofilms (Supplementary Figure 3). The CFW signal was also significantly correlated with the RFP of clinical isolates, and to a lesser extent, with the moxifloxacin MIC (Figures 2B,C). Since poly-N-glucosamine is synthesized by enzymes encoded by the icaADBC operon, we quantified the expression of icaA for all clinical isolates after 3 h of incubation, knowing that the expression of this gene is upregulated mainly during the early stages of biofilm development (Kot et al., 2018; Idrees et al., 2021). These expression levels were then compared to that of ATCC 33591, since this reference strain produced a biofilm containing a similar amount of viable bacteria as the clinical isolates (see Supplementary Figure 1B) and the same persister fraction as ATCC 25923 (RPF = 1). A significant correlation between icaA expression and CFW fluorescence was noticed (Figure 2D); p-values were slightly > 0.05 for correlations between icaA expression and the RPF of the isolates or moxifloxacin MIC (Figures 2E,F).

We then evaluated the effects exerted by moxifloxacin on bacterial viability and biomass in biofilms formed by clinical isolates in comparison with ATCC 25923 and 33591. To this effect, 24 h-old biofilms were exposed to this antibiotic over 24 h, using a broad range of concentrations in order to study the entire concentration-effect relationships (Figures 3A,B). The insure graph readability, isolates were also stratified here according to their phenotype [susceptible with low RPF (S-LP); susceptible with high RPF (S-HP), or resistant with high RPF (R-HP)] (Figures 3C–H). Except for biomass in biofilms developed by resistant isolates (R-HP), a quadratic Hill function with slope factor set to 1 could be fitted to all data (Figure 3 and Table 1 for the definition and values of the corresponding pertinent pharmacodynamic parameters used in our analysis). Considering ATCC 25923 first, moxifloxacin caused a concentration-dependent decrease in both CFUs and biomass, with EC₅₀ (i.e., inflection point of the Hill curve) and C_–1log or C_20% values (indicators of the relative potency) being both at low multiples of the moxifloxacin MIC. At high concentrations, the maximal reduction in viability (E_max) reached −2.7 log₁₀ CFU and biomass was reduced to 34% of its control value (biofilms incubated in the absence of antibiotic). Against ATCC 33591 (Table 1), moxifloxacin was as effective (similar E_max) as against ATCC 25923 regarding CFUs but less against biomass; it was also was considerably less potent for both criteria (higher C_–1log and C_20% values). For susceptible clinical isolates, and focusing first on the effect exerted by moxifloxacin on viability (Figures 3C,D), the relative maximal efficacy was almost reached at the highest concentration tested and not different between S-LP and S-HP subpopulations [E_max: −2.6 to −2.9 log₁₀ CFU for S-LP and S-HP isolates, respectively, see also Supplementary Figure 4 and Supplementary Table 2 where data are not normalized as difference from the control (not exposed the antibiotic), illustrating the fact that the actual counts of CFU are higher in clinical isolates]. In contrast, the relative potency of moxifloxacin was markedly decreased since 28–46-fold larger concentrations than for ATCC 25923 were needed to decrease the CFU counts to 10% (1 log₁₀ decrease) of their original value, with again no significant difference between susceptible isolates with low (S-LP) or high (S-HP) RFP. For resistant isolates with high RFP (R-HP), the relative potency of moxifloxacin was low, so that only the upper part of the “best-fit” Hill function could be documented by actual data points, making the determination of the maximal relative efficacy value uncertain (Figure 3E). Considering the effect of moxifloxacin on biomass (Figures 3B,F–H), only minor reductions were seen for biofilms formed by susceptible isolates, and no significant effect for biofilms formed by resistant isolates. Restricting the analyses to changes observed at a moxifloxacin concentration corresponding to its average human C_max [4 mg/L (Sullivan et al., 1999); see the black thin vertical line on the graphs], the antibiotic caused only marginal decreases in bacterial counts (less than 0.5 log₁₀ CFU decrease against susceptible isolates and no reduction against resistant isolates) and no effect on biomass in all cases.

In a next step, we measured the concentrations reached by moxifloxacin in the deepness of biofilms, again comparing those formed by isolates belonging to each the three phenotypic groups to that formed by ATCC 25923. To this effect, biofilms were examined in confocal microscopy, taking advantage of the intrinsic fluorescence of moxifloxacin, and using the reduction of 5-cyano-2,3-ditolyl tetrazolium chloride (CTC) to formazan to detect living, metabolically-active, bacteria (Figure 4A). Biofilms were exposed to a concentration of 20 mg/L moxifloxacin to allow for its reproducible detection by fluorescence in the biofilm (Siala et al., 2014). Moxifloxacin concentration was high in biofilms from the reference strain and from S-LP isolates, reaching values well above the MIC of these isolates. Moxifloxacin concentration was lower in biofilms from isolates with high RPF, but still above the MIC for S-HP isolates, which was no more the case for R-HP isolates (Figures 4B,C).

Using these data together with those from Figures 1, 3, we then examined the relationships between moxifloxacin antibacterial activity in the biofilm (reduction of CFU at 20 mg/L) and its concentration in the biofilm. We observed a significant correlation between these two parameters (Figure 4D). In addition, the concentration of moxifloxacin in the biofilm was significantly inversely correlated with the biomass of each individual biofilm, and even more with the fluorescence of CFW (Figures 4E,F).

The persister character of the isolates had been determined in stationary phase cultures. We therefore wondered whether this persister character was maintained when bacteria were encased into biofilms and exposed to antibiotic pressure in this environment. As persisters are characterized by a non-dividing status (Roostalu et al., 2008), we followed bacterial multiplication within the biofilm, using previously constructed transformants of a S-LP strain (RN4220) and of a S-HP strain (1214) having the same susceptibility to moxifloxacin (MIC, 0.064 mg/L) and expressing GFP under the control of a tetracycline-inducible promotor (Nguyen et al., 2020b). These two strains also produce similar biofilms regarding biomass as well as CFU counts in control conditions (10^7.3–7.4 CFU/mL, both in the presence or in the absence of tetracycline; CV absorbance: 6.7–6.9 or 6.1–6.3 in the absence of or in the presence of tetracycline, respectively).

We first compared the kinetics of activity of moxifloxacin at 100 mg/L against stationary phase cultures and biofilms of these two strains (Figure 5A). As previously described (Nguyen et al., 2020b), a plateau was observed for both strains after 5–10 h of incubation with moxifloxacin in stationary-phase cultures, with no further killing (up to 24 h), which is one of the key hallmarks of persisters (Balaban et al., 2004; Brauner et al., 2016). The percentage of persisters was approx. 0.01 and 0.04% of the initial population for the S-LP and the S-HP isolates, this difference being small but nevertheless statistically significant (p-value: 0.01; t-test). Flow cytometry profiles (Figure 5B) recorded at the beginning of the experiment (time 0) and after 24 h incubation with moxifloxacin at 100 × its MIC (6.4 mg/L) or at 100 mg/L showed only a minor decrease in fluorescence (minimal shift of the normalized fluorescence profile) for strain RN4220 (S-LP) and no decrease for strain 1214 (S-HP). Thus, there was only minimal or no dilution of the tracer, confirming that survivors only minimally divided or did not divide during the incubation time.

Moving now to data obtained with the same strains in biofilms, we see that killing occurred much more slowly (after a drop of approx. 1 log₁₀ CFU occurring in about 1–2 h) than for bacteria in stationary phase culture, proceeding at a constant rate during the rest of the 24 h incubation period and without significant difference between the two strains (Figure 5A). Bacteria collected from 24 h-biofilms before addition of moxifloxacin (time 0) and after 24 h incubation with moxifloxacin were also examined in flow cytometry [Figure 5C; the GFP inducer was present during biofilm formation but removed at the time of addition of moxifloxacin (time 0)]. At time 0, most of the bacteria were highly fluorescent due to the maintenance of the inducer during biofilm growth. A partial dilution of the signal was nevertheless noted, especially in the S-HP isolate, which we attribute to a probable suboptimal penetration of the inducer in the growing biofilm. After 24 h incubation in the presence of 100 mg/L moxifloxacin, the fluorescence signal of the survivors was modestly reduced for the S-LP strain RN4220-pALc2084 and even more slightly for the S-HP strain 1214-pALc2084, indicating a limited division capacity in moxifloxacin-exposed biofilms, especially for the HP isolate.

The tolerant character of a strain toward an antibiotic can be estimated by the determination of the minimum duration of incubation with an antibiotic to achieve a predetermined extent of killing (MDK; Fridman et al., 2014). The table in Figure 5 shows that MDKs for a 90, 99, and 99.9% reduction of CFU in the presence of moxifloxacin at 100 mg/L were considerably shorter for bacteria in stationary cultures than for those in biofilms. In stationary cultures, MDK were short because they were calculated for reductions in CFU that still correspond to the elimination of the bulk of the population (preceding the plateau with infinitively long MDK against the persister subpopulation). In biofilms, MDKs were longer, demonstrating antibiotic tolerance even for small reductions in CFUs. In addition, the rate of killing as well as the reduction in CFUs reached at 24 h were similar for both strains.