Four S. aureus strains (15981, ISP479r, IPLA1, and Sa9) were selected to test the antibiofilm activity of phage-derived proteins. S. aureus 15981 and ISP479r have a clinical origin (Valle et al., 2003), while S. aureus IPLA1 and Sa9 were isolated from food environments (García et al., 2007; Gutiérrez et al., 2012). As standard culture conditions, strains were grown in TSB (Tryptic Soy Broth, Scharlau, Barcelona, Spain) at 37°C with shaking, and frozen stocks (stored at -80°C) were plated onto TSB supplemented with 2% agar (TSA) and incubated for 24 h.

Phage-derived proteins LysH5 (García et al., 2010), CHAP-SH3b, HydH5-SH3b (Rodríguez-Rubio et al., 2012b), and Dpo7 (Gutiérrez et al., 2015) were purified as previously described. Prior to biofilm treatment assays, the buffer was exchanged to 50 mM sodium phosphate (NaPi) buffer (pH = 7.4) using the Kit “Zeba^TM Spin Desalting Columns, 7K MWCO, 5 mL” (Thermo Fisher Scientific, Madrid, Spain) following the supplier’s recommendations; afterward, the supernatants were filtered using 0.22 μm PES membrane filters (VWR, Spain). The amount of protein was quantified by using the Quick Start^TM Bradford Protein Assay kit (Bio-Rad, Madrid, Spain). Quantification of the lytic activity of proteins LysH5, CHAP-SH3b, and HydH5-SH3b was performed by the turbidity reduction assay against planktonic S. aureus Sa9 cells (Obeso et al., 2008) and by MIC (minimal inhibitory concentration) determination assay which was carried out by the conventional broth microdilution technique in TSB (CLSI, 2015). The specific lytic activity of the lytic proteins was calculated as the decrease in the absorbance (OD₆₀₀ nm) per μM of protein and per minute (ΔOD × μM^-1 × min^-1); the MIC was set at the concentration (μM) where no bacterial growth was observed and calculated as the mode of three independent biological replicates.

Biofilm formation was carried out as described previously (Gutiérrez et al., 2016a). Briefly, standardized grown cultures were diluted down to 10⁷ CFU/ml in fresh TSBG broth (TSB supplemented with 0.25% glucose). Then, 100 μl of this suspension were poured into 16-well E-plates (∼10⁶ cells/well), which were then connected to the xCelligence RTCA-DP (ACEA Biosciences Inc., San Diego, CA, United States) holder pre-warmed at 37°C. Biofilm formation was monitored by recording impedance measurements of the CI every 10 min for 8 h. Biofilm formation was performed in three independent biological replicates.

The effect of phage-encoded proteins over 8 h-preformed biofilms was determined after addition of 100 μl of increasing concentrations of the four phage-derived proteins (0.04–12 μM) diluted in TSBG. Monitoring of the CI variations at 37°C was recorded every 10 min for an extra period of 16 h. A sample with NaPi buffer 50 mM pH = 7.4 diluted in TSBG was also included as a control. The RTCA software 1.2.1 (ACEA Biosciences Inc.) was used for further analysis of the data obtained. First, a time-point normalization of the CI was performed 10 min after starting the treatment (at this time point, the value of the “normalized CI” is 1). These normalized CI values were used to calculate the percentage of biofilm removal referred to the control values after 16 h of treatment. In a next step, all data were referred to the control value (“baseline CI”) by subtracting the normalized CI of each sample from the normalized CI of the control; thus, the value of the “baseline normalized CI” for the control is always 0. From these data, the specific antibiofilm activity was determined as the decrease in the baseline normalized CI per mM of protein and per minute (Δbaseline normalized CI × mM^-1 × min^-1) in the linear range of the curve. Moreover, these baseline normalized CI values were also used to calculate a dose–response curve (DRC) of the baseline normalized CI at a particular time vs concentration (% of baseline normalized CI × time vs μM of protein). The time was set according to the following two criteria: first, there should be a clear difference in the effect of the concentrations under study, and, second, the DRC should fit R² > 0.98 at the selected time point. The resultant sigmoidal DRC was calculated at this time for the three biological replicates independently; the MBEC₅₀ was finally determined by the RTCA software and expressed as the mean ± standard deviation. The MBEC₅₀ was defined previously as the concentration causing 50% reduction in the biofilm metabolic activity (Budzynska et al., 2011); herein, MBEC₅₀ represents the minimum biofilm eradicating concentration of protein that decreases the normalized CI × time by 50%. Moreover, the LOABE value was defined as the lowest concentration of protein tested that produced a detectable antibiofilm effect. This value is equivalent to the LOAEL (lowest observed adverse effect level) calculated in epidemiological or toxicological animal studies (Jeffery et al., 2004; Valdés et al., 2015).

Additionally, traditional staining with crystal violet was carried out to determine the total biomass adhered to the gold microelectrodes of the E-plate wells after protein treatment (16 h) with the modifications described previously (Gutiérrez et al., 2016a). The percentage of biofilm removal was calculated in reference to the control wells.

The SPSS Statistics for Windows V. 22.0 (IBM Corp.) package was used to perform two assessments: (1) within each strain, antibiofilm differences among phage-derived proteins, and (2) within each phage-derived protein, antibiofilm differences among strains. The data were expressed as the mean ± standard deviation and the differences were determined by one-way analysis of variance (ANOVA) followed by the Student–Newman–Keuls test for comparison of means at a level of significance p < 0.05.

In addition, the LOABE value was calculated for each protein using the values obtained from DRC at the selected time point by comparing (two by two, by one-way ANOVA) the baseline normalized CI values obtained for consecutive protein concentrations; the LOABE was defined as the first point that showed statistical differences (p < 0.05).

On the other hand, linear regression equations among different numeric parameters (percentage of biofilm reduction calculated using normalized CI vs absorbance 495 nm, and absorbance 495 nm vs specific antibiofilm activity) were calculated in order to obtain the coefficients of determination (R²) which show how well data fit to the linear regression equations.

Based on our previous experience, strain S. aureus 15981 was chosen for the screening of the effective concentration range of four phage-derived proteins as antibiofilm agents: the three lytic proteins LysH5, CHAP-SH3b and HydH5-SH3b and the polysaccharide depolymerase Dpo7. Indeed, this strain is a strong biofilm former and had previously shown a good correlation between RTCA data and other methods (crystal violet staining and viable cell counting; Gutiérrez et al., 2016a).

To establish the concentration range for the four proteins under study, we monitored S. aureus 15981 biofilm disruption with the xCelligence RTCA system. To do that, biofilms were grown in the E-plates for 8 h until early stationary phase was reached. At this point, increasing concentrations (from 0.04 to 12 μM) of the four purified proteins were added to the wells. Immediately after adding the proteins, there is a noticeable increase in the impedance signal which seems to be proportional to the protein concentration. However, this increase was actually due to the buffer used for protein storage (NaPi, 50 mM pH = 7.4), and not to protein activity. Indeed, treatment of 8 h old preformed biofilms with different dilutions of the buffer alone in TSB led to an initial increase in the impedance signal, followed by a stabilization of the values after 10 min (data not shown). As a result, to avoid the effect of the buffer and ensure that all the values obtained are protein dependent, normalization of the impedance signal was performed after 10 min. Thus, at this time (10 min) data were transformed to “Baseline normalized CI” and measurements after this point, were used to calculate the activity of the different proteins (Supplementary Figure S1).

The treatment of 8-h biofilms of S. aureus 15981 with increasing concentrations of the phage-derived proteins resulted in a dose-dependent reduction of the baseline normalized CI due to biofilm removal (Figure 1A). The minimum value was reached after 2 h for LysH5 and CHAP-SH3b and 7–8 h for HysdH5/SH3b and Dpo7; these lowest baseline normalized CI values, indicating the lowest remnant biofilm biomass, correspond with the highest protein concentration. After this time point, all impedance values remained unaltered.

To compare the efficacy of the four proteins, a numerical parameter from the baseline normalized CI vs time curves was established after defining the best time point for each protein (dashed line, Figure 1A). Then, at this time, the baseline normalized CI values were represented as a function of the protein concentration resulting in a DRC (Figure 1B). The best time point to obtain the DRC was set according to two criteria: first, the effect shown by the protein concentrations should be clearly differentiated, and, second, the resulting DRC should fit R² > 0.98. The obtained results confirmed that an increase in protein concentration led to a gradual reduction in the normalized CI values and that this decrease was higher at higher concentrations of the compounds tested. Therefore, the DRCs clearly demonstrate a dose-dependent antibiofilm effect of the proteins tested against 8 h-biofilms formed by S. aureus 15981. These curves subsequently allowed the calculation of the MBEC₅₀ and the LOABE values corresponding to the four proteins (Figure 1B and Table 1). MBEC₅₀ values ranged from 3.5 ± 1.1 to 6.6 ± 0.5 μM, obtained respectively for LysH5 and HydH5-SH3b, and showed statistically significant differences among the four proteins tested (p < 0.05). Interestingly, the biofilm removal activity of the exopolysaccharide depolymerase Dpo7 and CHAP-SH3b was similar to that observed for LysH5 (p ≥ 0.05). Moreover, within the studied concentrations range, the LOABE values were similar for all four proteins (∼1.5 μM) (Figure 1B and Table 1). This result seemed to indicate that concentrations below the LOABE values did not alter the biofilm structure, neither lysing the cells in the case of the lytic proteins, nor removing the polysaccharide matrix in the case of Dpo7.

To study a potential relationship between antimicrobial and antibiofilm activities for the phage lytic proteins (LysH5, CHAP-SH3b, and HydH5-SH3b), we determined the specific lytic activity of each protein against S. aureus 15981. The results of this assay revealed that LysH5 and CHAP-SH3b showed the highest specific lytic activity (2.5 and 1.7 units, respectively), while HydH5-SH3b exhibited a lower specific activity (0.6 units) (Table 1). This result is in good agreement with the lower antibiofilm activity observed by RTCA for this protein in comparison with the other two. Moreover, MIC values were further calculated and were surprisingly similar to the MBEC₅₀ values obtained for each protein (Table 1).

Our results, so far, confirmed the potential of the xCelligence RTCA system for screening antibiofilm compounds. In a step forward, this procedure was used to assess the spectrum of activity of the selected antibiofilm agents for other S. aureus isolates. Three additional staphylococcal strains with different ability to form biofilms were used: ISP479r is a strong biofilm producer and IPLA1 and Sa9 are weak biofilm producers. The ability of these strains to produce stronger (15981 and ISP479r) or weaker (Sa9 and IPLA1) biofilms was established after measuring the CI for 8 h. In general, the maximum CI values for strong biofilm producers in stationary phase were over 2 whereas the values obtained for weak biofilm producers were around 1 (data not shown). Preformed biofilms (8 h-old) of S. aureus 15981, ISP479r, IPLA1, and Sa9 were treated with each protein for 16 h at 37°C using a protein concentration (∼7 μM) which was twice the MBEC₅₀ for LysH5, CHAP-SH3b and Dpo7; in the case of HydH5-SH3b, the MBEC₅₀ value (6.6 μM) was used due to the impossibility to purify a more concentrated stock of this last protein. The four assayed staphylococcal strains turned out to be sensitive to all proteins since a decrease in the baseline normalized CI can be observed in all cases (Figure 2). The baseline normalized CI data were then used to define the antibiofilm specific activity (Δbaseline normalized CI × mM^-1 × min^-1), which resulted in values ranging from 0.08 ± 0.01 to 0.79 ± 0.02 units (Table 2). These values reflect the sensitivity of a given strain to each protein along time, allowing comparison of the activity spectrum among different proteins. Moreover, taking together the data obtained with different strains, we could establish similarities and differences in the activity spectra of the four proteins. When the data obtained for the four strains were combined, differences among the treatment with the four proteins (p < 0.05) were easily denoted. For instance, LysH5 showed the highest specific antibiofilm activity against all strains, while the lytic proteins derived from the VAPGH HydH5 possessed a medium specific antibiofilm activity. In contrast, the lowest specific antibiofilm activity against all strains was observed for the exopolysaccharide depolymerase Dpo7 (Supplementary Figure S2). Interestingly, the activity of the proteins against the biofilms was not dependent on the robustness of the biofilm since similar values of activity were obtained against strong and weak biofilms for each protein (Table 2).

Finally, to confirm a correlation between the data obtained with the RTCA system and the standard crystal violet staining procedure, the percentage of biofilm reduction was calculated at the point corresponding to 16 h of treatment. On the one hand, the calculation was performed using the normalized CI values before adjusting the parameters to the control baseline (Figure 3A) and, on the other hand, the total biomass was determined by crystal violet staining measuring the absorbance at 595 nm (Figure 3B). When comparing both results, in general, there were no statistically significant differences between them (p ≥ 0.05) indicating that the RTCA system can be used to calculate the percentage of biofilm reduction, similarly to the standard crystal violet method currently used. Indeed, both methods are measuring the amount of remnant biomass after treatment with phage-related proteins at a given point. In fact, there was a good correlation between the data obtained with the two systems (Supplementary Figure S3). However, as expected, there was not a good correlation between the percentages of biofilm reduction calculated with crystal violet staining and the specific antibiofilm activity (data not shown). This is not surprising since the latter is representative of death kinetics where three parameters are taken into account: the decrease in impedance, the concentration of protein used in the experiment and the time spent to reach the lowest impedance value.