Staphylococcus aureus possesses the ability to form biofilms and is responsible for chronic infections which are hard to treat and cause significant morbidity and mortality.

Biofilms are communities of bacteria which adhere to surfaces and are encapsulated in a self-produced extracellular polysaccharide matrix. They constitute an important strategy implemented by microorganisms to survive in harsh environmental conditions (Donlan and Costerton, 2002). Biofilms are responsible for chronic, recurrent infections and are known to survive very high concentrations of antibiotics (Lewis, 2008; Lebeaux et al., 2014). One hypothesis to explain the lower activity of antimicrobial drugs on biofilms is the high prevalence of persister cells in biofilms (Lewis, 2008; Singh et al., 2009). These persisters, unlike resistant bacteria which are genetically modified, consist of clones of bacteria expressing a different but reversible phenotype which allows them to transiently escape the effects of antibiotics (Lewis, 2008).

The antibiotic therapies currently used against biofilm infections are often associated with poor clinical responses and frequent relapses (Davies, 2003). For several years, different solutions have been proposed to eradicate biofilm bacteria such as phages, quorum sensing inhibitors or physical methods (Ivanova et al., 2017). However, although highly innovative strategies still need to be developed to deal with severe infections by both tolerant and multi-resistant bacteria, the method which can most rapidly and easily be implemented in patients at present is to combine existing drugs or to modify their therapeutic regimen (dose, frequency, and mode of administration).

In the case of suspected S. aureus infection, vancomycin therapy is often initiated in patients to provide antibacterial activity against both Methicillin-Sensitive S. aureus (MSSA) and Methicillin-Resistant S. aureus (MRSA) (Deresinski, 2009). However, although vancomycin can kill planktonic bacteria, its activity against Biofilm-Embedded Bacteria (BEB) is quite low. Lebeaux et al. (2015) showed that after exposure to a very high, constant concentration of vancomycin (5000 mg/L) for 24 h, the percentage of bacteria surviving in a 24 h-old S. aureus biofilm exceeded 20% and was even close to 100% for 2 of the 4 tested strains. Singh et al. (2009) reported similar results and found no statistically significant difference between the bacteria remaining in a non-treated S. aureus biofilm or in a biofilm exposed for 24 h to vancomycin concentrations equal to or higher than those clinically achievable. Post et al. (2017) demonstrated that vancomycin is able to eradicate a mature biofilm of S. aureus from metal implants by using a static concentration of 200 mg/L over 28 days. Nevertheless, such a concentration profile cannot be achieved by systemic administration or local delivery vehicles currently available. To overcome this poor activity on biofilms, an aminoglycoside is often combined with vancomycin. Synergistic activity between vancomycin and aminoglycosides had already been demonstrated on S. aureus (Watanakunakorn and Glotzbecker, 1974; Cokça et al., 1998) but these studies were performed by exposing planktonic bacteria for no more than 24 h to constant antibiotic concentrations whereas in the in vivo situation, antibiotic concentrations continuously fluctuate over several days. The effects of a combination of gentamicin and vancomycin on S. aureus were more rarely tested under dynamic in vitro conditions with varying antibiotic concentrations or in animal models of infection. No significant synergy was observed in two studies where low inocula of S. aureus were exposed to the two drugs (Backo et al., 1999; Aeschlimann et al., 2000). Another study on large inocula of MRSA and MSSA, representative of a biofilm- associated infection, was performed in an in vitro simulated endocardial vegetation model. The effect of vancomycin alone was statistically significant compared to the control after 3 days but the activity of vancomycin on MSSA or MRSA was not improved by adding gentamicin (LaPlante and Woodmansee, 2009). However, in this study, the vancomycin concentrations tested were almost two times higher than the free and active concentrations routinely obtained in patients because no correction was performed for the 45% plasma protein binding of vancomycin (Liu et al., 2002; Butterfield et al., 2011).

To propose new treatment optimizations, the predictivity of in vitro experiments needs to be improved, for example by exposing both planktonic and BEB in parallel over the complete duration of treatment (several days), to drug concentrations identical to those that would be encountered under clinical conditions in patients.

In this study, we studied the effects of amikacin, an aminoglycoside, and vancomycin on planktonic and biofilm-embedded S. aureus by using an in vitro dynamic model, the Hollow-Fiber (HF) infection model, which mimics the fluctuations of antibiotic concentrations over time, as would occur in the plasma of patients during a 5-day treatment. The HF model was recently labeled by the European Medicines Agency (European Medicines Agency, 2015; Gumbo et al., 2015) for drug dosage optimization in the treatment of tuberculosis. We have further adapted this model to explore drug activity not only on planktonic but also on biofilm-embedded S. aureus. Indeed, in previous studies conducted in HF (Nicasio et al., 2012; Ferro et al., 2015), the bacteria were systematically exposed to drugs during the exponential phase of growth, when there was no time for biofilm development, whereas in this study, the biofilm was allowed to form for 3 days before drug exposure. The killing effects of drugs and the potential selection of resistance were assessed both on planktonic bacteria over time and on BEB at the end of exposure. We first compared monotherapy and combinations of amikacin and vancomycin at the currently recommended dosing regimens, i.e., 1g vancomycin twice a day and 15 mg/kg amikacin once a day for 5 days. Such therapeutic regimens are considered sufficient to achieve the PK/PD indices classically expected to obtain drug efficacy. For aminoglycosides, the most predictive PK/PD index is the Maximal Concentration (C_max) divided by the Minimal Inhibitory Concentration (MIC) ratio (Moore et al., 1987) and a value from 8 to 10 is usually recommended to ensure efficacy against the pathogen (Toutain et al., 2002). For vancomycin, the best predictive index is the AUC over 24 h divided by the MIC (AUC_24h/MIC) (Nielsen et al., 2011), and value of 400 is recommended to achieve clinical effectiveness (Rybak et al., 2009; Jung et al., 2014; Song et al., 2015).

We then explored the effects of a slight deviation from these standard dosages by simulating an increased dose of amikacin, which is a concentration-dependent antibiotic, (Frimodt-Møller, 2002) and by modifying the mode of administration (infusion vs. bolus) of vancomycin, which is a time-dependent antibiotic (Waineo et al., 2015).

Due to the refractoriness of S. aureus biofilm infections to antibiotic treatments, there is an urgent need to optimize the use of currently available drugs to ensure bacterial killing and the prevention of resistance. In this study, we developed an innovative use of the HF model by delaying exposure to the antibiotics and studied the effects of a combination of vancomycin and amikacin both on planktonic bacteria and on BEB in conditions representative of clinical situations. Different concentration profiles of the drugs were tested, and bacteria were subjected to the fluctuating concentrations that might be encountered in patients during a complete treatment. These experimental conditions should have greater predictive value than simple static assays in which bacteria are exposed to a fixed concentration over time. Moreover, due to the lack of medium renewal in static assays, such experiments are often conducted over 24 h whereas longer periods are needed to assess the selection of resistance by antibiotics (Drusano, 2017). Compared to animal models, which may exhibit very different pharmacokinetics to humans and in which some human pathogens cannot develop, all bacteria can be cultured in the HF model and exposed to drug concentration profiles that mimic the range of human profiles (Toutain et al., 2010). For example, as vancomycin is eliminated much faster in mice (half-life = 32 min) than in humans (Knudsen et al., 2000), dosage regimens tested in mice can hardly be extrapolated to humans. Obviously, the main weakness of static or dynamic in vitro assays is the absence of the immune system which can cooperate with antibiotics to clear an infection.

Several in vitro studies in dynamic systems including the HF model (Nicasio et al., 2012; Lenhard et al., 2016) have investigated the antibacterial activity of drugs combined with vancomycin against planktonic S. aureus. However, the use of dynamic in vitro systems, such as the CDC biofilm device or others, to study the effects of combinations on biofilm is rarely reported. To our knowledge, the present study is the first to use a HF model to conduct experiments on a 3-day old biofilm of a single S. aureus strain to assess the activity of drugs combination, over 5 days, on both planktonic bacteria and BEB. The HF model had already been used to simulate in two distinct studies the free concentration-time profiles of amikacin or vancomycin that can be achieved in patients receiving the recommended doses (Nicasio et al., 2012; Ferro et al., 2015). In our study, exposure to different dosage regimens of a susceptible strain of S. aureus with MICs of 1 μg/mL for amikacin and vancomycin led to equal or higher values of the PK/PD indices than those classically expected to obtain drug efficacy (Zelenitsky et al., 2003; Rybak et al., 2009; Song et al., 2015). For aminoglycosides, for which the most predictive PK/PD index is the C_max/MIC ratio (Moore et al., 1987), we targeted C_max/MIC values from 70 to 130 in the HF model whereas a value from 8 to 10 is usually recommended to ensure efficacy against the pathogen (Toutain et al., 2002). For vancomycin, for which the best predictive index is the AUC over 24 h divided by the MIC (AUC_{24 h}/MIC) (Nielsen et al., 2011), we targeted the value of 400 recommended to achieve clinical effectiveness (Rybak et al., 2009; Jung et al., 2014; Song et al., 2015) and obtained AUC_{24 h}/MIC values ranging from 372 to 417 for the bolus of vancomycin in the HFIM and 480 for the constant infusion. Even though these targeted values of the PK/PD indices were attained for both drugs, almost no bactericidal activity was observed on the 3-day old biofilm or on the co-existing planktonic bacteria when amikacin or vancomycin were administered alone for 5 days. These results are in agreement with previous studies which demonstrated the low activity of vancomycin on large bacterial inocula (LaPlante and Rybak, 2004; LaPlante and Woodmansee, 2009) and on biofilms (Hogan et al., 2016). One study involving a HF model showed that a peak concentration as high as 80 mg/L was needed to achieve bactericidal activity against a large inoculum of a MRSA strain with a MIC of 1 μg/mL for vancomycin (Lenhard et al., 2016). One proposed explanation for the inoculum effect and reduced efficacy of vancomycin is that bacteria at high density are in a stationary growth phase with low dividing rate and low cell wall synthesis (Brown et al., 1988; Lamp et al., 1992). Another explanation is that vancomycin may be sequestrated by S. aureus on peptidoglycan layers, thus reducing the free vancomycin concentrations surrounding the bacteria (Srinivasan et al., 2002; Ekdahl et al., 2005; Yanagisawa et al., 2009). Finally, a reduced penetration of vancomycin through S. aureus and S. epidermidis biofilms has also been described (Doroshenko et al., 2014; Singh et al., 2016) and, even worse than the lack of efficacy, low concentrations of vancomycin were reported to stimulate biofilm formation in some clinical isolates of S. epidermidis (Cargill and Upton, 2009). In this study on S. aureus, our results were concordant as the biofilm which was exposed to vancomycin alone contained 10 times more bacteria than the control.

The lack of efficacy of the drugs used in monotherapy in this study supports the clinical recommendation to associate an aminoglycoside with vancomycin for the treatment of S. aureus biofilm infection (Deresinski, 2009). Compared with the absence of activity of amikacin or vancomycin alone, exposure to combinations of vancomycin and amikacin for 5 days in the HF model had a synergistic bactericidal effect on the planktonic bacterial populations. However, despite this synergy, the planktonic bacteria remaining after 5 days of exposure to the combination (D7) still exceeded 2.5 log₁₀ CFU/mL. We therefore investigated the ability of other dosage regimens of amikacin and vancomycin to improve the antibacterial efficacy against this planktonic population. Contrary to our expectations, given the concentration-dependent activity of aminoglycosides, increasing the C_max of amikacin 1.8-fold (from 70 to 130 μg/mL) did not increase the efficacy on planktonic bacteria. For vancomycin, the efficacy of the combination seemed to be slightly decreased by constant rate infusion, especially on planktonic bacteria, but there were not enough replicates to draw a definitive conclusion. Contrary to the planktonic population, the addition of vancomycin (as a bolus or constant infusion) to amikacin did not result in an additional bacterial reduction on S. aureus biofilm, and no synergy between the two drugs was observed. The distinct activity of the combination on planktonic bacteria and BEB confirmed the different phenotypes of these two populations of bacteria and that the drugs were less active on BEB. Indeed, biofilms are supposed to contain more persister bacteria which have lower growth rates and are therefore less affected by antibiotic drugs (Singh et al., 2009; Conlon et al., 2015). Moreover, no dosing regimen tested in this study, even if it exceeded the recommended PK/PD index values, was able to fully eradicate the planktonic bacteria co-existing with a biofilm, which could suggest that some planktonic bacteria were continuously released from the biofilm. As our study is the first one focusing on the biofilm in the HF, microscopy imaging will be further needed to investigate the distribution of the biofilm in the HF cartridge, which could be influenced, among others, by the shear forces in the extracapillary space. It should also be kept in mind that our system was characterized by an absence of the immune system and the presence of a rich medium – more favorable to bacterial growth -, that both limit the efficacy of antibiotic treatments compared to the in vivo situation. However, our in vitro results are in agreement with the reported lack of efficacy of systemic antibiotic treatments in patients for whom additional treatments, such as mechanical removal of biofilms or very high local antibiotic concentrations, are advised whenever possible (McConoughey et al., 2014; Wu et al., 2015).

In addition to efficacy, we assessed the ability of the combination to reduce the selection of resistant bacteria in planktonic and biofilm populations. The absence of resistance to vancomycin in this study was in accordance with other experiments conducted on S. aureus (LaPlante and Rybak, 2004). Conversely, bacteria (approximately 10^-6) able to grow on agar supplemented with 6 μg/mL (sixfold MIC) of amikacin were systematically present in the planktonic and biofilm populations before drug exposure, implying that small proportions of such bacteria are spontaneously present in large populations, as previously reported (Ferro et al., 2015). Since similar proportions were also found at the end of the control experiments, it suggests that the growth and survival rates of less-susceptible and fully susceptible bacteria were similar in the absence of drugs. After 5 days of antibiotic exposure, the MIC of amikacin for these bacteria able to grow on agar supplemented with amikacin and termed “less-susceptible,” never exceeded the resistance breakpoint (>16 μg/mL). These bacteria showed an intermediate amikacin-susceptibility with regard to the EUCAST breakpoints, implying that the administration of amikacin to patients infected by these bacteria would have an uncertain therapeutic effect (Rodloff et al., 2008), but it should also be stressed that the initial MIC of the tested strain was low (1 μg/mL). This suggests that the same selection phenomenon occurring on a strain with a two or four-fold higher MIC would lead to the selection of “true” resistant bacteria. The selection of less-susceptible bacteria, which represented the main population of planktonic bacteria and BEB after exposure for 5 days to amikacin in monotherapy, could be explained by an inducible mechanism of resistance, known as adaptive resistance, in which thickening of the cell wall results in less penetration of amikacin into the bacterial cell (Yuan et al., 2013). Interestingly, the addition of vancomycin to amikacin considerably reduced the proportions of these less-susceptible bacteria in both planktonic bacteria and BEB compared to amikacin alone, especially when vancomycin was administered in boluses. These results suggest that vancomycin was able to limit the growth of these bacteria less-susceptible to amikacin and prevent their selection. The vancomycin administered by CRI associated with amikacin seemed to limit the selection of less-susceptible bacteria to a lesser extent, but these differences require more thorough investigation.

By studying planktonic bacteria and BEB in parallel and by mimicking the fluctuations in antibiotic concentrations over 5 days, as can occur in vivo after daily administrations, we demonstrated the increased efficacy of a combination of amikacin and vancomycin on planktonic bacteria but not on BEB. However, even though vancomycin did not increase the killing activity of amikacin on BEB, it reduced the selection of bacteria less-susceptible to amikacin, which could help to maintain the efficacy of this drug during treatments. Even if these results need to be further confirmed with clinically relevant strains of MSSA and MRSA, they highlight the importance of selecting combination therapies not only based on efficacy but also on resistance selection endpoints by taking into account the 2 co-existing populations of planktonic bacteria and BEB.

Equations:

With HF being the Hollow-Fiber, CR the Central Reservoir and ECS the Extra-Capillary Space.

DB, AF, FW, FE, P-LT, and AB-M: substantial contributions to the conception or design of the work. DB, ML, AF, P-LT, and AB-M: acquisition, analysis, or interpretation of data for the work. DB, ML, FW, FE, P-LT, AB-M, and AF drafting the work or revising it critically for important intellectual content. Final approval of the version to be published. Agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.