Octenidine’s Efficacy: A Matter of Interpretation or the Influence of Experimental Setups?

With its broad antimicrobial spectrum and non-specific mode of action via membrane disruption, any resistance to octenidine (OCT) seems unlikely and has not been observed in clinical settings so far. In this study, we aimed to investigate the efficacy of OCT against Escherichia coli and mutants lacking specific lipid head groups which, due to altered membrane properties, might be the root cause for resistance development of membrane-active compounds. Furthermore, we aimed to test its efficacy under different experimental conditions including different solvents for OCT, bacterial concentration and methods for analysis. Our primary goal was to estimate how many OCT molecules are needed to kill one bacterium. We performed susceptibility assays by observing bacterial growth behavior, using a Bioscreen in an analogous manner for every condition. The growth curves were recorded for 20 h at 420–580 nm in presence of different OCT concentrations and were used to assess the inhibitory concentrations (IC100%) for OCT. Bacterial concentrations given in cell numbers were determined, followed by Bioscreen measurement by manual colony counting on agar plates and QUANTOMTM cell staining. This indicated a significant variance between both methods, which influenced IC100% of OCT, especially when used at low doses. The binding capacity of OCT to E. coli was investigated by measuring UV-absorbance of OCT exposed to bacteria and a common thermodynamic framework based on Bioscreen measurements. Results showed that OCT’s antimicrobial activity in E. coli is not affected by changes at the membrane level but strongly dependent on experimental settings in respect to solvents and applied bacterial counts. More OCT was required when the active was dissolved in phosphate or Hepes buffers instead of water and when higher bacterial concentration was used. Furthermore, binding studies revealed that 107–108 OCT molecules bind to bacteria, which is necessary for the saturation of the bacterial surface to initiate the killing cascade. Our results clearly demonstrate that in vitro data, depending on the applied materials and the methods for determination of IC100%, can easily be misinterpreted as reduced bacterial susceptibility towards OCT.


Introduction
In contrast to antibiotics, antiseptics are currently thought to occupy a unique position in causing no antimicrobial resistance due to their unspecific mode of action. However, this assumption has been challenged in recent years. In particular, the widely described clinical manifestations of resistance (including even cross-resistance to antibiotics) to chlorhexidine (CHG), one of the most important antiseptic agents worldwide, has raised serious concerns among healthcare professionals about its use for preventive measures in infection control. Therefore, concerns about the existence of such antimicrobial resistance to octenidine (OCT), which has been clinically used for more than 30 years in Europe, and more recently, in  [19][20][21] d12f2 F-, proA23, lac-28, tsx-81, trp-30, his-51, rpsL173(strR), rfa-31, rfa-1, ampCp-1

OCT's Antimicrobial Activity Is Influenced by the Buffer System and Bacterial Concentration in Relation to a Given Dose of Antiseptic
To the best of our knowledge, it is unknown whether the buffer system in which OCT is dissolved has the potential to influence the activity profile of OCT. Therefore, we dissolved the antiseptic in the most frequently used fluids in bacterial in vitro research, namely water (OCT water ), phosphate-buffered saline (OCT PBS ) and Hepes buffer (OCT Hepes ) and assessed its antimicrobial efficacy against wildtype E. coli. Growth curves of 1 × 10 6 CFU/mL E. coli were recorded in the absence and presence of OCT ranging from 0.1 to 16 mg/L ( Figure 1A; for clarity only the curves using 0.5-4 mg/L OCT are shown). OCT water completely abolished growth of E. coli at 1 mg/L, whereas 2 mg/L of OCT PBS and 4 mg/L of OCT Hepes were required to observe the same inhibition. Figure 1B clearly shows that E. coli cells exposed to 1 mg/L OCT exhibit normal growth behavior when dissolved in both buffering systems, while in contrast, no growth was observed for OCT water . . MIC values represent the lowest concentration of OCT that prevented the growth of 1 × 10 6 E. coli. (C) Total inhibitory concentration (IC100%) for OCT dissolved in the indicated buffer systems against E. coli of different concentrations ranging from 1 × 10 6 -2.5 × 10 8 CFU/mL, as determined by Bioscreen at 420-560 nm. All results are representative data of at least three independent experiments. . MIC values represent the lowest concentration of OCT that prevented the growth of 1 × 10 6 E. coli. (C) Total inhibitory concentration (IC 100% ) for OCT dissolved in the indicated buffer systems against E. coli of different concentrations ranging from 1 × 10 6 -2.5 × 10 8 CFU/mL, as determined by Bioscreen at 420-560 nm. All results are representative data of at least three independent experiments.

Determined IC100% Values Are Influenced by Different Methods for Estimating Bacterial Cell Counts
In the next step, we investigated whether the OCT solvent has an impact on its antimicrobial activity in the presence of several bacterial concentrations ( Figure 1C). Interestingly, we found that higher bacterial cell counts required more OCT molecules to completely inhibit microbial growth. With increasing bacterial concentrations from 1 × 10 6 Antibiotics 2022, 11, 1665 5 of 14 to 2.5 × 10 8 CFU/mL, IC 100% for OCT water changed from 1 mg/L up to 4 mg/L, for OCT PBS from 2 mg/L to 8 mg/L and for OCT Hepes from 4 mg/L to 8 mg/L, respectively. These results show that calculated IC 100% values differ significantly between the tested OCT solutions and are even further influenced by the initial bacterial cell count. Thus, especially in the presence of very low OCT concentrations, some cells may not contact a sufficient number of antiseptic molecules and consequently, bacterial growth could be misinterpreted as reduced susceptibility or tolerance.

Determined IC 100% Values Are Influenced by Different Methods for Estimating Bacterial Cell Counts
The so called "inoculum-effect", described first for antibiotics, e.g., quinolones or penicillin [30], is a well-known phenomenon that affects the activity profile of antimicrobial peptides by changing the concentration of bacterial cell density [31][32][33]. Our previous studies [9,10] indicated that a specific number of OCT molecules per bacteria might be necessary to induce the killing cascade and consequently disrupt the membrane. As shown in Figure 1C, an increase in antimicrobial activity of OCT water (from 1/2/2 to 4 mg/L) correlated with elevated bacterial concentrations (1 × 10 6 /1 × 10 7 /1 × 10 8 /2.5 × 10 8 CFU/mL). To follow up on these observations, we first speculated about the accuracy of our standard method of determining E. coli cell count, which is routinely measured by the optical density (OD) of the cell suspension and recalculated to a CFU/mL value.
We compared results from wildtype E. coli by using the most common methods, which are: (1) absorbance/OD followed by direct recalculation of OD 600 = 1 to 8 × 10 8 CFU/mL, (2) manual counting of colonies on agar plates after incubation for 24 h at 37 • C and (3) measurement of total cells according to QUANTOM TM methodology. In fact, results from the 3 methods differ in our laboratory by 25-50% (Table 2). Based on these findings, we conducted a series of experiments using manual colony counting on agar plates as well as QUANTOM TM total cell staining to investigate the antimicrobial activity of OCT water in the presence of different wildtype E. coli cell counts. We performed microdilution assays with bacterial concentrations ranging from 1 × 10 7 to 1 × 10 8 CFU/mL (or cells/mL), respectively, with or without antiseptic at constant OCT concentrations of 0.5, 1 and 2 mg/L. When the inoculum concentration was determined by QUANTOM TM technology (Figure 2a-c), a slight delay in growth of E. coli was observed in the presence of 0.5 mg/L OCT (data not shown) which became more significant at 1 mg/L OCT. Here, bacterial growth was close to the control (at ≥4 × 10 7 cells/mL), shifted drastically at 3 × 10 7 cells/mL and was completely abolished when 2 × 10 7 cells/mL or less bacteria were used (Figure 2b). At 2 mg/L OCT no bacterial growth was observed ( Figure 2c). However, when the inoculum concentration was estimated according to the colony counting approach (Figure 2d-f), growth curves recorded in presence of 1 mg/L OCT also showed a substantial impact but never caused complete growth inhibition ( Figure 2e). This was achieved first at an initial 4 × 10 7 CFU/mL in the presence of measured directly by QUANTOM technology (a-c) or optical density following the recalculation of CFU/mL by colony counting on agar plates (d-f), respectively. Bacterial growth in the absence (a,d) and presence of indicated OCT concentrations (b,c,e,f) over 20 h as recorded by Bioscreen at 420-560 nm. The results are representative data of at least three independent experiments.

Low Numbers of OCT Molecules Are Taken Up by Bacteria
In order to estimate the maximum number of OCT molecules which are able to bind to a single bacterium, we performed two calculations based on published procedures, as described more precisely in the Materials and Methods (Section 4.4). Both approaches enable calculation of how many OCT molecules bind to a single cell ( Figure 3). As depicted in Figure 3A, the first calculation results from measuring the absorbance of OCT at 281 nm after incubation with bacteria of defined concentration (insert), showed an estimated 1-2 × 10 8 OCT molecules are required per cell to achieve IC 100% . The second calculation ( Figure 3B) is derived by modelling IC 100% values obtained from microdilution assays [33] and resulted in approximately 1 × 10 7 OCT molecules per cell. Based on these calculations, it appears that OCT molecules remain associated with the cell after triggering cell death. The difference between these two estimations is due to the applied methodology: the first approach "counts" the number of bound OCT molecules per cell after 5 min of incubation, whereas the second approach gives the number of bound OCT per cell that cause bacterial growth inhibition.
to a single bacterium, we performed two calculations based on published procedures, as described more precisely in the Materials and Methods (Section 4.4). Both approaches enable calculation of how many OCT molecules bind to a single cell ( Figure 3). As depicted in Figure 3A, the first calculation results from measuring the absorbance of OCT at 281 nm after incubation with bacteria of defined concentration (insert), showed an estimated 1-2 × 10 8 OCT molecules are required per cell to achieve IC100%. The second calculation ( Figure 3B) is derived by modelling IC100% values obtained from microdilution assays [33] and resulted in approximately 1 × 10 7 OCT molecules per cell. Based on these calculations, it appears that OCT molecules remain associated with the cell after triggering cell death. The difference between these two estimations is due to the applied methodology: the first approach "counts" the number of bound OCT molecules per cell after 5 min of incubation, whereas the second approach gives the number of bound OCT per cell that cause bacterial growth inhibition.
Interestingly, in both scenarios a larger number of unbound antiseptic molecules is estimated and only a smaller portion of OCT appears to be taken up by bacterial cells. It is important to note that this observation strongly depends on both cell count and OCT concentration. It is also important to note that overestimating the cell concentration will result in an underestimation of the number of OCT per cell, and vice versa. Figure 3A, for example, shows that in the IC100% range (about less than 2 mg/L [3.2 μM] at 1 × 10 6 CFU/mL and 1 × 10 7 CFU/mL) this portion increases up to 30-50% of total applied OCT, and diminishes only when an excess of OCT, above the IC100% value, is used. Since in these conditions, most of the OCT remains free in solution at IC100% concentration, it can be assumed that a significant degree of OCT saturation outside the cells is necessary to kill the bacteria. . The unbound fraction of OCTwater as estimated by 281 nm after exposure to 1 × 10 6 and 1 × 10 7 cells/mL dissolved in PBS. OCT was applied at concentrations in the range where it exerts antimicrobial activity against E. coli and above the IC100% (IC100% for 1 × 10 6 CFU/mL and 1 × 10 7 CFU/mL in PBS buffer is 2 mg/L). The insert shows absorbance spectra of OCT at 281 nm. The lowest spectra correspond to the lowest OCT concentrations and their absorbance intensity increased with increasing OCT concentrations. In the presence of bacterial cells, the slightly lower absorbance in all spectra was observed indicating association of OCT with bacteria. Measurements were performed at least three times. To obtain the number of molecules per cell, the bound (partitioned) fraction is calculated by subtracting the Figure 3. Estimating the number of OCT molecules binding to an E. coli cell. (A). The unbound fraction of OCT water as estimated by 281 nm after exposure to 1 × 10 6 and 1 × 10 7 cells/mL dissolved in PBS. OCT was applied at concentrations in the range where it exerts antimicrobial activity against E. coli and above the IC 100% (IC 100% for 1 × 10 6 CFU/mL and 1 × 10 7 CFU/mL in PBS buffer is 2 mg/L). The insert shows absorbance spectra of OCT at 281 nm. The lowest spectra correspond to the lowest OCT concentrations and their absorbance intensity increased with increasing OCT concentrations. In the presence of bacterial cells, the slightly lower absorbance in all spectra was observed indicating association of OCT with bacteria. Measurements were performed at least three times. To obtain the number of molecules per cell, the bound (partitioned) fraction is calculated by subtracting the unbound (free) fraction from the overall OCT concentration. In this case, the bound fraction is proportional to bacterial concentration, and results in an estimate of 1-2 × 10 8 OCT molecules per single cell at IC 100% . (B) IC 100% values as a function of cell concentration. MIC or IC 100% corresponding to OCT s concentration where no bacterial growth was observed, were calculated from growth curves recorded in MHB for OCT water in presence of different bacterial concentrations (1 × 10 6 , 1 × 10 7 , 1 × 10 8 and 2.5 × 10 8 CFU/mL). To obtain the number of molecules per cell, the overall OCT concentration is divided in bound and unbound fractions. Free OCT is independent of cell concentration and is revealed at low bacterial presence (about less than 2 mM [1.2 mg/L]). The steep increase of IC 100% instead is due to the bound OCT fraction. It is linear with cell concentration and enables the extraction of OCT molecules bound on a single cell (it is around 1 × 10 7 ). Solid lines are the best data fits, and dashed lines enclose the area accounting for the level of confidence.
Interestingly, in both scenarios a larger number of unbound antiseptic molecules is estimated and only a smaller portion of OCT appears to be taken up by bacterial cells. It is important to note that this observation strongly depends on both cell count and OCT concentration. It is also important to note that overestimating the cell concentration will result in an underestimation of the number of OCT per cell, and vice versa. Figure 3A, for example, shows that in the IC 100% range (about less than 2 mg/L [3.2 µM] at 1 × 10 6 CFU/mL and 1 × 10 7 CFU/mL) this portion increases up to 30-50% of total applied OCT, and diminishes only when an excess of OCT, above the IC 100% value, is used. Since in these conditions, most of the OCT remains free in solution at IC 100% concentration, it can be assumed that a significant degree of OCT saturation outside the cells is necessary to kill the bacteria.

Discussion
Antimicrobial resistance is rising to dangerous levels worldwide, threatening treatment options for successful patient outcomes. In that context, antiseptics constitute a suitable tool, not only in the field of infection prevention by reducing the transmission of (multi)drug resistant microorganisms in healthcare facilities, but also for local therapy. However, for some widely used antiseptics, reduced bacterial susceptibility and even (cross-)resistance to antibiotics have been reported recently.
Generally, the accurate dosage and application interval of antimicrobial agents plays an important role, as the formation of resistance is closely related to their administration at sublethal doses and/or prolonged timeframes. Using concentrations inadequate for microbial killing may lead to the survival of mutants and the induction of antimicrobial resistance [34]. Parameters such as MIC or MBC are used to describe antimicrobial susceptibility in vitro, and the obtained values can, if reliably assessed, help quantitatively describe the efficacy of a specific antimicrobial. However, the credibility of these values is dependent on multiple factors, including a determination method and the interpretation of results regarding clinical relevance [35].
Recent in vitro studies reported an increased tolerance to the clinically widely used antiseptic OCT in Gram-negative Enterobacterial species that is related to (i) efflux pump [12,13]; (ii) membrane remodeling by preventing the production of the anionic phospholipid phosphatidylserine (PS) and phosphatidylglycerol (PG) [13] and (iii) sequestering of cationic OCT molecules by components of the growth media or assay plates [17]. Of note, in all these experiments, only extremely low concentrations of OCT were used, far away from those used for clinical application (0.05-0.1%). Other studies excluded efflux pumps as a possible resistance mechanism for OCT in Gram-positive bacteria [5,14] as well in Enterobacterial species [15]. Furthermore, recent studies from our group [9,10], which indicated a non-selective phospholipid targeting of OCT, challenge the assumption by Bock et al. [13] that lipid remodeling may affect OCT s activity to such an extent that it might induce bacterial resistance. We showed that depletion of any lipid identified in B. subtilis cell envelope as lipoteichoic acid, glucolipid, lysophosphatidylglycerol, cardiolipin, phosphatidylethanolamine or PS did not strongly affect OCT s antimicrobial activity. At a molecular level, we also showed that OCT was most effective in disrupting membranes composed of neutral phospholipids, but it was also potent in breaking anionic membranes or membranes with different properties as found in mixed phospholipid membranes [9]. We concluded that OCT favours hydrophobic over electrostatic interactions, and this hypothesis is further strongly supported by the data presented in this manuscript: mutants of E. coli defective in the production of major anionic membrane lipids, truncated LPS or CL and PG are not able to impair OCT s activity at all (Table 1). Cationic compounds, such as the antimicrobial peptide PMAP23, generally show increased activity towards mutants lacking anionic lipids [36]. Therefore, the most probable explanation is that anionic lipids preclude their association to and perturbation of the cytoplasmic membrane [37]. Nevertheless, in the case of OCT, electrostatic interactions seem not to be the driving force for its mode of action, as E. coli BKT29 mutant lacking both anionic phospholipids, PG and CL, is equally susceptible to OCT as the wildtype strain ( Table 1). For that reason, in respect to lipid depletions and in relation to membrane remodelling that might occur upon defective lipid metabolism in bacteria [37,38], it is less conceivable that OCT s activity might be affected upon membrane remodeling than has been reported previously [13]. Taken together, our results clearly show OCT s capability to kill bacteria of different membrane compositions and to disrupt membranes of different composition. It is worth mentioning in this context that cytoplasmic membranes of E. coli and other Enterobacterial species such as P. aeruginosa do not significantly differ in their phospholipid composition [16,38], and certain variations of membrane composition in respect of OCT s action are less recognized.
Although standardized procedures for MIC and MBC assessment exist, mainly for antibiotics, in case of antiseptics and antimicrobial peptides, there are still some limitations that are encountered [16]. In particular, different growth media compositions and various methods for assay monitoring further complicate an accurate correlation of in vitro effects. As a result, data obtained from individual laboratories are often not compatible, especially when experiments are performed with amphipathic molecules like OCT, which can interact with many components of utilized research materials [39,40]. Indeed, an earlier publication demonstrated the importance of media composition and plate material on antiseptic activity in vitro [17]. Brill et al. [40] also noticed that a low uptake of OCT due to its binding to components of the growth media apparently led to a reduced killing rate. Altogether, these studies further highlight the need for optimized testing methods to obtain reliable and comparable results under laboratory conditions. Indeed, we have encountered and showed herein that under the very same experimental conditions, MIC values vary between 1 and 4 mg/L and even further increase when higher E. coli concentrations are used, depending on the respective buffer used to dissolve OCT (water, PBS or Hepes buffer; Figure 1). The critical micelle concentration of OCT in ultrapure water was reported at 3.79 mM (= 2 g/L = 0.2%) [41], which is about 1,000 times higher than its activity profile [4,10]. Interestingly, in preliminary tests we observed a strong potential of OCT to precipitate in different buffers, usually at concentrations above 0.01% (data not shown), at which large white floccules were visible to the eye. The same observation was made when 1% OCT stock solution in water was stored at 4 • C.
In that context, recent studies by Rzycki et al. on the simulation of artificial bacterial membranes composed of only phospholipids revealed that OCT first forms aggregates before it interacts with the membrane: the occupied area of OCT in the membrane requires 0.5 nm 2 , which is similar to the area of a standard phospholipid (0.55-0.6 nm 2 ) [42]. The authors assumed that an aggregation of OCT molecules is needed to reduce the entropy of the system. According to estimation derived from our binding studies, 1 × 10 7 bound OCT per cell can cover the outermost bacterial surface entirely (i. e.~1-2 OCT per nm 2 ) and potentially exceed the number of LPS (~1-2 OCTs per LPS), located on a surface of a typical E. coli bacterium (assuming a cylinder of 1 µm diameter and 2 µm length). Indeed, after 5 min of incubation, we determined that the number of OCT molecule per cell was about 10 times the number required to saturate the bacterial surface and cause cell death. This phenomenon was already investigated in more detail by Stella s group and others for different membrane-active antimicrobial peptides (AMPs) in E. coli as well as in artificial model membranes ( [16,[43][44][45]). These studies showed that a high membrane coverage is necessary for membrane perturbation. In any case, it was concluded that microbicidal mechanism in vitro can only be achieved when the membrane-active compound reaches the threshold concentration necessary to kill the bacteria. From this it is also plausible that the affinities to other components such as cell debris [36], or aggregation of OCT molecules affected by (salt) ions [42] might determine OCT s local concentration on the membrane for bactericidal activity. In summary, our predictive model based on theoretical calculations describes the trend observed for the cell density dependence of MIC values for membrane-active compounds, mainly based on two assumptions: (1) a threshold of cell-bound OCT molecules is needed for bacterial killing, and (2) OCT association to bacteria is approximately described by a partition equilibrium. Most importantly, our results demonstrate that the OCT dosage plays a crucial role in controlling its use.
Following our assumption that a defined quantity of OCT molecules per single bacterial cell might be necessary for killing, we further provide important evidence that various standard methods for estimating bacterial cell counts result in different values of OCT's inhibitory concentration (Figure 2). Importantly, the difference between those methods is not only in the range of a few cells, but it consists of 60 million cells/mL. Unfortunately, this is often not considered in experimental settings. As a matter of fact, this is of particular relevance when OCT is utilized at low concentrations. Our binding studies revealed that immediately after exposure of the cell to OCT at IC 100% , only approximately a quarter of the total amount of OCT molecules appear to interact directly with bacteria. However, even this number (1 × 10 8 OCT/cell) exceeds the total coverage of all lipid molecules (not only LPS) in one E. coli cell (4.5 × 10 7 lipids/cell). This once again implies that, similar to studies by Stella et al. with AMPs [31,32,36,46], a threshold for OCT concentration is needed for membrane perturbation and hence, bacterial killing. It is therefore plausible that another 60 million bacterial cells/mL require more OCT molecules to exhibit the same bactericidal effect. The apparent error of 25 to 50% associated with two different measurements of cell counts can thus drastically influence results under laboratory conditions. This might depend on the bacterial concentration exposed to OCT, the sequestering of OCT molecules by other present compounds, or the influence of potential OCT aggregation as reported previously [9,10,17,41,42]. In this context, simulation of bacterial tolerance by in vitro assays, whereby a certain volume of bacterial culture without exact determination of bacterial concentration is reinoculated over several passages and exposed to antimicrobial agents, should be very critically questioned. Caution should be taken when interpreting those data to be the result of the development of resistance, tolerance or adaptation to an investigated compound. In general, measures in global healthcare have to be taken, since recent reports indicate growing resistances also for the last-resort antibiotics like carbapenems [47,48].
In summary, our results presented herein provide evidence that OCT s antimicrobial activity in E. coli is not affected by changes at the membrane level. Mutants defective in production of major constituents of bacterial membranes, LPS, PG and CL are susceptible to OCT, as it is the wildtype. For its potent bactericidal activity, a high OCT coverage of the bacterial surface needs to be accompanied with a high proportion of OCT molecules in solution that do not directly bind to the cells. This important threshold concentration of OCT, necessary for initiating the killing cascade, might be further affected in vitro by different ions (as present in PBS and Hepes buffer), bacterial cell density or other reported factors such as the aggregation of OCT on the bacterial surface or binding to other components (e. g. growth media, plate material, cell debris). Consequently, bacterial growth in vitro, especially in the presence of very low concentrations of OCT, could be misinterpreted as tolerance or adaption. Altogether, in our view, due to its unspecific mode of action targeting critical parts of bacterial membranes, the development of resistance towards OCT is unlikely.

Preparation of Bacterial Cultures
Overnight cultures were made from freshly grown E. coli cells in Müller Hinton broth (MHB, Carlroth, Austria) under shaking at 37 • C. All strains used in the study are shown in Table 1. E. coli mutants d21, d21f2, BKT12 and BKT29 were kindly gifted by Lorenzo Stella and Maria Luisa Mangoni (University of Roma, Italy). The cells were further reinoculated in fresh MHB to OD 600 nm = 0.05 and grown for 3.5-4 h to reach the mid-log phase. Optical density of the culture was then measured at 600 nm, and cells were washed two times using PBS and their concentration was determined as described in Section 4.2.

CFU/mL by Literature Obtained Values
Colony forming units (CFU/mL) were recalculated from the absorbance measured by OD 600 nm, whereby according to the literature, an OD of 1 corresponds to 8 × 10 8 CFU/mL [49].

CFU/mL by Counting Colonies
Bacterial culture of desired optical density was plated on diagnostic agar plates and incubated overnight at 37 • C. CFU/mL was estimated by manual counting of the colonies of the respective culture.

Cells/mL Using QUANTOM ™ Total Cell Staining
The number of cells/mL was determined using the QUANTOM ™ Total Cell Staining Kit (Biocat, Austria) according to the manufacturer s instructions. The QUANTOM ™ Total Cell Staining Dye is membrane permeable and fluoresces upon binding to nucleic acids in living and dead bacterial cells. The total cell number was obtained directly from microbial cell counter QUANTOM ™ (Biocat, Austria) allowing the counting of single cells.

Assessment of Bacterial Survival in the Presence of OCT by Micro Dilution Assay
Stock solutions of bacteria were made from washed mid-log E. coli cells in fresh MHB media to the required concentrations, ranging from 1 × 10 6 to 2.5 × 10 8 CFU/mL. The bacterial stock solutions are equal to term inoculum. 90 µL of bacterial culture was incubated with 10 µL OCT stock solution to final concentration of 0.1 mg/mL to 16 mg/L OCT. Note that the final concentration of inoculum corresponds to 9 × 10 5 CFU/mL but was set to 1 × 10 6 in our experiments. Alternatively, cultures ranging from 1 × 10 7 to 1 × 10 8 cells/mL were exposed to a constant OCT concentration of either 0.5, 1 or 2 mg/L. The growth was observed at 420-560 nm over 20 h at 37 • C under shaking in Bioscreen C (Oy Growth Curves Ab Ltd., Helsinki, Finland) plate reader. The lowest concentration of OCT, that resulted in total inhibition of growth was taken as the IC 100% which is equivalent to MIC (minimum inhibitory concentration). IC 100% values obtained from the above-described microdilution assay were modelled with a previously published approach [31]. Briefly, the total dose of OCT is decoupled in unbound ([U]) and bound ([B]) molar concentrations. The latter is proportional to bacterial concentration (n cell ) and hence leads to the number of bound OCT per cell: [B] = N OCT × n cell × 10 3 /N A , where N OCT is the number of OCT per single cell and N A = 6.022 × 10 23 /mol is the Avogadro number (see [33] for further details). The factor 10 3 is needed for conversion from milliliter to liter.

Calculation of OCT Molecules Bound on Cells as Measured by Absorption of OCT at 281 nm after Incubation of OCT with Bacteria
OCT solutions of defined concentrations from 1 to 16 mg/L dissolved in water, which corresponds to 1.8 to 29 µM, were incubated without and with washed mid-log grown cells in PBS buffer of 1 × 10 6 cells/mL, 1 × 10 7 cells/mL, 1 × 10 8 cells/mL and 1 × 10 9 cells/mL for 5 min at 37 • C. Prior to measurement at 281 nm, samples were harvested at 16,000 g for 5 min and only the supernatant was taken for measurements. The spectra were recorded on JASCO V-630 spectrofluorometer (Biolab, Vienna, Austria) combined with JASCO Spectra Menager software for V-600 series (VWS-771, Biolab, Vienna, Austria). The baseline was corrected by subtraction of buffer samples. Of note, the spectra reordered for 1 × 10 8 cells/mL or higher exhibited two spectra indicating higher interference coming from cells or cellular components, which were not considered for calculations. The absorbance of OCT at 281 nm is proportional to OCT concentration in the supernatant, i.e., the unbound concentration

Conclusions
With regard to the experimental design, even seemingly trivial parameters like the method of cell determination or the buffer system used can significantly influence the interpretation of bacterial susceptibility towards an antiseptic agent. Concerning OCT, no antimicrobial resistance has been reported in clinical isolates to date, although this molecule has been routinely applied in a large field of topical applications for more than 3 decades. Therefore, data from in vitro studies must be critically questioned in terms of their practical relevance in "real world" clinical settings.
Author Contributions: N.M. is responsible for conceptualization, investigation, supervision, methodology, funding acquisition of the project and for writing of the original draft. A.Ö. and A.I. performed formal analysis with Bioscreen. D.V. did investigation, formal analysis, validation, data curation, supervision and helped by writing-review and editing. E.F.S. analyzed the data for binding studies. All authors have read and agreed to the published version of the manuscript.