Atomic-scale simulations of reactive oxygen plasma species interacting with bacterial cell walls

In recent years there has been growing interest in the use of low-temperature atmospheric pressure plasmas for biomedical applications. Currently, however, there is very little fundamental knowledge regarding the relevant interaction mechanisms of plasma species with living cells. In this paper, we investigate the interaction of important plasma species, such as O3, O2 and O atoms, with bacterial peptidoglycan (or murein) by means of reactive molecular dynamics simulations. Specifically, we use the peptidoglycan structure to model the gram-positive bacterium Staphylococcus aureus murein. Peptidoglycan is the outer protective barrier in bacteria and can therefore interact directly with plasma species. Our results demonstrate that among the species mentioned above, O3 molecules and especially O atoms can break important bonds of the peptidoglycan structure (i.e. C–O, C–N and C–C bonds), which subsequently leads to the destruction of the bacterial cell wall. This study is important for gaining a fundamental insight into the chemical damaging mechanisms of the bacterial peptidoglycan structure on the atomic scale.

Currently, one of the main applications of LTAPP sources is the sterilization of medical tools, i.e., the treatment of heat-sensitive materials or disinfection from both gram-negative and gram-positive bacteria (see e.g. [4,6,[10][11][12][13][14][15]). Many experimental studies have been performed on the decontamination of bacteria by altering the external plasma process parameters (e.g. gas mixture composition, driving frequency and voltage). However, gaining control over the processes occurring in the plasma itself and especially in the plasma region in contact with the bio-organism still remains a challenge. Solving this issue solely by means of experiments is extremely difficult. Indeed, even simple plasma characteristics such as temperature and density cannot be measured easily with common measuring tools such as probes, due to the small scale of LTAPPs. Even if such measurements can be carried out, they lead to possible disturbances of the plasma from the measuring tool [16]. On the other hand, non-invasive methods (such as optical emission spectroscopy) can be used instead, in order to measure plasma characteristics (e.g. the electron density, see [17]). However, these methods are limited in the information they can provide and cannot give useful information, especially regarding the processes occurring in the contact region between the plasma and the bio-organism.
Computer simulations, however, are ideally suited to tackle these issues. Indeed, plasma simulations and simulations of plasma-surface interactions can provide fundamental information about the processes occurring in the plasma and at the surface of living cells. However, until now very limited effort has been spent on modeling the plasma itself and in particular on its interaction with living organisms, such as bacteria [16]. Moreover, simulating the interaction of plasma with bacteria is very difficult compared to modeling solid 3 state materials. However, if a proper interatomic potential can be constructed for accurately describing all relevant interatomic interactions, molecular dynamics (MD) simulations can provide atomic scale insights into them.
Based on these considerations, we investigate the interaction of plasma species with bacterial peptidoglycan (PG), as a model system for plasma-bacteria interactions. For this purpose, reactive MD simulations based on the reactive force field (ReaxFF) potential are carried out [18], using the ReaxFF glycine-force field, as developed by Rahaman et al [19]. The plasma species under study are O 2 and O 3 molecules, as well as O atoms, which were previously identified as being biomedically important (see [11,[20][21][22][23][24]). In this work, we investigate the interaction of the above-mentioned species with PG, which is the outer part of the cell wall of gram-positive bacteria such as Staphylococcus aureus.
It should be mentioned that our simulations are performed under 'perfect' (or 'optimized') conditions. In reality, there might be many interactions between plasma species and liquids surrounding the bacteria-e.g. skin, wounds and other biological conditions. However, in this paper we do not take this into account. Our next study will, however, be devoted to the interaction of plasma species with PG surrounded with water molecules. Indeed, the moist environment could prevent the PG destruction in reality, since oxygen species (O, O 2 and O 3 ) might be less exposed to the PG in the bacterial cell wall. On the other hand, it is known that under some conditions, the plasma jet can blow away (part of) the liquid film [25]. Thus, the present simulations provide detailed insights into the mechanisms of bacterial cell wall damage by plasma species when the liquid film is not present or is blown away, or when it is of little importance.
More information about the PG structure and an overview of the computational details are presented in sections 2 and 3, respectively. The results are discussed in section 4, and finally the conclusion is given in section 5.

Structure of the peptidoglycan
PG, also known as murein, is an important component of the bacterial cell wall. It is found on the outside of almost all bacterial membranes [26][27][28]. It forms a mesh-like layer composing the cell wall and serves as a protective barrier in bacteria. The PG layer is substantially thicker in grampositive bacteria than in gram-negative bacteria. For instance, in the gram-positive bacterium S. aureus, the PG structure is typically 20-30 nm thick, whereas in a gram-negative bacterium, such as Escherichia coli, it is only 6-7 nm thick [29][30][31].
As mentioned above, in this paper we consider the interaction of plasma species with the gram-positive bacterium S. aureus PG. Its chemical structure can be found in [31][32][33]. A schematic picture of the PG structure is presented in figure 1. It is assembled from repeating units consisting of a disaccharide, a stem and a bridge. The disaccharide is composed of β(1-4) linked N-acetylglucosamine and N-acetylmuramic acid (GlcNAc-MurNAc), the stem is the pentapeptide l-alanine-d-iso-glutamine-l-lysine-d-alanine-d-alanine (l-Ala 1 -d-iso-Gln 2 -l-Lys 3 -d-Ala 4 -d-Ala 5 ), and the bridge is a pentaglycine (Gly 1 -Gly 2 -Gly 3 -Gly 4 -Gly 5 ) interpeptide. The pentaglycine bridge, branching off the ε-amino group of the l-Lys of the stem peptide, connects one PG chain to the d-Ala 4 of a neighboring chain (figure 1). It is important to note that this composition is often found in nascent PG, whereas the last (fifth) d-Ala 5 residue is lost in the mature macromolecule [28]. Hence, in this paper we assume a tetrapeptide stem instead of a pentapeptide stem (see figure 1). Unfortunately, the exact three-dimensional structure (or tertiary structure) of the PG is still unclear and remains elusive, although many experimental investigations have been performed to define the chemical structure and to study the physical properties of the PG. One of the reasons for this is that it is not possible to distinguish the strands and peptides with conventional electron microscopy due to the similarity in dimensions of these structures and the occurring artifacts of this technique [29,34]. However, various experimental studies have partially characterized the PG structure (see e.g. [32]). Moreover, based on these experimental studies, several models have been proposed for the three-dimensional structure of the PG, such as the vertical scaffold and the horizontal layered models [28][29][30][34][35][36]. However, these models can explain only certain properties of the PG.
In our simulations, we model the PG structure as consisting of only two disaccharides with tetrapeptide stems (see figure 1, left and right sides), connected with one pentaglycine interpeptide (see figure 1, center). Note that with this construction we are able to take into account all possible atomic bonds in the PG structure.

Interatomic potential
In an MD simulation, the trajectories of all atoms in the system are calculated by integrating the equations of motion. Forces on the atoms are derived from the ReaxFF potential [18]. This force field has now been successfully applied to describe nearly half of the periodic table of the elements and their compounds, including hydrocarbons [18,37], metals and metalcatalyzed reactions [38,39], metal oxides [40], metal hydrides [41] and silicon and silicon dioxide [42][43][44]. Recently, it has also been used for organic molecules, such as glycine [19,45], as well as for complex molecules, such as DNA [46]. The ReaxFF parameters are optimized to obtain good general agreement with quantum mechanical calculations for reaction energies, barriers and structures (in that order of importance).
The ReaxFF potential uses the bond order/bond distance relationship formally introduced by Abell [47]. The total system energy is the sum of several partial energy terms that include lone pairs, undercoordination, overcoordination, valence and torsion angles, conjugation and hydrogen bonding. Moreover, non-bonded interactions, namely Coulomb and van der Waals energy terms, are also taken into account. These interactions are calculated between every pair of atoms, such that the ReaxFF potential is capable of describing not only covalent bonds, but also ionic bonds and the whole range of intermediate interactions. The electronegativity equalization method [48,49] is used to calculate charge distributions based on geometry and connectivity. A detailed description of the force field parameters used in this study can be found elsewhere [19].

Simulation method
In our simulations, the PG structure is placed in a box with dimensions ∼75 Å × 88 Å × 51 Å. Periodic boundary conditions are not applied to any plane as no infinite surface is needed in the simulations. However, a few atoms in the structure are fixed to keep the structure from drifting. The fixed atoms are chosen such that they are positioned at the periodically repeating parts of the PG structure, i.e. O and H atoms in d-Ala 4 , MurNAc and GlcNAc, as well as two H atoms in l-Lys (see figure 1, red dashed circles).
Prior to the particle impacts, the structure is equilibrated at room temperature (i.e., 300 K) as follows. First, the structure is thermalized in the isothermal-isobaric ensemble (i.e., NPT dynamics) for 100 ps to equilibrate the temperature of the system and to obtain a structure at zero stress. The obtained structure is subsequently equilibrated in the NVT ensemble using the Berendsen heat bath [50] for 40 ps. Finally, the resulting structure is relaxed in the microcanonical ensemble (i.e., NVE dynamics) for 20 ps. The temperature relaxation constant is set to 0.1 ps in all temperature controlled simulations, i.e. during the thermalization, as well as during the particle impact simulations. In all simulations, we used a time step of 0.1 fs.
In all simulations, the impacts of the plasma species are performed as follows. Ten incident particles (e.g. ten O 3 molecules) are randomly positioned at a minimum distance of 10 Å around the PG structure and also from each other. This distance ensures that there is initially no interaction between the plasma species and the PG structure due to the long distance interactions (i.e. Coulomb and van der Waals interactions). The initial energy of impinging plasma species corresponds to room temperature and their velocity directions are chosen randomly. To study all possible damaging mechanisms of the PG and to obtain statistically valid results for 6 bond-breaking processes, we performed 50 runs for each plasma species (i.e. for O 2 , O 3 , as well as for the O atoms). Every simulation trajectory lasts 300 ps, corresponding to 3 × 10 6 iterations. This time is long enough to obtain a chemically destroyed PG structure, at least if a critical bond in the structure is broken (see below). Thus, at the end of the simulation all plasma species interacted with the structure, possibly resulting in the breaking of various bonds as described below in section 4.

Results and discussion
It is clear from figure 1 that the most important bonds in the PG structure are C-C, C-N and C-O. As far as the structural integrity of the PG is concerned, the C-O bonds are of importance only in disaccharides (i.e. in MurNAc-GlcNAc, see figure 1). If these C-O bonds or any of the C-C or C-N bonds break, this will lead to the destruction of the PG. Note that there are also C-O bonds in other parts of the PG, but we do not take these into account as they are less important in damaging the PG structure.
It should also be mentioned that there are no bond cleavage events observed in the case of O 2 impacts. These molecules are found to have only weak attractive non-bonded interactions with the PG structure. Therefore, we do not consider O 2 molecules in our further investigation. In the following studies, examples of the bond-breaking mechanisms will only be shown for oxygen atoms, though similar mechanisms have also been observed for O 3 molecules. Note that in the case of O 3 , the first reaction to occur is invariably with hydrogen from the PG, resulting in the formation of an O 2 molecule and an OH radical.

Dissociation of C-N bonds
One of the cleavage mechanisms of C-N bonds is presented in figure 2, where d-Ala, connected to the pentaglycine bridge, is being broken by an impinging oxygen atom (see figure 1, righthand side). It is clear from figure 2(a) that oxygen (encircled by the red dashed line) is reacting with hydrogen initially bound with carbon, i.e. a hydrogen-abstraction reaction is taking place. Note that there is another oxygen (shown by the dashed green circle) that has already abstracted a hydrogen atom from another carbon atom. As we concentrate on the breaking of C-N bonds in this section, an explanation for this process will be given later. As shown in figure 2(b), the distance between the C-C bonds starts to decrease after the hydrogen-abstraction reaction (cf bond distances in figures 2(a) and (b)). Because of the hydrogen abstraction, a primary alkyl radical is generated in d-Ala. Since this radical is not stable [51], a double C-C bond is created by homolytic cleavage of the C-N bond, with the formation of a resonance-stabilized amide radical (see figure 2(c)). Note that the resonance-stabilized amide radical is more stable than the primary radical, which could therefore be the driving force for the C-N bond-breaking process. It should also be noted that also a secondary alkyl radical is created after the hydrogen abstraction (see dashed green circle in figure 2(a)). Since secondary alkyl radicals are more stable than primary alkyl radicals, no bond breaking takes place in this case.
After dissociation of the C-N bond, the two carbon atoms form a double bond (see figure 2(c)). The average C-C bond length does not change further until the end of the simulation and is found to be about 1.33 Å, i.e. a typical value for the bond length of a double C-C bond [51], as shown in figure 3 (see red dashed line).  In figure 4 the time evolution of the average number of C-N bonds is shown. Note that the average number of C-N, C-C and C-O bonds is calculated from 50 independent runs for each incident species. It is clear from the figure that the average number of C-N bonds decreases during the simulation, i.e., the plasma species effectively break the C-N bonds in the PG structure. It is also obvious from figure 4 that most of the C-N bonds break due to oxygen impacts, rather than due to O 3 molecules. When oxygen is used as the impacting species, dissociation of C-N bonds is observed in 26% of the simulations cases. When O 3 is used, on the other hand, this value decreases to 8% (see table 1). Moreover, it is also observed that most of the C-N bond dissociations occur only in alanines. In the case of impacting O atoms, for instance, almost 80% of the C-N bond dissociations occur in alanines. Note, however, that the dissociation of C-N bonds occurs only due to the hydrogen abstraction from a methyl group in the alanines (see below). The calculated bond length distribution of the carbon atoms is shown in figure 5 for different simulation times. At the beginning of the simulation (t = 0 ps), there is only one peak centered at ∼1.56 Å, indicative of single C-C bonds [51]. After 150 ps, the amplitude of this peak decreases and a new maximum appears at ∼1.34 Å, indicating the presence of double C-C bonds. In the subsequent 150 ps, this maximum continues to increase slightly, indicating that the number of double C-C bonds is increasing. Note that the formation of double C-C bonds can only be due to hydrogen abstraction by plasma species. As demonstrated above, this can lead to the breaking of neighboring bonds in some cases (e.g. C-N bonds in alanines, see above).
It should also be mentioned that the breaking of C-N bonds in alanines depends on the position of the hydrogen abstraction taking place in the alanine. Our observations show that if the hydrogen abstraction primarily takes place in the methyl part of the alanine, this eventually leads to the formation of a double C-C bond and the destruction of the neighboring C-N bond (see e.g. figure 2). On the other hand, if the hydrogen abstraction primarily takes place in the central part of the alanine, it cannot lead to a cleavage of the C-N bond, even if a double C-C bond is formed. For instance, it is obvious from figure 6 that after two hydrogen-abstraction  reactions in l-Ala (see figure 1, left side of the PG), a double C-C bond is created without any bond breakage (see figure 6(c)). Indeed, the first hydrogen abstraction occurs in the central part of the alanine (see figure 6(b)) with the formation of a carbon radical. This radical is stabilized by its amide neighbors due to electron delocalization effects [52] and is therefore more stable than the primary radical mentioned above (see figure 2(a)). Hence, after the second hydrogen abstraction, a primary radical is also created and these two radicals subsequently form a stable double C-C bond (see figure 6(c)).
Note that breaking of C-N bonds is also observed in other parts of the PG. Our calculations show that the mechanism of the C-N bond dissociation in these parts is similar to the cleavage mechanism in alanines as described above, i.e. it occurs due to hydrogen abstraction and subsequent formation of double C-C bonds. However, no C-N bond-breaking events were observed in the pentaglycine interpeptide, even when hydrogen abstraction took place. This The OH radical abstracts another hydrogen atom, which is connected to C 1 (cf C 1 from (b)), and a water molecule is created (t = 45.6 ps). Subsequently, some double C-O and C-C bonds are formed, which lead to breaking of the C-O ether bonds (cf the bonds between numbered atoms from (b) and (c), and see text for a detailed description of the mechanism). is probably due to the resonance-stabilized structure of pentaglycine. We should also mention that the examples of the C-N bond-breaking mechanisms were demonstrated only for oxygen atoms. As mentioned above, similar mechanisms have been observed for O 3 molecules.

Dissociation of C-O bonds
The structurally important C-O bonds are found both in the disaccharides (MurNAc-GlcNAc) as well as between them. There are four oxygen atoms that are linked with two carbon neighbors in one disaccharide (see figure 1), while the other oxygen atoms in the structure have either hydrogen neighbors or they are connected with carbon atoms by double bonds. Therefore, we do not consider the breaking of these C-O bonds in this study, as they do not play a significant role in the destruction of the PG structure (see figure 1).
We observe several C-O bond-breaking mechanisms in our simulations. In all of these mechanisms, the dissociation again occurs due to the hydrogen-abstraction reaction. After the hydrogen abstraction, a double C-O bond may be formed (more specifically, one C-O bond is broken and another becomes a double bond), or a double C-C bond may be formed, or both double bonds are formed. In most cases, the hydrogen-abstraction reaction leads to a cascade of C-O bond cleavage events, i.e. the breaking of three important C-O bonds: one in MurNAc, one in GlcNAc and one between them (see figure 1). Note that here we only denote ether bonds as 'important C-O bonds', i.e., as one of two bonds between two carbon atoms and one oxygen atom in disaccharides (see figure 7(a), black dashed circles).
One of the observed consecutive breaking mechanisms of three ether bonds is presented in figure 7. Note that this mechanism is observed most in our simulation for all impinging species. In figure 7(a), the oxygen atom abstracts a hydrogen atom from GlcNAc with the formation of an OH radical (see first red dashed circle). After a few ps, a second oxygen atom abstracts a hydrogen atom from MurNAc and a second OH radical is formed (see figure 7(b), red dashed circle). For clarity, we numbered the atoms that participate in the dissociation of the ether bonds (see figure 7(a), black dashed circles). In figure 7(c), the OH radical abstracts another hydrogen, which is connected to C 1 , with the formation of a water molecule. Consequently, a radical is created at C 1 , which results in a cascade of homolytic cleavage and double bond formation reactions starting with the creation of a double C 1 -O 2 bond, which in turn leads to cleavage of the C 3 -O 2 bond. The latter subsequently leads to the creation of a double C 3 -O 4 bond. Moreover, this leads to the cleavage of the C 5 -O 4 bond, which in turn yields the formation of the double C 5 -C 6 bond. This subsequently results in the dissociation of the C 6 -O 7 bond and the formation of the double C 8 -O 7 bond. Finally, the cascade process leads to cleavage of part of the molecule, which in turn eventually results in damage to the bacterial cell wall.
In figure 8, the time evolution of the average number of ether bonds is shown. As mentioned above, the average number of ether bonds is calculated from 50 runs for each incident species. It is clear from figure 8 that the average number decreases during the simulation, i.e. the C-O bonds are broken by the impacting plasma species. Moreover, most of the ether bond cleavage events occur due to oxygen impacts, rather than due to O 3 molecules (see figure 8). From the 50 simulations that were carried out for each type of incident species, the dissociation of the ether bonds occurs with oxygen atoms in 39 cases (i.e., 78%), whereas this value is 28 for O 3 (56% of the cases), i.e. about 1.4 times less (see table 1). We should also mention that most of the bond cleavage events occur in disaccharides, rather than in other parts of the PG structure. Thus, the dissociation of the important ether C-O bonds in disaccharides is observed much more frequently than the dissociation of C-N or C-C bonds in other parts of the PG structure. The values for C-N bond breaking are shown above and those for C-C bonds breaking are given below in section 4.3. Our study on C-O bond dissociation shows that if the hydrogen-abstraction reaction is initiated at a carbon atom located near one of the three important ether bonds in disaccharides (see figure 7(a), black dashed circles), this can eventually lead to consecutive cleavage of at least two of these three C-O bonds. Moreover, we should also mention that the carbon, at which the hydrogen-abstraction reaction takes place and subsequently the dissociation of the important C-O bonds occurs, should not necessarily be a first neighbor of the participating O atom in a ring of MurNAc or GlcNAc. Indeed, the dissociation of the ether bonds in MurNAc, GlcNAc or between them can also occur when the hydrogen-abstraction reaction takes place from another carbon atom, which is the second or the third neighbor of the participating O atom in a MurNAc or GlcNAc ring. Furthermore, the cleavage of the important C-O bonds (and even cleavage of all of three) can be observed when the hydrogen abstraction takes place from a carbon or oxygen, which are positioned in the residue close to the O atom in a ring of MurNAc or GlcNAc (see figure 7(b), residues connected to C 1 and C 6 ).
The dissociation of a fourth ether bond occurs mostly due to the hydrogen abstraction from a methyl residue of MurNAc, which is close to this bond (see figure 9(a), black dashed circle). Indeed, after the hydrogen abstraction from the methyl residue, an unstable primary alkyl radical is created, which leads to the formation of a double C-C bond and this, subsequently, will induce the breaking of the ether bond and the formation of the double C-O bond (see figure 9(b)); the bond lengths are indicated for the sake of clarity.
It should also be mentioned that after breaking of any of the important C-O bonds in disaccharides, a double bond remains instead of two single bonds. It is also clear from figure 10 that at the beginning of the simulation, the number of single and double C-O bonds in the PG is approximately the same (see figure 10, light gray curve). However, after 150 ps, i.e. in the middle of the simulation, the number of single bonds has already decreased significantly, whereas the number of double C-O bonds has increased (see figure 10, gray curve). Finally, at the end of the simulation, the peak of the double C-O bonds becomes approximately twice as high as the

Dissociation of C-C bonds
As is clear from figure 1, bonds between carbon atoms are found in all parts of the PG structure. However, our investigation shows that the dissociation of C-C bonds occurs only in disaccharides and in most cases after the breaking of three ether bonds (see above).
As is the case for the C-N and the C-O bonds, a number of different bond-cleavage mechanisms are observed for C-C bond dissociation. The most frequently observed mechanism is illustrated in figure 11. In this process, the C-C bond in GlcNAc is broken after the homolytic cleavage of three ether bonds. Again, for clarity, we numbered the C atoms participating in the mechanism (see figure 11). It can be seen in figure 11(a) that after the initial hydrogenabstraction reaction an O radical is created, which is connected to C 2 . For the O radical to form a double C-O bond, one of the two C-C bonds (i.e. bonds between C 1 -C 2 or C 2 -C 3 ) needs to be homolytically broken. This bond cleavage is found to occur between C 1 -C 2 . This results in the formation of two molecules: one with a conjugated system (see figure 11(b), red dashed circle) and one with a resonance-stabilized radical. Note that the radical in the conjugated system is also resonance stabilized, although this radical is not important for the above-mentioned breaking mechanisms since we also observed the same mechanisms when there was no radical at that position (see figure 11, C 5 ). Figure 11. Snapshots from MD simulations, showing the breaking mechanism of a C-C bond upon oxygen atom impact. (a) After the hydrogen abstraction, the distance between the oxygen radical and C 2 begins to decrease (t = 81.6 ps). (b) Oxygen radical forms a double bond with C 2 and this leads to the cleavage of the C 1 -C 2 bond (t = 82.8 ps). As a result, a conjugated system is formed (see red dashed circle). The time evolution of the average number of C-C bonds is depicted in figure 12. As was mentioned above, the average number of C-C bonds is calculated from 50 runs for each incident particle. It is clear from figure 12 that the average number of C-C bonds decreases during the simulation, i.e., the C-C bonds are effectively broken by impinging plasma species. Again, most of the C-C bond-breaking events occur due to oxygen impacts, rather than due to O 3 molecules. Furthermore, our investigation shows that from the 50 simulations performed for each impinging species, oxygen atoms lead to dissociation in 19 cases (38%), whereas O 3 leads to dissociation in 13 of these cases (26%, see table 1).
It should be mentioned that for O 3 molecules the same C-C bond-breaking mechanisms were observed as for the O atoms.

Conclusion
In this work, we investigated the interaction of important plasma species-i.e., O 3 and O 2 molecules, and O atoms-with the PG structure by means of reactive molecular dynamics simulations. We demonstrated that such atomic-scale simulations give valuable insights into the details of the microscopic events that occur in the plasma treatment of bacteria.
We found that O 2 molecules do not result in structural damage to the PG. They were found to only have weak non-bonded interactions with the PG. The other investigated plasma species, i.e., O 3 and especially O atoms, were found to break structurally important bonds of the murein (i.e. C-O, C-C and C-N bonds), which subsequently leads to structural damage of the bacterial cell wall. This is in qualitative agreement with the literature [11,[20][21][22][23][24], where it is indeed reported that oxygen-based active species such as atomic oxygen and ozone have a strong germicidal effect. Moreover, our atomic-level investigations revealed the important role these species play as plasma-borne disinfectants.
We found that the dissociation of C-N, C-O and C-C bonds upon O atom impacts occurred in 26, 78 and 38% of the simulation cases, respectively. Note that these percentages have been calculated from 50 simulations performed for each kind of plasma species and it is quite possible that in one simulation case impinging O atoms could give rise to the dissociation of both a C-N and a C-O bond, for instance, so the sum could be higher than 100%. For the O 3 molecules, the cleavage of C-N, C-O and C-C bonds occurred in 8, 56 and 26% of the cases, respectively. Hence, it is clear that the O atoms are more effective in bond cleavage than the O 3 molecules, and the C-O bonds seem to break up more easily, followed by the C-C bonds and C-N bonds.
The mechanisms of the important bond-breaking processes in PG were studied thoroughly. In all bond-cleavage mechanisms, the dissociation of important bonds (i.e., C-O, C-C and C-N bonds) occurs due to the hydrogen abstraction by incident particles. As mentioned above, these bonds are most frequently broken due to interaction with atomic oxygen, rather than by O 3 species. It was also found that most of the bond-breaking processes occur in disaccharides, rather than in other parts of the PG structure, i.e., the dissociation of the ether bonds in disaccharides is observed much more frequently than the dissociation of C-N or C-C bonds in other parts of the PG structure, as also mentioned above. Moreover, no bond-breaking events were observed in the pentaglycine interpeptide, even when hydrogen-abstraction reactions took place. We explain this by the resonance-stabilized structure of the pentaglycine.
C-N bond dissociations are found to occur most often in alanines, rather than in other parts of the PG structure. These dissociations are again invariably initiated by a hydrogen-abstraction reaction, removing a hydrogen atom from the methyl part of the alanines. The hydrogenabstraction reaction which leads to the dissociation of the ether bonds in disaccharides, most often also leads to a cascade of additional ether bond-breaking events. This results in the cleavage of three important C-O bonds, i.e., the ether bonds in MurNAc and GlcNAc, and the ether bond between them. Moreover, the dissociation of a fourth important C-O bond occurs mostly due to the hydrogen abstraction from the methyl residue of MurNAc, which is close to this bond. It was also found that the cleavage of the C-C bonds occurs only in disaccharides and in most cases after the breaking of three ether bonds.
It is known that some plasma sources create higher concentrations of oxygen species than others (see e.g. [2,21]), and they might therefore be more efficient for inactivating bacteria, since the probability of PG destruction might be higher. Our findings could therefore be useful for the development of plasma sources for enhanced antibacterial properties.
It should, however, be realized that our simulations were performed under 'perfect' conditions. In reality, bacteria are typically surrounded by a liquid film, which affects the plasma species. The effect of this liquid film has not yet been included in our study, but will be investigated in the near future. Nevertheless, we believe that our present results already provide some atomic-level insight into the mechanisms of plasma species interacting with bacteria, and in plasma disinfection in general.