At the beginning, the positions of centre of mass of a dimeric HD5 along a permeation path (+ Z ◊ -Z) and exerted forces on HD5 are calculated and shown in Fig. 2A and 2B. For both simulations, the double force wells are observed from Z ~ 0–2 nm whose positions are in the outer leaflet (Fig. 2A). Two exerted force peaks (a major peak at Z ~ 2 nm (F1) and a minor one at Z ~ 0–1 nm (F2)) are derived from the strong electrostatic interactions between HD5 and lipid A head groups (KDO and Pi) (Fig. 2A). Then, HD5_1 and HD5_2 arrive a hydrophobic core at 100 ns and 136 ns, corresponding to Z = 0 nm for HD5_1 and − 1 nm for HD5_2. These different Z positions observed are due to the curvature of LPS membrane (Fig. 2C). Besides, a drop of Z positions before 100 ns for HD5_1 (Z from 1◊0 nm) and 136 ns for HD5_2 (Z from 0◊–1 nm) indicates a quick move from the outermost layer into a membrane core (Fig. 2B). This drop suggests the less afford required to dive into a centre of hydrophobic core. To pass through the inner leaflet (Z from 0 ◊~ -4 for HD5_1 and Z from − 1 ◊ -4 for HD5_2), two comparable sizes of exerted forces are observed (F3 and F4 in Fig. 2A). These two applied forces refer to the barriers for dragging HD5 out of a hydrophobic core (F3) and phospholipid head groups (F4). The force profiles clearly show that passing through a LPS surface into a membrane core requires much larger exerted force than going out of the inner leaflet due to strong charge-charge interactions between HD5 and lipid A head groups. Seemingly, the electrostatic interactions between HD5 and lipid A head groups cause a charged LPS surface to be a facilitator for the HD5 adhesion, but hindrance for the passage of HD5 at the same time. In addition, the HD5 translocation causes a water leakage in both simulations (Fig. 3B). As seen earlier, the curvature of a LPS membrane is found in both simulations, although HD5_2 shows a higher degree of a membrane curvature. A pulling force seems to influence a LPS membrane in HD5_2, however both HD5_1 and HD5_2 can provide a similar pattern of force profiles (Fig. 2A). This implies our SMD work can be repeated. Moreover, in Fig. 2C, HD5 in both simulations employ only one chain to adhere on a LPS membrane. In this work, only chains b is used to stick on a host membrane.
Comparing to X-ray structure, the protein conformation of HD5_1 seems to be more preserved than HD5_2. Overall, the conformational change of T1 in chain b in all cases is observed (Fig. 2C). T1 region is thus the most flexible part where higher T1 flexibility is found in HD5_2 (Fig. 2C and S1 in supplementary information). T1 in chain b starts to move away from its β-stranded core region in both simulations, while this move allows the dimer dissociation at the end of simulation in HD5_2 (Fig. 2C). Such T1 dynamics may be important to facilitate the HD5 adsorption and permeation because this T1 reorientation is preserved throughout an additional conventional MD work (unpublished work). To observe how HD5 passage affects the outer leaflet, the density maps of LPS membrane are computed in Fig. 3A. It is clear in Fig. 3A that a dimeric HD5 interrupts the distribution of polar moietires on a LPS surface which can destabilise the outer leaflet. Different sizes of water-filled pores generated by each simulation are shown in Fig. 3A (green circles). Different pore sizes are formed due to the different alignments of a HD5 dimer inside a membrane (Fig. 3B). Although both systems employ chain b to land on a membrane interface, inside a membrane, HD5_1 seems to align perpendicularly to the z axis, while HD5_2 is in parallel (Fig. 3B) which causes various degrees of water leakage during a HD5 travel.
Considering the HD5-LPS interactions, a dimeric HD5 strongly interacts with both KDO sugars and phosphate (Pi) moieties of lipid A, but the higher number of KDO interactions are found (Fig. 3C). Upon arrival, a high number of HD5-KDO interactions is formed (~ 15 hydrogen bonds), but they are dropped when HD5 translocates close to a hydrophobic region (after 75 ns). Nonetheless, a loss of such hydrogen bonds is compensated by HD5-Pi (outer) interactions (Fig. 3C). As expected, HD5 gains more hydrogen bonds with Pi (inner) of phospholipid head groups when it travels close to a surface of the inner leaflet (Fig. 3C). Apparently, a dimeric HD5 is continuously wrapped by polar groups of lipid A along its permeation pathway. It means that the lipid A head groups are dragged down into a membrane core by a cationic HD5. During a HD5 insertion, the lipid A head groups (KDO and Pi) are not only pulled down into a membrane core, but they also carry water molecules with them. KDO sugars and Pi are found to escort a comparable number of water molecules down into a membrane centre (Figure S2 in supplementary information). Besides, a presence of constant KDO- and Pi(outer)- water contacts suggests the existence of a water-filled pore (Figure S2 in supplementary information). Altogether, both pulled lipid A head groups and permeated HD5 induce the entry of water into a hydrophobic region resulting in a membrane disruption and water leakage along a membrane axis. This highlights a key role of electrostatic interactions between a cationic HD5 and LPS surface on the pore-forming activity.
To further investigate the behaviour of a HD5 dimer, sets of distances between two monomers and a pair of key residues (V19-E21) for dimerization are measured in Fig. 4A and 4B. In HD5_2, The shifted distances between two chains (chains a and b) and V19-E21 at ~ 136 ns obviously illustrates the dimer separation (Fig. 4A and 4B). A V19-E21 interaction was reported to be important for dimerization [8, 11, 24]. Thus, the breakdown of V19-E21 pair in HD5_2 serves as a sign for a dimer dissociation (Fig. 4B). The orientations of V19 and E21 can be seen in Fig. 4C. In contrast, a dimeric HD5_1 is more stable owing to the ability to maintain a V19-E21 interaction (Fig. 4B and 4C). In addition, hydrogen bonds of HD5-membrane and HD5-water are also computed. HD5_1 and HD5_2 can form interactions with both membrane and water along a permeation path (Fig. 4D and 4E). This confirms the ability of a dimeric HD5 to drag polar moieties of lipid A and water molecules down into a membrane core. In Fig. 4D, a number of HD5-membrane hydrogen bonds are gradually increased until ~ 75 ns due to a close contact to a hydrophilic membrane surface and then reduced after 100 ns when approaching a hydrophobic region. A sharp peak of HD5-water interactions is captured at ~ 136 ns in HD5_2. This is because, at 136 ns, the breakdown of HD5 causes a sudden loss of contacts with KDO and Pi at the outer leaflet which requires more water interactions to compensate (Fig. 3C and 4E). Nonetheless, after 136 ns, a number of water contacts drop down since HD5 gains more Pi(inner) contacts from the inner leaflet (Fig. 3C and 4E).
In addition, sets of hydrogen bonds between each chain and its environment are investigated (Fig. 5). Because a dimeric HD5 utilises chain b to anchor on a LPS membrane, a high number of chain b-membrane hydrogen bonds are captured (~ 15 hydrogen bonds in Fig. 5B). Only ~ 7.5 chain a-membrane interactions are found (Fig. 5A). However, a comparable degree of water interactions is observed in both chains. Expectedly, the water contacts are maximized at the membrane surface in both simulations (Fig. 5C and 5D). Continuous interactions with water in Fig. 5C and 5D indicate a presence of water-filled pore. A sudden shift of chain a-water hydrogen bonds at Z ~ -1 nm in HD5_2 corresponding to a dimer separation as explained earlier (Fig. 5C).
For further analysis, the residue-membrane hydrogen bonds are computed (Fig. 6). Upon binding, chain b of a dimeric HD5 employs its T1 (residue 6–14) and β3 (residue 25–32) regions to interact with a membrane surface via strong interactions with arginines in the active region (R6, R9, R13, R25, R28, and R32) (Figs. 1, 2 and 6). Not only arginines, but other polar residues also contribute to the adsorption. In case of chain a, it can interact with a membrane using R13, R32, A1, and T2 (Fig. 6). Seemingly, HD5 requires one main chain to stick on a cell surface, whereas the other chain acts as an assistant. When HD5 approaches the inner leaflet, an additional set of interactions are formed (after 100 ns). A strand β3 of chain a in both simulations flips down and align normal to a membrane axis and forms hydrogen bonds with C3, S23, R25, Y27, R28 (Fig. 6). In HD5_1, chain b employs a part of β2 to additionally interact with a membrane (L16, S17, G18, I22), while the extra interactions with A1 and C5 on β1 are found in HD5_2 (Fig. 6). In chain b, the conformational change of T1 region is found (Fig. 2C). This region contains three arginines (R6, R9, and R13) which form strong interactions with charged moieties of lipid A (Fig. 6). Such interactions may trap T1 and cause the change in T1 orientation. The effect of lipid interactions on a defensin conformation is also seen in other α -defensins [25, 26]. Unlike chain a in HD5_1, more residue-membrane hydrogen bonds are found in chain a of HD5_2 (Fig. 6). Such interactions allow the tighter binding of chain a to lipid A layer which can retard a HD5 passage and lead to the split of a dimer in HD5_2.
Overall, our results show that all arginines in the active region contribute to charge-charge interactions between HD5 and a LPS surface. Persistent arginine-membrane hydrogen bonds observed in Fig. 6 demonstrate the role of arginine residues in both adsorption and permeation. Especially, R13 and R32 of both chains in all simulations are found to form interactions with a membrane (Fig. 6). This implies the important role of both R13 and R32 in HD5’s activity. The double mutation of R13 and R32 to alanine was previously reported to reduce the antibacterial activity against gram-negative bacteria [27]. Based on our results, a loss of R13 and R32 contacts can diminish the binding affinity of HD5 to host cell surface resulting in a decrease in antibacterial activity as seen in a previous work. Besides, R9 and R28 were reported to be important for killing gram-positive strains, especially B. cereus [27]. In our case, insignificant interactions with R9 and R28 are observed (Fig. 6A). This demonstrates a minor role of R9 and R28 in interacting with our gram-negative membrane model. Comparing among six defensins, HD5 contains the highest number of arginines (Figure S3 in supplementary information). This may be one of factors that allow HD5 to show high level of bacteria-killing activity [7]. Arginine residues appear to be crucial for interacting with LPS head groups which facilitate a HD5 adhesion and disrupting a membrane structure. This finding can be used to explain why more arginines can enhance HD5 activity as seen in a previous work [28]. Seemingly, the HD5 function is dependent on its cationicity. However, the substitution of arginine to lysine was found to disturb the bacterial killing ability of HD5 [27, 29]. For example, the double mutation of arginine to lysine (R1332K and R928K) causes the reduction of E.coli killing, but only R928K affects the bacterial activity against S.aureus [27]. It appears that the arginine-to-lysine substitution affects the bacterial killing activity selectively. Although it was reported that the different degrees of hydrophobicity and ability for electrostatic interactions between arginine and lysine cause the selective antibacterial activity [27, 30, 31], the understanding of how lysine influences the antibacterial activity remains unclear.