Potential of Mean Force
The simulation protocol includes two sets of simulations: aSMD for 45 ns divided in 9 steps and 25 replicas per step, followed by 3 replicas of the relaxation step, consisting in 100 ns of cMD each. This experimental design was applied to investigate the behaviour of 3 control CPPs (Arg9, MAP, and TP2) in 3 different membranes. Subsequently, the same procedure was run for the DynA WT and the three clinical variants (L5S, R6W, and R9C).
Potential of the Mean Force (PMF) indicates the resistance of the peptide to cross each bilayer, as illustrated in Fig. 1. The overall PMF for each bilayer (Fig. 1) indicates that bilayer complexity is positively correlated with higher PMF values. Cholesterol has been associated with reduced efficiency in CPP translocation, –a phenomenon previously discussed by Pae and colleagues [57]. Arg9 and MAP show the most difficulty to cross any bilayer, regardless of the composition. TP2 is capable of crossing neutral bilayers (even with cholesterol), but shows higher PMF in DPPC:DOPC:DPPS:DOPS:CHOL bilayers. In the case of DynAs, DynA R9C stands out as having the highest energy requirement for bilayer crossing, similar to Arg9 and MAP values. In contrast, the remaining two clinical variants and DynA WT display similar average PMF values, emerging, among these, DynA L5S as the peptide with the lowest average energy. DynA WT, L5S and R6W show similar behaviour to TP2, with low energy barriers in the first two membranes and higher in the most complex membrane.
Our analysis reveals a notable correlation between membrane complexity and increased PMF, as demonstrated by the averages (Fig. 1). In the DPPC membrane, peptides exhibit, on average, a relatively low energy requirement to traverse the bilayer. The introduction of cholesterol to the membrane results in an overall increase in PMF. In parallel, the addition of unsaturated fatty acids (DOPC) should enhance the internalization of CPPs and lower the PMF [58], but this effect seems to be insufficient to compensate the influence of cholesterol. Finally, in the most complex membrane composition, we observe the highest resistance to bilayer crossing, likely due to strong electrostatic interactions between the positive charge of the peptides and the negatively charged phospholipids head groups in the upper leaflet.
Simulation results
After the aSMD simulation, the molecular distribution is similar for all cases (Fig. 2A): the peptide has been steered into the lower part of the bilayer and is close to the polar heads of the lipids in the lower part of the bilayer. Some polar heads of the upper leaflet and waters (Fig. 2B) have been dragged along with the peptide during the steering process, in agreement with the previously described Defect Assisted by Charge (DAC) phenomenon [59]. As a consequence of the peptide pulling across the membrane, a pore is formed at the end of the aSMD simulation, allowing water molecules to flow across the bilayer (Fig. 2B). We define the pore as a large defect in the membrane that allows for a continuous water flow between the upper and lower leaflets.
Once the peptide has successfully penetrated the lower region of the lipid bilayer, we conduct conventional MD simulations to relax the system and allow it to return to equilibrium (Video S1 and Video S2 for CPPs and DynA variants, respectively). During this phase, we observed four possible behaviours for the peptides: (1) Translocation: the peptide stays on the lower part of the bilayer and ultimately achieves translocation across the membrane; (2) Pore formation: peptide remains in the middle portion of the bilayer, leading to formation of pores of different radius in the membrane; (3) Insertion: peptide stays in the middle region of the bilayer, but without leading to pore formation; (4) Upper Part Relocation or Return: peptide returns to the upper part of the bilayer. A summary of these behaviours, observed across all peptides and membrane compositions, is presented in Fig. 3.
A more detailed summary is presented in Table 3, indicating the occurrence ratio of each phenomenon, and differentiating pore sizes. In most cases, the three replicas exhibit a similar behaviour, with only one exception: the DynA L5S simulation in the DPPC:DOPC:DPPS:DOPS:CHOL membrane. DynA L5S shows translocation in two of the replicas, while it remains in the middle part of the membrane forming a pore in the third one. This behaviour suggests a close relationship between translocation and pore formation, implying that translocation may involve an initial quick and transient pore formation.
Table 3
Simulation results for all the peptides in the 3 membrane compositions. The numbers in parenthesis indicate the ratio of this behaviour observed in the replicas’ runs.
Peptide | DPPC | DPPC:DOPC CHOL | DPPC:DOPC DPPS:DOPS:CHOL |
Arg9 | Translocation (1) | Medium pore (1) | Return (1) |
MAP | Small pore (1) | Insertion (1) | Insertion (1) |
TP2 | Insertion (1) | Return (1) | Return (1) |
DynA WT | Big pore (1) | Small pore (1) | Return (1) |
DynA L5S | Big pore (1) | Small pore (1) | Translocation (0.66) Medium pore (0.33) |
DynA R6W | Big pore (1) | Return (1) | Return (1) |
DynA R9C | Return (1) | Return (1) | Return (1) |
When examining the behaviour of control CPPs, we observe distinct patterns: Arg9 successfully translocates the DPPC membrane, whereas MAP induces pore formation, and TP2 gets inserted in the bilayer. Focusing on Arg9, which exhibits translocation capabilities, we notice an interesting trend: as the membrane complexity increases, its translocating capacity is lost and is only capable of forming a pore in DPPC:DOPC:CHOL membrane, again, suggesting that the addition of cholesterol to a membrane hinders the ability to translocate, as discussed by Lorents et al. [60]. Furthermore, these authors emphasize that the incorporation of negative polar heads lipids is not directly correlated with the internalization efficiency of the cationic CPPs. In fact, our study indicate that negatively charged lipids are, in general, associated with a loss of translocation capacity. Likewise, MAP loses its pore-forming ability when the membrane complexity is increased, resulting in insertion without the formation of a pore. In parallel, TP2 does not show pore-forming or translocation capacity in any membrane (only insertion in DPPC), which can be related to the fact that TP2 in monomeric form enters the cell via spontaneous membrane translocation, rather than the pore formation mechanism [61, 62].
On the other hand, we do not observe any translocation of the DynAs in the DPPC membrane. However, DynA WT and two clinical variants (L5S, and R6W) form a pore. DynA pore-formation behaviour has been described in previous experiments [63], suggesting that the pathogenic effects of DynA WT and mutants may be related to transient pore formation. Furthermore, Perini and collaborators also suggested that the pores formed can present different sizes [64]. In the second membrane, two of these peptides (DynA WT, L5S) are still able to induce pore formation, like Arg9. Lastly, DynA L5S is the only capable peptide of translocating in the DPPC:DOPC:DPPS:DOPS:CHOL membrane, which may indicate the highest CPP potential across all DynA peptides. In average, DynA WT and R6W show similar behaviour, forming a big pore in DPPC, and returning to the upper part in the negatively charged membrane, as observed by Alvero-Gonzalez et al [22]. Besides, in DPPC:DOPC:CHOL, DynA WT only forms a small pore, which can be more transient and not as stable as the larger pores, leading to early pore closure, as seen in other simulations. In this regard, Björnerås and collaborators concluded that R6W induces approximately as much leakage as DynA WT [65]. In turn, DynA R9C, even though it also shows leakage, does not show membrane disrupting capacity in the aSMD simulations, implying that these two processes may occur through different mechanisms. On the other hand, while DynA L5S does not show leakage in experiments [65], in our simulations it shows the highest membrane-disrupting capacity. This suggests that the L5S translocation may occur rapidly and does not provide enough time to cause the ion influx, as previously described with Tat and other peptides [66, 67].
Overall, DynA WT shows close behaviour to Arg9, since they lose translocation/pore-forming capacity as the membrane complexity increases, DynA R6W is similar to TP2, as they stay inserted in DPPC membrane (DynA R6W through pore formation) and return in the other two membranes. Conversely, MAP, DynA L5S, and DynA R9C present distinct behaviours: MAP is the only peptide that is able to get inserted in the three membranes (forming a small pore in DPPC), DynA L5S forms a pore in the first two membranes, and is the only peptide that gains translocation capacity when negatively charged polar heads are added to the membrane, and DynA R9C is the only peptide that does not get inserted or translocates in any membrane.
Further insight can be obtained from the electron density analysis in the DPPC membrane (Fig. 4). This analysis provides valuable information regarding the water and phospholipid displacement (DAC) produced by the pore formation in DynA WT, L5S and R6W simulations. Regarding the remaining peptides, we did not observe considerable differences despite the disparity of their behaviours: translocation in the case of Arg9, pore formation in MAP, insertion in TP2 and return in DynA R9C. Nonetheless, a closer examination of the electron density of MAP reveals a relatively subtle DAC effect, as seen in the raw data (not shown). A more rational comparison can be drawn among Arg9, TP2, and DynA R9C, since in Arg9 and DynA R9C end in one interphase of the membrane, and TP2 gets inserted but without the formation of a pore. Similar results have been seen for DPPC:DOPC:CHOL (Figure S2) and DPPC:DOPC:DPPS:DOPS:CHOL (Figure S3).
This analysis does also show the considerable differences between the large pore seen in DynA WT, L5S, and R6W, and the small pore formed in MAP, as in Table 3. This phenomenon is discussed further in the next section. The secondary structure was analysed, but no significant differences were observed, probably because of the time scale (ns) of the simulation.
Pore analysis
The molecular configuration of the system in the various scenarios is illustrated in Fig. 5. This figure emphasizes the contrast in bilayer arrangement between representations of big (A) and small (B) pores. Figure 5C shows the final snapshot of Arg9 translocation (left) and DynA R9C upper part relocation (right), highlighted by the absence of a water pore.
Furthermore, as outlined in Table 4, we discern a notable disparity in radii between big and small pores: the larger pore has a minimum radius of approximately 17 Å, whereas the smaller has a minimum radius smaller than 1 Å. Moreover, we have defined a third configuration as medium pore, characterized by pore sizes ranging from 1 to 6 Å. This description is provided on Table 3. An evolution of the pore size along the 100 ns cMD is shown on Figure S4.
Table 4
Mean radius size (Å) of the last 80 ns of the relaxation.
Peptide | DPPC | DPPC:DOPC CHOL | DPPC:DOPC DPPS:DOPS:CHOL |
Arg9 | 0.19 ± 0.03 | 6.30 ± 0.04 | 0 |
MAP | 0.71 ± 0.03 | 0 | 0 |
TP2 | 0 | 0 | 0 |
DynA WT | 16.6 ± 0.1 | 2.42 ± 0.04 | 0 |
DynA L5S | 18.8 ± 0.1 | 1.25 ± 0.03 | 3.71 ± 0.05 |
DynA R6W | 16.5 ± 0.1 | 0 | 0 |
DynA R9C | 0 | 0 | 0 |
Residues’ occupancy
Lastly, we have analysed the occupancy of peptide residues by the polar heads of the phospholipids in the upper and lower leaflets (Fig. 6). Occupancy is defined as the percentage of simulation time during which the peptide residue is in contact with the polar head of the phospholipids.
Comparing all three CPPs we see, in average, a higher interaction ratio for Arg9 (several residues have 100% occupancy), which can be explained due to the polycationic nature of this CPP, strongly attracted to the negatively charged polar heads of the lipids. Similarly, MAP, which has alternating positive (K) and hydrophobic (L, A) residues, preferably interacts with the lipids in the positive residues, that is, K1, K5, K9, K12, and K16. Besides, in some cases, neighbouring residues also show high occupancy, due to spatial proximity. TP2 contains only two charged residues, R6 and R9, which are prone to interact with the polar heads of the lipids and show high occupancy across the three bilayers. However, the N- and C-terminal parts are also interacting with the polar heads in three and two bilayers, respectively. In DPPC, the peptide is inserted into the membrane and stretched, thus interacting with a leaflet in each end. In the second bilayer, the C-terminal residues seem to interact due to the spatial proximity to R9. Besides, the N-terminal residues (specially P1) shows a high occupancy, which we hypothesize that can be explained by the positive charge in the N-terminal residue.
In parallel, when comparing the occupancies across all three bilayers, there are noteworthy differences between: (1) the case where the peptide that translocates and has a higher occupancy in the lower leaflet (Arg9 in DPPC), (2) the peptides that form a pore and interact with the polar heads in upper and lower leaflets (MAP in DPPC, and Arg9 in DPPC:DOPC:CHOL), (3) the peptides that get inserted into the bilayer and also interact with both leaflets (TP2 in DPPC, MAP in DPPC:DOPC:CHOL), and (4) the peptides that are relocated to the upper leaflet and thus only interact with the polar heads in this leaflet (TP2 in DPPC:DOPC:CHOL, Arg9 and TP2 in DPPC:DOPC:DPPS:DOPS:CHOL). Nonetheless, there is an exception, that is MAP in the third membrane composition, which get inserted in the bilayer but only interacts with the upper leaflet, that may be because the peptide is coiled and does not reach the polar heads in the lower leaflet. Interestingly, Arg9 in DPPC:DOPC:DPPS:DOPS:CHOL interacts more intensely with the polar heads in PS lipids than PC lipids. Being a highly cationic peptide, this can be explained by the strong attraction between the side chains and the negatively charged lipids [66].
Figure 7 shows the DynA WT and its clinical variants occupancy by the polar heads of the lipids. DynA WT has five positively charged residues: R6, R7, R9, K11, and K13. In DPPC, R7, R9, K11, and K13 interact with the polar heads, in DPPC:DOPC:CHOL these are R6, R7, K11, and K13; and in the third membrane composition all five residues are key in the interaction with the polar heads. Moreover, the surrounding residues also show high occupancy due to the closeness to R or K residues. Intriguingly, there is a high interaction zone from residues Y1 to F4, which is maintained across all three bilayers, which has been shown in NMR experiments by Lind et al. [68]. We hypothesize that these interactions are explained by a similar mechanism as for TP2: the positive charge in the N-terminal favours these interactions. Besides, the only residue with negative charge (D15) seems that does not favour the interaction with the polar heads and the occupancy is reduced with respect to the surrounding residues.
Comparing the effect of the clinical variants, the mutation L5S seems to impact the behaviour of the peptide compared to the WT. The replacement of a hydrophobic residue by a polar one facilitates the peptide global interactions with the PC polar heads in DPPC and DPPC:DOPC:CHOL membranes. In the DPPC:DOPC:DPPS:DOPS:CHOL membrane the peptide global contacts are shifted towards PS polar heads, resulting in a fast escape from the hydrophobic core facilitating the translocation of the L5S peptide. In the cases of DynA R6W and R9C, both mutations entail the loss of a positively charged residue (R). In the first case, the interactions in residues 2–10 are maintained in all membrane compositions. This phenomenon may be explained by the presence of a neighbouring arginine in position 7 in R6W, which could be mitigating the charge loss. In R9C, on the other hand, the lack of arginine in position 9 is not mitigated by another neighbouring residue and translates into a general loss of interactions in residue 9 compared to the other variants, which are only regained in the third membrane.
Regarding the occupancy across all membranes, in DPPC DynA WT, L5S and R6W have a higher interaction with the upper leaflet, but they also interact with the lower leaflet, which allows the pore formation. This phenomenon aligns with the process explained by Herce and colleagues, wherein the peptide forms ‘complexes’ with the phospholipids due to the interaction between positively charged residues and the negatively charged phosphate groups of polar heads [66]. Water molecules solvate these charged groups, leading to the connection of waters molecules across both spaces and subsequent pore formation. Notably, in most cases, it is the positively charged residues (K, R) and their neighbours that exhibit a higher occupancy by the polar heads. On the other hand, DynA R9C does not interact with the lower leaflet and is rapidly relocated to the upper part of the bilayer.
In the other two membranes, there is a difference between the peptides that form a pore and interact with the polar heads in both leaflets (DynA WT, L5S in DPPC:DOPC:CHOL), the peptides that only interact with the upper leaflet and return to the upper part (DynA R6W, R9C in DPPC:DOPC:CHOL, and DynA WT, R6W, R9C in DPPC:DOPC:DPPS:DOPS:CHOL), and DynA L5S in the third membrane, that only interacts with the lower leaflet owing to the translocation. Interestingly, R6W and R9C peptides have a similar behaviour in DPPC and DPPC:DOPC:CHOL, indicating that these contacts are key for the interaction with bilayers. Furthermore, all four peptides lose interactions with the polar heads of PC (DPPC/DOPC) lipids in the more complex membrane, since they also interact with PS lipids (DPPS/DOPS), as previously seen for Arg9. This change in interactions is not correlated with higher translocation capacity (expect for DynA L5S), as the addition of a negative lipid is not directly correlated with the internalization efficiency of CPPs [59].
Altogether, we see how interactions with the upper leaflet strongly influence the ability of the peptides to interact with the lower leaflet and, consequently, their ability to form pores or translocate. The peptides that form pores interact with the polar heads in both leaflets, as they do the peptides that get inserted into the bilayer (except MAP in DPPC:DOPC:DPPS:DOPS:CHOL). Besides, the addition of negatively charged lipids causes a higher interaction with these polar heads instead of PC lipids in the case of Arg9 and all DynAs.