By measuring their biological activity using a human glioblastoma U87 cell line stably transfected for constitutive expression of FLuc/NLuc reporter genes, we observed for both peptides C6M1 and C6M1-L a lower silencing activity (60% and 35%) compared to both WRAP peptides (80%-90%) encapsulating in each case the same amount of siRNA (10 nM) (Figure 1B). Considering the close similarity in amino acids content between the two families of peptides and differences in the primary sequence, this result suggested that the amount of leucine and arginine residues were important for the silencing activity of the resulting PBNs. This could be the consequence of the higher number of arginine in the C6M1 series (7) compared to the WRAP series (4). Therefore, we could formulate objectively the hypothesis that a higher content of arginine is unfavorable for a strong biological response. Indeed, in our biological assay, we measured a global effect, ranging from the PBNs formation till the siRNA effects at the RISC system. In between, the PBNs have to internalize the cells and siRNAs have to be properly decaged from the PBNs. A lower decaging for the C6M1 PBNs could explain a weaker release and consequently a lower biological response of siRNA molecules because of the higher content of ionic interactions. In line with this, we clearly observed the direct influence of arginine residues in the decaging rate of siRNAs from PBNs made of WRAP peptides with different amount of arginine residues (see below).
Another noticeable difference was the lower number in leucine residue for the C6M1 peptide series (6 or 7) compared to the WRAP series (8). A lower number of leucine residues seemed to correlate with a lower biological response. Along this line, we observed that the deletion of one leucine in the C6M1-L peptide compared to the native C6M1 peptide seemed to further reduce the silencing property of the PBNs (Figure 1B). Therefore, we investigated whether the N- and C-terminal and internal leucine doublets in our WRAP peptide series were indispensable all along the different steps from the nanoparticle formation towards the recorded gene silencing activity.
Dissecting the role of the leucine doublets within the WRAP sequence.
To confirm this hypothesis, we synthesized analogues deleted of one leucine located at both N- and C-terminal regions of the WRAP peptides (W1-2L and W5-2L, respectively). Additionally, we synthesized the W5-2Lm peptide to analyze the role of the intra-sequential leucine doublets and the W5-4L peptide with all possible leucine doublet deletions (Table 1).
Both parental peptides, W1 and W5, were compared to their respective DLeu-analogues by agarose shift assay in order to verify first their ability to complex siRNAs (Figure 2A and 2B). Without the peptide addition, siRNAs migrated into the agarose gel (= 100% signal). But when complexed with an increased molar ratio of W1 and W5 peptides, siRNA migration was prevented in a molar ratio-dependent manner as previously reported  (Figure 2A and 2B). W1-2L, W5-2L, W5-2Lm and W5-4L clearly complexed siRNAs similarly to the parental peptides, with a complex formation starting at a molar ratio of 10 (R = 10 for >50% of complexed siRNAs). However, we decided to use the molar ratio R = 20 for PBN formulations since we showed that PBNs were thus more stable [16,17,24].
Additionally, biological assays were performed using the same PBN solutions studied by CD and DLS experiments. This allowed us to directly correlate the PBN formation with the biological activity, thus avoiding potential artifacts due to the preparation of different formulations.
As reported previously , WRAP peptide solutions were nearly unstructured on their own (W1 = random coiled and W5 = turn conformation). They adopted an α-helical conformation once incubated with siRNAs. Circular dichroism (CD) measurements were performed for all peptide analogues, both alone and in the presence of siRNAs to verify whether leucine doublet deletions would influence any conformational switch (Figures 2C and 2D and Additional files 1 and 2). W1-2L revealed a conformational switch showing the tendency of a helical structure formation (increased maxima at 195 nm and induced minima at 202 nm or 227 nm) (Figure 2C). In contrast, W5-2L, W5-2Lm and W5-4L remained mainly unstructured with the presence of the two minima (203 and 227 nm) suggesting a turn conformation as for W5 (Additional file 2 A-D). Upon incubation with siRNAs, the tryptophan cluster contribution, corresponding to the minima around 227 nm, was maintained for all mutants but the lack of leucine gave rise to a small maximum at 210 nm (Figure 2D). To better understand these differences, we performed in silico 3D structure prediction of WRAP and analogue peptides alone (Figure 2E). After computation, the peptide models generated by PEPstrMOD [38,39] in a hydrophilic environment revealed that W1, W5, W1-2L were able to adopt α-helical structure. W5-2Lm and W5-4L had no helical structure, confirming the CD measurements. It is noteworthy to mention that the leucine-deleted peptides W5-2L was predicted to adopt a short helix, which is in contradiction to the CD evaluation. However, in case of discrepancies between the measured and the expected results about the helical content within the peptides, it is worth noticing that we were more confident with the results provided experimentally by CD measurements than with those predicted by a theorical modelling system.
Afterwards, colloidal features of WRAP:siRNA complexes (R = 20) were characterized by Dynamic Light Scattering (DLS) to determine the nanoparticle size and polydispersity of size distribution. Intensity measurements (%) revealed that W1 and W5 formed PBNs with diameters of 80-100 nm with polydispersity indexes (PdI) around 0.3 (Table 1) as reported previously . A comparable result could be observed for W1-2L, showing a mean size of 111 nm with a PdI of 0.42. In contrast, W5-2L, W5-2Lm and W5-4L showed mean sizes higher than 1,000 nm indicating that these three peptides were not able to form PBNs in the presence of siRNAs.
Evaluating the role of arginine residues within the WRAP sequence.
Since the C6M1 peptide harbored arginine residues at both extremities, we wanted to evaluate the effect of arginine residue addition at both N- and C-terminal ends of the WRAP sequences by synthesizing analogues with one additional arginine residues at both ends of the peptide (Arg = 6 for W1-6R and W5-6R). However, because the comparison between C6M1 and WRAP peptides suggested the importance to keep an identical amount of arginine residues in order to maintain a good biological effect (Figure 1B), we designed peptides with two additional arginine residues at both extremities, but with two arginine residues less within the sequence to keep the same charge number as the parental peptides (Arg = 4 for W1, W5, W1-4R and W5-4R) (Table 1).
As performed above for the DLeu-analogues, we first evaluated by agarose gel shift assay the ability of these Arg-analogues to form PBNs in the presence of siRNAs depending on the used molar ratio required for an optimal siRNA complexation (Figure 3A). Increasing the number of arginine residues clearly improved complex formation compared to the parental peptides (molar ratio R = 7.5 for W1-6R and W5-6R versus R = 10 for W1 and W5, respectively). This was likely related with the higher number of positive charges in peptides containing 6 arginine residues. In contrast, the simultaneous N-and C-terminal arginine addition together with the internal arginine deletion seemed to slightly impact negatively nanoparticles formation as revealed by the higher molar ratio required to fully complex the siRNA (R = 12.5 for W1-4R and W5-4R). This could be related with the lower biological effect observed for the C6M1 peptides compared to the WRAP peptides (Figure 1B) and to the hypothesis of a weaker decaging ability for peptides containing a higher number of arginine residues. On the other hand, the CD spectra showed in both cases that W1-6R and W5-6R adopted in the presence of siRNAs a helical conformation comparable to those recorded with the parental peptides (Figure 3C and 3D). For the W1-4R siRNA-loaded complexes, we observed some slight changes in the helical structure while the lack of arginine inner residues of W5-4R do not give any conformational change compared to W5 in the presence of siRNA. Structure predictions by PEPstrMOD [38,39] in a hydrophilic environment revealed that the potential helix was shorter for W1-4R and W5-4R compared to the other peptides (Figure 3E). These behaviors could indeed impair with the capacity of these peptides to form stable PBNs with the siRNA.
As expected from the results above, all Arg-analogues complexed with siRNAs formed PBNs with mean sizes in the 100 nm range as measured by DLS (Table 1). However, we observed for W1-4R slightly bigger (441 ± 175 nm) and more dispersed PBNs (0.50 ± 0.05).
In conclusion, adding two arginine residues to the N- and C-terminal end of WRAP sequences seems to induce PBN formation at lower molar ratio without perturbing importantly the size of the formed PBNs. Displacing the two internal arginine residues at the N- and C-terminal end of WRAP sequences (W1-4R and W5-4R - same number of arginine residues as the parental peptides) resulted in minimal conformational changes and PBN formation, with the exception of W1-4R which formed four times bigger PBNs.
Cellular activity of the different WRAP PBNs was evaluated on a luciferase positive human glioblastoma U87 cell line. To this aim, PBNs solutions used for CD and DLS measurements were diluted to siRNA concentrations of 5 nM, 10 nM and 20 nM and directly applied on cells to perform the luciferase assay. W1:siRNA and W5:siRNA gave impressive silencing activities for the three siRNA concentrations (Figures 4A, 4B and Table 1). We noticed that these inhibitions were higher than inhibition levels previously reported . For instance, we obtained before around 60% luciferase activity remaining with 5 nM siRNA concentration compared here to 10%. Interestingly, we found out that this disparity resulted from the differently formulated PBNs. The knock-down efficiency of siRNA-loaded WRAP PBNs appeared more efficient when PBNs were first formulated at high concentration and then afterwards diluted, compared to those formulated directly at the desired concentrations (Additional file 3). At this moment, we do not have any rational explanation for understanding this phenomenon and we are performing studies to understand this factual result.
However, because all compared PBNs were formulated and diluted in the same way, the silencing efficiency could be directly compared. W1-2L:siRNA showed a reasonable luciferase knock-down activity even if the effect is not so pronounced compared to W1:siRNA. In contrast, none of the three W5 DL-analogues induced any luciferase silencing (Figure 4A).
All W1-6R, W5-4R and W5-6R siRNA-loaded PBNs showed a dose-dependent luciferase silencing, but slightly less important compared to the parental WRAP PBNs (Figure 4B). In contrast, with a bigger PBN size compared to parent peptide, W1-4R:siRNA had no activity at all used siRNA concentrations.
By looking in details all measured characteristics, we found out that the length of the peptide helix formed in the presence of siRNAs could be a favorable indicator for the PBN silencing activity (Table 1). Knowing that a typical a-helix contains 3.6 amino acids per helical turn , we simply calculated the amount of helix turns in the helix depending on the number of amino acids implicated in the helix formation (highlighted in red in the primary sequence in Table 1). Interestingly these numbers directly correlated with the level of the luciferase silencing. No luciferase silencing activities were measured for peptides forming less than two helical turns (W1-4R and W5-2L). However, if more than two helical turns were present in the peptide, we observed an important luciferase silencing activity suggesting that this minimal helix length was crucial for stable PBN formation and efficient cellular translocation.
However, the overall length of the helical structure in the PBNs could not fully highlight a clear correlation with an optimal silencing efficacy. For example, W1-6R:siRNA with nearly four helical turns showed less silencing activity than the parental W1 peptide (79% vs 91% of knock-down, respectively). A similar effect was observed for W5-6R:siRNA, with a number of helical turns (3.1) comparable to those of the parental peptide W5. In this latter case, the silencing activity dropped down from 88% for W5 to 57% for W5-6R, respectively. To confirm that arginine residue addition had a negative impact on PBN activity, we performed heparin competition experiments to check the stability of the different PBNs. Heparin are sulfated polysaccharide molecules, highly present on the extracellular matrix of cells, which could be able to interact with positive charges of peptides contributing thereby to the siRNA decaging. The heparin sensitivity of PBNs provoking their instability (Additional file 4) could explain the lack of silencing activity. Compared to the parental peptides W1 and W5, only W5-2L and W1-4R showed a higher level of destabilization in good correlation with the lack of silencing activity (Table 1). Whether this lower transfection efficacy was the result of a lower decaging of the siRNA at the cell membrane level or within the cell remains to be fully assessed.
Evaluating the role of the helix formation in the WRAP sequences.
To evaluate the importance of the helical structure, and more particularly of its length within the WRAP sequences on PBN formation, we synthesized a new peptide set by integrating one or two proline residues in the middle of their sequences. Indeed, proline residues are unable to form hydrogen bonds within an alpha-helix structure because of the lack of hydrogen on their amide nitrogen. Therefore, proline residues are well-known perturbators of helical structures .
As expected, both proline-containing peptides (W1-1P and W5-2P) showed nearly no structural features, whether the analysis was performed using molecular 3D structure prediction (PEPstrMOD) or using circular dichroism analyses (Figures 5A, additional file 1 and additional file 2). In details, CD analyses did not show any major structural changes whether analyses were performed on the peptides alone or associated with siRNAs (Additional file 1-F and additional file 2-G). Once mixed with siRNAs, both peptides W1-1P and W5-2P still formed PBNs, but with diameters of 151 ± 28 nm and 242 ± 60 nm, respectively, slightly bigger than those of the PBNs formed with the corresponding parental peptides (80-100 nm range, see Table 1). When evaluating their luciferase knock-down activity, W1-1P and W5-2P were shown to be 6 to 9 times less efficient, respectively, compared to the parental peptides (Figure 5B). Whether the slight increase in PBN size was responsible for this significant reduction luciferase silencing induced by the siRNA-loaded PBNs remains difficult to be clearly distinguished.
Because the central “RLLRSL” motif of the CADY peptide sequence initiated the helical structure in all hydrophobic media , we evaluated if the central leucin doublet could have also an impact on the W1 helix formation by synthetizing a W1 analogue having only one leucin residue in the middle of the sequence (W1-Lm). This D-L peptide analogue showed nearly no structural features as shown by CD measurements (Additional file 1-C) even if the PEPstrMOD prediction revealed a helix of 2.2 turns (Figure 5A). As already exposed, in case of discrepancies between the measured and the expected results about the helical content within peptides, it is worth noticing that we could be more confident in experimental results obtained from CD measurements than in those theoretically predicted. Nevertheless, the formed W1-Lm:siRNA complexes with a mean size of 127 ± 7 nm could induce a luciferase activity silencing of 68% revealing that the deletion of the internal leucin doublet could impact the size (higher) and the activity (lower) of the PBNs.
Finally, we also synthesized W1-mix and W5-mix analogues in which the tryptophan residues were displaced regarding their location in the parental peptides in order to evaluate their influence on the helix formation. Even, if the helicity of the peptides in the presence of siRNA was fully maintained (Additional file 1-G and additional file 2-H), we observed a slightly higher mean size diameter for these PBNs compared to the parent peptides. Concerning the biological effects of these W-mixed peptides, we were surprised to observe differences: W1-mix:siRNA revealed a luciferase knock-down activity of 80% compared to W5-mix:siRNA having only 63% (Figure 5B, Table 1). These results implicated that 4 tryptophan residues could be dispatched along the peptide sequence but that 3 tryptophan residues should be preferably grouped to obtain a comparable luciferase silencing property.
WRAP nanoparticles compared to lipid-based reagents for siRNA transfection
Finally, we compared our WRAP-PBNs with other commercially available reagents for siRNA transfection such as RNAiMAX (Life Technologies), INTERFERin (PolyPlus), DharmaFect (Dharmacon) and HiPerFect (Qiagen) selected from their publication score in PubMed.
First, we performed luciferase knock-down experiments strictly under the same conditions. We evaluated silencing properties of the WRAP-PBNs compared to the lipid-based reactants using the classic incubation conditions without removing the transfection solution (1.5 h in serum free-medium + addition of serum-containing medium for 36 h). Unfortunately, compared to the WRAP-based conditions, we observed an important cell mortality (>80%) for most of the lipid-based conditions making the comparison impossible (Data not shown).
Therefore, to avoid cell toxicity, we changed slightly the incubation protocol by removing the whole transfection solution after 1.5 h and adding fresh serum-containing medium for 36 h (Figure 6A). Using these conditions, cytotoxic effects of the lipid-based transfection reagents were reduced under 20% as measured by LDH assay. More importantly, we showed that all transfection reagents (peptide- or lipid-based) were identically active with a specific luciferase silencing of about >80% using a siRNA concentration of 20 nM (Data not shown) or 50 nM siRNAs (Figure 6A). Again, no silencing was recorded with the scrambled siRNA version.
To evaluate the potential long-term effect (overs 14 days) of the peptide- or lipid-based transfection reagents, we performed clonogenic assays. This cell survival assay evaluated all modalities of cell death based over a period of two weeks on the ability of a single cell to grow as a colony. In analogy to the luciferase screening, U87 cells were incubated with the different transfection reagents using the same protocol (Figure 6A). We determined a viability threshold corresponding to 100% ± 20% (dotted line in Figure 6B). Under this condition, siRNA alone, W1 and W5 alone or complexed with siRNAs (siLuc or siSCR) have no impact on cell viability with a cell proliferation equal to the non-treated cells.
In contrast, we could highlight important cytotoxic effects for RNAiMAX and DharmaFect, alone or complexed with siRNA (viability ≤20%). For INTERFERin, no cytotoxicity for the transfection reagent alone was recorded, but once complexed to siRNAs, only 30% of cell colonies survived. The only lipid-based transfection reagent without any deleterious effect on cell division was HiPerFect, whether used alone or complexed to siRNAs.
In conclusion, we evaluated 6 different peptide- or lipid-based transfection reagents for their luciferase silencing and long-term cell viability. Only W1, W5 and HiPerfect revealed the same silencing properties (>80% using 50 nM siRNA) without any long-term cytotoxicity compared to RNAiMAX, INTERFERin or DharmaFect. Therefore, W1 and W5 could be used as alternative transfection reagents for inducing an efficient siRNAs internalization and protein knock-down. Once the siRNA delivered into cells, peptides are expected to be fully degraded by proteases and recycled potentially by the cells.