Control of pH responsive peptide self-association during endocytosis is required for effective gene transfer

Abstract

Cationic amphipathic histidine rich peptides demonstrate differential nucleic acid binding capabilities at neutral and acidic pH and adopt conformations at acidic pH that enable interaction with endosomal membranes, their subsequent disordering and facilitate entry of cargo to the cell cytosol. To better understand the relative contributions of each stage in the process and consequently the structural requirements of pH responsive peptides for optimal nucleic acid transfer, we used biophysical methods to dissect the series of events that occur during endosomal acidification. Far-UV circular dichroism was used to characterise the solution conformation of a series of peptides, containing either four or six histidine residues, designed to respond at differing pH while a novel application of near-UV circular dichroism was used to determine the binding affinities of the peptides for both DNA and siRNA. The peptide induced disordering of neutral and anionic membranes was investigated using ²H solid-state NMR. While each of these parameters model key stages in the nucleic acid delivery process and all were affected by increasing the histidine content of the peptide, the effect of a more acidic pH response on peptide self-association was most notable and identified as the most important barrier to further enhancing nucleic acid delivery. Further, the results indicate that Coulombic interactions between the histidine residues modulate protonation and subsequent conformational transitions required for peptide mediated gene transfer activity and are an important factor to consider in future peptide design.

INTRODUCTION

The challenge of delivering therapeutic nucleic acids to mammalian tissues effectively and without causing undue cytotoxicity has led to the development of a number of non-viral vectors [1]. Although overshadowed for some time by cationic lipid and polymer based delivery systems, the emergence of cationic, secondary amphipathic peptides that can effectively deliver therapeutic siRNA not only in vitro [2, 3] but also in vivo [4] suggest that such peptides have strong potential for development as components of delivery systems which may find applications in human health. Consequently, increasing the efficiency of nucleic acid delivery mediated by such systems will ultimately increase their in vivo efficacy and their chances of being adopted as part of a therapeutic strategy.

Within the class of cationic amphipathic vector peptides, our recent interest has focussed on peptides that contain a pH responsive element, in particular histidine residues (LAH peptides) [5, 6] or 2,3-diaminopropionic acid [7], the objective being to take advantage of the pH changes that accompany endocytosis that modify the properties of the peptides.

The LAH peptides have proven to be effective at delivering plasmid DNA to a variety of cell lines and have siRNA delivery efficiency in vitro that is comparable to that of other transfection agents, including Lipofectamine [8] or the CADY peptide [9]. Further, the LAH4 peptide has found application in the delivery of protein based vaccines adjuvanted with Toll-like receptor 9 agonist CpG oligonucleotide leading to a therapeutic anti-tumour effect in mice [10]. The delivery afforded by pH responsive peptides is robust and substantially greater than that offered by analogous cationic but non-pH responsive peptides [7]. Thus, incorporation of histidine or imidazole has proven to be an efficient strategy to increase delivery to cells via endocytotic pathways [11]. However, our attempts to increase the efficiency of the original LAH4 peptide by introducing various modifications resulted in only modest improvements [12-14].

A better understanding of the mechanism of pH responsive peptide mediated delivery has revealed that a large amount of peptide is expected to be released from the nucleic acid-peptide complex during endosomal acidification [15] and that subsequently the interaction of the cationic peptide with the endosomal membrane, particularly anionic lipids, plays an important role in promoting release of cargo into the cell [12]. Increasing the histidine content of the LAH peptides was expected to increase the amount of peptide released from the peptide-DNA complex and improve disordering of anionic lipids in membranes leading to a substantial enhancement of nucleic acid delivery.

Here, since the expected enhancement was not obtained [13, 14], we have investigated the underlying physico-chemical properties of the peptides, in particular their interactions with polyanions including DNA and siRNA, their disordering of membranes of varying composition and their self-association in solution, all as a function of pH. We have investigated whether Coulombic interactions, that are likely to exist between histidine residues located close to each other in space, and modulated by the immediate environment, are the key to understanding the differing pH responses of LAH peptides. We identify the main barriers to increasing the efficacy of nucleic acid delivery and unlock the potential for further enhancement of the delivery capabilities of pH responsive peptides.

MATERIALS AND METHODS

Materials

The peptides (Table 1) were purchased from either EZBiolab (Carmel, IN, USA) or Pepceuticals Ltd (Nottingham, UK) as desalted grade and, for biophysics experiments, were further purified using water/acetonitrile gradients using either a Waters Symmetry™ C8, 5 μm, 7.8 × 100 mm column or a Waters SymmetryPrep™ C8, 7 μm, 19 × 300 mm column. DNA duplex, as described previously [16], (AAA TAC ACT TTT GGT and complement, Mw duplex 9141.1, ε = 233,145.9 L.mole⁻¹.cm⁻¹) and siRNA duplex (AAU CGA AGU ACU CAG CGU AAG and complement, Mw duplex 13355.0) were from Integrated DNA Technologies (Coralville, IA). The lipids 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphatidylcholine (POPC), 1-palmitoyl_d31-2-oleoyl-sn-glycero-3-phosphatidylcholine (POPC-d31), 1-palmitoyl_d31-2-oleoyl-sn-glycero-3-phosphatidylserine (POPS-d31) and cholesterol were obtained from Avanti Polar Lipids, Inc. (Alabaster, AL) and used without further purification. All other reagents were analytical grade or better.

Table 1.

Sequences of LAH4 or LAH6 derivatives used in this study. Lysine, alanine and leucine residues are in normal text, histidine and tryptophan residues are in bold and proline and phenylalanine are marked in bold and underlined. The pK_α quoted for the peptides is the midpoint of the main conformational transition detected in solution using far-UV circular dichroism.

Peptide	Sequence	Length	Average Hydro- phobicity (H)*	pKα at 37°C
LAH4-L1	KKALLAHALHLLALLALHLAHALKKA	26	0.050	5.10 ± 0.15
LAH(4)-P10	KKLAHALHLPALLWLHLAHALKKA	24	0.002	5.29 ± 0.05
LAH4-X1	KKLALHALHLLALLWLHLAHLALKK	25	0.058	4.67 ± 0.05
LAH4-X1F1	FKKLALHALHLLALLWLHLAHLALKK	26	0.079	4.87 ± 0.02
LAH4-X1F2	FFKKLALHALHLLALLWLHLAHLALKK	27	0.099	4.37 ± 0.01
LAH6-X1	KHKALHALHLLALLWLHLAHLAKHK	25	−0.016	5.21 ± 0.07
LAH6-X1L-W	KHKLLHLLHLLALLWLHLLHLLKHK	25	0.028	3.65 ± 0.18
LAH6-X1L	KHKLLHLLHLLALLALHLLHLLKHK	25	0.024	3.43 ± 0.01
LAH6-X1L-26	KHKLLHLLHLLALLALHLLHLLAHKK	26	0.032	3.52 ± 0.09
LAH6-X1L-P13	KHKLLHLLHLLAPLALHLLHLLKHK	25	0.000	5.95 ± 0.07

^*As determined by the Eisenberg (1982) Consensus scale (Ile, 0.73; Phe, 0.61; Val, 0.54; Leu, 0.53; Trp, 0.37; Met, 0.26; Ala, 0.25; Gly, 0.16; Cys, 0.04; Tyr, 0.02; Pro, −0.07; Thr, −0.18; Ser, −0.26; His, −0.40; Glu, −0.62; Asn, −0.64; Gln, −0.69; Asp, −0.72; Lys, −1.1; and Arg, −1.8).

Cell culture and DNA transfection

Dulbecco’s modified Eagle medium (DMEM; Gibco-BRL) was supplemented with 2 mM L-glutamine, 100 units/ml penicillin, 100 μg/ml streptomycin and 10% of fetal calf serum (FCS; HyClone). 120,000 human fetal lung fibroblasts (MRC-5) or 100,000 SV40 transformed human fetal lung fibroblasts (MRC-5 V2) per well were plated in 24-well plates (Costar) one day before transfection. SMD2-LucΔITR (7.6 kb) is an expression plasmid encoding the firefly luciferase gene under the control of the human cytomegalovirus (CMV) immediate-early promoter. 4 μg of plasmid DNA and the desired amount of peptide were each diluted in 100 μl of 150 mM NaCl and gently mixed. DNA complexes were prepared by varying the peptide/DNA w/w ratios in order to find out which ratio allows for maximal reporter gene expression. Optimal ratios were between 6:1 and 15:1 for the LAH4 and LAH6 peptides.

Luciferase assay

Cells were harvested in 250 μl of lysis buffer (8 mM MgCl₂, 1 mM dithiothreitol, 1 mM EDTA, 1% Triton X-100, 15% glycerol, and 25 mM Tris-phosphate buffer pH 7.8). The cell lysate was then transferred into Eppendorf tubes and centrifuged for 5 minutes at 10,000 g to pellet debris. Luciferase light units were measured in a 96-well plate format with a luminometer (Perkin Elmer) from an aliquot of the supernatant (50 μl). The measurement was done over 10 s after automatic injection of 100 μl assay buffer (lysis buffer without Triton X-100 supplemented with 2 mM ATP) and 100 μl of luciferin solution (167 μM in water; Molecular Probes). Luciferase background was subtracted from each value and the protein content of the transfected cells was measured by Bradford dye-binding using the BioRad protein assay. The transfection efficiency is expressed as light units/10 s/mg protein and the values are the means of the duplicates. Statistical analysis was by ANOVA with Bonferonni post-hoc t-test, one tailed.

Circular dichroism experiments

Peptides were dissolved in 5 mM Tris amine without pH adjustment at a final concentration of between 30 and 50 μM. The samples were titrated down by adding 0.3% or 1% (v/v) HClO₄ solution in micro-litre amounts. ClO ⁻₄ is optically transparent hence the substitution of HCL (which has substantial absorbance below 200 nm) by HCLO4 makes a significant improvement on spectral quality in the far-UV region. CD spectra were acquired on a Chirascan™ Spectrometer (Applied Photophysics, Leatherhead, UK) with samples maintained at 310 K. For pH titration experiments, far-UV CD spectra were obtained with the samples titrated down by adding 0.3% or 1% (v/v) HClO₄ solution in micro-litre amounts. Spectra were recorded from 260 to 180 nm using a 0.5 mm path length and were processed using Chirascan software where a spectrum of the peptide free solution was subtracted and Savitzky-Gorlay smoothing with a convolution width of 5 points applied. For nucleic acid/peptide binding measurements, both near and far-UV CD spectra were obtained. Near-UV measurements were performed in a 10 mm path length cell, recording from 340 to 220 nm; far-UV measurements were performed on the same sample in 1 mm cells recording from 260 to 185 nm. DNA or siRNA was prepared at a concentration of approximately 1.6 μM in 5 mM Tris/piperazine buffer at either pH 7 or pH 5 and was titrated through the addition of small volumes of 0.4 mM peptide stock solutions prepared in the same buffer. UV absorbance spectra, acquired simultaneously with the CD measurements, are presented without correction in the supplementary materials to show the effects of light scattering and here, after being zeroed at 340 nm, to allow the absorbance at 260 nm to be followed more readily. The CD intensities at 272 nm (DNA) or 263 nm (siRNA) were plotted as a function of the change in peptide to DNA weight ratios. The structural response of peptide bound to the DNA duplex to changes in pH was also probed using far-UV circular dichroism. DNA solutions were prepared in 5 mM Tris amine without pH adjustment and loaded with peptide at 37°C as per the peptide-DNA binding experiment described above. The preparation of the complex was monitored using near-UV CD as above and when the fraction of unbound peptide reached 0 the sample was transferred to a 1 mm path length cell and pH titrations was conducted as above. Results for all CD assay measurements are expressed as an average of two or more independent repeated experiments.

Solution NMR measurements

Around 1.5 mg of peptide was solubilized in 600 μl D₂O/H₂O (10%/90%) containing 5 mM Tris buffer, pH 8.4. Titration was as above, Spectra were recorded on a 500 MHz Bruker Avance spectrometer equipped with a cryoprobe, using a standard 1D WATERGATE pulse sequence (p3919gp), at 310 K. 16 scans were accumulated, with a recycle delay of 1 second.

Sample Preparation for Solid-state NMR

For solid-state NMR, samples with the lipid composition POPC/POPC-d31/cholesterol (85:15:30) or POPC/POPS-d31/cholesterol (85:15:30) were prepared. A total of around 5 mg lipids per sample were dissolved and mixed in chloroform and dried under rotor-evaporation at room temperature. In order to remove all organic solvent, the lipid films were exposed to vacuum overnight. The films were then rehydrated with 4 ml of a suspension of LAH4-L1 or LAH6-X1L in 10 mM Tris/piperazine buffer at various pH at room temperature. For wideline ²H NMR experiments LAH4-L1 or LAH6-X1L derivatives were added to the lipids at 1.5% by mol. Samples were subjected to five rapid freeze-thaw cycles for further sample homogenization, generating multi-lamellar vesicles, and then centrifuged at 21000g for 30 min at room temperature. Previous determination of a binding constant of 4.2 × 10³ [17] for LAH4 and PC vesicles at pH 7.5 indicates that very little peptide will be free in solution after this process with binding for peptides to anionic membranes, particularly at acidic pH, likely to be stronger. The pellets, containing lipid vesicles and associated peptides were transferred to Bruker 4 mm MAS rotors for NMR measurements. Lipid vesicles were also prepared in this way in the absence of peptide.

Solid-state NMR

²H quadrupole echo experiments [18] for samples containing either POPC-d31 or POPS-d31 were performed at 61.46 MHz on a Bruker Avance 400 NMR spectrometer using a 4 mm MAS probe, spectral width of 100 KHz and with recycle delay, echo delay, acquisition time and 90° pulse lengths of 0.25 s, 100 μs, 2.6 ms and 3 μs respectively. The temperature was maintained at 310 K to keep the bilayers in their liquid-crystalline phase. During processing the first 10 points were removed in order to start Fourier-transformation at the beginning of the echo. Spectra were zero filled to 1k points and 50 Hz exponential line-broadening was applied. Smoothed deuterium order parameter profiles were obtained from symmetrised and dePaked ²H-NMR powder spectra of either POPC-d31 or POPS-d31 using published procedures [18-21]. Order parameters were averaged along the length of the acyl chain and the difference in average order between peptide free and peptide containing liposomes was used to calculate a pK_mem for the peptide induced disordering of the deuterated reporter lipids.

RESULTS

Peptide design

A series of ten cationic amphipathic peptides were used in the present study (Table 1) and are based on the most potent nucleic acid delivery peptide that we have generated over previous studies, LAH4-L1 [12]. When the peptides adopt an amphipathic helix, the histidine residues are segregated on one face and, in an idealised helical wheel representation, describe an angle of 80° [12]. This angle is maintained in each of the ten peptides used here. Five of the peptides contain four histidine residues and are primarily distinguished by their differing average hydrophobicites with increasing hydrophobicity conferred by the addition of phenylalanine residues at the N terminus. All peptides are amidated at the C-terminus conferring a nominal charge of +5 to all peptides at neutral pH. Of the five peptides containing six histidine residues, four are designed to adopt uninterrupted α-helix conformations. These peptides are differentiated by only small changes in the positioning of the lysine residues or the presence or absence of tryptophan which has been incorporated to allow measurements of tryptophan fluorescence which are not discussed here. Of note however is one peptide, LAH6-X1, which is substantially more hydrophilic than the other LAH6 peptides. This increase in hydrophilicity is caused by replacing two leucines in the LAH4-X1 sequence with histidine to generate LAH6-X1. The remaining LAH6 peptides have had the average hydrophobicity of the template LAH4 peptides restored by substituting a number of the alanine residues with leucine. Finally, two peptides, LAH(4)-P10 (based on the LAH template [7]) and LAH6-X1L-P13, incorporate a single proline residue which is intended to disrupt the formation of α-helix conformation which is related to the self-association of analogous antimicrobial peptides [22].

Delivery of plasmid DNA to lung fibroblast cells

The efficacy at delivering plasmid DNA to a variety of cell lines was investigated for each peptide. LAH6-X1L-W, LAH6-X1L and LAH6-X1L-26 all have similar transfection capabilities [9] hence only the former is considered in the present study, while LAH-P10 performed similarly to its parent peptide LAH whose capabilities are reported elsewhere [7]. Here data for 6 peptides are shown: four peptides have 4 histidines while the remaining two have 6 His residues. Robust gene delivery capabilities to both cell lines demonstrated by LAH4-L1 were not replicated by all peptides tested (Fig. 1). In particular, the effectiveness of LAH4 peptides at delivering reporter gene to MRC-5 cells decreased with increasing peptide hydrophobicity while LAH6 peptides also had inferior performance. MRC-5 foetal human lung cell fibroblasts and MRC-5 V2 are closely related with the latter derived from the former by transformation by SV40 virus [23]. Of a number of changes induced in this cell line by SV40 transformation [23], we have shown that this cell line also has increased levels of phosphatidylserine in the outer leaflet of the cell membrane [24]. MRC-5 V2 cells were more amenable to gene transfer by all of the pH responsive peptides than the MRC-5 cells. Gene delivery mediated by hydrophobic LAH4 peptides was observed for MRC-5 V2 cells where low or none was observed for MRC-5. Of the two LAH6 peptides, LAH6-X1L-W proved more effective at delivering to the transformed cell line than LAH6-X1.

Figure 1.

Transfection of human lung fibroblasts (MRC-5) or SV40 transformed human lung fibroblasts (MRC-5 V2) with a luciferase reporter gene by a variety of pH responsive peptides. Luciferase activities are shown adjusted for protein content (A) and relative to the most effective gene delivering peptide (LAH4-L1) for both cell lines (B).

Peptide conformation and self association in solution

CD spectra with characteristic negative bands at 220 and 210 nm and positive bands at 193 nm reveal that LAH4-L1 adopts an α-helix conformation when dissolved in neutral or slightly basic aqueous solution (Fig. 2A). When the solution is titrated with acid, a shift to a disordered conformation is observed and the intensity of the positive band at 193 nm and negative bands at 220 nm are reduced with the negative band at 210 shifting to shorter wavelength. The CD intensity at 220 nm is representative of the α-helix content of the peptide and, when plotted as a function of pH (Fig. 2B), a pK_α for the conformational transition of 5.10 ± 0.15 is determined. Previous studies using dynamic light scattering to characterise LAH4-L1 in solution have established that at neutral pH the peptide exists in small particles of around 30 nm diameter whereas at acidic pH a hydrodynamic diameter of only around 3 nm is found [25]. Taken together, this indicates that the α-helix conformation is stabilised when the peptide exists in a self-associated state and that the CD studies in solution monitor both secondary structure and peptide self-association. Though not well suited to the study of peptide aggregates, NMR has shown that the histidine side chains in LAH peptides become protonated with a pK_a close to the pK_α for the conformational transition [6, 26, 27]. The pK_α value determined is notable since it is lower than that often quoted for histidine. We reasoned that this lower value is a consequence of Coulombic interactions that would exist between the histidine side chains that are located close to each other through space [27], in particular when the α-helix conformation is adopted. According to this theory, if modifications to the peptides promote self-association and stabilise the α-helix conformation then the pK_α observed would be depressed. Similarly, increasing the number of histidine residues in the sequence would also increase the inter residue Coulombic interactions, hindering protonation, and would also be manifested in a more acidic conformational transition. The conformational response to pH changes of LAH4 peptides of increasing hydrophobicity (Fig 2C/D; Supp. Fig.1) indicates that the more hydrophobic peptides do indeed respond at a more acidic pH. More notable however is the effect of increasing the number of histidines. LAH6-X1L-W adopts an α-helix conformation at neutral pH and also responds to decreasing pH by adopting an increasingly disordered conformation (Fig. 3A). However, the midpoint of this transition is at a much more acidic value of 3.65 ± 0.18 (Fig. 3D), when compared with LAH4-L1. Similar results were obtained for two closely related peptides, LAH6-X1L and LAH6-X1L-26 (Table 1). LAH4-L1 and LAH6-X1L-W are of similar hydrophobicity and it is notable that when the hydrophobicity is not adjusted, following incorporation of the two further histidines, the more acidic conformational response is not observed. At first sight LAH6-X1 also appears to adopt an α-helix conformation at neutral pH but the red shift of the negative band around 227 nm and the altered ratio of positive (193 nm) and negative (c 210 nm) band intensities is suggestive of ordered turn conformations. Nevertheless, as with the other peptides studied, LAH6-X1 also responds to decreasing pH by adopting a disordered conformation (Fig. 3B). The conformational transition however has a midpoint at a more basic 5.21 ± 0.07 supporting the theory that both self-association and concomitant α-helix formation is required for the effects of increasing the histidine content on the conformational transition to be fully registered. The role of Coulombic interactions in determining the pH response of histidine containing peptides was further tested by investigating the effect of anionic charges. LAH6-X1L-W was solubilised in the presence of the anionic detergent sodium dodecyl sulphate (SDS) which stabilises an α-helix conformation at both basic and acidic pH (Fig. 3C). The CD spectra are of sufficient quality however that the pH response is readily detected when the change in the CD signal at 220 nm is plotted as a function of pH. The peptide responds at a substantially more basic pH in the presence of SDS when compared with the corresponding titration in aqueous conditions. This is consistent with the negative charges of SDS mitigating the influence of the inter-side chain Coulombic interactions and also application of the Gouy-Chapman-Stern model that would predict a greater concentration of protons in the proximity of the negatively charged micelle surface [28].

Figure 2.

Circular dichroism reveals the differential pH response of cationic amphipathic peptides containing four histidine residues. Far-UV CD spectra of LAH4-L1 (A) in aqueous environments indicate the peptide undergoes a helical to unordered transformation as the pH is lowered from pH 7. The change in ellipticity at 220 nm is plotted as a function of pH to determine the pK_α for this transition for LAH4-L1 (B) and three further peptides (C). The determined pK_α is compared with the average hydrophobicity for each peptide establishing a trend where increasing hydrophobicity leads to a more acidic pH response.

Figure 3.

Circular dichroism reveals the differential pH response of cationic amphipathic peptides containing six histidine residues. Far-UV CD spectra of LAH6-X1L-W (A) and LAH6-X1 (B) in aqueous environments confirm both peptides undergo a helical to unordered transformation as the pH is lowered from pH 7. A similar experiment performed for LAH6-X1L-W in the presence of 50 mM SDS (C) indicates α-helix conformations are stabilised at both acidic and neutral pH but a pH dependent conformational change can nonetheless be detected. When the change in ellipticity is plotted in response to changes in pH (D) a low apparent pK_α for LAH6-X1L-W in an aqueous environment of 3.65 ± 0.18 is determined which is dramatically raised to 6.98 ± 0.02 in the presence of the anionic detergent.

Effect of proline on secondary structure, pH response and gene delivery

Since the pK_α results obtained for peptides in aqueous solution indicated that a combination of both self-association and increased histidine content is required for very acidic responses to pH changes, we investigated whether disruption of α-helix secondary structure would cause the pK_α to revert to a more basic value that might be achieved during the earlier stages of endocytosis. One final peptide was therefore prepared incorporating a proline residue since a further parallel study combining far-UV CD spectroscopy with gel filtration chromatography has shown that the self-associated state of related peptides can be readily disrupted by such a feature [22]. The peptide was found to be amenable to study by both ¹H NMR (Fig. 4A) and CD (Fig. 4C) in aqueous solution. The resonances attributable to the ε1 and δ2 protons of the six histidine residues of LAH6-X1L-P13 were identified between 6.8 and 8.8 ppm and could be followed during pH titration. Although resonances were not assigned to individual histidine residues, the chemical shift changes were plotted as a function of pH (Fig. 4B) and the pK_a of these residues were found to lie between 6.04 ± 0.04 and 6.46 ± 0.06 when adjusted for the presence of 10% deuterium (the pK_a in the presence of deuterium, pK_a* ≈ pK_a - 0.4 [29]). Far-UV CD spectra were obtained as above for the peptide in aqueous solution and although spectral features suggestive of some α-helix conformation were detected at basic and neutral pH, the low intensity of the positive band at 193 nm and two negative bands at 210 and 220 nm indicated that a more disordered average conformation with substantially lower α-helix content is adopted by the proline containing peptide when compared with the proline free analogues. Furthermore, plotting the change in CD at 220 nm as a function of pH (Fig. 4D) reveals the pK_α, of the structural transition to an almost completely disordered state, was 5.95 ± 0.07, almost two and a half units higher than that for the analogous proline-free peptide, LAH6-X1L. The gene delivery capabilities of LAH6-X1L-P13 were also assessed but the peptide had poor transfection capabilities and was observed to induce substantial cell cytotoxicity (data not shown).

Figure 4.

The introduction of a proline residue in the LAH6-X1L sequence restores an elevated pH response. Titrations of LAH6-X1L-P13 in solution monitored by ¹H NMR (A) or circular dichroism (C) reveal the protonation state of the histidine residues and the secondary structure are sensitive to changes in pH in the same region. Analysis of the chemical shift changes in the ε1 and δ2 protons of the six histidine residues (C) produces pK_a in the range 6.04 to 6.46 when adjusted for the presence of deuterium, while the CD response at 220 nm (D) pinpoints the main conformational transition at pH 5.95 ± 0.07.

Peptide binding to nucleic acids

Following the observation that the more hydrophobic LAH6 peptides responded at a very acidic pH when suspended in solution but that this was strongly influenced by the presence of negative charges, we investigated the relative nucleic acid binding capabilities of LAH4-L1 and LAH6-X1L-W at neutral and at an acidic pH that might be achieved during endocytosis. A peptide-nucleic acid binding assay was devised based on the conformational response of either duplex DNA or siRNA to the addition of peptide. We have shown previously that the response in the near-UV CD spectra obtained for duplex DNA can be related to the relative binding of cationic peptides [16]. Representative near-UV CD spectra obtained for duplex DNA are consistent with that expected for B-form DNA (Fig. 5A). Condensation of the DNA through the addition of peptide leads to an initial collapse of the positive band at 273 nm which stabilises at a point where further addition of peptide induces no further change to this region of the spectrum, indicating all peptide binding sites are all occupied. Similar experiments were performed for siRNA incubated with e.g. LAH4-L1 (Fig. 5C). Here we have used the change in the CD signal, at 273 nm for duplex DNA and 260 nm for siRNA, to obtain an EC₅₀ for the binding of peptide to nucleic acid (Fig. 5B/D). The near-UV CD spectra obtained in the absence of peptide represent the situation where all possible peptide binding sites are vacant. When the above signals are plotted as a function of the peptide to nucleic acid weight ratio, a sigmoidal response is detected. The point at which no further change in nucleic acid signal is detected is considered to represent nucleic acid with all peptide binding sites occupied and the mid-point in this transition is the binding capacity EC₅₀.

Figure 5.

Typical near UV circular dichroism spectra of 15 base pair duplex DNA (A) or 21 base pair duplex siRNA (C) incubated with increasing amounts of LAH4-L1 at room temperature. The nucleic acid concentration was typically between 1 and 2 μM and peptide to nucleic acid weight ratios are shown. The CD response at 273 nm for DNA (B) or 263 nm for siRNA (D) is used here as a measure of the fraction of unbound peptide remaining and allows comparison of the binding characteristics of the peptides at neutral and basic pH.

The qualitative information provided by this technique allows for a direct comparison of the relative binding affinities of the LAH4-L1 and LAH6-X1L-W peptides at pH 7 and pH 5. The summarised results (Table 2) show the peptide to nucleic acid weight ratios at which half the available peptide binding sites are occupied. This allows comparison with the weight ratios used in the in vitro transfection experiments. When molar ratios are analysed, the indication is that a 15 base pair stretch of DNA can accommodate between four and ten peptides bound directly to the nucleic acid depending on the pH and the peptide. Importantly, both peptides have a substantially greater affinity for both types of nucleic acid at acidic pH. The results obtained for LAH4-L1 using the CD binding assay indicate that around half as much peptide can be accommodated by the nucleic acids at acidic pH as at neutral pH, consistent with results obtained for LAH4 by calorimetric methods [15], and confirming that substantial amounts of peptide can be expected to be released from the peptide-nucleic acid complex during endosomal acidification. Crucially, the same characteristics are observed for LAH6-X1L-W and hence substantial amounts of this peptide can also be expected to be released during endocytosis despite the very acidic conformational response observed above for the peptide in solution.

Table 2.

The pH responsive behaviour of LAH4-L1 and LAH6-X1L-W. The pK_α quoted for the peptides in solution and bound to the DNA duplex is the midpoint of the main conformational transition detected using far-UV circular dichroism. EC₅₀ values are a measure of the binding capacity derived from the structural transitions of the nucleic acid in response to peptide binding and are based on peptide to nucleic acid weight to weight ratios. Values are presented ± standard deviation of two or more independent repeats. All pK_α values were determined at 37°C whereas binding experiments were conducted at room temperature.

Peptide	pKα [solution]	pKα [DNA]*	EC50 DNA pH 7	EC50 DNA pH 5	EC50 siRNA pH 7	EC50 siRNA pH 5
LAH4-L1	5.10 ± 0.15	4.88 ± 0.04	3.25 ± 0.01	1.61 ± 0.03	3.88 ± 0.32	1.70 ± 0.37
LAH6-X1L-W	3.65 ± 0.18	4.03 ± 0.05	2.26 ± 0.11	1.32 ± 0.18	6.19 ± 0.07	3.57 ± 0.72

^*The limitations of pK_α values determined for peptides bound to DNA are discussed in the results.

Membrane activity

The membrane disordering capabilities of pH responsive peptides at acidic pH have previously been linked to gene delivery activity [12]. Here, by preparing samples at pH values between 8 and 4 we have been able to identify whether the more acidic response of LAH6-X1L is also manifested in the disordering capabilities of the peptides. Furthermore, by incorporating either POPC-d31 or POPS-d31 as the deuterated reporter lipid the influence of anionic lipids in the membrane on the pH response can be assessed. ²H echo spectra of multi lamellar vesicles incorporating deuterated lipids (Supp. Fig. 4/5) comprise quadrupolar splittings for the deuterated groups located at increasing depth in the membrane. Groups located closer to the tail of the lipid and hence closer to the hydrophobic core of the membrane are more disordered and give rise to splittings of lower magnitude. Disordering induced by peptides is manifested by reduced splittings and calculated order parameters and is again plotted as a function of pH (Fig. 6). The pH response of LAH6-X1L, in terms of membrane disordering, was consistently lower than that of LAH4-L1 in both membrane types while the presence of anionic lipids caused both peptides to respond at a more basic pH (Fig. 6B). Despite the observed differences in membrane activity, both peptides however did respond to changes in pH that would be expected during endocytosis and at a pH much higher than that observed for the structural transition in solution. This supports the view that most of the observed peptide induced disordering occurs as a result of the structural re-alignment [17, 26], and not from alterations in binding affinity.

Figure 6.

Solid-state NMR of chain deuterated lipids reveals the effect of pH on membrane disordering induced by pH responsive peptides. Average order parameters (S_CD) are shown for the deuterated acyl chains in membranes comprising POPC/POPC-d31/cholesterol (A) or POPC/POPS-d31/cholesterol (B) in the absence or presence of 1.5 mol % peptide. In the presence of peptide, increasing disordered membranes are observed at acidic pH. The samples were maintained at 37°C.

DISCUSSION

Barriers to improving gene delivery mediated by histidine containing pH responsive peptides

The existing paradigm for LAH peptide mediated nucleic acid transfer involves condensation of the nucleic acid by peptide at neutral pH, adhesion to the cell membrane and stimulation of cellular uptake via endocytosis pathways. During endocytosis, the acidification of the endosomal lumen to around pH 5.9 [30] and concomitant increase in the cationicity of the pH responsive peptides causes a substantial amount of peptide to be released from the peptide-nucleic acid complex [15]. Peptides released from the complex are able to access the surface of the endosomal membrane and act to disorder or disrupt the membrane facilitating release of the endosomal contents into the cytosol.

The present study incorporates a combination of biophysical techniques, including a novel application of near-UV CD spectroscopy to characterise peptide-nucleic acid binding, and indicates that modifications to peptide primary structure will affect at least three stages of this process, namely the binding and release of peptide from the nucleic acid, the conformation and aggregation state of the peptide in solution and the disordering of the membrane. Importantly, the pH at which the peptides respond will influence each of these stages. Indeed, if the pH response occurs at a too acidic pH then the peptides will not be activated at a sufficiently early stage and cargo will be transferred from early and late endosomes to lysosomes and will be degraded.

The LAH6 peptide with a net hydrophilic character, LAH6-X1, undergoes a conformational transition in solution at a pH close to that of the original LAH4-L1 peptide yet has lower gene transfer capabilities. Analogous antibacterial peptides operate optimally within a narrow window of hydrophobicity [31] with increasing hydrophobicity and/or hydrophobic moment leading to increased activity against mammalian cells [32, 33]. By analogy, it can be expected that any reduction of the hydrophobicity or hydrophobic moment in the design of pH responsive peptides is likely to substantially impair the ability to partition into the membrane and affect the entry of cargo.

Despite introduction of two additional histidines, whilst maintaining the hydrophobicity and hydrophobic moment of LAH4-L1 in the same range, the variants LAH6-X1L and LAH6-X1L-W displayed gene transfer capabilities which remained lower than those of LAH4-L1 [9]. Therefore, a detailed comparison of the biophysical properties of LAH4-L1 and these LAH6 variants (Table 2/3) was required to identify the most likely factor limiting the activity of the LAH6 peptides. A parallel study has compared the size of peptide/nucleic acid complexes comprising either siRNA or DNA and each of LAH4-L1, LAH6-X1L and LAH6-X1L-W at peptide to nucleic acid ratios related to the in vitro transfection experiments. While siRNA containing complexes formed with the LAH6 peptides were significantly smaller than corresponding LAH4-L1 containing complexes, differences were smaller and often not significant between LAH6 peptides and LAH4-L1 for plasmid DNA containing complexes [9]. In the present study, LAH4-L1 and LAH6-X1L-W are shown to share similar pH dependent DNA and siRNA binding characteristics in that the nucleic acids accommodate a much lower amount of both peptides at pH 5 compared with pH 7 ensuring a substantial amount of peptide would be released during endosomal acidification. LAH4-L1 and LAH6-X1L also share similar membrane disordering capabilities and, although LAH6-X1L consistently responds at a more acidic pH, both would be expected to be capable of exerting disordering effects on membranes during the acidification of the endosomes. In contrast, the pK_α for the structural transition observed for the peptides in solution is markedly different, with that observed for the two closely related LAH6 peptides occurring almost one and a half pH units below that observed for LAH4-L1 and substantially below the pH that would be expected to be achieved during the early stages of endocytosis. Accordingly, release from acidic compartments will only be achieved when very acidic pH is obtained. Consistent with this, the intra-cellular distribution of LAH4 and LAH6 peptides was found to differ considerably when used in the parallel study to deliver fluorescently labelled siRNA to MCF-7 breast cancer cells [9]. LAH6-X1L/siRNA complexes were shown to co-localise with lysosomes to a much greater degree than LAH4-L1/siRNA complexes, supporting a model where the low pK_α causes insufficient disruption of earlier compartments.

Table 3.

The pH responsive behaviour of LAH4-L1 and LAH6-X1L. The pK_α quoted for the peptides in solution is the midpoint of the main conformational transition detected using far-UV circular dichroism. The pK_mem quoted for the two membrane systems is the midpoint of the peptide induced disordering of the deuterated lipid. EC₅₀ values are a measure of the binding capacity derived from the structural transitions of the nucleic acid in response to peptide binding and compare peptide to DNA weight ratios. Values derived from NMR are presented ± standard error while the remaining values are presented ± standard deviation of two or more independent repeats. All pK values were determined at 37°C whereas binding experiments were conducted at room temperature.

Peptide	pKα [solution]	pKmem [POPCd31]	pKmem [POPSd31]	EC50 DNA pH 7	EC50 DNA pH 5
LAH4-L1	5.10 ± 0.15	5.95 ± 0.19	6.75 ± 0.28	3.25 ± 0.01	1.61 ± 0.03
LAH6-X1L	3.43 ± 0.01	5.49 ± 0.28	6.40 ± 0.59	1.51 ± 0.13	1.15 ± 0.02

The α-helix conformation adopted by the peptide in solution is indicative of a self-associated state, generating particles of around 30 nm in diameter [25]. The disordered conformation identified at more acidic pH corresponds to a smaller particle [25]. In light of recent studies with analogous cationic anti-cancer peptides, the persistence of the self-associated state for hydrophobic LAH6 peptides at low pH suggests a possible mechanism for the inhibition of peptide mediated nucleic acid transfer. The high cationic charge density of the peptides will promote electrostatic binding to membrane constituents including anionic lipids but also glycosaminoglycans, in particular heparan sulphate proteoglycans (HSPGs) [34-36]. The role of negatively charged HSPGs in the uptake of cationic complexes is unclear since clathrin and caveolae independent, proteoglycan-dependent uptake pathways have been demonstrated for complexes [37] that have also been shown to enter cells via either clathrin or caveolae dependent pathways [38]. Nevertheless, while also likely contributing to the first step in internalisation of pH responsive peptides and their cargoes, the HSPGs and other components of the outer cell envelope may also constitute a barrier, preventing access of some peptides to the membrane itself. Cationic antimicrobial peptides with anti-tumour activity are considered to exert their cytotoxic activity through interactions with the plasma membrane. The anti-cancer activity of such peptides has been shown to be inhibited by HSPGs [39]. Interestingly, when the anti-tumour peptides were shortened, their cytotoxic activity was shown to be insensitive to the presence of cell surface HSPGs [40]. These studies suggest a hypothesis that access to the cell membrane by anti-tumour peptides is controlled by HSPGs, with larger peptides or peptide aggregates having a greatly reduced ability to reach the target membrane and exert a disordering effect. Analogously, if access to the early endosomal membrane for pH responsive peptides is controlled in a similar way either by the persistence of clathrin or HSPGs then, by persisting in a self-associated form, even at rather acidic pH, hydrophobic LAH6 peptides would be less able to access the membranes of early endosomes and aid escape of nucleic acid cargo.

The potential for delivery selected by membrane charge

The influence of an anionic environment on the pH response of the histidine rich peptides is notable. At one extreme, the presence of the anionic detergent SDS causes a shift in the conformational pH response of LAH6-X1L-W of over three pH units. Consistently more basic pH responses were also detected in model membranes containing anionic phosphatidylserine when compared with membranes composed of zwitterionic phosphatidycholine. Taken together, this indicates that the presence of anionic charges in the target endosomal membrane may mitigate the acidic shift in pH response caused by introducing further histidine residues into the sequence. Furthermore, since LAH peptides have been shown to be particularly effective at disordering anionic lipids in such membranes [12], the presence of such lipids in the target membrane may promote selective delivery to cells with this characteristic. Some evidence for this is provided by the greater relative gene transfer to MRC-5 V2 over MRC-5 cells mediated by LAH6-X1L-W. However the self-associated configuration of this and related peptides will largely prevent such advantageous interactions.

Overcoming barriers to improved nucleic acid transfer

Here, a combination of biophysical techniques, including a novel application of near-UV CD spectroscopy to characterise peptide-nucleic acid binding, has revealed the barriers to improved peptide mediated nucleic acid transfer but has also suggested means by which these barriers may be overcome. Since the self-associated solution conformation adopted by LAH6-X1L-W and related peptides hinders the peptide from reaching the endosomal membrane where it may enjoy favourable interactions with anionic lipids, strategies that seek to promote a conformational change at a less acidic pH [6] may succeed in increasing both the efficacy and selectivity of gene transfer.