Conformation of oligodeoxynucleotides associated with anionic liposomes

Abstract

There has been significant progress in the development of antisense therapeutics for a wide range of medicinal applications. Further improvement will require better understanding of cellular internalization, intracellular distribution mechanisms and interactions of oligodeoxynucleotides with cellular organelles. In many of these processes interactions of oligodeoxynucleotides with lipid assemblies may have a significant influence on their function. Divalent cations have been shown to assist cellular internalization of certain oligodeoxynucleotides and to affect their conformation. In this work we have investigated conformational changes of phosphorothioate oligodeoxynucleotides upon divalent cation-mediated interaction with 1,2-dipalmitoyl-sn-glycero-3-phosphatidylglycerol (DPPG) liposomes. For the sequences investigated here the native conformation underwent significant change in the presence of anionic DPPG liposomes only when divalent cations were present. This change is sequence-specific, ion-selective and distinct from previously reported changes in oligodeoxynucleotide structure due to divalent cations alone. The conformation of one oligodeoxynucleotide in the presence of calcium and DPPG yields circular dichroism spectra which suggest C-DNA but which also have characteristics unlike any previously reported spectra of liposome-associated DNA structure. The data suggest the possibility of a unique conformation of liposome-associated ODNs and reflect a surprisingly strong tendency of single-stranded DNA to retain a characteristic conformation even when adsorbed to a surface. This conformation may be related to cellular uptake, transport of oligodeoxynucleotides in cells and/or function.

INTRODUCTION

Antisense oligodeoxynucleotides (ODNs) are being developed into therapeutics for a wide range of pathological conditions, such as inflammation, cancer and AIDS (1). With the development of potent nuclease-resistant analogs there has been a growing interest in enhancing penetration into target cells, such as those of the central nervous system or tumor tissues (2). Since ODNs are charged molecules of high molecular weight, one would expect that transport across cell membranes or lipid bilayers should be difficult (3). Mechanisms underlying cellular uptake of ODNs are now being resolved. For many years it was speculated that cellular internalization of ODNs took place primarily by receptor-mediated uptake involving scavenger receptors (4). However, a recent study using knockout mice has demonstrated that uptake characteristics are virtually identical in animals with or without the receptor (5). Others have shown that ODN uptake in certain cell culture systems is not influenced by pretreatment of cells with inhibitors of receptor-mediated endocytosis (chloroquine or monensin) (6). It has been shown that receptor-independent uptake is possible, but the mechanism is still unclear. It has been demonstrated that in some cell cultures divalent cations can facilitate cellular internalization of ODNs (7). However, the precise role of divalent cations in this process has still not been demonstrated. Divalent cations may induce structural perturbations in the lipid bilayer and/or may affect the secondary structure of the ODNs and thereby facilitate the penetration of ODNs across cell barriers.

Interactions between ODNs and membranes are complex. The importance of elucidating the mechanisms of ODN uptake into cells and the interaction of ODNs with certain cellular organelles is becoming apparent. The present work involves investigation of the interaction of ODNs with anionic liposomes. Liposomes of specific composition have been used as simple models of cell membranes to understand mechanisms of cellular transport (8). Membrane permeation of many small molecules and ions has been extensively studied using phospholipid vesicles (9).

The secondary structure of double-stranded (ds) DNA is changed significantly upon interaction with 1,2-dipalmitoyl-sn-glycero-3-phosphocholine (DPPC) liposomes (10). Calorimetric measurements have shown that the melting points of certain polynucleotides increase in the presence of phospholipid membranes (10). The process of polynucleotide adsorption is regarded as a surface phenomenon and has been reported not to change the overall morphology of the liposome, based on measurement of leakage of entrapped components from the vesicles (11,12). Infrared spectroscopy data suggest partial penetration of adenylic rings into the hydrophobic domain of lipid bilayers (13). Thus, both nucleotide bases and ionic phosphate groups could play important roles in DNA interactions with anionic or neutral lipids. Other studies (11,14,15) provide strong evidence that divalent cations mediate interactions between single-stranded (ss) ODNs or dsDNA and anionic membranes, presumably by formation of salt bridges between the phosphate residues of the polynucleotides and the phosphate groups in the phospholipids. Mg²⁺ and Ca²⁺ have been reported to stabilize complexes of polynucleotides with phosphatidylcholine or with mitochondrial lipids (11). These ionic interactions are reversible and can be dissociated by metal ion chelators such as EDTA or by competition with monovalent cations (Na⁺) (11). High resolution nuclear magnetic resonance studies of dsDNA [complexes of poly(dA·dU)] with colloidal DPPC or 1,2-dimyristoyl-sn-glycero-3-phosphocholine vesicles in the presence of Mg²⁺ also confirmed the presence of divalent cation-mediated salt bridges (12).

Cations are also known to affect the conformation of DNA (16). Tetraplex G-DNA helices, with Hoogsteen hydrogen bonding between guanine bases, are stabilized by divalent cations which can occupy an electronegative central core formed by guanine C6 carbonyl groups (17). This is most commonly observed for sequences which have four or more consecutive guanines, but has also been observed with G₃-containing sequences (18).

Cationic liposomes have emerged as promising non-viral vectors for the delivery of DNA molecules (19) and many studies have considered the interaction of DNA and cationic liposomes composed of sphingosine (20), 1,2-dioleoyl-3-trimethylammonium propane (21) or dimethyldioctadecylammonium bromide (22). Ionic interaction of DNA with cationic liposomes results in dehydration and condensation of the DNA, inducing a structural transition from native B-form to C-form DNA (23,24). C-form DNA may further collapse into a tertiary structure known as the chiral Ψ-DNA phase under extreme dehydration and high condensation (25). ODNs may also induce aggregation and fusion of cationic vesicles (26). The ratio of ODN to cationic lipid plays a crucial role in determining the physicochemical properties that govern their interaction.

Relatively little is known about interaction of ODNs and anionic lipids. Rhodes and Liu (27) demonstrated that the phase behavior of anionic 1,2-dipalmitoyl-sn-glycero-3-rac-phosphoglycerol (DPPG) monolayers at the air/water interface is affected by a Ca²⁺- or Mg²⁺-mediated interaction with ODNs. These effects were specific to the divalent cation and to the ODN sequence. We have also shown that changes in ODN conformation upon interaction with divalent cations were specific, depending on the divalent cation and the ODN sequence (18). In the present study we have used circular dichroism (CD) to investigate conformational changes of ODNs upon binding to DPPG liposomes. Our data suggest that the conformation of ODNs bound to anionic liposomes through divalent cation mediators is distinct from that of the ODN in buffer or complexed with divalent cations. This work is the first to use CD to study ion-mediated conformational changes in ODNs upon interaction with anionic liposomes.

MATERIALS AND METHODS

All salts and buffer components were purchased from Sigma (St Louis, MO). Lipids were purchased from Avanti Polar Lipids (Alabaster, AL) and were used without further purification. All water used was distilled and then deionized using a Barnstead Nanopure water filtering system. Phosphorothioate ODNs were generously donated by Isis Pharmaceuticals (Carlsbad, CA) and were used without further purification. Two sequences were used: 1, 5′-GTGGGCCATGATGATGGAAGG-3′; 2, 5′-GGGGTTGGGG-3′. HEPES buffer (10 mM, pH 7.4) was used throughout these experiments.

Spectra were collected using a Jasco model 715 CD spectrometer at 20°C, using a quartz cuvette from Starna Cells (Atascadero, CA) with a 10 mm path length in a thermostatted holder. CD spectra were collected from 350 to 200 nm in 0.1 nm increments. Three scans were accumulated and averaged to improve the signal-to-noise ratio. Stock aqueous solutions (0.2 mM) of phosphorothioate ODNs were stored at 4°C. Concentrations of the stock solution were verified by measuring absorbance at 260 nm at 20°C and estimating the molar extinction coefficient by base composition. Although concentrations for a given ODN were quite reproducible, there was some variability from one molecule to another due, apparently, to variation in water content in the dry samples.

For optical spectroscopy with liposomes it is important to minimize interference from light scattering, so sonicated vesicles were used in this work. Multilamellar vesicles of DPPG (1 mg/ml in buffer) were sonicated to clarity, yielding small unilamellar vesicles (SUVs). The average diameter of the SUVs was determined by light scattering and electron microscopy to be ∼100 nm.

After obtaining a blank CD spectrum with sonicated liposomes in buffer, an aliquot of ODN stock was added to the cuvette to yield a final concentration of 2.5 µM and actual concentrations were determined by measuring absorbance at 260 nm. Divalent cation (chloride) stocks were made gravimetrically in polycarbonate labware using ultrapure water. Divalent cations were added to ODN samples by removing a small aliquot of liposome–ODN suspension from the cuvette, replacing it with an equal volume of the divalent cation stock and mixing. CD data were collected for the cation-containing solution and the process continued to a final concentration of ∼5 mM divalent cation chloride for lipid-containing samples or ∼28 mM divalent cation chloride for other samples. All data were collected in duplicate or triplicate and representative scans are presented here. All measurements were corrected for any dilution due to addition of the divalent cation chloride solutions. An instrumental baseline correction was applied, if necessary, using data from a region of the spectrum which had negligible ellipticity. The mean ellipticity was calculated for data from 305 to 350 nm and this value used to calculate an offset correction, which was applied to the entire spectrum. Raw data (ellipticity, θ, in millidegrees) were corrected for molar concentration of the ODN (C, molar) and path length (l, cm) and were converted to molar ellipticity (Δɛ, dm³ mol^–1 cm^–1) using Δɛ = θ/32980·C·l.

RESULTS

The ODNs in solution showed distinct CD spectral characteristics. Although it does not have any G₄ sequences, 1 is a G-rich sequence. It has a guanine triplet near the 5′-end and a six-purine sequence at the 3′-end and five others in the central portion of the sequence. The CD spectrum of 1 in the absence of added divalent cation featured a broad positive band at 270 nm and a negative band at 240 nm (Fig. 1, trace a). The magnitude of Δɛ indicated that the ODN had a well-defined conformation in solution. This may represent a mixture of B-DNA and parallel G-DNA conformations. Addition of divalent cations caused distinct changes in the CD spectrum of 1 (see also 18). Addition of even small amounts of Ca²⁺ (250 µM) shifted the spectrum to a positive band at 262 nm and a negative band at 242 nm, indicative of parallel G-DNA tetraplex formation (Fig. 1, trace b). Further addition of Ca²⁺ (Fig. 1, traces c–f) slightly increased the molar ellipticity of the 262 nm band. The positions of the maxima of these bands suggest that the Ca²⁺ ions stabilized parallel G-tetraplex conformations. Addition of other divalent cations (Mg²⁺, Sr²⁺ or Ba²⁺) also appeared to result in formation of a parallel G-tetraplex, but the overall increases in Δɛ differed (spectra not shown). In particular, following the initial spectral shift to a 262 nm band, 27.8 mM Ca²⁺ increased the molar ellipticity of 1 by 23%, whereas similar concentrations of Sr²⁺ or Ba²⁺ resulted in increases of 44 and 34%, respectively.

Figure 1.

CD spectra of ODN 1 with CaCl₂ concentrations of 0 (a), 0.25 (b), 1.5 (c), 3.98 (d), 12.15 (e) and 27.8 mM (f).

The CD spectrum of 2 was similar to previously reported spectra from G-tetraplex ODNs (18). As seen in Figure 2 (trace a), 2 in buffer appears to have negative bands at 237 and 277 nm, a positive band at 255 nm and a small positive band at 295 nm. Addition of divalent cations such as Ca²⁺ led to a change in the CD spectrum (Fig. 2, traces b–f) which indicated stabilization of parallel G-DNA, as reported previously (18). The bands at 277 and 295 nm were diminished and those at 237 and 255 nm were enhanced. In the final spectrum ([Ca²⁺] = 27.8 mM) the major positive band was red-shifted to 260 nm and the negative band to 238 nm. The increase in Δɛ for the 260 nm band was 51%. Other divalent cations, Mg²⁺, Sr²⁺ and Ba²⁺, produced similar transitions in the spectra with increases in Δɛ of the major positive band of 45, 115 and 115%, respectively (data not shown).

Figure 2.

CD spectra of ODN 2. The curves represent CaCl₂ concentrations of 0 (a), 0.25 (b), 1.50 (c), 3.98 (d), 12.15 (e) and 27.80 mM (f).

Glycerophospholipids have at least one chiral center and thus may be expected to have a CD signal. However, Figure 3 (trace a) shows that the sonicated SUVs of DPPG do not have any significant ellipticity in the wavelength range used in this study. An apparent positive ellipticity was observed at wavelengths <240 nm, but it is believed that this was an artifact due to light scattering. The phase behavior of DPPG liposomes is known to depend on the concentration of divalent cation (27), but the ellipticity of the sonicated vesicles did not change upon addition of divalent cations. The apparent light scattering artifact is also observed in mixtures of DPPG SUVs and 2 (Fig. 3, trace b), but only at short wavelengths. At λ > 245 nm the spectrum from this mixture is quite similar to that of 2 alone (Fig. 3, trace c). The CD spectrum of a mixture of DPPG SUVs and 2 (Fig. 3, trace b) is nearly identical to a simple sum of the CD spectra of the individual components. Previous measurements of the phase behavior of lipids in monolayers indicated that DPPG and ODNs do not interact in the absence of divalent cations (27). Figure 3 shows that DPPG liposomes induced no significant change in the conformation of 2. Similar measurements were carried out with DPPG:DNA ratios (mol lipid:mol DNA base) as high as 40:1, but no appreciable conformational change was detected.

Figure 3.

CD spectra of DPPG SUVs (a), a mixture of DPPG SUVs and ODN 2 (b) and ODN 2 alone (c).

The CD spectrum of 1 in the presence of DPPG SUVs (Fig. 4, trace a) was not significantly different from that of the ODN alone, with a positive band at 270 nm and a negative band at 240 nm. Addition of small amounts of Ca²⁺ (≤0.25 mM) did not alter the CD spectrum (Fig. 4, traces b and c). Upon further increasing the Ca²⁺ concentration to 0.5 mM, the positive band shifted to 275 nm and the negative band to 244 nm (Fig. 4, trace d). As the Ca²⁺ concentration was increased to 1.8 mM the spectrum gradually shifted such that the positive band was at 270 nm and the negative band was at 242 nm (Fig. 4, traces e–g). At 2.7 mM Ca²⁺ precipitation and light scattering made interpretation of the CD spectrum more difficult (data not shown).

Figure 4.

CD spectra of ODN 1 with DPPG SUVs and added CaCl₂. CaCl₂ concentrations are 0 (a), 0.08 (b), 0.25 (c), 0.5 (d), 0.75 (e) 1.0 (f) and 1.83 mM (g).

Addition of up to 0.25 mM Sr²⁺ to a solution of 1 in the presence of DPPG SUVs had relatively little effect on the CD spectrum (Fig. 5, traces a–c). At 0.5 mM Sr²⁺ the band at 241 nm became more negative and shifted to 244 nm and the broad band at 270 nm sharpened and became more positive (Fig. 5, trace d). Increasing the Sr²⁺ concentration to 1.25 mM simply increased the magnitude of the ellipticity of these bands, albeit not proportionately (Fig. 5, traces e–f). However, at [Sr²⁺] = 1.74 mM the spectrum suddenly shifted to a strongly negative ellipticity dominated by major negative bands at 240 and 290 nm with Δɛ values of –435 and –240 dm³ mol^–1 cm^–1, respectively (Fig. 5, trace g). This CD spectrum is characteristic of Ψ-DNA. At higher concentrations of Sr²⁺ precipitation made accurate measurements impossible due to scattering.

Figure 5.

CD spectra of ODN 1 with DPPG SUVs and added SrCl₂. SrCl₂ concentrations are 0 (a), 0.08 (b), 0.25 (c), 0.5 (d), 0.83 (e) 1.25 (f) and 1.74 mM (g).

Addition of small amounts of Ca²⁺ to a solution of 2 in a suspension of sonicated DPPG SUVs did not produce a significant change in the CD spectrum (Fig. 6, traces a–c). However, with 0.5 mM Ca²⁺ the CD spectrum dramatically changed from a small negative band at 237 nm, a positive band at 255 nm and a negative band at 277 nm to a large negative band at 246 nm and a positive band at 279 nm (Fig. 4, trace d). Further increasing the Ca²⁺ concentration (to 1.75 mM) enhanced these features (Fig. 6, trace e), but Ca²⁺ concentrations >4 mM led to precipitation and the CD spectrum was not considered reliable. Other divalent cations, such as Mg²⁺, produced similar transitions, but the concentration at which the effects were observed differed. For example, addition of Mg²⁺ resulted in small, gradual changes in the spectrum up to ∼0.5 mM Mg²⁺. At 1.75 mM Mg²⁺ significant shifts in the spectrum were observed, similar to those observed with 0.5 mM Ca²⁺. Sr²⁺ and Ba²⁺ induced aggregation and precipitation at concentrations of ≤4 mM, but only minor spectral shifts were observed at lower divalent cation concentrations.

Figure 6.

CD spectra of ODN 2 with DPPG SUVs and added CaCl₂. CaCl₂ concentrations are 0 (a), 0.08 (b), 0.25 (c), 0.5 (d) and 1.75 mM (e).

DISCUSSION

Adsorption of dsDNA to the surface of cationic liposomes typically results in the conversion of B-DNA into C-DNA (25), characterized by CD spectra with a strong negative band at 240 nm and a small positive band at 270 nm. The structure of C-DNA is only slightly different from that of B-DNA. (The vertical rise per base pair is 1.8% smaller and the tilt of the base pairs is –8° instead of –6°.) The structure of ssODNs is thought to be less well defined because of the lack of base pairing, especially in the case of short ODNs. Thus, association of ODNs with liposomes would be expected to have a significant impact on the ODN conformation. However, it is uncertain, as the lipid bilayer and ODN interact, to what extent the structures of each will be affected. Moreover, due to the unique constraints on a liposome-associated ODN compared to those on an ODN in free solution, it is not known whether the conformation of an adsorbed ODN should be a conformation found in any solution condition.

Due to the presence of some sequential guanines and its purine-rich sequence, 1 can form a parallel G-DNA tetraplex structure. However, some of the sequence in this molecule probably has a secondary structure other than a parallel G-DNA tetraplex (e.g. B-DNA), since the overall CD spectrum apparently consisted of contributions from each of these components. Since the major positive bands in the CD spectra of B-DNA (260 nm) and parallel G-DNA tetraplex (275 nm) are separated by only 15 nm, the two components may not be well resolved, resulting in the broad band at 270 nm. However, addition of Ca²⁺ or other divalent cations resulted in a CD spectrum characteristic of a parallel G-tetraplex.

In the presence of DPPG SUVs and Ca²⁺ the CD spectrum of 1 was characterized by a negative band at 240 nm and a positive band at 270 nm (Fig. 4, traces a–c). This spectrum from the ODN is relatively unchanged for Ca²⁺ concentrations up to 0.25 mM, perhaps because ion binding by the liposome anions decreases the effective free concentration of Ca²⁺. Although the positive band is somewhat similar to that expected for C-DNA, the negative band is not at 240 nm, as would be expected. It is possible that this is a distorted conformation with some similarity to C-DNA.

The sequence of 2 is G₄T₂G₄ and, due to the presence of sequential guanines, this ODN can easily assemble into parallel G-DNA tetraplexes stabilized by Ca²⁺ or Sr²⁺. The presence and stabilization of these structures was confirmed by their characteristic CD spectra (Fig. 2). In the presence of DPPG liposomes, however, Ca²⁺ effected a complete change of the conformation (Fig. 6). At [Ca²⁺] > 0.5 mM the CD spectrum featured a negative band at 246 nm and a positive band at 279 nm. The 246 nm band is close to that expected for C-DNA (240 nm), but the 279 nm band is clearly not characteristic of C-DNA. The closest correspondence is to bands for B-DNA. It is possible that the CD spectrum for a 2–Ca²⁺–DPPG complex is a B-DNA-like conformation, but other conformations are possible. Any of these proposed conformations would require that the ionic interactions between the ODN phosphate, the divalent cation and the lipid phosphate can overcome the affinity of the carbonyl-rich electronegative core of the G-tetraplex at divalent cation concentrations >0.5 mM.

It is also important to emphasize that the effects described here are cation-specific. As shown in Figures 4 and 5, Ca²⁺ and Sr²⁺, ions which have very similar chemical and biochemical characteristics, have very different effects when used to mediate the interaction between ODNs and anionic lipids. In this case, ODNs associated with liposomes through Ca²⁺ were in the novel ODN conformation described in this work, while ODNs associated through Sr²⁺ were in the previously described Ψ-DNA conformation.

The conformations observed for these ODNs adsorbed to liposomes are apparently unique. Based on the CD spectra, these are conformations distinct from: (i) conformations in free solution (either with or without added divalent cation); (ii) conformations of liposome-associated dsDNA (25); (iii) even conformations of ODNs associated with cationic liposomes (unpublished results). Thus far, efforts to simulate these spectra through combinations of ‘model’ spectra (CD spectra typical of specific, known conformations) have not indicated that the CD spectra observed here result from a mixture of previously known conformations. These conformations are cation-specific and sequence-specific and we speculate that they may be related to the divalent cation dependence of ODN uptake in cell culture.

CONCLUSIONS

These data show that divalent cations can mediate specific interactions between anionic membranes and ODNs. This interaction can lead to a significant change in the ODN conformation and the structure of the adsorbed ODN may be a novel secondary structure. This conformation may be related to divalent cation-mediated internalization of ODNs.