An Inverted Hexagonal Phase of Cationic Liposome-DNA Complexes Related to DNA Release and Delivery

See allHide authors and affiliations

Science  03 Jul 1998:
Vol. 281, Issue 5373, pp. 78-81
DOI: 10.1126/science.281.5373.78


A two-dimensional columnar phase in mixtures of DNA complexed with cationic liposomes has been found in the lipid composition regime known to be significantly more efficient at transfecting mammalian cells in culture compared to the lamellar (Lα C) structure of cationic liposome–DNA complexes. The structure, derived from synchrotron x-ray diffraction, consists of DNA coated by cationic lipid monolayers and arranged on a two-dimensional hexagonal lattice (HII C). Two membrane-altering pathways induce the Lα C → HII C transition: one where the spontaneous curvature of the lipid monolayer is driven negative, and another where the membrane bending rigidity is lowered with a new class of helper-lipids. Optical microscopy revealed that the Lα C complexes bind stably to anionic vesicles (models of cellular membranes), whereas the more transfectant HII C complexes are unstable and rapidly fuse and release DNA upon adhering to anionic vesicles.

Complexes consisting of DNA mixed with oppositely charged cationic liposomes (CLs; closed bilayer membrane shells of lipid molecules) mimic natural viruses in their ability to act as synthetic carriers of extracellular DNA across outer cell membranes and nuclear membranes for gene delivery (1–3). The use of nonviral rather than viral methods for gene delivery has several advantages, including nonimmunicity and the potential for transferring and expressing (transfecting) large pieces of DNA into cells. Partial sections of first-generation human artificial chromosomes (HACs) on the order of 1 Mbp can be transferred into cells by means of CLs, although extremely inefficiently (4). The low transfection efficiencies of nonviral delivery methods may be improved through insights into transfection-related mechanisms at the molecular and self-assembled levels.

The efficiency of transfection mediated by mixtures of cationic lipids and so-called neutral “helper-lipids” varies widely and unpredictably (1, 3, 5). The choice of the helper-lipid has been empirically established to be important; for example, transfection of mammalian cells in culture is efficient in mixtures of the univalent cationic lipid DOTAP (dioleoyl trimethylammonium propane) and the neutral helper-lipid DOPE (dioleoyl phosphatidylethanolamine), and not in mixtures of DOTAP and a similar helper-lipid, DOPC (dioleoyl phosphatidylcholine) (6). We recently showed that mixing DNA with CLs consisting of DOPC and DOTAP leads to a topological transition into condensed CL-DNA complexes with a multilamellar structure (Lα C), with DNA monolayers sandwiched between cationic lipid bilayers (7) in a manner similar to the schematic in Fig. 1 (left).

Figure 1

Schematic of two distinct pathways from the lamellar Lα C phase to the columnar inverted hexagonal HII C phase of CL-DNA complexes. Along pathway I, the natural curvature C o = 1/R o of the cationic lipid monolayer is driven negative by the addition of the helper-lipid DOPE. This is shown schematically (center top); the cationic lipid DOTAP is cylindrically shaped whereas DOPE is conelike, leading to the negative curvature. Along pathway II, the Lα C → HII C transition is induced by the addition of helper-lipids consisting of mixtures of DOPC and the cosurfactant hexanol, which reduces the membrane bending rigidity.

Using synchrotron small-angle x-ray scattering (SAXS) and optical microscopy, we found a completely different columnar inverted hexagonal HII C liquid-crystalline state in CL-DNA complexes (Fig. 1, right). The commonly used helper-lipid DOPE induces the Lα C → HII C structural transition by controlling the spontaneous curvature C o = 1/R o of the lipid monolayer, whereR o is the natural radius of curvature (Fig. 1, pathway I). We also identified a class of helper molecules that control the membrane bending rigidity κ and give rise to a second pathway to the HII C phase (Fig. 1, pathway II). The CL-DNA complexes containing DOPE that are empirically known to transfect exhibit the HII C phase, rather than the Lα C structure found in complexes containing DOPC. Optical imaging showed that complex interactions with model cell membranes mimicking the early stages of transfection are structure-dependent.

Synchrotron SAXS scans of positively charged CL-DNA complexes for ρ = 3 are shown in Fig. 2A as a function of increasing ΦDOPE in the DOPE-DOTAP CL mixtures along pathway I (8). [Here and below, ρ denotes the DOTAP/DNA weight ratio, ΦDOPE is the weight fraction DOPE/(DOPE + DOTAP), and ΦDOPC is the weight fraction DOPC/(DOPC + DOTAP).] The complexes are charged positively for ρ > 2.2 and negatively for ρ < 2.2, which indicates that charge reversal occurs when complexes are stoichiometrically neutral (with one positive lipid per negatively charged nucleotide base). At ΦDOPE = 0.41, SAXS scans of the lamellar Lα C complex (Fig. 2A) show sharp peaks at q 001 = 0.099 Å−1 and q 002 = 0.198 Å−1 from the lamellar periodic structure (d = 2π/q 001 = 63.47 Å) with DNA intercalated between cationic lipid bilayers (Fig. 1, left). Because the DOPE-DOTAP bilayer thickness at ΦDOPE = 0.41 is δm = 40 Å (9), the water gap between bilayers δw = d – δm = 23.47 Å is just large enough to accommodate a hydrated monolayer of DNA. The middle broad peak at q DNA = 0.172 Å−1 is due to the one-dimensional (1D) array of DNA chains, with the spacing between the DNA strandsd DNA = 2π/q DNA. This structure, found in CL-DNA complexes with ΦDOPE < 0.41, is analogous to that reported recently (7, 10).

Figure 2

Synchrotron SAXS patterns of the lamellar Lα C and columnar inverted hexagonal HII C phases of positively charged CL-DNA complexes. (A) SAXS scans of CL-DNA complexes as a function of increasing ΦDOPE along pathway I of Fig. 1. At ΦDOPE = 0.41, the SAXS pattern results from a single phase with the Lα C structure shown in Fig. 1 (left). At ΦDOPE = 0.75, the SAXS scan results from a single phase with the HII C structure shown in Fig. 1 (right). At ΦDOPE = 0.65, the SAXS shows coexistence of the Lα C (dotted line) and HII C phases. At ΦDOPE = 0.87, SAXS shows coexistence of the HII C phase and the inverted hexagonal HIIphase of pure DOPE (arrows). SAXS patterns of complexes made from extremely dilute DNA (0.01 mg/ml) and lipid (0.1 mg/ml) solutions are plotted as solid lines for ΦDOPE = 0.41 and 0.75. (B) SAXS scans of CL-DNA at a constant ΦDOPC with no hexanol (a cosurfactant) and at a hexanol/total lipid molar ratio of 3:1 along pathway II of Fig. 1. With no hexanol (filled squares), the structure is lamellar Lα C, whereas the complexes with hexanol (open squares) exhibit the hexagonal HII C structure. (C) SAXS scans of CL-DNA complexes with ΦDOPC= 0. The complexes remain in the Lα C phase with and without added hexanol.

For 0.7 < ΦDOPE < 0.85, the peaks of the SAXS scans of the CL-DNA complexes are indexed perfectly on a 2D hexagonal lattice with a unit cell spacing of a = 4π/[(3)0.5 q 10] = 67.4 Å for ΦDOPE = 0.75. We observed Bragg peaks up to the seventh order, which indicates a high degree of regularity of the structure. At ΦDOPE = 0.75, the first- through fourth-order Bragg peaks of this hexagonal structure occur at q 10 = 0.107 Å–1, q 11 = 0.185 Å–1, q 20 = 0.214 Å–1, and q 21 = 0.283 Å–1 (Fig. 2A). This is consistent with a 2D columnar inverted hexagonal structure (Fig. 1, right), which we refer to as the HII C phase of CL-DNA complexes. The DNA molecules are surrounded by a lipid monolayer, with the DNA-lipid inverted cylindrical micelles arranged on a hexagonal lattice. The structure resembles that of the inverted hexagonal HII phase of pure DOPE in excess water (11), with the water space inside the lipid micelle filled by DNA. The greater electron density of DNA with respect to water leads to the relative suppression of the (11) and (20) Bragg peak intensities compared with that in the lipid HIIphase (9). If we assume an average lipid monolayer thickness of 20 Å, the diameter of the micellar void in the HII C phase is near 28 Å, again sufficient for a DNA molecule with approximately two hydration shells. For 0.41 < ΦDOPE < 0.7, the Lα C and HII C structures coexist as shown at ΦDOPE= 0.65 and are nearly epitaxially matched with ad. For ΦDOPE > 0.85, the HII Cphase coexists with the HII phase of pure DOPE, which has peaks at q 10 = 0.0975 Å–1,q 11 = 0.169 Å–1, andq 20 = 0.195 Å–1 (arrows in Fig. 2A at ΦDOPE = 0.87) with a unit cell spacing ofa = 74.41 Å.

SAXS scans of CL-DNA complexes at the concentration (0.01%) typically used in cell transfection studies (6) are also plotted in Fig. 2A (solid lines at ΦDOPE = 0.41 and 0.75). The complexes have their first-order Bragg peaks at exactly the same positions as in the corresponding more concentrated (1%) samples. Thus, the internal structures of the complexes are independent of the overall DNA and lipid concentrations.

The Lα C → HII C phase transition can be induced along a second pathway (Fig. 1, pathway II) with a new type of helper-lipid mixture. Complexes containing mixtures of DOPC and DOTAP exhibited the lamellar Lα Cstructure (7), as shown by the SAXS scan in Fig. 2B (bottom; ΦDOPC = 0.7) with an interlayer spacing ofd = 2π/q 001 = 66.84 Å. As a function of increasing ratio of hexanol (a membrane-soluble cosurfactant) to DOPC, we found a structural transition to the HII C phase. The first four SAXS peaks for complexes containing DOPC, DOTAP, and hexanol (ΦDOPC = 0.7; hexanol/total lipid molar ratio, 3:1) can be indexed on a hexagonal lattice with a unit cell size a = 62.54 Å. In CL-DNA complexes of pure DOTAP (Fig. 2C), hexanol addition did not induce the transition and we always found the Lα Cstructure. In this case, the only effect of the addition of hexanol is to thin the cationic bilayer membrane (consisting of hexanol:DOTAP at a 3:1 molar ratio) from d = 57.91 Å to 54.17 Å. The interaxial DNA-DNA spacing d DNA was also observed to increase from 27.1 to 28.82 Å, consistent with a decrease in the membrane charge density with the addition of hexanol.

To understand the Lα C → HII Ctransition qualitatively along the two pathways in Fig. 1, we consider the interplay between the electrostatic and membrane elastic interactions in the complexes. Pure electrostatic interactions alone are expected to favor the HII C phase, which minimizes the charge separation between the anionic groups on the DNA chain and the cationic lipids (1, 12). The electrostatic interaction may be resisted by the Helfrich elastic cost (per unit area) of forming a cylindrical monolayer membrane around DNAEmbedded Image(1)where κ is the lipid monolayer rigidity, R is the radius of curvature, andR o is the natural radius of curvature. Along pathway I (Fig. 1), cationic DOTAP with 1/R o DOTAP = 0 favors the flat lamellar Lα phase, but DOPE has a negative natural curvature 1/R o DOPE < 0; that is, DOPE has a larger area per two chains than area per head group (Fig. 1, center top) and forms the inverted hexagonal HII phase (11). Thus, along pathway I, as a function of increasing ΦDOPE (with 1/R o = ΦDOPE V/R o DOPE, where ΦDOPE V is the volume fraction of DOPE), we expect a transition to the HII C phase because the membrane elastic energy favors a curved interface (Fig. 2A).

Along pathway II, the membrane bending rigidity κ is reduced significantly because of the addition of the membrane-soluble cosurfactant molecule hexanol. Cosurfactant molecules cannot stabilize an interface separating hydrophobic and hydrophilic regions, but when mixed with longer chain “true” surfactants they can lead to large changes in interface elasticities. The addition of hexanol to membranes of lamellar phases with a molar ratio of 2 to 4 will lead to a decrease of κ from ∼20 k B T(where k B T is the thermal energy) to between 2 and 5 k B T (13). Simple compressional models of surfactant chains show that κ scales with chain length ln (∝ δm, membrane thickness; n = number of carbons per chain) and the area per lipid chain A L as κ ∝l n 3/A L 5(14), and hexanol both decreases ln and increases A L (13, 14) (Fig. 1, center bottom). The addition of cosurfactant results in a reduction in κ, and hence a reduction in the elastic energy barrier to the formation of the HII C phase favored by the electrostatic interactions. The transition to the HII Cphase along pathway II occurs only in CL-DNA complexes with low enough charge density, where DOTAP/DOPC < 0.5 (9). In this regime where the Lα C structure is retained in complexes with pure DOTAP with and without added hexanol (Fig. 2C), the SAXS data are consistent with theory, which predicts a renormalized increase in κ with increasing surface charge density (15).

In the absence of DNA, mixtures of DOPC and DOTAP studied in this work with or without hexanol formed stable lamellar Lα phases (with 1/R o = 0) with no tendency to form the inverted HII phase (9). However, DOPE-DOTAP-water mixtures formed coexisting HII and Lα phases.

In both condensed phases, the complexes appear as highly dynamic birefringent aggregates when viewed with video-enhanced optical microscopy in differential interference contrast (DIC) and fluorescence configurations, as shown in Fig. 3A for HII CDOPE = 0.73) and Fig. 3B for Lα CDOPE = 0.3) complexes along pathway I (16). The positive complexes (with ρ = 3) formed aggregates of connected blobs; with increasing complex charge, the aggregates became smaller and eventually dissociated into individual blobs. The Lα C phase formed linear structures, whereas the HII C phase formed predominantly branched aggregates, which indicated that the shape of HII Ccomplexes is inherently anisotropic (9). The observed overlap of lipid and DNA distributions in the two fluorescence modes (Fig. 3, A and B) shows that the complexes are highly compact objects with a close association of lipid and DNA.

Figure 3

Video microscopy images of positively charged CL-DNA complexes in the (A) HII C and (B) Lα C phases, and interacting with negatively charged G-vesicles (C and D). In all cases, complexes were viewed by DIC (left), lipid fluorescence (middle), and DNA fluorescence (right). In (C) the Lα C complexes simply stick to the G-vesicle and remain stable for many hours, retaining their bloblike morphology. The blobs are localized in DIC as well as in both fluorescence modes. In (D), the HII C complexes break up and spread immediately after attaching to G-vesicles, indicating a fusion process between the complex and the vesicle lipid bilayer. The loss of the compact structure of the complex is evident in both fluorescence modes. Scale bar for DIC images: 3 μm (A and B), 20 μm (C and D). Scale bar for fluorescence images: 6 μm (A and B)], 20 μm (C and D).

To understand the effect of structure on the early stages of transfection, we studied the interaction of CL-DNA complexes with giant anionic vesicles (G-vesicles), which are models of cellular membranes (for example, plasma membrane; anionic endosomal vesicles). The main entry route to mammalian cells is believed to be endocytosis, where a local inward deformation of the cell plasma membrane leads to budding off of an internal vesicle forming the early-stage endosome (17). Thus, at the early stages of cell transfection, an intact CL-DNA complex may be captured inside an anionic endosomal vesicle.

The positively charged HII C and Lα C complexes interacted very differently with model anionic lipid membranes, even when both types of structures contained DOPE. Typical micrographs of positively charged (ρ = 4) complexes immediately after mixing with G-vesicles are shown in Fig. 3, C and D. Lα C complexes attached to the fluid membranes of the G-vesicles and remained stable (Fig. 3C); no fusion occurred between the complex and the G-vesicle. Lα Ccomplexes containing DOPC (7) showed the same behavior. HII C complexes attached to the G-vesicle and rapidly fused with it, spreading and losing their compact structure (Fig. 3D, left). Because the amount of lipid in the complex was comparable with that in the G-vesicle and the fusion occurred rapidly, this produced multiple free lipid lamellae. The loss of the compact complex structure and the subsequent desorption of DNA molecules from membrane and their Brownian motion between the lamellae were seen in fluorescence images (Fig. 3D, center and left). This behavior is expected after fusion, which results in the mixing of cationic lipid (from the HII C complex) with anionic lipid (from the G-vesicle), effectively “turning off” the electrostatic interactions (which gave rise to the compact CL-DNA complexes) and releasing DNA molecules inside the space between the lamellae and the G-vesicle bilayer. Because the geometry is the inverse of that of CL-DNA complexes inside anionic endosomal vesicles, we expect that upon fusion DNA should be released and expelled outside the endosome within the cytoplasm (18). On longer time scales (a few hours), we observed lipid transfer between the Lα C complexes and G-vesicles (9). Thus, the observation that DOPE-DOTAP Lα C and DOPC-DOTAP Lα C complexes do not fuse with G-vesicles reveals a kinetic rather than equilibrium effect. In principle, it may be possible to design Lα C complexes with a lower kinetic barrier to fusion. Moreover, the reported behavior of complexes containing univalent cationic lipids may be different from that of multivalent cationic lipids.

Our findings correlate the self-assembled structure of CL-DNA complexes and transfection efficiency: The empirically established transfectant complexes containing DOPE in mammalian cell cultures exhibit the HII C structure rather than the Lα C structure found in complexes containing DOPC (7). We have also found the HII C phase in two other negatively charged polyelectrolytes: polyglutamic acid (PGA), a model polypeptide, and polythymine (poly-T), a model of single-stranded oligonucleotides used in antisense delivery applications (19, 20). Optical microscopy reveals a likely reason for why the different structures transfect cells with varying efficiency: In contrast to Lα C complexes, HII C complexes fused and released DNA when in contact with anionic vesicles, which are cell-free models of cellular membranes, in particular, anionic endosomal vesicles.

  • * Present address: Sektion Physik der Ludwig-Maximilians-Universität München, Geschwister-Scholl-Platz 1, 80539 München, Germany.

  • Present address: Physikdepartment, Technische Universität München, Institut für Biophysik (E22), James Franck-Strasse, 85747 Garching, Germany.


View Abstract

Navigate This Article