Structure of DNA-Cationic Liposome Complexes: DNA Intercalation in Multilamellar Membranes in Distinct Interhelical Packing Regimes

See allHide authors and affiliations

Science  07 Feb 1997:
Vol. 275, Issue 5301, pp. 810-814
DOI: 10.1126/science.275.5301.810


Cationic liposomes complexed with DNA (CL-DNA) are promising synthetically based nonviral carriers of DNA vectors for gene therapy. The solution structure of CL-DNA complexes was probed on length scales from subnanometer to micrometer by synchrotron x-ray diffraction and optical microscopy. The addition of either linear λ-phage or plasmid DNA to CLs resulted in an unexpected topological transition from liposomes to optically birefringent liquid-crystalline condensed globules. X-ray diffraction of the globules revealed a novel multilamellar structure with alternating lipid bilayer and DNA monolayers. The λ-DNA chains form a one-dimensional lattice with distinct interhelical packing regimes. Remarkably, in the isoelectric point regime, the λ-DNA interaxial spacing expands between 24.5 and 57.1 angstroms upon lipid dilution and is indicative of a long-range electrostatic-induced repulsion that is possibly enhanced by chain undulations.

Somatic gene therapy depends on the successful transfer and expression of extracellular DNA to the nucleus of eucaryotic cells, with the aim of replacing a defective or adding a missing gene (1). Viral-based carriers of DNA are presently the most common method of gene delivery, but there has been a tremendous activity in developing synthetic nonviral vectors (2). In particular, cationic liposomes (CLs) (3), in which the overall positive charge of the cationic liposome-DNA (CL-DNA) complex enhances transfection by attaching to anionic animal cells, have shown gene expression in vivo in targeted organs (4), and human clinical protocols are ongoing (5). Cationic liposome transfer vectors exhibit low toxicity, nonimmunogenicity, and ease of production, but their mechanism of action remains largely unknown; transfection efficiencies vary by up to a factor of 100 in different cell lines (26).

This unpredictability, which is ubiquitous in gene therapy (7) and in particular in synthetic systems, may be attributed to a lack of knowledge regarding the interactions between DNA and CLs and the resulting structures of CL-DNA complexes. DNA-membrane interactions might also provide clues for the relevant molecular forces in the packing of DNA in chromosomes and viral capsids. Studies show regular DNA condensed morphologies induced by multivalent cations (8) and liquid-crystalline (LC) phases at high concentrations of DNA both in vitro (9) and in vivo in bacteria (10). More broadly, the nature of structures and interactions between membranes and polymers, either adsorbed (11) or tethered to the membranes (12), is currently an active area of research.

Felgner et al. (2, 3) originally proposed a “bead-on-string” structure of the CL-DNA complexes and pictured the DNA strand decorated with distinctly attached liposomes. Electron microscopy (EM) studies have reported on a variety of structures including stringlike structures and indications of fusion of liposomes in metal-shadowing EM (13), oligolamellar structures in cryo-transmission EM (14), and tubelike images possibly depicting lipid bilayer-covered DNA observed in freeze-fracture EM (15).

We have carried out a combined in situ optical microscopy and x-ray diffraction (XRD) study of CL-DNA complexes. On semi-macroscopic length scales, the addition of linear or circular plasmid DNA to binary mixtures of CLs induces a topological transition from liposomes into collapsed condensates in the form of optically birefringent LC globules with sizes on the order of 1 μm. The solution structure of the globules was revealed on the 1- to 100-nm length scale by high-resolution synchrotron XRD studies. Unexpectedly, the complexes consist of a higher ordered multilamellar structure with DNA sandwiched between cationic bilayers.

We have discovered distinct interhelical packing regimes for linear λ-phage DNA above, below, and at the isoelectric point of the complex by varying the concentrations of DNA and the lipid components comprising the complex. Remarkably, in the isoelectric regime of the CL-DNA complex, the DNA interaxial distance dDNA increases from 24.5 to 57.1 Å as a function of lipid dilution and is quantitatively consistent with an expanding one-dimensional (1D) lattice of DNA chains. Thus, the DNA chains confined between bilayers form a novel 2D smectic phase.

DNA molecules can be readily labeled and imaged by fluorescence microscopy (16). Free λ-DNA in aqueous solution appears as a highly dynamic blob of ∼1 μm in diameter, in agreement with a classical random coil configuration, whereas the contour length of λ-phage DNA is 16.5 μm. The CLs consisted of binary mixtures of lipids that contained either DOPC (dioleoyl-phosphatidylcholine) or DOPE (dioleoyl-phosphatidylethanolamine) as the neutral co-lipid and DOTAP (dioleoyl trimethylammonium propane) as the CL (17). The DOTAP/DOPC and DOTAP/DOPE CLs had a size distribution ranging between 0.02 to 0.1 μm in diameter, with a peak around 0.07 μm (18). We used highly purified linear λ-phage DNA (48,502 bp) in most of the experiments but some were carried out with Escherichia coli DNA and pBR322 plasmid DNA (4361 bp); the latter consisted of a mixture of nicked circular and supercoiled DNA (19). Condensation of CLs with λ-DNA was directly observed by using differential interference microscopy (DIC) and fluorescence microscopy (20).

We show in Fig. 1A a series of DIC images 30 min after preparation in CL-DNA mixtures as a function of the total lipid to λ-DNA weight ratio L/D, where L = DOTAP + DOPE denotes the weight of lipid and D is the weight of DNA. Similar images were observed with λ-DNA replaced by the pBR322 plasmid DNA or E. coli DNA, or DOPE replaced by DOPC. At low DNA concentrations (Fig. 1A, L/D = 50), in contrast to the pure liposome solution where no objects >0.2 μm were found, 1-μm globules were observed. The globules coexisted with excess liposomes. As more DNA was added, the globular condensates formed larger chainlike structures (Fig. 1A, L/D = 10). The Brownian motion of these globules suggests their linkage by an invisible thread. At L/D ≈ 5 the chainlike structures flocculated into large aggregates of distinct globules. For L/D <5, the complex size was smaller and stable in time again (Fig. 1A, L/D = 2), and coexisted with excess DNA. Fluorescence-labeled DNA and lipid could be detected on each globule, indicating that the globules are DNA-lipid condensates (21). Polarized microscopy also showed that the distinct globules were birefringent, indicative of their LC nature.

Fig. 1.

(A) High-resolution DIC images of CL-DNA complexes forming distinct condensed globules in mixtures of different lipid to DNA weight ratio (L/D) [see text and (17, 19); scale bar is 10 μm]. (B) Average size of the lipid-DNA complexes measured by dynamic light scattering (18).

The size dependence of the complexes as a function of L/D (Fig. 1B) was independently measured by dynamic light scattering (18). The large error bars represent the broad polydispersity of the system. The size dependence of the aggregates can be understood in terms of a charge-stabilized colloidal suspension. The charge of the complexes was measured by their electrophoretic mobility in an external electric field (22). For L/D >5 (Fig. 1A; L/D of 50 or 10) the complexes are positively charged, while for L/D <5 (Fig. 1A; L/D of 2) the complexes are negatively charged. The charge reversal is in good agreement with the stoichiometrically expected charge balance of the components DOTAP and DNA at L/D ≈ 4.4, where L = DOTAP + DOPE in equal weights. Thus, the positively and negatively charged globules at L/Ds of 50 and 2, respectively, repelled each other and remained separate while, as L/D approached 5, the nearly neutral complexes collided and tended to stick due to van der Waals attraction. Remarkably, the size of the globules appears to be only weakly dependent on the length of the DNA in similar experiments carried out with E. coli DNA or pBR322 plasmid (4361 bp) (22).

The XRD experiments (23) revealed unexpected structures for mixtures of CLs and DNA. Small-angle x-ray scattering (SAXS) data of dilute (Φw = the volume fraction of water = 98.6 ± 0.3%) DOPC/DOTAP (1→1)-λ-DNA mixtures as a function of L/D (L = DOPC + DOTAP) (Fig. 2A) are consistent with a complete topological rearrangement of liposomes and DNA into a multilayer structure with DNA intercalated between the bilayers (24) (Fig. 3A). Two sharp peaks at q = 0.0965 ± 0.003 and 0.193 ± 0.006 Å−1 correspond to the (001) peaks of a layered structure with an interlayer spacing d (= δm + δw), which is in the range 65.1 ± 2 Å (Fig. 2B, open squares). The membrane thickness and water gap are denoted by δm and δw, respectively (Fig. 3A). The middle broad peak qDNA arises from DNA-DNA correlations and gives dDNA = 2π/qDNA (Fig. 2B, solid circles). The multilamellar structure with intercalated DNA is also observed in CL-DNA complexes containing supercoiled DNA both in water and in Dulbecco's modified Eagle's medium used in transfection experiments in gene therapy applications (25). This novel multilamellar structure of the CL-DNA complexes protected DNA from being cut by restriction enzymes (26).

Fig. 2.

(A) A series of SAXS scans of CL-DNA complexes in excess water as a function of different lipid to DNA weight ratio (L/D). The Bragg reflections at q001 = 0.096 Å−1 and q002 = 0.1.92 Å−1 result from the multilamellar Lα structure with intercalated monolayer DNA (see Fig. 3A). The intermediate peak at qDNA is due to the DNA-interaxial spacing dDNA as described in the text. (Inset) SAXS scan of an extremely dilute (lipid + DNA = 0.014% volume in water) λ-DNA-DOPE/DOTAP (1→1) complex at L/D = 10, which shows the same features as the more concentrated mixtures and confirms the multilamellar structure (with alternating lipid bilayer and DNA monolayers) of very dilute mixtures typically used in gene therapy applications. (B) The spacings d = δm + δw and dDNA (see Fig. 3A for notation) as a function of L/D show that (i) d is nearly constant and (ii) two distinct regimes of DNA packing, one where the complexes are positive (L/D >5, dDNA ≈ 46 Å) and the other regime where the complexes are negative (L/D <5, dDNA ≈ 35 Å). (C) SAXS scans of the lamellar Lα phase of DOPC/DOTAP (cationic)-water mixtures done at lower resolution (rotating-anode x-ray generator). A dilution series of 30% (d = 57.61 Å), 50% (d = 79.49 Å), and 70% (d = 123.13 Å) H2O by weight is shown.

Fig. 3.

(A) Schematic picture of the local arrangement in the interior of lipid-DNA complexes (shown at two different concentrations in Fig. 1A and in (B) below. The semiflexible DNA molecules are represented by rods on this molecular scale. The neutral and cationic lipids comprising the membrane are expected to locally demix with the cationic lipids (red) more concentrated near the DNA. Micrographs of DNA-lipid condensates under (B) bright light and (C) crossed polarizers showing LC-like defects.

In the absence of DNA, membranes comprised of mixtures of DOPC and the cationic lipid DOTAP (1→1) exhibited strong long-range interlayer electrostatic repulsions that overwhelm the van der Waals attraction (27, 28). In this case, as the volume fraction Φw of water was increased, the Lα phase swelled and d could be obtained from the simple geometric relation d = δm/(1 − Φw) (27). The SAXS scans in Fig. 2C show this behavior with the (001) peaks moving to lower q as Φw increases. From d (= 2π/q(001)) at a given Φw, we obtain δm = 39 ± 0.5 Å for DOPC/DOTAP (1→1). Liposomes made of DOPC/DOTAP (1→1) with Φw ≈ 98.5% did not exhibit Bragg diffraction in the small wave vector range covered in Fig. 2A.

The DNA that condenses on the CLs strongly screens the electrostatic interaction between lipid bilayers and leads to condensed multilayers. The average thickness of the water gap δw = d − δm = 65.1 Å − 39 Å = 26.1 ± 2.5 Å is just sufficient to accommodate one monolayer of B-DNA (diameter ≈ 20 Å) including a hydration shell (29). We see in Fig. 2B that d is almost constant, as expected for a monolayer DNA intercalate (Fig. 3A). In contrast, as L/D decreased from 18 to 2, dDNA suddenly decreased from ∼44 Å in the positively charged regime just above L/D = 5 (near the stoichiometric charge neutral point) to ∼37 Å for the negatively charged regime (Fig. 2B). In these distinct regimes, lamellar condensates coexist with excess giant liposomes in the positive regime and with excess DNA in the negative regime (30).

The driving force for higher order self-assembly is the release of counterions. DNA carries 20 phosphate groups per helical pitch of 34.1 Å, and due to Manning condensation 76% of these anionic groups are permanently neutralized by their counterions, which leads to a distance between anionic groups ≈ the Bjerrum length = 7.1 Å (31). During condensation, the cationic lipid tends to fully neutralize the phosphate groups on the DNA, in effect replacing and releasing the originally condensed counterions (both those bound to the 1D DNA and to the 2D cationic membranes) in solution.

To improve on the signal-to-background intensity ratio, the synchrotron XRD experiments were carried out at concentrations (lipid + DNA ≈ 1.4 ± 0.3% volume in water), which, although dilute (24), were nevertheless greater than the concentrations used in the microscopy work (17, 19). A typical SAXS scan in mixtures at the optical microscopy concentrations (Fig. 1A) is shown in Fig. 2A (inset), which exhibits the same features and confirms that the local multilayer and DNA structure (Fig. 3A) is unchanged between the two concentrations. The x-ray samples consisted of connected yet distinct globules (Fig. 3B). What is remarkable is the retention of the globule morphology consistent with what was observed at lower concentrations in DIC (Fig. 1A). Under crossed polarizers (Fig. 3C), LC defects, both focal conics and spherulitic defects (32), resulting from the smectic-A-like layered structure of the DNA-lipid globules, are evident. The globules at the lower concentrations (Fig. 1A) show similar LC defects.

We further probed the nature of λ-DNA packing within the lipid layers by conducting a lipid dilution experiment in the isoelectric point regime of the complex. The total lipid (L = DOTAP + DOPC) was increased while the charge of the overall complex, given by the ratio of cationic DOTAP to DNA, was kept constant at DOTAP/DNA = 2.45 ± 0.15 (33). The SAXS scans in Fig. 4A (arrow points to the DNA peak) show that dDNA = 2π/qDNA increased, with lipid dilution from 24.54 to 57.1 Å as L/D increased with lipid dilution between 2.45 and ≈9 (Fig. 4B). The most compressed interaxial spacing of 24.55 Å at L/D = 2.45 approaches the short-range, repulsive hard-core interaction of the B-DNA rods containing a hydration layer (29).

Fig. 4.

(A) A series of SAXS scans of CL-DNA complexes at DOTAP/DNA = 2.4 ± 0.1 (approximately the isoelectric point) which shows the DNA peak (arrow) moving toward smaller q as L/D increases (that is, increasing the DOPC to DOTAP ratio at a constant DOTAP/DNA; L = DOTAP + DOPC, D = DNA). (B) dDNA and d = δm + δw from (A) plotted as a function of L/D (see Fig. 3A for notation). Circles are synchrotron data, and triangles are rotating anode. The solid line is the prediction of a packing calculation (with no adjustable parameters) where the DNA chains form a space-filling 1D lattice. (C) The average domain size of the 1D lattice of DNA chains derived from the width of the DNA peaks shown in (A) [corrected for resolution and powder averaging broadening effects; for example, see (27, 38)].

The DNA interaxial spacing can be calculated rigorously from simple geometric considerations. If we assume that all of the DNA is adsorbed between the bilayers and that the orientationally ordered DNA chains separate to fill the increasing lipid area as L/D increases, while maintaining a 1D lattice (Fig. 3A), then Embedded Image(1)

Here, ρD = 1.7 (g/cc) and ρL = 1.07 (g/cc) denote the densities of DNA and lipid, respectively; δm the membrane thickness; and AD the DNA area [AD = Wt(λ)/(ρDL(λ)) = 186 Å2, Wt(λ) = weight of λ-DNA = 31.5 × 106/(6.022 × 1023) g and L(λ) = contour length of λ-DNA = 48502 × 3.4 Å]. The solid line in Fig. 4B is then obtained from Eq. 1 with no adjustable parameters and shows a remarkable agreement with the data over the measured interaxial distance from 24.5 to 57.1 Å. The observed deviation from linear behavior both in the data and the solid line arises from the slight increase in δm as L/D increases (34). The existence of a finite-sized ordered lattice is made unambiguous from the linewidths of the DNA peaks (Fig. 4A), where we find that the 1D lattice of DNA chains has a correlated domain size extending to near 10 unit cells (Fig. 4C). Thus, the DNA chains form a 1D ordered array adsorbed between 2D membranes; that is, they form a novel finite-sized 2D smectic phase. Beyond L/D >10 we found multiphase behavior (35).

The lattice expansion at the isoelectric point covering interaxial distances with negligible short-range hydration forces (29) (B-DNA diameter ≈ 20 Å) is indicative of a long-range repulsion. The distribution of the counterion lipid (DOTAP) concentration according to the Poisson-Boltzmann equation along the top and bottom monolayer that bound the DNA molecules (Fig. 3A) will lead to a long-range electrostatic-induced interhelical interaction from the counterion lipid pressure (due to the expected local demixing of the cationic and neutral lipids) and the electric field. Preliminary salt-dependent experiments (35), which show shifts in the DNA peak, indicate that long-range electrostatic-induced interactions are present. Additionally, because of the semiflexible nature of λ-DNA [consisting of between 170 and 340 persistence lengths (ξp) in dilute solution (ξp ≈ between 500 and 1000 Å)], we expect the long-range repulsions to be further enhanced by chain-undulation interactions. A similar enhancement has been observed in a hexagonal lattice of DNA (29, 36). This phase of 1D DNA chains is the lower dimensional analog of 2D fluid membranes in that it may either be dominated by electrostatic-induced forces (27, 28) or the interplay between electrostatics and undulations (3739).

Further experiments are needed to elucidate the precise nature of the intermolecular forces and the interplay between electrostatic and chain undulation interactions (40). Future studies may also reveal regimes with 3D correlations between the DNA chains from layer to layer in analogy to recent theoretical findings in highly condensed DNA phases (41). The observed quantitative control over the structural nature of the DNA packing in CL-DNA complexes may lead to a better understanding of the important structural parameters relevant to transfection efficiencies in gene therapy (26); in particular, they should be directly relevant to our understanding of the interactions of the complex with cellular lipids and the mechanism of DNA transfer across the nuclear membrane.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 30.
  31. 31.
  32. 32.
  33. 33.
  34. 34.
  35. 35.
  36. 36.
  37. 37.
  38. 38.
  39. 39.
  40. 40.
  41. 41.
  42. 42.
View Abstract

Stay Connected to Science

Editor's Blog

Navigate This Article