THE JOURNAL OF GENE MEDICINE
`J Gene Med 2005; 7: 739–748.
`Published online 31 January 2005 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/jgm.717
`
`R E S E A R C H A R T I C L E
`
`New multivalent cationic lipids reveal bell curve for
`transfection efficiency versus membrane charge
`density: lipid–DNA complexes for gene delivery
`
`Ayesha Ahmad1
`Heather M. Evans1
`Kai Ewert1
`Cyril X. George2
`Charles E. Samuel2
`Cyrus R. Safinya1*
`
`1Departments of Materials, Physics,
`and Molecular, Cellular and
`Development Biology, University of
`California, Santa Barbara, Santa
`Barbara, CA 93106-5121, USA
`2Molecular, Cellular and
`Developmental Biology Department,
`Biomolecular Science and Engineering
`Program, University of California,
`Santa Barbara, Santa Barbara, CA
`93106, USA
`
`*Correspondence to:
`Cyrus R. Safinya, Materials Research
`Laboratory, UC Santa Barbara,
`Santa Barbara, CA 93106, USA.
`E-mail: safinya@mrl.ucsb.edu
`
`Received: 22 July 2004
`Revised: 9 September 2004
`Accepted: 20 September 2004
`
`Copyright  2005 John Wiley & Sons, Ltd.
`
`Abstract
`
`Background Gene carriers based on lipids or polymers – rather than on
`engineered viruses – constitute the latest technique for delivering genes into
`cells for gene therapy. Cationic liposome–DNA (CL-DNA) complexes have
`emerged as leading nonviral vectors in worldwide gene therapy clinical trials.
`To arrive at therapeutic dosages, however, their efficiency requires substantial
`further improvement.
`
`Methods Newly synthesized multivalent lipids (MVLs) enable control of
`headgroup charge and size. Complexes comprised of MVLs and DNA have
`been characterized by X-ray diffraction and ethidium bromide displacement
`assays. Their transfection efficiency (TE) in L-cells was measured with a
`luciferase assay.
`
`Results Plots of TE versus the membrane charge density (σM, average
`charge/unit area of membrane) for
`the MVLs and monovalent 2,3-
`dioleyloxypropyltrimethylammonium chloride (DOTAP) merge onto a
`universal, bell-shaped curve. This bell curve leads to the identification of three
`distinct regimes, related to interactions between complexes and cells: at low
`σM, TE increases with increasing σM; at intermediate σM, TE exhibits saturated
`behavior; and unexpectedly, at high σM, TE decreases with increasing σM.
`
`Conclusions Complexes with low σM remain trapped in the endosome.
`In the high σM regime, accessible for the first time with the new MVLs,
`complexes escape by overcoming a kinetic barrier to fusion with the
`endosomal membrane (activated fusion), yet they exhibit a reduced level
`of efficiency, presumably due to the inability of the DNA to dissociate from
`the highly charged membranes in the cytosol. The intermediate, optimal
`regime reflects a compromise between the opposing demands on σM for
`endosomal escape and dissociation in the cytosol. Copyright  2005 John
`Wiley & Sons, Ltd.
`
`Keywords gene therapy; cationic lipids; transfection efficiency; membrane
`charge density
`
`Introduction
`
`Cationic liposome–DNA (CL-DNA) complexes are attracting consider-
`able attention as gene vectors due to their safety and other inher-
`ent advantages over viral delivery methods [1,2]. These advantages
`
`Moderna Ex 1006-p. 1
`Moderna v Protiva
`
`

`

`740
`
`A. Ahmad et al.
`
`lack of
`include ease and variability of preparation,
`immunogenicity, and a capacity for DNA of unlimited size,
`allowing for delivery of human artificial chromosomes
`[3]. Recent setbacks in clinical trials with viral vectors, in
`particular a fatality induced by a severe inflammatory
`response [4] and insertional mutagenesis caused by
`retroviral vectors [5], have further promoted a diligent
`effort in developing efficient nonviral methods.
`Currently,
`lipofection is a prevalent nonviral gene
`transfer technology used in clinical
`trials worldwide
`[6]. CLs for transfection typically consist of a mixture
`of cationic and neutral (helper) lipid. Numerous lipids
`with varied chemical and physical properties have been
`synthesized [7–9] to improve the transfection efficiencies
`of CL-DNA complexes to the level of viral vectors. These
`include multivalent lipids, which have been described as
`superior to their monovalent counterparts [10,11].
`Despite this abundance of different cationic lipids,
`unifying themes and a comprehensive understanding
`of
`the interactions between CL-DNA complexes and
`mammalian cells are lacking. In order to rationally design
`and improve lipid-based delivery systems, however,
`such an understanding is essential. In particular, it is
`necessary to identify the interactions between the CL-DNA
`complexes and the cells along the transfection pathway
`to overcome the biological
`impediments to optimal
`transfection by directed alteration and optimization of
`CL-DNA complex formulations.
`In part, the lack of mechanistic understanding of gene
`delivery by CL-DNA complexes is due to the large num-
`ber of parameters involved. Few investigations to date
`include a complete examination of lipid performance as
`a function of lipid-bilayer composition and lipid/DNA
`charge ratio (ρchg). Even in comparative studies [12],
`typically only one or two data points per lipid are eval-
`uated, allowing the ideal lipid composition (the ratio of
`neutral to cationic lipid) or cationic lipid/DNA ratio to be
`overlooked [10,11].
`In previous experiments with commercially available
`lipids, Lin et al. identified the membrane charge density,
`σM, as a universal parameter for transfection by lamellar
`CL-DNA complexes, but the scope of these investigations
`was limited by the lipids used [13]. The membrane
`charge density is the average charge per unit area of
`the membrane. It is controlled by the ratio of neutral to
`cationic lipid in the liposome formulation. On the other
`hand, the lipid/DNA charge ratio, ρchg, is the number of
`charges on the cationic lipid divided by the number of
`charges on the DNA. In our experiments, we keep both
`the amount of DNA and ρchg (and thus the number of
`charges contributed by the cationic lipid) constant. Thus,
`σM is varied solely by changing the amount of neutral
`lipid per transfection assay, ‘‘diluting’’ the cationic lipid in
`the membrane.
`The work reported here presents both a confirmation
`and a significant extension of
`the earlier findings.
`We have synthesized a set of new multivalent lipids
`(MVLs) through methodical variation of headgroup size
`and charge and have examined the dependence of
`
`transfection efficiency (TE) on two key parameters,
`lipid composition and lipid/DNA charge ratio, ρchg. The
`MVLs [14], which form lamellar DNA complexes alone
`and when mixed with neutral 1,2-dioleoyl-sn-glycero-3-
`phosphatidylcholine (DOPC), enabled us to systematically
`probe very high membrane charge densities for the
`first time. For all DNA complexes of the MVLs as well
`as monovalent 2,3-dioleyloxypropyltrimethylammonium
`chloride (DOTAP), TE plotted versus σM fits the same, bell-
`shaped curve, confirming σM as a universal parameter.
`The curve shows three distinct regimes of TE and a
`clear maximum of TE at an optimal charge density,
`∗. Here, the TE rivals that of hexagonal CL-DNA
`σM
`∗ of the universal curve shifts
`complexes. The optimal σM
`systematically with ρchg, with an increase inρ chg resulting
`∗. Only the new, highly charged
`in higher values for σM
`lipids investigated here have permitted unambiguous
`identification of the universal maximum in TE and its shift
`with ρchg, as well as the discovery of a third regime of TE,
`where TE decreases (not saturates) with increasing σM.
`
`Materials and methods
`
`Materials
`
`The multivalent lipids (MVLs), MVL2 (molecular weight
`(MW) = 884.2 g/mol), MVL3
`(MW = 977.8 g/mol),
`MVL5 (MW = 1552.7 g/mol),
`and TMVL5 (MW =
`1253.0 g/mol) (Table 1), were synthesized according
`to the procedure previously described [14]. Nα,Nδ-
`Bis(Boc)-ornithine and Nδ-Boc-ornithine (Novabiochem)
`were used as the starting materials for MVL2 and
`MVL3, respectively. 2,2(cid:2)-(Ethylenedioxy)diethylamine for
`TMVL5 was purchased from Aldrich. For all lipids except
`MVL2, the final deprotection was performed by dissolving
`of the protected lipid in trifluoroacetic acid (Fisher),
`incubating at room temperature for 30 min and drying
`in vacuum,
`in deviation from the published protocol
`[14]. As described below, cationic liposomes were
`prepared containing these lipids as well as the cationic
`lipid 2,3-dioleyloxypropyltrimethylammonium chloride
`(DOTAP, MW = 698.55 g/mol), in combination with the
`lipids DOPC (MW = 786.13 g/mol) and 1,2-
`neutral
`dioleoyl-sn-glycero-3-phosphatidylethanolamine (DOPE,
`MW = 744.05 g/mol), all
`from Avanti Polar Lipids.
`CL-DNA complexes were formed from these cationic
`liposomes and the appropriate DNA. For X-ray samples,
`ethidium bromide (EtBr) experiments, and transfection
`assays, highly purified λ-phage DNA (New England
`Biolabs), highly polymerized calf thymus DNA (Amersham
`Life Sciences) and pGL3 plasmid DNA containing the
`luciferase gene (Promega Corp.) were used, respectively.
`
`Liposomes
`
`Lipid mixtures were prepared volumetrically by combin-
`ing chloroform/methanol (4 : 1) solutions of cationic and
`
`Copyright  2005 John Wiley & Sons, Ltd.
`
`J Gene Med 2005; 7: 739–748.
`
`Moderna Ex 1006-p. 2
`Moderna v Protiva
`
`

`

`Lipid–DNA Complexes for Gene Delivery
`
`741
`
`same manner as above, but, prior to transfer to cells, the
`complexes were mixed with 1.8 µl of a 6 mg/ml chloro-
`quine solution and incubated for 10 min. The normal
`transfection protocol was then resumed. To compensate
`for the variation in cell behavior over time, the data
`for DOTAP/DOPC complexes was normalized using data
`taken at the same time as TE data for MVL5.
`
`X-ray diffraction (XRD)
`
`CL-DNA complexes were prepared by mixing 75 µg
`of λ-phage DNA at 5 mg/ml with liposome solutions
`(20 mg/ml) in an Eppendorf centrifuge for approximately
`3 h. Samples were prepared at ρchg = 2.8. After storage
`for 3 days at 4 ◦C, allowing the samples to reach
`equilibrium, they were transferred to 1.5 mm diameter
`quartz X-ray capillaries (Hilgenberg, Germany). The
`high-resolution XRD experiments were carried out at
`the Stanford Synchrotron Radiation Laboratory. Two-
`dimensional powder diffraction images were obtained
`using an image plate detector (Mar Instruments).
`
`EtBr displacement assay
`
`Samples were prepared in a 96-well plate. Each well
`contained 2.4 µg of DNA, 0.28 µg of EtBr and the
`appropriate amount of cationic lipid. Water was added to
`each well, achieving a final volume of 200 µl per well.
`Fluorescence was measured on a Cary Eclipse fluorescence
`spectrophotometer.
`
`Results and discussion
`
`MVL design and structures
`
`Table 1 shows the chemical structures and maximum
`charges of the MVLs used in this study. These lipids
`were designed to achieve a systematic variation of the
`headgroup charge with minimal change in the chemical
`structure. Only a single aminopropyl unit per cationic
`charge was added to the headgroup, starting from the
`ornithine unit of MVL2. The oleyl chains provide strong
`anchoring in the membrane and miscibility with DOPC.
`TMVL5 has a slightly longer (triethylene glycol) spacer
`than the other MVLs.
`
`X-ray characterization of MVL-DNA
`complexes shows a lamellar phase
`
`We used XRD to determine the structure of MVL-
`DNA complexes. For all MVLs, at all
`investigated
`cationic/neutral
`lipid compositions of 0–90% DOPC,
`C) phase, the
`MVL-DNA complexes form the lamellar (Lα
`highly prevalent of the two known complex structures
`[15–17]. Moreover, XRD shows no evidence of phase
`separation, indicating that the complexes contain both
`
`Structure
`
`2 Cl-
`
`+
`NH3
`
`N
`H
`
`O
`
`O
`
`N
`H
`
`H2N+
`
`+
`NH3
`
`+
`NH3
`
`H
`N
`
`H
`N
`
`O
`
`O
`
`-
`3 Cl
`
`O O
`
`O O
`
`+
`NH3
`
`H
`N+
`
`+
`NH3
`
`+
`NH3
`
`+
`NH3
`
`O
`
`H2N+
`
`NH
`
`NH
`
`O
`
`-
`5 CF3CO2
`
`O O
`
`N+H
`
`+
`NH3
`
`+
`NH3
`
`+
`NH3
`
`O
`
`H2N+
`
`NH
`
`O
`
`2
`
`-
`
`5 Cl
`
`NH
`
`O
`
`OO
`
`Table 1.
`
`Name
`(max. chg.)
`
`MVL2
`(+2)
`
`MVL3
`(+3)
`
`MVL5
`(+5)
`
`TMVL5
`(+5)
`
`neutral lipids. The solvents were evaporated, first under a
`stream of nitrogen and subsequently in a high vacuum to
`ensure complete removal of the solvents. The dried lipid
`mixtures were hydrated at 37 ◦C for at least 6 h with the
`appropriate amount of deionized water of 18.2 M (final
`concentration of 20 mg/ml for X-ray samples; final con-
`centration of 0.5 mg/ml for transfection, and EtBr assay
`samples), sonicated to clarity with a VibraCell from Sonics
`and Materials Inc., and filtered through a 0.2 µm Teflon
`filter (Whatman). The obtained liposome solutions were
`stored at 4 ◦C.
`
`Cell Transfection
`
`Mouse fibroblast L-cells were cultured in Dulbecco’s mod-
`ified Eagle’s medium (DMEM, Gibco BRL) supplemented
`with 1% (v/v) penicillin-streptomycin (Gibco BRL) and
`5% (v/v) fetal bovine serum (Gibco BRL) at 37 ◦C in
`a humidified atmosphere with 5% CO2, reseeding the
`cells every 2–4 days to maintain subconfluency. The
`cells were transfected at 60–80% confluency in 24-well
`plates (7 mm diameter per well). Liposome (0.5 mg/ml)
`and DNA (1 mg/ml) stock solutions were diluted with
`DMEM to a final volume of 0.1 ml and complexes, con-
`taining 0.4 µg of pGL3-DNA per well, were prepared at
`the desired cationic-to-anionic charge ratio (ρchg). The
`cells were incubated with the complexes for 6 h, rinsed
`three times with phosphate-buffered saline (PBS, Gibco
`BRL), and incubated in supplemented DMEM for an
`additional 24 h (sufficient for a complete cell cycle) to
`allow expression of the luciferase gene. Luciferase gene
`expression was measured with the luciferase assay sys-
`tem from Promega Corp., and light output readings were
`performed on a Berthold AutoLumat luminometer. Trans-
`fection efficiency, measured as relative light units (RLU),
`was normalized to the weight of total cellular protein
`using the Bio-Rad protein assay dye reagent. For exper-
`iments with chloroquine, complexes were formed in the
`
`Copyright  2005 John Wiley & Sons, Ltd.
`
`J Gene Med 2005; 7: 739–748.
`
`Moderna Ex 1006-p. 3
`Moderna v Protiva
`
`

`

`742
`
`A. Ahmad et al.
`
`the lipid bilayers and corresponds to a DNA interhelical
`spacing dDNA = 2π/qDNA [17]. The experimental values
`are dDNA = 31 ˚A for MVL3 and dDNA = 27 ˚A for TMVL5.
`In Figures 1D and 1E, XRD patterns of complexes
`containing 100% MVL3 (D) and TMVL5 (E) under
`otherwise identical conditions are shown. The narrow
`peaks and multiple harmonics show that at the salt
`concentrations present in DMEM, i.e. under conditions
`as in the transfection experiments, stable and well-
`defined complexes form even from membranes containing
`exclusively the highly charged multivalent lipids. The
`salt present in DMEM screens the electrostatic repulsions
`between the headgroups and enables formation of stable
`complexes. However, when prepared in deionized water,
`complexes without neutral lipids exhibit broadening of
`the first lamellar peak (at q001) and do not show higher
`harmonics of the first lamellar peak, indicative of smaller
`size multilamellar assemblies (results not shown).
`
`Estimation of lipid headgroup charge:
`EtBr displacement assay
`
`Figure 2 shows data from an EtBr displacement assay
`[18–21], performed to examine the ability of the MVLs to
`condense DNA within the CL-DNA complexes. The data
`was acquired by collecting fluorescence measurements
`at various weight ratios of MVL to DNA (with a fixed
`weight of DNA and EtBr per point) and normalizing the
`intensity to the fluorescence of DNA and EtBr in solution.
`EtBr fluoresces when intercalated between the base pairs
`of DNA, but self-quenches in solution. As the MVL
`C
`liposomes are introduced and self-assembly into the Lα
`phase occurs, EtBr is displaced and overall fluorescence
`decreases until all DNA has been incorporated into the
`MVL-DNA complexes at the isoelectric point. Thus, this
`method allows a quick and efficient assessment of the
`effective charge on the headgroup of these lipids. The
`
`Figure 2. Normalized fluorescence data from the EtBr displace-
`ment assay for MVL2, MVL3 and MVL5. The dashed lines
`are drawn at the isoelectric points, determined as described
`in the text. They result in valencies of ZMVL2 = 2.0 ± 0.1,
`ZMVL3 = 2.5 ± 0.1, ZMVL5 = ZTMVL5 = 4.5 ± 0.1
`
`Figure 1. (A) Schematic of the lamellar phase indicating the
`characteristic dimensions. Reprinted with permission from [15].
`(B, C) Typical X-ray diffraction (XRD) scans from lamellar
`C) CL-DNA complexes, containing 40 mol% MVL ((B) MVL3;
`(Lα
`(C) TMVL5) and 60 mol% DOPC, at a lipid/DNA charge ratio
`ρchg = 2.8 in the presence of DMEM. (D, E) Typical XRD scans
`C CL-DNA complexes containing 100 mol% MVL ((D)
`from Lα
`MVL3; (E) TMVL5), prepared at ρchg = 2.8 in the presence of
`DMEM (Reproduced in part with permission from reference 15)
`
`MVL and DOPC, as intended. A schematic of the lamellar
`C) and its characteristic dimensions is shown
`phase (Lα
`in Figure 1A. Figures 1B and 1C shows typical XRD
`patterns of complexes containing 40 mol% MVL (MVL3
`(B); TMVL5 (C)) and 60 mol% DOPC, at a lipid/DNA
`charge ratio of 2.8, prepared in the presence of DMEM.
`The sharp peaks, labelled q001, q002, q003, respectively,
`give the lamellar repeat distance, d, which is the sum
`of the membrane thickness (δm) and the thickness of
`a water/DNA layer (δw): d = δm + δw = 2π/q001. The
`labeled qDNA, results from one-
`diffuse weaker peak,
`dimensional ordering of the DNA sandwiched between
`
`Copyright  2005 John Wiley & Sons, Ltd.
`
`J Gene Med 2005; 7: 739–748.
`
`Moderna Ex 1006-p. 4
`Moderna v Protiva
`
`

`

`Lipid–DNA Complexes for Gene Delivery
`
`743
`
`isoelectric point was determined as the intersection
`point of
`linear fits to the data at high and low
`lipid/DNA ratio [22,23]. This gives headgroup charges of
`ZMVL2 = 2.0 ± 0.1, ZMVL3 = 2.5 ± 0.1, ZMVL5 = 4.5 ± 0.1.
`The dashed lines in Figure 2 indicate the corresponding
`isoelectric MVL/DNA weight ratios.
`
`Transfection efficiency as a function of
`lipid composition
`
`Figure 3 shows TE results for complexes transfecting
`mouse fibroblast cells at various MVL/DOPC ratios.
`Also included is data for the monovalent lipid DOTAP
`mixed with DOPC, a well-investigated reference system
`which constituted the starting point of our studies [13].
`The complexes were prepared at ρchg = 2.8, which Lin
`et al. have found to be the optimum charge ratio for
`DOTAP/DOPC complexes [24]. The amount of DNA and
`cationic lipid per sample was kept constant. Thus, only the
`amount of neutral lipid varies between data points. All
`MVL-DNA complexes form globular particles of around
`0.2 µm diameter in water, as previously reported for
`DOTAP [25] and MVL5 [14]. In DMEM, these particles
`form much larger aggregates due to screening of their
`electrostatic repulsion by the high salt concentration, as
`we have shown for MVL5 by optical and epi-fluorescence
`microscopy [14]. Figure 3A shows the TE data as a
`function of the molar fraction of cationic lipid. For all
`cationic lipids, a maximum in TE as a function of lipid
`composition is observed: at 65 mol% for MVL2, 70 mol%
`for MVL3, 50 mol% for MVL5, 55 mol% for TMVL5, and
`90 mol% for DOTAP. The optimal molar ratio results in a
`TE that is over two decades higher than that of the lowest
`transfecting complexes in these systems, and each data
`set fits a skewed bell-shaped curve.
`
`Membrane charge density is a
`universal parameter: three regimes of
`transfection efficiency
`
`Figure 3B shows the data from Figure 3A plotted versus
`the membrane charge density, σM. A notable simplification
`occurs and all the data points merge onto a single curve.
`This identifies σM as a universal parameter for transfection
`by lamellar CL-DNA complexes.
`the average charge
`As mentioned above, σM is
`per unit area of the lipid membrane; therefore, the
`headgroup areas of
`the lipids,
`their charge, and
`the molar
`fractions of cationic and neutral
`lipid
`required to calculate σM. We
`are the parameters
`calculated σM as described by Lin et al. [13], with σM =
`total charge/total membrane area = eZNcl/(Ncl Acl +
`Nnl Anl) = [1 − nl/( nl + r cl)]σcl, where Ncl and Nnl
`are the number of cationic lipids and neutral lipids in
`the complexes, respectively; r = Acl/Anl is the ratio of the
`headgroup areas of the cationic and the neutral lipid;
`σcl = eZ/Acl is the charge density of the cationic lipid
`
`with valence Z; and nl and cl are the molar fractions of
`the neutral and cationic lipids, respectively. For our data,
`we used Anl = 72 ˚A2 [26,27], rDOTAP = 1, rMVL2 = 1.05 ±
`0.05, rMVL3 = 1.30 ± 0.05, rMVL5 = 2.3 ± 0.1, rTMVL5 =
`2.5 ± 0.1, ZDOTAP = 1, ZMVL2 = 2.0 ± 0.1, ZMVL3 = 2.5 ±
`0.1, ZMVL5 = ZTMVL5 = 4.5 ± 0.1. The values for Z were
`obtained by the EtBr displacement assay as described
`above. The values for r can be considered as fitting
`parameters, but they yield physically reasonable values
`that agree with chemical intuition. (The values of r were
`determined based on agreement with the Gaussian fit.
`For the monovalent lipid DOTAP, r was assumed to be
`equal to 1, consistent with previous findings [13].) It
`is interesting to note that the optimal TE for MVL3 is
`found at a larger molar fraction of cationic lipid than for
`MVL2 despite the fact that MVL3 has a higher headgroup
`charge. This can be attributed to the significantly larger
`headgroup size of MVL3 and shows the importance of the
`parameter r.
`The resulting curve of TE vs. σM can be described
`empirically by a simple Gaussian (solid line in Figure 3B):
`TE = TE0 + A exp−[(σM − σM
`∗
`)/w]2
`(1)
`where TE0 = −(1.9 ± 5.6) × 107 RLU/mg protein; A =
`(9.4 ± 0.6) × 108 RLU/mg protein; w = (5.8 ± 0.5) ×
`10−3 e/˚A2. For the optimal charge density, the fit gives
`∗ = (17.4 ± 0.2) × 10−3 e/˚A2.
`σM
`Remarkably, in extension of previous results which
`showed the increase and a subsequent levelling-off of TE
`with σM [13], we see an entire bell curve of efficiency,
`including a decrease in TE at higher charge densities.
`Previously, without the series of new MVLs, Lin et al.
`were not adequately prepared to measure this range of
`charge densities.
`The new universal TE curve of lamellar complexes
`exhibits three well-defined regimes. Regime I (dark
`gray), corresponding to low σM, features an exponential
`increase in efficiency over three orders of magnitude.
`Regime III (light gray), corresponding to high σM, is
`characterized by a decrease in efficiency with increasing
`σM, suggesting that there also is an obstacle of electrostatic
`nature to successful DNA delivery by lamellar CL-DNA
`complexes. The competition of the two effects that give
`rise to regimes I and III leads to the existence of the
`intermediate regime II (white) as the region of optimal
`charge density, corresponding to the highest TE. This
`clearly demonstrates the importance of including neutral
`lipid in the formulation of CL-DNA complexes, particularly
`those of newer, multivalent lipids. Due to the universality
`of the curve, it should also be possible to estimate the
`optimal composition for any given lipid by performing
`the simple EtBr displacement assay and estimating the
`headgroup size of the lipid from its chemical structure.
`We will address the implications of the universal curve
`for the mechanism of transfection in more detail below.
`In addition to data for
`the lamellar complexes,
`Figure 3B shows TE data for
`the commonly used
`C)
`DOTAP/DOPE lipid system. The inverted hexagonal (HII
`
`Copyright  2005 John Wiley & Sons, Ltd.
`
`J Gene Med 2005; 7: 739–748.
`
`Moderna Ex 1006-p. 5
`Moderna v Protiva
`
`

`

`744
`
`A. Ahmad et al.
`
`Figure 3. (A) TE in RLU per mg total cellular protein plotted as a function of mol% DOPC for DNA complexes prepared with MVL2
`(green diamonds), MVL3 (red squares), MVL5 (blue triangles) and TMVL5 (purple inverted triangles) as well as DOTAP (gray
`circles). All data was taken at ρchg = 2.8, using the same amount of DNA for each data point. (B) The same TE data plotted against
`the membrane charge density, σM. Also included are data for DOTAP/DOPE complexes (gray open circles) and a Gaussian fit to the
`DOPC systems. The three regimes of transfection efficiency are indicated by different shading in the plot
`
`Figure 4. TE data taken in the presence of chloroquine (chlq) to
`assess the relevance of endosomal escape in the three regimes
`(regime I, dark gray; regime II, white; regime III, light gray).
`The relative increase in TE, TEchlq/TE-1, is plotted for MVL3
`(black bars) and MVL5 (white bars) in the three regimes. TE is
`enhanced by about a factor of three in regime I (10 mol% MVL),
`but not in regime II (60 mol% MVL) and regime III (100% MVL).
`This indicates that endosomal release is not a limiting factor for
`regimes II and III
`
`[15] DOTAP/DOPE complexes exhibit high TE even at
`low σM, due to their distinct mechanism of cellular entry,
`which relies on the fusogenic properties of DOPE [2,13].
`This has made DOPE a popular choice as a co-lipid, and
`complexes with DOPE are abundant in the literature.
`However, the vast majority of neutral
`lipids lead to
`lamellar CL-DNA complexes, and, even with PE-based
`lipids, the hexagonal phase is only observed in a small
`window of composition [15]. Furthermore, the in vivo
`performance of PE-based complexes is disappointingly
`poor and cholesterol, which leads to lamellar complexes, is
`increasingly used as a neutral lipid for in vivo applications
`[2,7]. Therefore, our result that the TE of optimized
`
`Figure 5. (A) Three-dimensional transfection phase diagram of
`lamellar complexes, combining data for MVL2, MVL3, and MVL5.
`Membrane charge density, σM, lipid/DNA charge ratio, ρchg, and
`TE are plotted along the x-axis, y-axis, and z-axis, respectively.
`∗ (from the data shown in (A)) is plotted
`(B) The optimal σM
`∗)
`against the charge ratio, ρchg. (C) The maximum TE (at σM
`plotted against ρchg. This TE essentially remains constant
`
`C
`C complexes rivals that of the highly transfecting HII
`Lα
`DOTAP/DOPE complexes is of great significance. It adds
`a compositional degree of freedom, because efficiently
`transfecting complexes can be prepared from a broad
`
`Copyright  2005 John Wiley & Sons, Ltd.
`
`J Gene Med 2005; 7: 739–748.
`
`Moderna Ex 1006-p. 6
`Moderna v Protiva
`
`

`

`Lipid–DNA Complexes for Gene Delivery
`
`745
`
`range of lipids without diminishing TE if σM is optimized.
`This includes lipids with specialized functions in the
`delivery process, such as peptide-lipids.
`
`Effect of chloroquine on transfection
`efficiency in the three regimes
`
`Endocytosis is the dominant mechanism of entry of
`lamellar CL-DNA complexes, as evident
`from recent
`confocal microscopy and transfection data [13] and work
`from other laboratories [28,29]. After cellular uptake
`via endocytosis, CL-DNA complexes must escape from
`endosomes in order for the DNA to progress toward the
`cell nucleus. Only a finite amount of time is available for
`this, since the endosomal pathway involves degradation of
`the contents of the endosome: initially through a lowering
`of the pH within the endosome and then through fusion
`with low-pH lysosomes.
`The importance of endosomal escape as a barrier to
`transfection can be assessed by performing transfection
`experiments in the presence of chloroquine (chlq),
`a well-established bio-assay known to enhance the
`release of material
`from endosomes by osmotically
`bursting the vesicle [30]. We have repeated efficiency
`experiments for MVL3 and MVL5 in the presence of
`chloroquine for data points characteristic of the three
`transfection regimes. The resulting data is shown in
`Figure 4, plotted as the relative increase in transfection
`efficiency, (TEchlq-TE)/TE = TEchlq/TE − 1. TE in regimes
`II (60 mol% MVL) and III (100% MVL) is not enhanced
`by the addition of chloroquine. However, an increase
`in TE is seen in regime I, at low σM (10 mol% MVL),
`where transfection efficiency is enhanced approximately
`threefold. These results suggest that complexes with lower
`σM remain trapped within endosomes, while complexes
`with higher σM are able to overcome this barrier.
`Therefore, the decrease in TE at highest σM, in regime
`III, is not related to endosomal entrapment but must be
`due to other effects.
`In regime I, a straight line fits the data of Figure 3B
`well, particularly for small values of σM. This regime was
`previously investigated by Lin et al. [13], who proposed
`that endosomal escape limits TE in this regime, consistent
`with the enhancement of TE by chloroquine. The
`escape from the endosome likely occurs via an activated
`fusion process of the oppositely charged membranes of
`endosome and complex [13]. The activation energy for
`this can be written as δE = aκ − bσM, where a and b are
`constants >0. The parameter κ is the bending rigidity of
`the membrane, which is mainly determined by the lipid
`tails and the area per lipid chain and therefore constant in
`our experiments. Bending of membranes, as required for
`fusion, results in an energy cost proportional to κ. Since
`the interacting membranes are oppositely charged, the
`activation energy decreases with increasing σM, making
`fusion with the endosomal membrane more likely. For
`endosomal entrapment being the main impediment to
`transfection as proposed by Lin et al., the activation
`
`energy for fusion directly relates σM to the transfection
`efficiency via an Arrhenius-type equation [13]:
`TE ∝ rate of fusion = 1/τ e
`−δE/kT
`
`(2)
`
`Here, 1/τ is the collision rate between the trapped CL-
`DNA particle and the endosomal membrane.
`At higher charge densities, in regimes II and III, TE
`is no longer limited by endosomal escape, as shown
`by the negligible effect of chloroquine. Here, the data
`of Figure 3B bends strongly away from a straight line
`but is well approximated by the empirical bell curve of
`Equation (1). Except for the lowest values of σM, this bell
`curve also fits the data in regime I, implying that the
`activation energy δE most likely contains contributions
`both linear and quadratic in σM.
`
`Effect of lipid/DNA charge ratio:
`transfection phase diagram
`
`All data shown in Figure 3 were taken at a fixed lipid/DNA
`charge ratio, ρchg, of 2.8. For other values of ρchg, we have
`observed similar, universal behavior. An efficiency phase
`diagram, shown in Figure 5A, summarizes these results.
`In this graph, σM varies along the x-axis, ρchg along the
`y-axis, and TE along the z-axis. At very low lipid/DNA
`charge ratios (i.e. below and near the isoelectric point,
`ρchg ≤ 1, where complexes are negatively charged or
`neutral), TE is low for all values of σM. This is to
`be expected, as an overall positive CL-DNA charge is
`required to promote initial electrostatic interactions with
`cell membranes [31–33]. As ρchg is increased to above
`unity, a maximum in TE, defining the optimal membrane
`∗, emerges and a bell curve of efficiency
`charge density σM
`∗ shifting to higher values
`is observed, with the optimal σM
`with increasing ρchg. This means that much more cationic
`lipid is required to achieve optimal TE at large lipid/DNA
`charge ratios. For clarity, this surprising trend is shown
`∗ against ρchg.
`in Figure 5B, which plots the optimal σM
`The descending part of the bell curve beyond the optimal
`∗ for ρchg >1 (i.e. regime III) cannot be seen for
`σM
`the highest lipid/DNA charge ratios. However, it seems
`likely that, given liposomes with an even higher charge
`density than the 27.17 × 10−3 e/˚A2 attainable with 100%
`MVL5, the decrease in the efficiency predicted by the bell
`curve shape would also be seen for the highest ρchg. As
`shown in Figure 5C, the maximum TE (TE at the optimal
`∗) does not change appreciably with ρchg. A relatively
`σM
`low lipid/DNA charge ratio, therefore, can be considered
`optimal since it allows for achievement of maximum TE
`with the least amount of cationic lipid. This is due to
`∗ with ρchg. Minimizing the
`the unexpected increase of σM
`amount of cationic lipid is desirable to reduce cost as well
`as potential toxic effects of the cationic lipid. In addition,
`achieving a given σM with fewer, more highly charged
`molecules should mean a smaller metabolic effort for the
`elimination of the lipids from the cell. This reasoning
`would favor multivalent over monovalent lipids. In this
`
`Copyright  2005 John Wiley & Sons, Ltd.
`
`J Gene Med 2005; 7: 739–748.
`
`Moderna Ex 1006-p. 7
`Moderna v Protiva
`
`

`

`746
`
`A. Ahmad et al.
`
`context, it is important to note that with the amounts
`of cationic lipid employed in our in vitro experiments,
`we find no toxic effects on the cells as judged by cell
`morphology and the amount of total cellular protein.
`
`A model of the intracellular CL-DNA
`complex pathway as a function of σM
`
`The schematic shown in Figure 6 summarizes our current
`C complexes in the
`understanding of the cellular fate of Lα
`three regimes. Initial attachment mediated by electrostatic
`attractions between CL-DNA complexes and negative
`charges at the cell surface (Figure 6a) is followed by
`endocytosis of the complex particle (Figure 6b), resulting
`in endosomal entrapment (Figure 6c) [13,31,34].
`If
`cellular attachment and uptake were limiting TE via a
`σM-dependent mechanism, a linear increase of TE with
`σM would be predicted. Thus, our data, which shows an
`exponential increase, excludes this possibility.
`∗ (regime I), trans-
`For complexes with low σM < σM
`fection is limit