Stable and monodisperse lipoplex formulations for gene delivery

Gene Therapy (1998) 5, 1272–1282 1998 Stockton Press All rights reserved 0969-7128/98 $12.00

http://www.stockton-press.co.uk/gt

Stable and monodisperse lipoplex formulations forgene delivery

O Zelphati1, C Nguyen1, M Ferrari2, J Felgner2, Y Tsai3 and PL Felgner1

1Gene Therapy Systems Inc, San Diego, CA; 2Vical Inc, San Diego, CA, USA; and 3Isis Pharmaceuticals, Carlsbad, CA, USA

A stable single vial lipoplex formulation has been of a T-connector. Homogenous cationic liposome prep-developed that can be stored frozen without losing either arations were prepared by extrusion in two different sizebiological activity or physical stability. This formulation was ranges of either 400 or 100 nm. Extruded liposomes pro-identified by systematically controlling several formulation duced more monodisperse and physically stable lipoplexvariables and without introducing either stabilizers or sur- formulations than unextruded liposomes, but the formu-factants. Analytical assays were used to unambiguously lations prepared with 100 nm liposomes were less activecharacterize the formulations. The critical formulation para- in in vitro transfection assays than either the 400 nm ormeters were: (1) the size of the cationic liposomes; (2) the unextruded liposomes. Low ionic strength and 5% sorbitolrate and method of DNA and cationic liposome mixing; and were required for the lipoplex formulations to survive freez-(3) the ionic strength of the suspending vehicle. The mixing ing and thawing. A frozen lipoplex formulation stored forconditions were precisely controlled by using a novel, spe- more than a year maintained its biological activity. Thesecially designed continuous flow pumping system in which results have broad implications for the pharmaceuticalthe DNA and liposome solutions were mixed at the junction development of lipoplex formulations for gene delivery.

Keywords: gene delivery; cationic lipid; extrusion; formulation; DMRIE

IntroductionCationic lipid/DNA complexes (lipoplexes) have beensuccessfully employed for in vitro and in vivo gene deliv-ery, and they are being evaluated in an increasing num-ber of human gene therapy clinical trials.1–6 However, thelack of control over the biophysical and molecular para-meters influencing lipoplex formation represents a limi-tation to consistently obtaining well-defined, stable andmonodisperse formulations with reproducible biologicalactivity.

When cationic liposomes and DNA are mixed, theyinteract electrostatically, reorganize and form lipoplexeswherein the DNA has been shown to induce aggregationand fusion of cationic liposomes.7–10 Many lipoplex prep-arations are physically unstable and their transfectionactivity can decrease with time. Such problems can beavoided for in vitro experiments by using low concen-trations of DNA and lipids, but clinical testing requiresmuch higher quantities of DNA. Consequently, for in vivoapplications the lipoplexes are usually prepared freshand used shortly after the DNA and cationic liposomesare mixed.1,2,11,12 Most of the clinical formulations haveto be prepared freshly by the physician at the bedsideprior to injecting them into patients. Thus, production ofstable lipoplex formulations in a single vial, which do nothave to be prepared immediately before use represents adesirable practical goal.

Correspondence: PL Felgner, Gene Therapy Systems Inc., 3525 John Hop-kins Court, San Diego, CA 92121, USAReceived 6 November 1997; accepted 23 March 1998

Recent publications have demonstrated increasingattention to the problems of developing and characteriz-ing physically stable lipoplex formulations. Quantitativeanalytical assays to characterize the physical properties oflipoplexes have just been reported.10,13 Other laboratorieshave focused on the development of stable lipoplex for-mulations and demonstrated that the addition of com-pounds such as spermidine, polyethyleneglycol14 or sur-factants15,16 can enhance the stability and prevent theaggregation of lipoplexes. Although very promising,these approaches used relatively low DNA concen-trations (200 mg/ml and below) and the added excipientsmay complicate the characterization, toxicity and manu-facturing of these formulations. Moreover, a moredetailed understanding of the real benefits of these addi-tives needs further investigation.

Here, we report a systematic survey to identify andcharacterize monodisperse stable lipoplex formulationsbased on quantitative biological and analytical physical–chemical assays. The results show that by combining twovery simple procedures, liposome extrusion and con-trolled mixing, aggregation problems can be avoided. Inthis way, well-defined (monodisperse) and stable lipo-plexes with high transfection activities can be preparedreproducibly without the addition of any other compo-nents. An isotonic vehicle that can be used to preparestable frozen single vial formulations without affectingphysical properties or biological activity is also described.

ResultsParameters influencing the efficacy of lipoplexes

Cationic lipid/DNA ratio effects on in vitro transfectionactivity: In order to optimize the transfection efficiency

Stable and monodisperse lipoplex formulationsO Zelphati et al

1273of the lipoplexes, several laboratories have examined theparameters that affect the level of transfection.1,2,17–19

From these studies, it appears that DNA and lipid con-centration, cationic lipid/DNA molar ratio, the presenceor absence of helper lipids (ie DOPE or cholesterol) andthe type of vehicle (ionic strength) used to make the lipo-plexes, are among the critical parameters. With DMRIE,one of the cationic lipids currently used in clinical trials,the optimum molar ratio (number of mole of cationic lip-ids per mole of DNA phosphate) for in vitro transfectionefficiency is between 0.25 and 1.517,20 (data not shown).As the cationic lipid/DNA molar ratio increased abovea value of 2, the activity of DMRIE progressivelydeclined, and with a large excess of lipids, toxicitybecame a limiting factor.

Cationic lipid/DNA ratio effects on physical stability: Theresults in Figure 1a illustrate the problem of lipoplex

Figure 1 (a) Particle size distribution of lipoplexes (prepared from unex-truded liposomes) as a function of the DMRIE:DOPE/DNA molar ratioand time of measurements. (b) Particle size determination of lipoplexesformulated at 0.5 DMRIE:DOPE/DNA molar ratio as function of theorder of addition. (c) Particle size determination of lipoplexes formulatedat 5 DMRIE:DOPE/DNA molar ratio as function of the order of addition.Size distributions were determined as mean of diameter on the basis ofvesicle volume.

physical stability at charge ratio near 1, wherein the netcharge on the lipoplex approaches 0. Freshly mixed lipo-plexes prepared by the classical method had a meandiameter of 200–500 nm whereas after 6 h of incubationat room temperature the formulations prepared near thecharge neutral condition aggregated to a particle diam-eter of 2000–2500 nm (Figure 1a). As the time course wasprolonged, the particles continued to increase in size(data not shown). These problems were greater whenhigher DNA concentrations were used. Under these con-ditions, severe aggregation resulted in low transfectionactivity2 (data not shown). Since 0.25 and 0.5DMRIE:DOPE/DNA molar ratios gave high in vitrotransfection activity17,20 with a relatively low tendency toaggregate, we have focused our efforts on the characteriz-ation of formulations prepared at these two ratios.

Order of addition: Another parameter influencing the par-ticle size of lipoplexes is the order of addition of DNAand lipids. Lack of attention to this preparation variablecan lead to extensive aggregation. In order to producethe smallest complexes at low cationic lipid/DNA chargeratios (,1), DMRIE:DOPE and DNA need to be mixedby adding the liposomes into the excess DNA (Figure 1b).Large aggregates were produced when the DNA wasadded into the lipid. When the complexes were preparedat a high cationic lipid/DNA charge ratio (.1), the opti-mal order of addition was reversed, and smaller com-plexes were produced when the DNA was added intothe excess cationic liposomes (Figure 1c). All vehicles andall cationic lipids tested obeyed these general rules (datanot shown). As mentioned previously, large aggregationresulted in lower transfection activity2 (data not shown).

The methods for reproducible preparation of physi-cally stable and homogenous lipoplex formulationsdescribed in this report were developed in recognition ofthese fundamental properties of lipoplex preparation.The continuous flow mixing method described hereavoids order of addition concerns by bringing the cat-ionic liposomes and DNA together at a constant ratioduring the entire mixing procedure. The mixing methodalso allows the rate of mixing to be precisely controlled.

Preparation of cationic liposomes with discrete sizesHand shaken or vortexed liposomes (large multilamellarvesicles, MLV) are heterogeneous and one would there-fore predict that lipoplexes formed from these cationicMLV liposomes might also be heterogeneous. The physi-cal properties and transfection activities of lipoplexes pre-pared from heterogeneous MLV were, therefore, com-pared with lipoplexes prepared from homogeneousextruded liposomes with well-defined sizes. However, inorder to accomplish this it was first necessary to demon-strate that it was possible to prepare cationic liposomeswith well-defined sizes.

The results in Figure 2 compare the particle size distri-butions of MLV DMRIE:DOPE liposomes with liposomesthat were extruded through either 400 or 100 nm mem-branes. The extruded liposomes were much more homo-geneous than the MLV and their diameter correspondedclosely to the pore diameter of the filter. Liposome prep-arations utilizing four different cationic lipids exhibitedsimilar behavior, and the same results were obtainedwith the three different vehicles, water, saline andsorbitol/sodium acetate buffer (S/A: 5% sorbitol/20 mm


1274

Figure 2 Particle size distribution of empty liposomes prepared in waterand produced by the extrusion method through polycarbonate membranes(400 or 100 nm) or produced by vortexing (MLV). The data show thepercentage of particles in the population for each size class as function ofthe particle diameter. Size distributions were the mean at each point basedon vesicle volume.

sodium acetate pH 6 buffer) (data not shown). The pro-duction of these liposomes was reproducible and no sig-nificant variability was observed between different inves-tigators.

A controlled mixing method to produce monodisperselipoplex formulationsSince the order of addition of DNA and cationic lipo-somes was shown to be critical for controling the sizeof the lipoplexes, we developed a method to mix bothcomponents simultaneously, at a fixed ratio and at a con-trolled speed. This novel method was designed to be sim-ple, scaleable and to minimize handling inconsistenciesamong different investigators.

Cationic liposomes and DNA were mixed togetherwith the aid of a syringe pump in order to control therate of mixing and obviate the need to control the orderof addition (Figure 3). DNA solution and cationic lipo-somes were loaded separately into tuberculin syringesconnected to an i.v. extension set. Then, the syringes wereplaced on the pump, and the plunger was adjusted onthe syringes to put the fluid level at the same positionfor both DNA and liposome preparations. The mixingtook place in the T-connector when the plunger of the

Figure 3 Schematic representation of the controlled mixing method. (1)Flow rate indicator; (2) flow rate control. (3) power switch: on, off andpause; (4) carriage of the syringe pump that pushes the plungers of thesyringes at the same time; (5) plungers of the syringes adjusted at thesame level; (6) two syringes loaded separately with the DNA solution andthe cationic liposome preparation; (7) i.v. extension set, luer-lock adaptersecured with four-way stopcock; (8) opening control (closed/open) of theT-connector; (9) T-connector site of mixing; (10) Receptacle, glass vial.

syringes was pushed simultaneously and at the samespeed (Figure 3).

The data in Figure 4 show the particle size distributionsof the DMRIE:DOPE/DNA complexes prepared with 400nm extruded liposomes at 0.25 molar ratio and at variousmixing rates. For the 400 nm extruded liposomes, theoptimal mixing rate was determined to be between 3.24and 8.1 ml/min (Figure 4a). At higher mixing rates aggre-gation was observed (Figure 4b and c). For the 100 nmextruded liposomes, the 15.7 ml/min rate was found tobe optimal for producing the smallest particle size distri-bution; at low (3.24 ml/min) and at high rates (.22.8ml/min) aggregation occurred (data not shown). Thus,the optimal mixing rate varied according to the startingliposome size. This semi-automated method, describedfor the first time, is a convenient way to mix cationic lipo-somes with DNA in a reproducible manner that elimin-ates the order of addition problem described earlier(Figure 1b and c). The mixing rate had no significant

Figure 4 Particle size distributions of DMRIE:DOPE/DNA complexesprepared from 400 nm extruded liposomes at 0.25 molar ratio and inwater. Concentration of DNA used was 1 mm. (a) 3.24 and 8.1 ml/min,(b) 15.7 ml/min; (c) 22.8 ml/min. The data show the percentage of particlesin the population for each class as a function of the particle size diameter.Size distributions were determined as mean of diameter based on vesiclevolume.


1275effect on the particle size distribution of the lipoplexesformed from MLV liposomes (data not shown). Thus, forall lipoplex formulations prepared by the controlled mix-ing method, the mixing rate used was 15.7 ml/min forthe 100 nm extruded and unextruded liposomes and 8.1ml/min for the 400 nm extruded liposomes.

The particle size distributions of the lipoplexes pro-duced from extruded liposomes and by the controlledmixing method were more homogeneous than thoseobtained from MLV (Figures 5a and 6). As the liposomesdecreased in size from MLV to 400 and 100 nm extrudedvesicles, the particle size distribution of the resulting lipo-plexes decreased as seen by the higher percentage of lipo-plexes in the small size range (, 200 nm). Indeed, withextruded liposomes, no lipoplexes were >1000 nmwhereas with the MLV a large percentage of the lipo-plexes exceed 1000 nm. In the same way, a small lipoplexpopulation (,200 nm) was observed only with theextruded liposomes. Moreover, the visual appearances ofthe formulations differed considerably. When MLV lipo-somes were used as a starting material, large particleswere always visible. Formulations prepared from 100 nmextruded liposomes were always clearer than formu-lations prepared with 400 nm extruded liposomes. It isimportant to appreciate that the dynamic laser light scat-tering instrumentation used here is not designed to accu-rately measure particles greater than about 1000 or 2000nm in diameter. Therefore, aggregated samples contain-ing visible particles are difficult to characterize by laserlight scattering because many of the particles in the

Figure 5 (a) Particle size distributions of DMRIE:DOPE/DNA com-plexes prepared from unextruded or extruded (100 or 400 nm) liposomesat 0.5 molar ratio and in S/A buffer. Concentration of DNA used was 1mm. (b) DMRIE:DOPE/DNA complexes prepared from 100 nm extrudedliposomes at 0.5 molar ratio in S/A buffer with different DNA concen-tration (0.01–1 mm). Size distributions were determined as mean of dia-meter based on vesicle volume.

Figure 6 Particle size distributions of DMRIE:DOPE/DNA complexesprepared from various liposomes at 0.25 molar ratio in saline buffer. Con-centration of DNA used was 1 mm. (a) MLV (unextruded liposomes); (b)extruded (400 nm) liposomes; (c) Extruded (100 nm) liposomes. The datashow the percentage of particles for each size class as function of the par-ticle size diameter. Size distributions were determined as mean of diameterbased on vesicle volume.

samples may be too large to be determined accuratelyby the instrument. However, the presence of these largeaggregates could be quantified with the filtration assay(see Figure 7).

The production of homogeneous and monodisperselipoplexes was DNA concentration dependent. By reduc-ing the quantity of DNA by 10- or 100-fold, a smallerlipoplex particle size distribution was obtained(Figure 5b). Low concentrations of DNA (0.1–0.5 mm) areacceptable for most in vitro transfection experiments, butfor in vivo studies and for clinical use, higher concen-trations (1–4 mm) are often necessary. A method for pro-ducing very small lipoplexes might involve producingthem under low concentrations and then concentratingthe final product. This approach to producing small par-ticle size lipoplexes at a high DNA concentrations is cur-rently being tested.

In addition, in order to be simple and scaleable, thisnovel method was designed to reduce handling inconsist-encies among different investigators. Thus, two differentinvestigators produced formulations of DMRIE:DOPE/DNA prepared at 0.25 molar ratio in saline. As show inFigure 6, the production of lipoplexes with extruded lipo-somes and the controlled mixing method was reproduc-ible and no significant variability was observed between


1276

Figure 7 Dependency of lipid and DNA recovery on filter pore diameter.Freshly prepared DMRIE:DOPE/DNA complexes, from unextruded orextruded (400 and 100 nm) liposomes, at 0.5 molar ratio in variousvehicles and under a controlled mixing rate, were unfiltered (a) or filteredthrough different pore size filter 600 nm (b) or 400 nm (c). Then, thepercentage of lipid recovered after filtration was determined as function ofthe liposomes starting material size and vehicle used. S/A, sorbitol/sodiumacetate buffer; H2O, distilled water.

different investigators (Figure 6b and c). In contrast,when MLV liposomes were used the lipoplex formu-lations differed significantly among preparations andbetween investigators (Figure 6a).

The methods for reproducible preparation of physi-cally stable and homogenous lipoplex formulationsdescribed in this report were developed in recognition ofthese fundamental properties of lipoplex preparation.The continuous flow mixing method described here avo-ids order of addition concerns by bringing the cationicliposomes and DNA together at a constant ratio during

the entire mixing procedure. The mixing method alsoallows the rate of mixing to be precisely controlled.

Physical characterization of lipoplexesThe stability of the lipoplexes formed from MLV orextruded liposome and by the controlled mixing method,was monitored by using analytical methods previouslydescribed.13 The filtration assay, which measures the per-centage of lipid and DNA recovered after filtration oflipoplexes through different pore size filters, was shownto be a sensitive way to quantitatively assess changes inparticle size and heterogeneity, and provided infor-mation on the extent of aggregation of the formulations.When samples contained large particles, the aggregatedproduct was retained on the filter and a correspondingreductions in DNA and lipid recoveries were determined.These assays also provided information about whetherthe formulations could be easily handled without loss ofmaterial due to adsorption. Without an analytical assay,this adsorption problem could easily have goneunnoticed and could have a negative impact on repro-ducibility.

Unextruded or extruded liposomes were mixed withDNA at 0.5 molar ratio and at the appropriate mixingrate. The resulting lipoplexes were filtered through 1000nm (data not shown), 600 nm or 400 nm membranes(Figure 7b and c), and the lipid and DNA recoveriesbefore and after filtration were determined. Without fil-tration, all the formulations exhibited comparable lipid(Figure 7a) and DNA recoveries (data not shown)approaching 100% of theoretical. The 600 nm filtration oflipoplexes prepared from MLV removed 40–80% of thelipid depending on the vehicle (Figure 7b). The S/Avehicle (5% sorbitol/20 mm sodium acetate pH 6 buffer)gave the highest recovery and saline gave the lowest. Theproduct recovery after filtration from lipoplexes preparedwith the extruded liposomes was greater than from lipo-plexes prepared with MLVs. This indicated that lipo-plexes prepared from extruded liposomes contained lessaggregated material than those produced from unex-truded liposomes. For MLVs prepared in saline, less than20% of lipid was recovered after 600 nm filtrationwhereas for the extruded liposomes 75% was recovered(Figure 7b).

The amount of lipid recovered after filtration was alsoaffected by the composition of the suspending vehicle.The lipid recoveries after filtration were greater when thelipoplexes were formulated in S/A buffer than when pre-pared in either water or saline (Figure 7b and c). Thisresult suggested that the lipoplex formulations preparedin S/A were more uniform and contained fewer largeparticles. The differences in homogeneity between MLVand extruded liposomes prepared in S/A, were apparentafter filtration through 600 nm membranes. Indeed,.95% of lipid was recovered with extruded liposomes,whereas MLV exhibited less than 65% recovery(Figure 7b). These differences were brought out moredramatically when the lipoplexes were filtered through a400 nm membrane (Figure 7c). Similar conclusions weremade for formulations prepared at 0.1 and 0.25DMRIE:DOPE/DNA molar ratios (data not shown). Thepercentage of DNA recovered was also monitored andthe results obtained were qualitatively similar (data notshown). However, since approximately 50% of the DNAis free in these formulations,13 the differences for DNA


1277recovery between the formulations were less pronouncedthan those with the lipid assay. Under all conditionsexamined, the formulations prepared by the controlledmixing method and with extruded liposomes, showedhigher recoveries (lipids and DNA) than lipoplexes pre-pared from unextruded liposomes. The filtration assay incombination with dynamic light scattering demonstratedthe advantages of using extruded liposomes and con-trolled mixing to produce monodisperse lipoplexes.

We then investigated which components of the S/Abuffer were responsible for producing these uniform andsmaller particles by monitoring lipid recoveries after fil-tration through 400 nm membranes (Figure 7c). When lip-ids and DNA were mixed in water or sorbitol vehicles,the resulting lipoplexes could not pass through the 400nm filters and all the lipids were lost. Sodium acetate wassufficient to produce a formulation that could be filteredthrough a small pore membrane. The results with 20 mmsodium acetate alone were comparable to those seen inthe complete S/A vehicle (Figure 7c). In summary, theseassays have demonstrated that lipoplexes produced fromextruded liposomes under controlled mixing conditionswere markedly more stable and less aggregated thanthose produce from MLV liposomes.

In vitro transfection efficiency of new lipoplexesThe production of stable and monodisperse lipoplex for-mulations is beneficial only if the biological activity canalso be retained. In this context, we tested the in vitrotransfection efficiency of lipoplexes formed from unex-truded or extruded liposomes using the controlled mix-ing method. The formulations were prepared in three dif-ferent vehicles (water, saline, S/A) and at 0.25 and 0.5DMRIE:DOPE/DNA molar ratios. The preparationsmade with 400 nm extruded liposomes were as active asMLV at the 0.5 molar ratio but less active than MLVwhen formulated at the 0.25 molar ratio (Figure 8). Theformulations prepared with 100 nm extruded liposomeswere the least active. Thus, lipoplexes prepared withlarger liposomes were more active than those preparedfrom smaller liposomes. It is difficult, however, to con-clude from these in vitro experiments which lipoplexeswill be the most active in vivo, and these studies areunderway. The relatively low activity of the smaller lipo-plexes produced from 100 nm extruded liposomes couldbe due to a sedimentation problem as suggested,18 sincethe larger DMRIE:DOPE/DNA complexes may facilitatecell surface contact by sedimentation.

The transfection efficiency of these new formulationswas also tested on different cell types (data not shown).The relative transfection activity rankings among the dif-ferent formulations were the same in three other cell lines(L293, UM449 and B16F10). In all of these cell lines, lipo-plexes formed from 100 nm extruded liposomes were lessactive than lipoplexes prepared from 400 nm extrudedliposomes or MLV (data not shown).

Production of frozen single vial formulations: stabilityand efficacyThe possibility of producing a physically stable, isotonic,single vial DMRIE:DOPE/DNA formulation was investi-gated. First, all formulations prepared by the controlledmixing method and tested previously (Figure 8) werefrozen quickly in ethanol/dry ice and then stored at−70°C. The results in Figure 9 show the in vitro transfec-

Figure 8 In vitro transfection activity of DMRIE:DOPE/DNA lipoplexesfreshly prepared. Preparation of lipoplexes from unextruded or extruded(400 or 100 nm) liposomes, at various positive/negative molar ratios, inthree different vehicles and transfection protocol were carried out asdescribed in Materials and methods. The data show the percentage of b-galactosidase expression in Renca cells as a function of the molar ratioand liposome used as the starting material. (a) Water; (b) saline; (c) S/A.Data are the mean ± standard error of two separate experiments.

tion activities of lipoplexes before (Figure 9a) and afterfreeze/thaw (Figure 9b). The data indicated that all lipo-plex formulations prepared in S/A buffer could be frozenand thawed without losing significant in vitro transfectionactivity. In contrast, the same formulations prepared ineither water or isotonic saline lost in vitro transfectionactivity after being frozen and thawed. All lipoplexes for-mulated without sorbitol (MLV or extruded) lost morethan 50% of their transfection activity (data not shown).The 20 mm sodium acetate was previously shown to berequired for optimal physical stability (Figure 8c). Thus,both the low ionic strength sodium acetate buffer andsorbitol were required to produce a stable single vial for-mulation that could be stored frozen.

The physical stability of frozen and thawed lipoplexformulations was examined. Lipoplexes prepared at 0.5DMRIE:DOPE/DNA molar ratio were frozen, stored at−70°C for 1 month, thawed and assayed for lipid andDNA recoveries before and after 600 nm filtration(Figure 10). Slight losses in lipid recovery from the unfil-tered saline formulations and from the MLV formulation


1278

Figure 9 In vitro transfection activity of DMRIE:DOPE/DNA lipoplexesfrozen and thawed in Renca cells. Preparation of lipoplexes from unex-truded or extruded (400 or 100 nm) liposomes, at 0.5 molar ratio, in threedifferent vehicles, and transfection protocol were carried out as describedin Materials and methods. (a) Freshly prepared lipoplexes; (b) frozen andthawed lipoplexes. Data are the mean ± standard error of two separateexperiments.

Figure 10 Percentage of lipid and DNA recovery after filtration through various pore sizes of frozen and thawed lipoplexes. DMRIE:DOPE/DNAcomplexes were prepared, from unextruded or extruded (400 and 100 nm) liposomes, at 0.5 molar ratio in various vehicles and under a controlledmixing rate, according to the procedure described in Materials and methods. After being thawed, lipoplexes were unfiltered (a and c) or filtered througha pore size filter of 600 nm (b and d). Then, percentage of lipid and DNA recovered after filtration was determined as a function of the liposomesstarting material size and vehicle used. (a–b) Percentage of lipid recovered before or after filtration; (c–d) percentage of DNA recovered before orafter filtration.

prepared in water were noted (Figure 10a). In contrast,the saline and water formulations were almost com-pletely depleted of lipids by the 600 nm filtration pro-cedure (Figure 10b). This result can be compared with theresults in Figure 6b which showed much higher lipidrecovery following 600 nm filtration of freshly preparedlipoplexes. However, the 600 nm filtration assays indi-cated that all of the S/A formulations were essentiallyunchanged after freezing and thawing (Figure 10b).Indeed, similar results were obtained for lipid recoveryfrom the freshly prepared lipoplexes in S/A buffer(Figure 7b). DNA recoveries revealed similar trends inthe abilities of the formulations to withstand freeze/thaw(Figure 10c and d). Filtration of lipoplexes prepared usingextruded liposomes showed .90% DNA recoveries afterfreeze/thaw in S/A buffer, but not in water and saline.These results indicated that formulations prepared ineither saline or water aggregated after freeze/thaw andwere, therefore, not suitable for use as frozen formu-lations. In contrast, the S/A buffer stabilized the formu-lations against freezing and thawing.

Dynamic laser light scattering measurements of thefrozen and thawed lipoplexes complemented the fil-tration data. Lipoplexes prepared in S/A buffer alsoshowed no significant differences between any of thefreshly prepared and frozen/thawed S/A formulations(Figures 11c and 2). In contrast, the water and saline for-mulations exhibited evidence of aggregation followingfreeze/thaw (Figure 11a and b). As mentioned pre-viously, largely aggregated samples are difficult tocharacterize by laser light scattering quantitatively. How-ever, the presence of large aggregates has been identified


1279

Figure 11 Particle size distributions of DMRIE:DOPE/DNA formu-lations after freeze and thaw. Lipoplexes were prepared from unextrudedor extruded (100 or 400 nm) liposomes at 0.5 molar ratio and variousbuffer. Concentration of DNA used was 1 mm. (a) Water; (b) saline; (c)sorbitol/sodium acetate buffer. Size distributions were determined as meanof diameter based on vesicle volume.

with the filtration assay (Figure 10). These results werealso in agreement with the biological activity of thefrozen and thawed lipoplexes (Figure 9). Taken togetherthese results suggested that the large aggregates pro-duced after freezing and thawing of the lipoplexes werenot biologically active in the in vitro transfection assay.

A study was initiated to determine the feasibility ofusing the S/A vehicle for long-term storage of lipoplexes.In vitro transfection activity of a formulation preparedwith MLV liposomes, frozen and thawed after 1 week, 1,3 or 12 months, exhibited similar in vitro transfectionactivity compared with a freshly prepared lipoplex for-mulation (Figure 12). Fast freezing in a dry ice/alcoholbath or −70°C freezer was preferred to a formulation thatwas frozen slowly at −20°C (data not shown), but oncefrozen, the samples could be stored at −20°C. The stab-ility of the individual compounds (DNA and lipids) wasanalyzed after 1-year storage at −20°C. As reflected bythe biological activity of these frozen and thawed formu-

Figure 12 In vitro transfection activity of long term storage lipoplex for-mulations. DMRIE:DOPE/DNA lipoplexes were prepared from unex-truded liposomes, at 0.1 molar ratio, in 5% sorbitol/20 mm sodium acetatepH 6.0 and freeze quickly at −70°C as described in Materials and methods.At indicated period of storage at −20°C, samples were thawed at roomtemperature and tested for their transfection activity on CV-1 cells. Theirtransfection efficiency was compared with freshly prepared lipoplexes. Thedata show the percentage of b-galactosidase expression remaining afterstorage as a function of time of storage. The percentage of b-galactosidaseexpression remaining represent the sum of pg b-galactosidase of thefreeze/thaw formulation relative to the sum of pg b-galactosidase of freshlyprepared lipoplexes.

lations (Figure 12), the DNA was still intact as analyzedby agarose gel electrophoresis (data not shown). Analysisof the lipid components by thin layer chromatographyrevealed .90% integrity of both the DMRIE and theDOPE (data not shown). The utility of the S/A buffer forthe long-term storage of the new lipoplex formulations iscurrently underway, as well as more detailed studies onDNA and lipid stability analyzed by HPLC,21 TLC andNMR.

DiscussionThese studies were intended to identify a simple and sca-leable method to reproducibly prepare stable, monodis-perse lipoplexes and to prepare a physically stable, sin-gle-vial, DMRIE:DOPE/DNA formulation that retainedbiological activity after freezing and thawing. The needto carefully optimize these methods was motivated bythe observation that better transfection activity wasobtained under conditions in which the complexesbecame increasingly less physically stable. Since hetero-geneous and aggregated materials are difficult to controland can lead to irreproducible results, these physicalstability issues are important from basic research, pro-duct development and clinical application perspectives.

In order to reproducibly prepare well characterized,homogeneous, and stable formulations with high trans-fection activity, a liposome extrusion method and a con-trolled mixing method were developed. These methodsallowed DMRIE formulations to be prepared at higherDNA concentrations (>330 mg/ml). The methods alsobypassed fundamental technical problems associatedwith controlling the mixing rate and order of addition ofthe cationic liposomes and DNA during complex forma-tion. The physical stability and in vitro transfection activi-ties of DMRIE:DOPE/DNA formulations prepared in thisway were compared. The use of MLV (prepared by vor-


1280 tex mixing) led to the production of large and hetero-geneous complexes with high in vitro transfection activitybut with poor physical stability (based on visual appear-ance, laser light scattering and lipid recovery afterfiltration). The use of extruded liposomes led to the pro-duction of more homogeneous, less aggregated com-plexes. The formulations prepared with 400 nm extrudedliposomes were as active as vortexed liposomes and moreactive than those prepared with 100 nm extruded lipo-somes. However, formulations prepared from 100 nmextruded liposomes were very physically stable, had alow tendency to aggregate and produced a transparentsolution that was easy to handle. Formulations preparedin S/A buffer were highly active in vitro and were morephysically stable than formulations prepared in eitherwater or saline. Recovery from the S/A vehicle wasgreater than from the saline or water vehicles indicatingthat the S/A formulations contained smaller and moreuniform particles. This effect of S/A buffer on physicalstability could be due to the controlled pH, the low ionicstrength, the monovalent buffer species, the sugar speciesor a combination of these. Under all conditions examined,the formulations prepared with extruded liposomesshowed better recoveries and were more homogeneousthan formulations prepared with MLV.

From the biological data reported here it appeared thatlarger (400–800 nm) lipoplexes were more active in invitro transfection experiments than the smaller ones (100–400 nm). This observation for DMRIE:DOPE has beenpreviously reported,17 and a recent study comparing thein vivo activity of lipoplexes produced from either MLVor SUV (small unilamellar vesicles) reached a similar con-clusion.22 However, for in vivo applications, the benefitsof using small versus large lipoplexes, may depend on theroute of administration, as well as on the intended cellor tissue targeted. Indeed, as reported by Liu et al22 theadvantage of using large lipoplexes for intravenousadministration seems to be when the lung is the primarytarget. In other biological models (tumor, spleen, liver orhematopoietic cells), the relative benefits of larger andsmaller lipoplexes need further investigation. It is alsoobvious from the freeze/thaw transfection experimentsthat very large lipoplexes produced after freeze/thaw insaline or water are not desirable since they lose most oftheir in vitro transfection activity. Nevertheless, it hasbeen shown by Templeton et al23 that the production ofhomogeneous lipoplex formulations by the extrusionmethod leads to improved systemic delivery.

A physically stable, single vial, frozen formulation wasidentified. The low ionic strength formulations (S/Abuffer) which contained a neutral non-reducing sugar,retained the biological activity and physical character-istics of the lipoplexes after freeze/thaw. One S/A for-mulation was stable for more than 1 year at −20°C. Incontrast, saline and water formulations were physicallyunstable to freeze/thaw. The importance of producingstable lipoplex formulations for gene delivery is becom-ing widely appreciated and several recent publicationshave claimed that stable lipoplexes can be produced byaddition of ‘helper’ compounds such as spermidine,polyethyleneglycol14 or surfactants.15,16 These papersdescribed lipoplexes that were formed in the presence ofexcess positive charge and the lipoplexes were preparedat lower DNA concentrations (<200 mg/ml) than thoseused in this study (from 330 mg/ml to 1.3 mg/ml). More-

over, previous studies have not thoroughly analyzed thephysical stability of the lipoplexes, and most of thesestudies have looked at relatively short-term storage (<3months). All these studies have emphasized the use ofadditional compounds, which were claimed to berequired to get stable formulations.14–16 The addition ofthese extra components increases the complexity of theformulations and accordingly makes their characteriz-ation, and manufacturing more difficult to control. Incontrast, our method demonstrates that DNA and lipidscan be mixed together to form a stable single vial formu-lation without the addition of other components and witha very simple and scaleable two-step procedure.

The reproducible production of well-characterizedlipoplex preparations is also critical for the understand-ing of the molecular, biophysical and biological mech-anisms of lipoplex formation and action. Recent studieswith oligonucleotides24,25 or plasmid DNA26–28 haveshown that the mechanism of cationic lipid-mediatednucleic acid delivery can be studied at the cellular leveland that the molecular and cellular barriers can becharacterized. The preparation of well-characterized,stable monodisperse lipoplexes will help to identifywhich structural features of lipoplexes are optimal forefficient DNA delivery and will aid in elucidating struc-ture–function relationships. This will also permit a morerational approach to further improving the efficiency ofthese self-assembled delivery systems.

In summary, the results of these studies provide a plat-form for characterizing and comparing the behavior ofsynthetic gene delivery systems. In addition, a physicallystable (over 1 year), isotonic, single vial DMRIE:DOPE/DNA formulation was described. Formulations of thiskind may provide more reproducible in vivo data andmay also be useful as practical clinical products.

Materials and methods

ReagentsCationic lipid 1,2-dimyristyloxypropyl-3-dimethyl-hyd-roxyethyl ammonium bromide (DMRIE) was synthesizedas described.17 Dioleoyl-phosphatidylethanolamine(DOPE) was purchased from Avanti Polar lipids(Alabaster, AL, USA). b-Galactosidase standard (gradeVIII from E. coli) and NaCl were purchased from Sigma(St Louis, MO, USA). Chlorophenol red pyranogalacto-side (CPRG) was obtained from Boehringer Mannheim(Indianapolis, IN, USA). All chemicals are reagent grade:Ultra pure agarose (Gibco BRL, Gaithersburg, MD, USA)Tris-HCl and EDTA (Fisher, Irvine, CA, USA). Sorbitoland sodium acetate were obtained from Aldrich. 0.9%Saline solution was purchased from Radix (Eau Claire,WI, USA). Distilled water (cell culture grade and endo-toxin screened) was purchased from Gibco BRL.

Cell cultureRenca (mouse renal carcinoma) cells were a generous giftfrom Dr Drew Pardoll at the Johns Hopkins University.B16F10 (mouse melanoma, CRL 6322) and CV-1 (AfricanGreen monkey kidney, CCL 70) were obtained fromAmerican Type Tissue Culture Collection (ATCC, Rock-ville, MD, USA). Mark Cameron and Dr Gary Nabelkindly provided UM449 (human melanoma) and L293cells (human embryonic kidney), respectively at the Uni-


1281versity of Michigan. All cells were grown in Dulbecco’sminimum essential medium containing 10% of fetal bov-ine serum and antibiotics. OptiMEM media (Gibco BRL)were used during the transfection procedure. Cells weremaintained in a 5–7% CO2 atmosphere at 37°C.

Plasmid DNAPlasmid DNA (VR1412) coding for b-galactosidase wasconstructed as described.29 DNA was produced and pur-ified according to Horn et al.30

Preparation of liposomesThe DMRIE:DOPE (1:1 molar ratio) lipid film was pre-pared as reported.17 For the production of MLV, the driedlipid film was rehydrated by adding the appropriatevehicle (ie distilled water, 0.9% saline or 5% sorbitol/20mm sodium acetate pH 6 buffer) in order to get a lipidconcentration of 1–4 m. Then, vials were vortexed con-tinuously for 2 min at the highest setting using a foamplatform attached to a Genie-2 vortexer (Fisher), to pro-duce the MLV. For the preparation of extruded cationicliposomes (large or small unilamellar vesicles, LUV orSUV), the MLV achieved in the previous step, were sub-jected to one freeze–thaw cycle before extrusion. Theliposomes produced were passed through an ‘extruder’(LipoFast; Avestin, Ottawa, Canada) mounted with a 100or 400 nm polycarbonate track-etched membrane (PCTEfilter) (Osmonics, Livermore, CA, USA) at room tempera-ture. The liposomes were extruded 19 times back andforth according to the manufacturer’s instruction. For the400 nm extruded liposomes, the extrusion was performedwith the aid of the ATI-Orion Sage syringe pump (CoastScientific, La Jolla, CA, USA) at 100% flow rate and notmanually (see below). Size determinations performed forother liposome formulations showed that liposomes for-med by this technique are primarily unilamellar.31

Preparation of lipoplexes

Classical preparation of lipoplexes: One volume of MLVcationic liposomes prepared in various vehicles wasmixed with one volume of DNA diluted in the samevehicle as the liposomes. Order of addition depends onthe charge ratio used (see Results). DNA and cationicliposomes were mixed together at the desiredDMRIE:DOPE/DNA molar ratio and used within 2 hafter mixing. The nomenclature used for the lipoplexmolar ratio corresponds to the number of mole of cationiclipids (DMRIE) per mole of DNA phosphate.

Controlled mixing method to prepare new lipoplex formu-lations: One volume of cationic liposomes and one vol-ume of DNA are mixed together with the use of the ATI-Orion sage syringe pump, in order to control the speedof mixing and eliminate the concern for the order ofaddition (Figure 3). DNA solution and cationic liposomeswere loaded into separate 1 ml tuberculin syringes whichwere then connected to an i.v. extension set equippedwith a luer-lock adapter (6 inch) secured with a four-waystopcock/T-connector (Baxter, Chicago, IL, USA). Then,the DNA and liposomes were loaded independently(adjust the openings of the T-connector correspondingly)into the i.v. extension. With the port closed, the tubercu-lin syringe was exchanged with a 3, 5 or 10 ml syringeleaving a head space between the plunger and the sol-

ution. The syringe was placed on to the cradle of thepump, secured with the metal arm, and the plunger wasadjusted on the syringes to put the fluid level at the sameposition for both DNA and liposome preparations. Themixing occurs when the carriage on the syringe pumppushes the plunger of the syringes simultaneously andat the same speed. The flow rate can be controlled eitherby using different size syringes and/or by adjusting thepercentage flow dial on the syringe pump.

Analytical assays to characterize the lipoplexThe percentage of DNA and lipid recovery, the determi-nation of the lipoplex physical stability by filtration, theamount of free DNA in the formulations and the nucleasesensitivity assay were determined as described.13 For thefiltration assay, 600 ml of lipoplex formulations or lipo-some control were filtered once (nonsequentially)through 0.4, 0.6 or 1 mm PCTE membranes (Osmonics)using a 13-mm stainless steel filter housing (FisherScientific). The housing was cleaned after each sample byflushing 10 ml methanol followed by 10 ml MilliQ waterthrough the filter housing and components and dryingusing filtered air. Total lipid and total DNA assays(fluorescamine and picogreen, respectively) were perfor-med on filtered samples as described.13

The DNA stability of the frozen lipoplexes was ana-lyzed by agarose gel electrophoresis. After being store at−20°C for various times (from 1 week to 1 year), the DNArecovered from the lipoplex formulations was loaded on0.8% agarose-0.1% SDS gel containing 0.5 mg/ml of ethid-ium bromide to visualize the DNA using a tris-acetate/EDTA buffer, pH 8. The lipid stability of thefrozen lipoplexes was analyzed by thin layer chromatog-raphy (silica gel plate) with 75% chloroform–25% meth-anol as the mobile phase followed by extraction.

Particle size was determined by laser light scattering.All dynamic light scattering measurements were perfor-med using a Malvern Zetasizer 4 (Malvern, PA, USA)equipped with a AZ104 cell at a scattering angle of 90°.The data were analyzed using the software supplied bythe manufacturer in the Contin analysis mode. Sampleswere prepared by mixing the appropriate volume of theformulation or liposome control with 1.5 ml of the diluentand briefly vortexing in a 15-ml conical tube. For the lipo-plexes produced by the classical method (Figure 1), thelipid concentration ranged from 19 to 37.5 mm and theDNA concentration varied from 1 to 50 mg/ml (3 to150 mm) depending on the charge ratio used. The coun-ting time for each sample was 180 s and the temperaturewas set at 25°C. The sample was diluted to give a 50–800(×1000) counts per second. For instrument standardiz-ation, latex nanosphere size standards (Duke Scientific,Palo Alto, CA, USA) of 96, 200, 300 and 400 nm weretested. One hundred microliters of standard was mixedwith 2.9 ml of water in a 15 ml conical tube, and furtherdiluted 10-fold for measurement.

In vitro transfection assayLipoplexes were prepared as described above at variousmolar ratios and in the three different vehicles. Theywere tested for their in vitro transfection efficiency in 96-well plates according to previously published protocols17

with the exception that the serial dilution into 96-wellplates of the DMRIE:DOPE/DNA complexes was doneas follows: 75 ml of the vehicle (water, 0.9% saline or 5%


1282 sorbitol/20 mm NaOAc pH 6) in which the formulationwas prepared, was distributed into each well of an empty96-well plate, then 75 ml of the lipoplexes were trans-ferred to the first well of each column 1 and 2× serialdilution was performed from column 1 to 12. Then, 75 mlof OptiMEM was dispensed into each well and 100 ml ofthis mixture was added to the cells. The sum of the pgb-galactosidase values obtained with the different formu-lations from the entire dilution series of the 96-well plateand expressed as the percentage of the pg sum value rela-tive to the pg sum value obtained with the MLV preparedat 0.5 molar ratio and in the respective vehicle.

AcknowledgementsWe would like to thank Susan Zimmerman for the cellculture technical assistance. We are also very grateful toVical production facilities for DNA and lipids supplies.We acknowledge Dr Suezanne Parker for helpful dis-cussion and comments.

References1 Gao X, Huang L. Cationic liposome-mediated gene transfer.

Gene Therapy 1995; 2: 710–722.2 Felgner PL et al. Improved cationic lipid formulations for in vivo

gene therapy. Ann NY Acad Sci 1995; 772: 126–139.3 Zelphati O, Szoka FC Jr. Liposomes as a carrier for intracellular

delivery of antisense oligonucleotides: a real or magic bullet? JControl Rel 1996; 41: 99–119.

4 Caplen NJ et al. Liposome-mediated CFTR gene transfer to thenasal epithelium of patients with cystic fibrosis. Nature Med1995; 1: 39–46.

5 Nabel GJ et al. Immune response in human melanoma aftertransfer of an allogeneic class I major histocompatibility com-plex gene with DNA–liposome complexes. Proc Natl Acad SciUSA 1996; 93: 15388–15393.

6 Rubin J et al. Phase I study of immunotherapy of hepatic metast-ases of colorectal carcinoma by direct gene transfer of an allog-enic histocompatibility antigen, HLA-B7. Gene Therapy 1997; 4:419–425.

7 Gershon H, Ghirlando R, Guttman SB, Minsky A. Mode of for-mation and structural features of DNA–cationic liposome com-plexes used for transfection. Biochemistry 1993; 32: 7143–7151.

8 Sternberg B, Sorgi FL, Huang L. New structures in complex for-mation between DNA and cationic liposomes visualized byfreeze–fracture electron microscopy. FEBS Lett 1994; 356: 361–366.

9 Lasic DD et al. The structure of DNA–liposome complexes. J AmChem Soc 1997; 119: 832–833.

10 Eastman SJ et al. Biophysical characterization of cationiclipid:DNA complexes. Biochim Biophys Acta 1997; 1325: 41–62.

11 Nabel GJ et al. Immunotherapy for cancer by direct gene transferinto tumors. Hum Gene Ther 1994; 5: 57–77.

12 Nabel EG et al. Safety and toxicity of catheter gene delivery tothe pulmonary vasculature in a patient with metastatic mela-noma. Hum Gene Ther 1994; 5: 1089–1094.

13 Ferrari ME et al. Analytical methods for the characterization ofcationic-lipid–nucleic acid complexes. Hum Gene Ther 1998; 9:341–351.

14 Hong K, Zheng W, Baker A, Papahadjopoulos D. Stabilizationof cationic liposome–plasmid DNA complexes by polyaminesand poly(ethylene glycol)-phospholipid conjugates for efficientin vivo gene delivery. FEBS Lett 1997; 400: 233–237.

15 Hofland HE, Shephard L, Sullivan SM. Formation of stable cat-ionic lipid/DNA complexes for gene transfer. Proc Natl Acad SciUSA 1996; 93: 7305–7309.

16 Liu F, Yang J, Huang L, Liu D. New cationic lipid formulationsfor gene transfer. Pharm Res 1996; 13: 1856–1860.

17 Felgner JH et al. Enhanced gene delivery and mechanism studieswith a novel series of cationic lipid formulations. J Biol Chem1994; 269: 2550–2561.

18 Boussif O, Zanta MA, Behr JP. Optimized galenics improve invitro gene transfer with cationic molecules up to 1000-fold. GeneTherapy 1996; 3: 1074–1080.

19 Van der Woude I et al. Parameters influencing the introductionof plasmid DNA into cells by the use of synthetic amphiphilesas a carrier system. Biochim Biophys Acta 1995; 1240: 34–40.

20 Wheeler CJ et al. Converting an alcohol to an amine in a cationiclipid dramatically alters the co-lipid requirement, cellular trans-fection activity and the ultrastructure of DNA–cytofectin com-plexes. Biochim Biophys Acta 1996; 1280: 1–11.

21 Felgner JH. Separation and quantification of cationic liposomecomponents by high performance liquid chromatography withevaporative light-scattering detection. Pharm Res 1997; 14:1269–1271.

22 Liu Y et al. Factors influencing the efficiency of cationic lipo-some-mediated intravenous gene delivery. Nature Biotechnol1997; 15: 167–173.

23 Templeton NS et al. Improved DNA:liposome complexes forincreased systemic delivery and gene expression. Nature Biotech-nol 1997; 15: 647–652.

24 Zelphati O, Szoka FC Jr. Intracellular distribution and mech-anism of delivery of oligonucleotides mediated by cationic lip-ids. Pharm Res 1996; 13: 1367–1372.

25 Zelphati O, Szoka FC jr. Mechanism of oligonucleotide releasefrom cationic liposomes. Proc Natl Acad Sci USA 1996; 93:11493–11498.

26 Zhou X, Huang L. DNA transfection mediated by cationic lipo-somes containing lipopolylysine: characterization and mech-anism of action. Biochim Biophys Acta 1994; 1189: 195–203.

27 Zabner J et al. Cellular and molecular barriers to gene transferby a cationic lipid. J Biol Chem 1995; 270: 18997–9007.

28 Xu Y, Szoka FC jr. Mechanism of DNA release from cationicliposome/DNA complexes used in cell transfection. Biochemistry1996; 35: 5616–5623.

29 Wheeler CJ et al. A novel cationic lipid greatly enhances plasmidDNA delivery and expression in mouse lung. Proc Natl Acad SciUSA 1996; 93: 11454–11459.

30 Horn NA, Meek JA, Budahazi G, Marquet M. Cancer gene ther-apy using plasmid DNA: purification of DNA for human clinicaltrials. Hum Gene Ther 1995; 6: 565–573.

31 Hope MJ, Bally MB, Webb G, Cullis PR. Production of largeunilamellar vesicles by a rapid extrusion procedure. Charac-terization of size distribution, trapped volume and ability tomaintain a membrane potential. Biochim Biophys Acta 1985; 812:55–65.

Documents

Stable and monodisperse lipoplex formulations for gene delivery