Doxorubicin physical state in solution and inside ...bigbro.biophys.cornell.edu/publications/139 Li...

Doxorubicin physical state in solution and inside liposomes loadedvia a pH gradient

Xingong Li a, Donald J. Hirsh a, Donna Cabral-Lilly a, Achim Zirkel b,Sol M. Gruner b, Andrew S. Jano¡ a, Walter R. Perkins a;*

a The Liposome Company, Inc., One Research Way, Princeton, NJ 08540, USAb Department of Physics, Cornell University, Ithaca, NY 14853, USA

Received 26 May 1998; accepted 9 September 1998

Abstract

We have examined doxorubicin's (DOX) physical state in solution and inside EPC/cholesterol liposomes that were loadedvia a transmembrane pH gradient. Using cryogenic electron microscopy (cryo-EM) we noted that DOX loaded to 200^300mM internal concentrations in citrate containing liposomes formed linear, curved, and circular bundles of fibers with nosignificant interaction/perturbation of the vesicle membrane. The individual DOX fibers are putatively comprised of stackedDOX molecules. From end-on views of bundles of fibers it appeared that they are aligned longitudinally in a hexagonal arraywith a separation between fibers of approx. 3^3.5 nm. Two distinct small angle X-ray diffraction patterns (oblique and simplehexagonal) were observed for DOX-citrate fiber aggregates that had been concentrated from solution at either pH 4 or 5. Thedoxorubicin fibers were also present in citrate liposomes loaded with only one-tenth the amount of doxorubicin used above(approx. 20 mM internal DOX concentration) indicating that the threshold concentration at which these structures form isrelatively low. In fact, from cryo-EM and circular dichroism spectra, we estimate that the DOX-citrate fiber bundles canaccount for the vast majority (s 99%) of DOX loaded via a pH gradient into citrate buffered liposomes. DOX loaded intoliposomes containing lactobionic acid (LBA), a monoanionic buffer to control the internal pH, remained disaggregated atinternal DOX concentrations of approx. 20 mM but formed uncondensed fibers (no bundles) when the internal DOXconcentration was approx. 200 mM. This finding suggests that in the citrate containing liposomes the citrate multianionelectrostatically bridged adjacent fibers to form the observed bundles. 13C-NMR measurements of [1,5-13C]citrate insideliposomes suggested that citrate `bound' to the DOX complex and `free' citrate rapidly exchange indicating that the citrate-DOX interaction is quite dynamic. DOX release into buffer was relatively slow (6 4% at 1 h) from liposomes containingDOX fibers (in citrate loaded to a low or high DOX concentration or in LBA liposomes loaded to a high internal DOXconcentration). LBA containing liposomes loaded with disaggregated DOX, where the internal DOX concentration was onlyapprox. 20 mM, experienced an osmotic stress induced vesicle rupture with as much as 18% DOX leakage in less than 10 min.The possible implications for this in vivo are discussed. ß 1998 Elsevier Science B.V. All rights reserved.

Keywords: Adriamycin; Remote loading; Cryo-electron microscopy; TLC D-99; Evacet

0005-2736 / 98 / $ ^ see front matter ß 1998 Elsevier Science B.V. All rights reserved.PII: S 0 0 0 5 - 2 7 3 6 ( 9 8 ) 0 0 1 7 5 - 8

Abbreviations: CD, circular dichroism; cryo-EM, cryogenic electron microscopy; DOX, doxorubicin; EPC, egg phosphatidylcholine;HBS, HEPES bu¡ered saline; LBA, lactobionic acid; LUV, large unilamellar vesicle ; MLV, multilamellar vesicle ; NMR, nuclearmagnetic resonance; SAXS, small angle X-ray di¡raction

* Corresponding author. Fax: (609) 520-8250; E-mail : [email protected]

BBAMEM 77484 1-12-98 Cyaan Magenta Geel Zwart

Biochimica et Biophysica Acta 1415 (1998) 23^40

1. Introduction

The major limitation of doxorubicin (DOX), oneof the most widely used anticancer agents, has beenits cumulative dose related cardiotoxicity [1]. Lipo-some encapsulation has been shown to reduce thecardiotoxicity of anthracyclines in animal models[2^5] and one formulation of liposomal DOX (Eva-cet, formerly TLC D-99) has been shown to be lesscardiotoxic in humans [6]. In this formulation, en-capsulation has been made highly e¤cient (s 95%)by taking advantage of DOX's accumulation to theliposome interior in response to an inside-acidictransmembrane pH gradient. The encapsulationprocedure has been referred to as remote loading[7^9].

Transmembrane pH gradients can be created di-rectly, as in the above case, by forming the liposomesin a well-bu¡ered solution of low pH (e.g., 300 mMcitrate at pH 4) and then adding a more basic sol-ution to raise the external solution pH [7]. pH gra-dients can also be created indirectly by an electricalpotential which drives protons to the vesicle interior[8] or from the electroneutral outward £ow of thecounterion to an entrapped acid (e.g., ammoniumsulfate [9]). With a pH gradient established, DOXaccumulates in the vesicle interior and the ideal dis-tribution of DOX in solution inside (Din) and outside(Dout) the liposomes is expected to be related to theinner and outer H� concentrations by

Din

Dout� V in

Vout

K � �H��inK � �H��out

� ��1�

where Vin and Vout are the internal and externalaqueous volumes and K is the dissociation constantfor DOX (pK = 8.22 [10]). However, we and others[11] have found that Din/Dout can exceed the valuepredicted by Eq. 1. One explanation for this en-hanced accumulation is that internalized DOX doesnot remain in solution. Precipitation of DOX, forexample, from the internal solution would facilitatemovement of additional DOX from the outside tosatisfy the equilibrium relationship of Eq. 1. Infact, Lasic et al. [12,13], employing an ammoniumsulfate gradient to load DOX into liposomes, haveshown that in these systems internalized DOX formsaggregates. Using cryo-electron microscopy, theywere able to visualize the internalized DOX aggre-

gates and found structures similar to those caused byprecipitation of DOX with sulfate anion in solution.

Aggregation could also explain the restriction ofDOX molecular motions noted by Cullis and co-workers who studied DOX entrapped by a trans-membrane pH gradient into citrate containing lipo-somes. Although Cullis and co-workers have sug-gested that citrate could cause precipitation of thisinternalized DOX [7,11], in more recent work theyhave proposed that the attenuation of the DOX nu-clear magnetic resonance (NMR) signal they ob-served was due entirely to its binding to the lipo-some's inner surface [14]. Moreover, in order toexplain the internalized structures they observed in-side DOX liposomes by cryo-electron microscopy[15] they proposed that DOX produces invaginationof the liposome bilayer membrane. Because imagesof these liposomes were similar in appearance tothose containing DOX-sulfate aggregates [13] andbecause we have noted that DOX forms a precipitatein citrate solutions, we believed the nature of DOX'sphysical state in citrate bearing liposomes to remainin question.

It was our desire here to study in detail the phys-ical state of DOX inside citrate containing liposomes.We examined DOX in solution and inside liposomesby various methods and found that DOX can formhighly organized bundles of ¢brous structures in thepresence of citrate. Our results indicate that s 99%of DOX internalized in liposomes at 20^300 mM ispredominantly self-associated and not bound to theinner membrane surface.

To assess the extent to which counterion valencyin£uences DOX organization, we also examined lip-osomes in which the monoanionic compound lacto-bionic acid (LBA) was used to bu¡er the vesicle in-terior. When a pH gradient was established and thesesystems were loaded with DOX (approx. 200 mMDOX inside liposomes), ¢brous structures were ob-served, but unlike the structures we found in citrateliposomes these ¢bers were disorganized and notcondensed into bundles. When these systems wereloaded with low internal DOX concentrations no ¢-bers were observed. Whether in this case DOX wasbound to the liposome inner surface remains unclear,but membrane invagination did not occur.

Because the physical state of DOX inside lipo-somes could be manipulated, we ventured to examine


X. Li et al. / Biochimica et Biophysica Acta 1415 (1998) 23^4024

its impact upon release. Leak pro¢les were not dis-tinguishable for DOX loaded at high internal con-centrations (approx. 200 mM DOX) into either cit-rate or LBA containing liposomes. For DOX loadedto low internal concentrations (approx. 20 mMDOX) where DOX was either in ¢ber bundles (cit-rate) or disaggregated (LBA), di¡erent leak rateswere noted. However, we found this di¡erence tobe due to an osmotic stress related rupture for theLBA liposomes. When osmotic stress was avoidedthe leak pro¢les were similar. Because in the LBAliposomes DOX was disaggregated at the low DOXconcentration, we speculate that ¢ber formation mayplay an important role in curtailing DOX loss invivo, an important concern considering the long res-idence time that DOX liposomes have in the circu-lation [16] and the fact that plasma constituents in-crease liposome leak [17].

2. Materials and methods

2.1. Materials

Egg phosphatidylcholine (EPC) and cholesterolwere obtained from Avanti Polar Lipids (Alasbaster,AL). DOX-HCl (Adriamycin RDF) was purchasedfrom Farmitalia Carlo Erba (Milan, Italy). ForNMR, DOX from Sigma (St. Louis, MO) wasused. Citric acid monohydrate, ammonium sulfate,chromiumoxalate (trihydrate) (potassium salt),4-oxo-TEMPO (TEMPONE) and 4-amino-TEMPOwere obtained from Aldrich (Milwaukee, WI). [1,5-13C]Citrate was purchased from Isotec (Miamisburg,OH). HEPES, cholesterol calibrator, and cholesteroldiagnostic kits were purchased from Sigma. Octa-ethylene glycol monododecyl ether (C12E8) was ob-tained from Fluka (Buchs, Switzerland). Tetra(sulfo-natophenyl)porphine was purchased from PorphyrinProducts (Logan, UT). Sodium carbonate anhydrouswas purchased from J.T. Baker (Phillipsburg, NJ).All the chemicals used were the highest purity avail-able.

2.2. Liposome preparation and lipid assay

Liposomes were made by extrusion of lipid suspen-sions ¢rst made by solvent evaporation. These struc-

tures were used instead of multilamellar vesicles(MLVs) because we wanted to avoid localized soluteexclusion [18,19]. Freezing and thawing the MLVswould not have attained an ideal solute distributionbecause the high concentration of solutes used herea¡orded these systems cryoprotection. For the initialliposome suspensions, chloroform stock solutions ofEPC and cholesterol (approx. 20 mg/ml lipid) weremixed at 55:45 (mole) ratio, accordingly. After add-ing 2.5 ml of either 300 mM citric acid (pH 4.0) or650 mM lactobionic acid (pH 3.6) into the chloro-form mixture (approx. 25 ml), the chloroform wasremoved by rotary evaporation at 40³C. When essen-tially all of the chloroform was judged to be removedan additional 2.5 ml of the appropriate bu¡er wasadded to the paste that had formed and the samplewas subjected to further rotary evaporation at 40³C.If necessary, the resulting suspensions were then ad-justed with distilled water to a total volume of 5 ml.These liposomes were then extruded through twostacked ¢lters of pore size 400 nm (one pass) andthen 200 nm (ten passes) at 40³C; ¢lters were poly-carbonate Nuclepore membranes from NucleporeCorporation (Pleasanton, CA). Liposome size wasdetermined by light scattering using a Nicomp Model270/370 Submicron Particle Sizer from Paci¢c Scien-ti¢c (Menlo Park, CA).

The phospholipid concentration of all liposomesamples was measured by a modi¢ed version of theprocedure of Chen et al. [20]. Cholesterol concentra-tions were determined using the cholesterol diagnos-tic kit from Sigma. Absorbances were recorded on anUV-2101PC UV-scanning spectrophotometer fromShimadzu Scienti¢c Instruments (Princeton, NJ).

2.3. Doxorubicin loading and leak assays

2.3.1. Loading of DOX into liposomesUnless stated otherwise transmembrane pH gra-

dients were created by adding an aliquot of sodiumcarbonate solution (17.6 mg/ml) into the liposomesamples. The pH of the aqueous solution outsidethe liposomes ranged from 7.8 to 8.1. This liposomesuspension was then mixed with a DOX saline sol-ution. The ¢nal overall DOX concentration was ei-ther 3.4 mM or 0.34 mM, whereas the lipid concen-tration was approx. 12.8 mM. An aliquot of 10 mM4-amino-TEMPO was added if the pH gradient


X. Li et al. / Biochimica et Biophysica Acta 1415 (1998) 23^40 25

across the lipid bilayer was to be measured. To ac-celerate loading, the sample was then heated to 55^60³C for 10 min and cooled to room temperature.

The loading e¤ciency was determined using gelchromatography to remove unencapsulated material.Aliquots of each sample were passed down columnspre-equilibrated with 10 mM HEPES, 150 mM NaCl(HBS) at pH 7.5. The liposome fraction was gatheredand this material and an aliquot not passed down thecolumn were adjusted to an equivalent volume withsaline and then assayed for lipid by phosphate anal-ysis. Each sample was also assayed for DOX contentby dissolution in methanol and measurement of theabsorbance at 490 nm. The percent entrapment wascalculated as that percent of DOX remaining withthe liposomes following elution; DOX concentra-tions were normalized using the lipid concentrations.

For LBA, we found that 650 mM LBA maintainedpH gradients (3 units) for several hours without anymeasurable decrease thus indicating that LBA's per-meability to these liposome membranes is low; allsubsequent experiments were performed well withinthis time frame. Additionally, 650 mM LBA allowedthe complete loading of DOX at 3.4 mM DOX and12.8 mM lipid. At 500 mM LBA, the bu¡er capacitywas insu¤cient to achieve complete DOX loading.

2.3.2. Doxorubicin leakage from liposomesTo compare DOX leak from liposomes, an aliquot

of liposomal DOX was diluted 300-fold by injectioninto a cuvette containing a solution of 10 mM HBS(pH 7.5). The £uorescence intensity of DOX wasmeasured continually with data collected every 10s. The excitation and emission wavelengths were480 nm and 590 nm, respectively. At the end ofeach measurement, C12E8 was added to dissolvethe liposomes and attain the 100% leakage value.The percentage of leakage of DOX from liposomesat a given time was calculated using Eq. 2:

% leakage � �I3I0��I1003I0�U100 �2�

where I was the £uorescence intensity at a given timeand I0 and I100 were intensities immediately after theinitial dilution into HBS or after addition of C12E8,respectively. DOX £uorescence intensity (without lip-osomes) decreased linearly with time for samplesmaintained above pH 7, due to DOX degradation

at higher pH values [10,21]. The magnitude of thisdecrease at pH 7 was small (6 2% per hour). There-fore, we did not adjust the leak equation to re£ectthis loss. At pH 4 no decrease in DOX £uorescenceintensity was noted up to 7 h.

2.3.3. Measurement of vpH and captured volumeTo measure internal pH, and thus vpH, we used

the distribution of an electron spin resonance (ESR)amine probe [22]. The ESR probe 4-amino-TEMPOwas added to liposomes at the time of DOX loading(¢nal concentration 200 WM). The sample was sepa-rated into fractions and these fractions were thendiluted equally with either a solution of bu¡er (10mM HBS at pH 7.5) alone or with bu¡er containingthe broadening agent chromiumoxalate (trihydrate).The solution containing broadening agent was ¢rstadjusted to the same osmotic strength as that of theliposome solution to avoid vesicle shrinkage or swell-ing. All osmolarity measurements were performedwith a 5500 Vapor Pressure Osmometer from Wescor(Logan, UT). The ESR spectra of both fractionswere then recorded and the vpH was calculated us-ing the following relationship:

vpH � log10Ain

Atot3AinU

Vout

V in

� ��3�

where Ain and Atot were the amplitudes of the ESRI = +1 resonance with and without the broadeningagent, respectively. Vin and Vout were the aqueousvolume inside and outside the liposomes, respec-tively. The values of Vin and Vout were obtained sep-arately from the measurement of captured volume.

The captured volume of the liposomes was deter-mined using an ESR procedure previously described[23] which is a slight modi¢cation of the method ofAnasi et al. [24]. The ESR spin probe TEMPONE (inthe appropriate bu¡er) was added to the sample inquestion which had ¢rst been adjusted to pH 7^8 byaddition of sodium carbonate (17.6 mg/ml) solution.Adjustment of the pH to above pH 5 was necessaryto prevent leakage of the chromiumoxalate broaden-ing agent that was to be added later. The liposomesample (38 mg/ml lipid) containing TEMPONE (1mM) was separated into 100 Wl fractions. To onefraction the same liposome bu¡er solution (100 Wl)was added and to another faction the chromiumox-alate broadening agent solution (100 Wl) was added.



The ESR signal amplitudes with and without broad-ening agent were then related to the inside and totalaqueous volumes, respectively. Captured volume (Wl/Wmole lipid) was then derived as previously describedusing a correction to account for the volume occu-pied by the lipid [23].

2.4. Microscopy

For confocal microscopy, £uorescence images wereacquired with a Biorad 1000 Confocal Microscopefrom Olympus (Hercules, CA), using a Kr/Ar laserwith Vex at 568 nm and Vem at 605 nm. For cryo-electron microscopy, frozen-hydrated samples wereprepared by placing 5 Wl of a liposomal suspensionon a copper grid with a holey carbon support. Eachsample was blotted to a thin ¢lm and immediatelyplunged into liquid ethane. The grids were thenstored under liquid nitrogen until used. The gridswere viewed on a Philips CM12 transmission electronmicroscope (Mahwah, NJ) operating at an accelerat-ing voltage of 120 kV. The microscope was equippedwith a Gatan 626 cryoholder (Warrendale, PA), andthe samples were maintained at 3177³C duringimaging. Electron micrographs were recorded underlow electron dose conditions at a magni¢cation of60 000U and 1.4^1.7 Wm defocus.

2.5. X-Ray di¡raction

Solution aggregates for X-ray di¡raction were pre-pared by mixing equal parts of a solution of DOX(15 mg/ml) and 600 mM (¢nal conc. 300 mM) citrateat the appropriate pH. Aggregates immediatelyformed and after 10 min the samples were concen-trated via centrifugation (14 000Ug) and the super-natant removed. For the heated samples, after aggre-gate formation the samples were heated to 55^60³Cand cooled back to room temperature prior to con-centration via centrifugation. For liposomes contain-ing DOX, loaded liposomes and empty liposomes (tocorrect for liposome scattering) were pelleted via cen-trifugation at 210 000Ug for 2 h. Pelleted sampleswere placed in 1.5 mm diameter glass X-ray capilla-ries which were then £ame sealed. X-Rays were pro-duced by a Rigaku RU-200 rotating anode generatorat a typical loading of 40 kV and 60 mA. The Cu KKline at 1.54 Aî was selected and focused by Ni ¢lter-

ing and double mirror Franks optics, which resultedin a typical £ux at the sampled of 7U107 X-rays/s.The sample temperature stage was thermoelectrical-ly adjusted to the desired temperature to withinþ 0.1³C. A two-dimensional X-ray detector basedon a 512U512 pixel Thomson CCD [25] was usedto acquire the data. The di¡raction patterns wereconcentric powder pattern rings which were azi-muthally integrated into one-dimensional plots of in-tensity versus scattering angle. Typical integrated ex-posures ranged from 20 to 60 min in duration.

Di¡raction signals were analyzed over a q-rangeof 0.1^0.8 Aî 31, where the wave vector q is de¢nedby

q � 4ZV

sina2

� ��4�

where a is the scattering angle and V is the wave-length of the incident radiation. Based on the cryo-EM data, we assumed that the X-rays basicallyprobed only two dimensions of the structure withthe third dimension (along the ¢ber's axis) beingtoo large to be seen. The di¡raction was analyzedin terms of a general oblique planar lattice in whichthe two unit cell basis vector lengths, a and b, are notconstrained to have the same length and in which thesmallest angle between them, Q+Z/2, is constrainedonly in that 06 Q6 Z/2. The general formula in twodimensions for the peak positions in reciprocal spacewith Miller indices h, k is given by:

q2

�2Z�2 �h2

a2 sin2 Q� 2hk cos Q

ab sin2 Q� k2

b2 sin2 Q�5�

2.6. Fluorescence and resonance light scattering

DOX £uorescence in solution was measured usingan Alpha Scan Spectro£uorometer from PTI (SouthBrunswick, NJ). Excitation and emission wave-lengths were set at 480 nm and 590 nm, respectively.For resonance light scattering measurements, excita-tion and emission wavelengths were kept identical asthe spectrum was scanned from 200 nm to 900 nm;this is essentially a U90³ light scattering experimentas a function of wavelength (see Pasternack et al.[26,27] for details). Tetra(sulfonatophenyl)porphineat pH 1 was used as a positive control for RLS (R.Pasternack, personal communication).



2.7. Spectrophotometric assays

2.7.1. Circular dichroism (CD) spectroscopyLiposomal DOX suspensions, or DOX aqueous

solutions/suspensions containing saline, water, cit-rate, or lactobionic acid were typically placed in a1 mm CD cell. The CD spectra were taken atroom temperature (20^23³C) from 700 nm to200 nm on a J-710 spectropolarimeter (Jas-co, Easton, MD) and typically the average of ¢vescans was reported with removal of high frequencynoise.

2.7.2. Turbidity and UV-Vis spectra measurementThe turbidity (absorbance) and UV-visible spectra

of di¡erent DOX aqueous solutions/suspensions weremeasured using a 1 cm pathlength quartz cell in aUV-2101 Spectrophotometer from Shimadzu Scien-ti¢c Instruments. For turbidity, absorbance wasmeasured at 800 nm. This relatively high wavelengthwas chosen to avoid any interference from DOX ab-sorption.

2.8. Di¡erential scanning calorimetry (DSC)

Calorimetry was performed using a MC2 Ultra-sensitive Scanning Calorimeter (MicroCal, North-hampton, MA). Heating scans were obtained from15³C to 85³C at a rate of 20³C/h.

2.9. NMR spectroscopy

2.9.1. NMR sample preparationFor NMR samples, liposomes were prepared as

described in Section 2.2 except that for citrate con-taining liposomes, [1,5-13C]citrate was mixed withunlabeled citrate to give either 17 or 47 mole %[1,5-13C]citrate. The ¢nal concentrations of citrateand LBA in which the liposomes were formed were270 mM and 650 mM, respectively. To establish apH gradient and to remove the outer bu¡er solu-tions, liposomes were passed down polyacrylamidecolumns (Pierce, Rockford, IL) equilibrated with20 mM HBS at pH 7.6. Only the ¢rst 2/3 or so ofthe eluting liposomes were used so as to guaranteeremoval of as much external bu¡er as possible andthe lipid concentration was then analyzed. DOX

(from Sigma) was then hydrated with saline and thenmixed with an aliquot of liposomes and additionalHBS such that the ¢nal lipid concentration was 12.8mM lipid and the ¢nal DOX concentrations wereeither 3.4 or 0.34 mM DOX (2 mg/ml or 0.2 mg/ml). Upon mixing, the liposomes were then heatedto 55^60³C for 10 min. The samples were then exam-ined by CD to con¢rm typical spectra and then usedfor NMR measurements. Samples were stored at ap-prox. 5³C. The NMR signal of the citrate bu¡eralone was used to gauge the amount of citratetrapped inside the empty liposomes. This amountof citrate was approx. 6.2 mM (total solution con-centration). At 3.4 mM overall DOX and assumingcomplete loading, this corresponds to a ratio of ap-prox. 1.8 citrate molecules per DOX molecule in theliposome interior.

2.9.2. 13C-NMR spectroscopyData were acquired on a Bruker AC 300 NMR

spectrometer using a 10 mm broadband probe fromNalorac Cryogenics Corporation. The resonantfrequency for 13C was 75.5 MHz. Free inductiondecays (FIDs) were acquired with proton decouplingusing a 20 kHz sweep width and 16K points. Nointernal ¢eld lock signal was used. Data for the T1

measurements were acquired using the saturation-re-covery method. Signal intensities (peak areas) wereplotted versus time and ¢t by the equation:I(t) = I(0)[13exp(3t/T1)]. Data for the T2 measure-ments were collected with the Carr-Purcell-Mei-boom-Gill pulse sequence. Where signal intensitiesof di¡erent samples were compared, the FIDs werecollected using a 14.8 Ws 90³ pulse, a recycle delayv 5UT1, and proton decoupling applied only duringthe data acquisition period. Other FIDs were col-lected with the minimum recycle delay allowed bythe spectrometer and the pulse width set to the Ernstangle for the spins of interest. At the end of dataacquisition for a particular sample, ethylene glycol,0.2% v/v, was added as an internal chemical shiftreference (64.0 ppm) and a ¢nal set of FIDs weresummed. The Fourier transform of this data wasused to assign the chemical shifts. Prior to Fouriertransformation, all FIDs were zero-¢lled to 32Kpoints and multiplied by an exponential decay corre-sponding to 3 Hz.



3. Results

To load DOX into liposomes, Cullis and co-work-ers have used transmembrane pH gradients(pHinW4, pHoutW7.5) created using either citrate[7] or glutamate [28] to bu¡er the vesicle interior.This work was the basis for a liposomal DOX for-mulation (Evacet, formerly TLC D-99) that is cur-rently being evaluated in clinical trials [29^31]. Withthis strategy, DOX can be loaded to high concentra-tions inside liposomes (200^300 mM internal concen-trations). This is far in excess of DOX's solubilitylimit in solution, which is 960 mM in water [12].The question then arises as to what physical struc-ture does DOX adopt or does it bind to the innermonolayer of the liposome. The purpose of ourstudy here was to investigate the physical state ofDOX inside liposomes.

3.1. Doxorubicin in solution

As indicated in Table 1, the physical state of DOXin solution is clearly in£uenced by the valency of thecounterion. When combined with the multianionscitrate or sulfate, DOX formed a viscous gel whichupon shaking yielded ¢brous structures (Fig. 1). Theresulting structures were similar in appearance tothose previously reported for DOX-sulfate aggre-gates [9]. No such structures were observed forDOX in water or in solutions of glutamate or themonoanionic bu¡er lactobionic acid (LBA). Why

glutamate, which has two carboxylic acid groups,did not initiate aggregate formation is unclear butit does also contain a cationic amine group whichmight obviate assemblage of the DOX structureselectrostatically. Because we were interested in whathappens to DOX inside citrate containing liposomeswe focused our attention upon citrate-doxorubicininteractions for the remainder of the studies dis-cussed here.

The DOX-citrate structures we observed in solu-tion were assembled into linear ¢brous-like struc-tures. Upon close inspection the larger strands werefound to be bundles of smaller ¢bers, similar in ap-pearance to micellar ¢bers resulting from porphyrinstacking/aggregation [32]. These ¢brous structuresdisappeared upon heating and reappeared with cool-ing. Calorimetric scans of 4 mg/ml DOX in 300 mMcitrate revealed a broad endotherm which was near43³C at pH 4 but which was shifted upward to 57³Cat pH 5; vH was 1.8 and 2.8 kcal/mole, respectively(data not shown). The melting of DOX-citrate pre-cipitate may be similar to the `melt' of hydrated mi-cellar crystals (Kra¡t point) that has been describedfor local anesthetics [33,34]. An exothermic peak wasalso observed in the scans of both samples at approx.75 and 65³C, respectively, which was absent uponsubsequent heating cycles. A visual check of heatedsamples con¢rmed that disaggregation was associ-ated with the endothermic transition and not theexothermic event. From TLC analysis, it appearedthat the exothermic transition represented loss ofDOX's amino sugar moiety (not shown). In thework we report here, samples in solution or invesicles were never heated to beyond 60³C.

3.2. Cryo-EM of DOX inside citrate liposomes

Shown in Fig. 2 are cryo-EM images of citratebearing liposomes loaded with DOX. In all casesDOX formed ¢brous aggregates inside the liposomesand the liposome membranes (both inner and outermonolayers) were well resolved with no observablemembrane invagination. The arrow in Fig. 2A pointsto an end-on view of a `U'-shaped bundle of DOX¢bers that are packed hexagonally. A similar U-shaped bundle is seen in pro¢le in Fig. 2B. FibrousDOX bundles were apparent in nearly all liposomesand were either linear, curved (some having a `U'-

Table 1Formation of DOX aggregates in solution

Counterion pH Aggregateformation

[DOX] at whichaggregationbeginsa (mM)

Citrate 4^7 yes 0.5^1.5Sulfate 4^7 yes 1.0LBA 4 no s 20b

Glutamate 4 no s 20b

aAggregation onset was measured by the increase in turbidity(A800 nm) of DOX at 4 mM in the indicated solutions.bDetermined visually not spectrophotometrically. DOX predis-solved in water was mixed 1:1 with solutions. Final counterionconcentrations were 300 mM for citrate, 650 mM for LBA, 110mM for sulfate, and 175 mM for glutamate. These counterionconcentrations were selected as they are the concentrations usedpreviously with DOX liposomes.



shape), or circular where the ¢bers apparently closedback upon themselves. For many of the longer linearDOX structures, liposomes were obviously elongatedto accommodate ¢ber growth. Many of the bundlesappeared to twist, as for example, the large circularstructure in Fig. 2A. From the hexagonal pattern inFig. 2A, the number of ¢bers in that particular bun-dle appeared to be approx. 50. This value wouldappear to vary since end-on views of other U-shapedstructures (not shown) revealed a range of 12^60¢bers/bundle, all in a hexagonal arrangement. Fromthis end-on perspective the average spacing between¢bers was estimated to be 3^3.5 nm. In some cases,striations could be observed along the length of thebundles that alternated with non-striated regions.For example, the arrows in Fig. 2C indicate repeat-ing striated regions in the DOX ¢brous bundles. Thedistance between the centers of these striated regionsappeared to be approx. 50 nm. Interestingly, meas-uring across the striations (perpendicular to bundle'slong axis) we estimated that the striated `lines' them-selves were approx. 3^3.5 nm apart suggesting that

these features are the side view of aligned rows of¢bers. Such a periodic alignment would occur if thehexagonally packed bundles twisted about their cen-ter. In which case, as viewed from the side the rowswould align with every 60³ rotation and give theappearance of striations. For some circular struc-tures, the striations appeared to persist. Perhaps thepitch of the rotation and the curvature weresynchronized or perhaps there was some interactionwith the inner monolayer that allowed alignmentwith that surface.

3.3. Small angle X-ray di¡raction of DOX aggregates

X-Ray di¡raction was used to better understandthe packing details of the aggregates. Although wecould not su¤ciently concentrate liposomes for peakanalysis by small angle X-ray di¡raction (SAXS) wewere able to examine the DOX-citrate aggregatespelleted via centrifugation. Fig. 3 shows the di¡rac-tion from samples produced at pH 4 and pH 5. ThepH 4 material at room temperature showed similar

Fig. 1. Confocal light microscopy of DOX-citrate ¢brous bundles. DOX in water was mixed 1:1 with a citrate solution for a ¢nalconcentration of 4 mM DOX and 300 mM citrate at pH 4. Similar DOX aggregates were observed in solutions of citrate at pH 5and ammonium sulfate at pH 4 and 7 (not shown). Image was colored to better visualize structures three-dimensionally. Bar length is5 Wm.



di¡raction both before heating (Fig. 3A) and after aheating protocol (see Section 2.5) meant to mimic thetemperature protocol used to load liposomes, e.g.,heat brie£y to 55^60³C and then cool back downto 20³C (data not shown). The arrows in Fig. 3Ashowed the expected peak positions of an oblique

lattice of a = 33.6 Aî , b = 29.2 Aî , and Q= 40³, as ob-tained by a least squares ¢t of Eq. 5 to the data.Because of the paucity of di¡raction peaks, and theabsence of certain peaks (which may simply be tooweak to be seen), the lattice assignment should betaken as possible, but not unequivocal. The pH 5material, before heating, yielded di¡raction very sim-ilar to the pH 4 material (data not shown). However,upon heating to 55^60³C and then cooling back to20³C, the di¡raction pattern changes to that seen inFig. 3B. An unconstrained ¢t to Eq. 5 yields a =33.7 Aî , b = 35.7 Aî , and Q= 61.48³. This is withinthe measurement errors of the simple hexagonal lat-tice (34.5 Aî basis length), indicated by the arrows inthe ¢gure.

A weak peak corresponding to a real space repeatof 10.1 Aî is also seen in the data, as shown in Fig.3A (see the peak in between the ones indexed with(2,0) and (3,3) in the ¢gure). This does not ¢t theplanar lattice assignment and might be due to corre-lations along the direction of the ¢bers, that is to sayperpendicular to the lattice plane.

From these data it would appear that doxorubicin-citrate ¢ber bundles can adopt two distinct packingarrangements. A simple hexagonal packing arosewhen both a pH 5 environment and a heating/cool-ing (recrystallization) step were employed for the un-encapsulated solution ¢bers. These conditions maybe met for liposomal DOX since the interior pH ofthe loaded vesicles increases to near pH 5 (see exam-ples in Table 3) and, as per our protocol, liposomesare always heated during loading (see Section 2.3).

We tried to determine whether the oblique struc-ture is present in the liposomes as well. As mentionedbefore, it was di¤cult to directly index peak posi-

Fig. 2. Cryo-EM images (three ¢elds) of DOX loaded into lipo-somes bu¡ered by citrate. The ¢nal DOX solution concentra-tion was 3.4 mM which is estimated to be 200^300 mM DOXinternally. This internal concentration was estimated basedupon DOX loading and the liposome captured volume (see Ta-ble 2). The total lipid concentration was 12.8 mM and the sam-ple was undiluted for microscopy. The arrow in A indicates theend of a ¢brous bundle in which the ¢bers are hexagonally ar-ranged with an approx. 3^3.5 nm separation. In B, a circular,U-shaped, and straight DOX-citrate complex are noted. The ar-rows in C indicate striated regions in the DOX ¢brous bundleswith a repeating distance of approx. 50 nm. Bar represents 100nm.6



tions for the DOX-citrate loaded liposomes ^ Fig. 3C(upper pattern) shows an example of the di¡ractionpattern of pelleted DOX loaded liposomes. Any sig-nal arising from a DOX-citrate lattice in the lipo-

somes would be superimposed on a signal arisingfrom the liposome shells. Hence we adopted the fol-lowing procedure: an image was taken of the DOXloaded liposomes as well as one of the empty lipo-somes (not shown). Assuming that the total scatter-ing arises from the product of the liposome formfactor and the structures factor arising from theDOX-citrate complex, we divided the di¡ractionfrom the DOX loaded liposomes by that of theempty liposomes (Fig. 3C, top pattern). Even withpoor statistics, one can see that the resultant peakscoincide with those of the DOX-citrate complexformed in solution at pH 4 (lower pattern of Fig.3C). This suggests that the oblique lattice may alsobe present in the liposomes. Note that the resolutionof the cryo-EM images does not allow us to discrim-inate between a simple hexagonal lattice and aslightly skewed hexagonal lattice.

3.4. Circular dichroism of doxorubicin in liposomes

As expected, unentrapped DOX that was mixedwith empty liposomes exhibited a CD pattern consis-tent with disaggregated dimeric DOX [35,36] (seeFig. 4). This spectrum was identical to that for

Fig. 3. Small angle X-ray di¡raction patterns of DOX-citratecomplexes in solution at (A) pH 4 and (B) pH 5 or in pelletedliposomes (C). For the pH 4 sample (A), the pattern is from anunheated sample: heating and cooling the sample did not ap-preciably change the di¡raction pattern. The arrows show theexpected peak positions for the oblique lattice described in thetext. Di¡raction from the samples at pH 5 before heatinglooked very similar to samples at pH 4, e.g., as shown in A.The pattern for pH 5 samples changed to that shown in Bupon heating and cooling. The arrows show the expected peakpositions for the hexagonal lattice described in the text. Theupper pattern of C is that derived for the DOX-citrate complexinside loaded liposomes (see text for details). Shown for refer-ence, the lower pattern of C is that for DOX-citrate complexpelleted from solution at pH 4.

Fig. 4. CD spectra of DOX loaded into citrate containing lipo-somes via a pH gradient where the ¢nal DOX concentrationwas either 3.4 mM (solid line, left y-axis) or 0.34 mM (dashedline, right y-axis). Internal DOX concentrations were estimatedto be 200^300 mM and 20^30 mM, respectively. Also shownis 3.4 mM DOX mixed with empty liposomes with no pHgradient (dotted-dashed line). Liposomes (approx. 100 nmLUVs) contained 300 mM citrate and lipid concentration was12.8 mM.



DOX in water without liposomes (data not shown).Loading of DOX into citrate containing liposomesdramatically changed its CD spectrum (Fig. 4). TheCD spectrum of loaded DOX indicated that a sig-ni¢cant change in DOX interactions had occurred,consistent with the observed structure formation.

When DOX was loaded to a ten times lower con-centration into liposomes the CD signal was nearlythe same (see Fig. 4). As shown in Fig. 5, cryo-EMimages of these loaded liposomes revealed similarinternalized structures to those found in liposomesloaded to the ten times higher internal DOX concen-tration. The dimensions of the internal DOX struc-tures were smaller in both length and breadth ascompared to those of Fig. 2 with fewer ¢bers perbundle. Because of these images and because thereseemed to be little contribution to the CD patternfrom disaggregated DOX we believe the complexedform is the predominant DOX species inside citratecontaining liposomes. In fact, for liposomes loadedto the higher internal concentration of DOX we es-timate from analysis of CD spectra that the DOX-¢ber species accounts for greater than 99% of thetotal DOX.

3.5. In£uence of citrate upon doxorubicin dimerization

To understand more about how citrate interactswith DOX we examined whether the citrate anionin£uences the association of DOX monomers intodimers. This association occurs at very low DOXconcentrations in water (from 5 to 20 WM) and onemight expect citrate to in£uence this conversion if thedimer is arranged such that the two positive chargesfrom the sugar moieties can be bridged by the citratemultianion. We compared the CD spectra of DOX ineither water or a solution of 300 mM citrate over therange of concentrations at which monomer to dimerconversion is expected (data not shown). At 5 WMDOX, CD spectra were similar to each other and tothat previously reported for monomeric DOX anddaunorubicin [35,36]. While there were changesfrom 5 WM to 1 mM that were consistent with mono-mer to dimer conversion [35^37], the spectra forDOX in water and citrate were similar to one anoth-er up to 1 mM DOX. However, at 2 mM DOX theCD patterns began to di¡er signi¢cantly, consistentwith our assessment of aggregation by turbidity (Ta-ble 1). From these data it does not appear that cit-rate in£uences DOX dimerization. This might sug-gest that dimerized DOX molecules are orientedsuch that the cationic amines cannot be bridged bythe citrate multianion. This ¢nding, along with ourother data, also suggests that citrate induces aggre-gation by coupling DOX dimers/oligomers (dimersinto oligomers and/or oligomers into side-by-sidepacked ¢ber bundles).

3.6. 13C-NMR measurements of citrate

To better understand the interactions betweenDOX and the citrate counterion inside liposomes,we used 13C-NMR to examine internalized citrate.The terminal carboxylic acid carbons of citrate, car-bons 1 and 5, are chemically identical and, at anyparticular pH, give rise to a single resonance in the13C-NMR spectrum. In our NMR studies of citratebu¡ered liposomes, the 13C-NMR signal arising fromthe terminal carboxylic acid carbons was enhancedby the presence of [1,5-13C]citrate. To simplify ourdescription of the NMR results, we will refer to theresonance arising from these terminal carboxylic acidcarbons as the citrate resonance.

Fig. 5. Cryo-EM image of DOX loaded into citrate containingliposomes where the ¢nal DOX concentration was 0.34 mM.The internal DOX concentration was estimated to be 20^30mM. The lipid concentration was 12.8 mM. Bar represents 100nm.



In citrate bu¡ered liposomes, the citrate resonancefrom inside the liposome was shifted and broadenedby the loading of DOX (Fig. 6). The chemical shiftsobserved are consistent with an increase in the inter-nal pH of the DOX loaded liposomes (see Table 3).We believe that the broadening of the citrate reso-nance is due to a distribution of internal pH valueswithin the liposomes, produced by DOX loading.The modest decrease in T2 of the citrate resonancewhich occurs with DOX loading (Table 2) cannotaccount for the magnitude of the line broadeningobserved. In LBA bu¡ered liposomes, loading withDOX induced similar changes in the chemical shiftand line width of the resonance arising from the

carboxylic acid carbon of internal LBA, (data notshown).

Because cryo-EM images show that DOX formed¢brous structures inside citrate bu¡ered liposomes,one might expect that some fraction of the citratewould also be immobilized. If so, this citrate mightnot be observed in the solution state spectrum due tolifetime broadening or dipolar broadening of its res-onance. Assuming a charge ratio of 3:1, citrate toDOX, and complete loading of the DOX into theliposomes, one would expect that approx. 18% ofthe citrate would be associated with DOX throughelectrostatic interactions alone. However, the signalintensity from citrate is, to within the uncertainty inour measurement, unchanged by the loading of DOX(approx. 200 mM inside) (Table 2). What we do ob-serve is a modest decrease in the T1 and T2 of thecitrate signal (Table 2), suggesting that citrate`bound' to DOX ¢bers may be in rapid exchangewith `free' citrate.

3.7. LBA liposomes

The DOX structures inside liposomes that formedwith citrate were similar in appearance to thoseformed with the divalent sulfate anion [13]. To de-termine whether these internalized DOX structuresrequired a multivalent counterion in order to form,we next examined liposomes loaded with DOX usingthe monoanionic bu¡er LBA to control pH; the ini-tial internal pH was approx. 4 and the outside pHwas raised to approx. 7.4. Cryo-EM of LBA lipo-somes containing DOX at internal concentrations

Fig. 6. E¡ect of DOX loading on the 13C chemical shift andline width of the citrate resonance. 13C-NMR spectra of EPC/cholesterol liposomes with an initial pH gradient created by270 mM citrate, pH 4.0, inside and 20 mM HBS at pH 7.5 out-side. After the exchange of external citrate bu¡er for 20 mMHBS, pH 7.0, some residual citrate remained outside the lipo-somes. This resonance is labeled Co in the ¢gure above. Theresonance due to internal citrate is labeled Ci. Subsequent tocreation of the pH gradient the liposomes were loaded with:(A) 0 mM, (B) 0.34 mM (20^30 mM internal concentration)and (C) 3.4 mM DOX (200^300 mM internal concentration).The lipid concentration was 12.8 mM. Internal pH values asdetermined from the chemical shifts were (A) pH 4.0, (B) pH4.2^4.0, and (C) pH 5.5^4.1, consistent with the values of Table3.

Table 2T1, T2, and normalized intensity of the 13C resonance from[1,5-13C]citrate inside liposomes

[DOX](mM)

T1 (s) T2 (s) Normalized intensitya

0.0 6.1 þ 0.5 0.16 þ 0.03 ^0.34 4.9 þ 0.2 ^ ^3.4 2.6 þ 0.4b 0.09 þ 0.03c 1.01 þ 0.08c

4.5 þ 1.0aIntensities were normalized to that of citrate with 0 mMDOX.bThe internal citrate gave rise to two broad resonances in theseliposomes. The ¢rst T1 corresponds to the more `down¢eld'peak.cThese numbers re£ect the integrated area of both resonances.



of 20^30 mM DOX did not reveal any noticeablestructures (Fig. 7A); an occasional ¢ber was the ex-ception. Again no membrane invagination wasnoted. The CD pattern for DOX loaded into LBAliposomes at this low concentration was consistentwith disaggregated dimeric DOX (Fig. 8). At internalDOX concentrations of approx. 200^300 mM, far inexcess of DOX's solubility limit, what appeared to benumerous uncondensed ¢bers were observed in theliposomes (Fig. 7B). The CD pattern for these un-condensed DOX ¢bers was similar to that for DOX¢ber bundles in citrate bearing liposomes except that

the shoulder at 520 nm was decreased relative to theother peaks (Fig. 8); this shoulder could be sensitiveto `side-by-side ¢ber' interactions.

Apparently, DOX ¢ber formation occurs in LBAliposomes only at DOX concentrations in excess ofits aqueous solubility limit, which you will recall isapprox. 60 mM in water [12]. LBA is incapable ofbridging charges so the ¢bers that formed above thesolubility limit of DOX remained uncondensed. Thisimplies that the DOX ¢ber bundles observed insidecitrate containing liposomes are the result of the cit-rate multianion electrostatic bridging between sepa-rate ¢bers. Although citrate may have simply elimi-nated charge repulsion, thus allowing an otherwisefavorable association to occur, this seems less likelyto be the reason for condensation since LBA couldhave also served this purpose but did not.

3.8. E¡ect of DOX physical state upon liposomalrelease

Having established that the DOX aggregationstate inside the liposome can be manipulated, wenext ventured to make comparable liposomes con-taining either citrate or LBA to establish the e¡ectof DOX's physical state upon the release rate. Thephysical properties of these liposomes are listed inTable 3. Liposomes contained either 650 mM LBA

Fig. 7. Cryo-EM images of DOX loaded into liposomes bu¡-ered by LBA. (A) Liposomes contained LBA (650 mM) asbu¡er and the solution concentration of DOX was 0.34 mM,which would give an internal [DOX] of approx. 20^30 mM. (B)LBA liposomes loaded such that the ¢nal solution concentra-tion of DOX was 3.4 mM, which we estimate to be 200^300mM DOX internally. The lipid concentration was 12.8 mM andsamples were examined undiluted for microscopy. Bar repre-sents 100 nm.

Fig. 8. CD spectra of DOX loaded into liposomes containingLBA as bu¡er. Final solution DOX concentrations were 3.4mM (solid lines, left y-axis) or 0.34 mM (dashed lines, righty-axis). The ¢nal lipid concentration was 12.8 mM. For thelower concentration of DOX in LBA liposomes, the CD signa-ture indicated that DOX was predominantly disaggregateddimers.



or 300 mM citrate, both at pH approx. 4 initially,and were loaded to low (16^29 mM) or high (160^290 mM) internal concentrations of DOX. (ForLBA, 650 mM was required to achieve completeDOX loading; 500 mM LBA was insu¤cient (notshown).) The liposomes were diluted into hepes bu¡-ered saline and leak monitored as described in Sec-tion 2.3.

In all cases where the internalized DOX formed¢bers, the leak pro¢les were indistinguishable withless than approx. 4% leak after 60 min (data notshown). For those LBA liposomes in which DOXwas loaded to a low internal concentration (no ¢-bers), the leak pro¢les were biphasic and variablewith, in some cases, a signi¢cant loss of material inthe ¢rst few minutes (see leak values at 10 min ^Table 3).

When samples were diluted into a hyperosmotic(approx. 1060 mOsM) solution of bu¡er containing20% (w/v) sucrose the liposomes displayed similarleak pro¢les (6 4% leak at 60 min) for all four sys-tems of Table 3 (¢bers and no ¢bers) (data notshown). This indicated that osmotic stress was in-volved for the LBA containing liposomes with a

low internal DOX concentration (no ¢bers). Thiswas not too surprising since the expected osmoticdi¡erential for the LBA containing liposomes dilutedinto hepes bu¡ered saline is approx. 565 mOsMwhich is close to the rupture threshold di¡erentialof 600 mOsM reported by Mui et al. [38] for spher-ical EPC/cholesterol liposomes of similar size andcomposition. Considering this, the variability inleak for DOX from LBA containing liposomes (no¢bers) was not too surprising either since those lip-osomes may have been teetering on the thresholdamount of stress required to cause rupture.

The reason why DOX leak from the LBA lipo-somes loaded to a high DOX internal concentration(¢bers) did not also display a rapid component isunclear, but in those liposomes DOX did form ¢bersthat may have been restricted sterically from escap-ing the vesicle. Alternatively, it may be that only theLBA liposomes containing the lower amount of dis-aggregated-DOX were osmotically stressed. That is,the cationic DOX ¢bers that formed at higher DOXinternal concentrations may have bound a signi¢cantportion of the LBA molecules, and thus reduced theosmotic di¡erence.

Table 3Properties of loaded liposomes

Bu¡er Physical state ofDOX inside

[DOX](mM)

Internal [DOX](mM)

Cap. volume(Wl/Wmole)

Mean vesicle size(nm)

FinalvpH

InternalpH

% leak at10 min

Citrate ¢bers (bundles) 3.4 290 0.9 124 2.5 4.9 0.83.4 290 0.9 99 1.8 5.6 1.1

¢bers (bundles) 0.34 29 0.9 124 2.9 4.7 0.70.34 29 0.9 124 ^ ^ 0.10.34 29 0.9 99 2.6 4.6 0.10.34 29 0.9 99 2.6 4.8 0.1

LBA ¢bers 3.4 260 1.0 120 1.0 6.4 0.83.4 250 1.1 120 0.6 6.8 0.13.4 160 1.6 94 2.4 5.0 6 0.13.4 200 1.3 115 2.4 5.0 1.3

no ¢bers 0.34 24 1.1 120 2.4 5.0 1.30.34 16 1.6 94 3.2 4.2 1.70.34 16 1.6 94 3.2 4.2 0.50.34 20 1.3 115 3.1 4.3 13.10.34 20 1.3 115 ^ ^ 18.40.34 20 1.3 115 3.1 4.3 9.7

Liposomes were made in either 300 mM citrate or 650 mM LBA at pH 4. Vesicle sizes, captured volumes, and pH measurementswere performed as outlined in Section 2. The listed DOX concentrations were the overall solution values prior to dilution; internalconcentrations were estimated based on captured volume, lipid concentration, and loading e¤ciency (for all cases s 95%) (see text).vpH and internal pH values were determined for vesicles following DOX loading. Leakage experiments were performed immediatelyafter DOX loading (6 1 h) and the data represent the percent leakage at the 10 min point along the leak curves (not shown) for sam-ples diluted into HEPES bu¡ered saline.



4. Discussion

In solution, citrate, like sulfate, caused DOX ag-gregation to occur at concentrations 100 times lowerthan its aqueous solubility limit. Inside citrate con-taining liposomes, DOX loaded via a pH gradientwas found to exist in ¢ber bundles by cryo-EMwith no observable membrane invagination. Whilethe straight ¢ber bundles were somewhat similar inappearance to the DOX-sulfate aggregates observedby Lasic and co-workers [12,39], we also notedcurved and circular ¢ber bundles. These appearedto result from the bending of the linear structuresthat had grown in length beyond what the liposomewas capable of accommodating. End-on views ofthese DOX-citrate ¢ber bundles indicated that the¢bers were packed longitudinally in a hexagonal ar-rangement with a separation of approx. 3^3.5 nmbetween ¢bers. Two distinct di¡raction patternswere also observed by SAXS for DOX-citrate ¢berbundles in solution, an oblique and a simple hexag-onal lattice. Which of the two, or if both, are presentinside liposomes remains unclear since cryo-EM does

not have the resolution to distinguish between a sim-ple hexagonal and a slightly skewed hexagonal pack-ing. Even so the basis lengths noted by SAXS likelycorrespond to the inter¢ber distance estimated bycryo-EM, both approx. 30^35 Aî . The repeating stria-tions observed for ¢ber bundles suggests that theentire bundle twists 60³ approximately every 50 nm.This is consistent with a twisting hexagonal latticesince, as viewed from the side, the rotation of a hex-agonal array a¡ords spatial alignment every 60³ (seeFig. 9). The fact that the peaks of Fig. 3B do notcorrespond exactly to a hexagonal pattern may beconsistent with our assumption that we have aslightly tilted hexagonal structure that is twisting.

Because for DOX-sulfate aggregates only straightbundles (rods) were noted by cryo-EM [12,13], theappearance of curved and circular bundles herewould seem to indicate that our DOX-citrate ¢berbundles are more £exible. Since only individual ¢bersformed in LBA liposomes (Fig. 2B) we believe thatbundles formed in citrate containing liposomes dueto inter¢ber crosslinking by the citrate multianion.This would then suggest that the di¡erence between

Fig. 9. Schematic representation of twisting hexagonally arranged ¢bers. In this proposed model each ¢ber is comprised of stackedDOX molecules. The lower illustration depicts how the stacked DOX molecules (circles) would appear as viewed end-on. As viewedfrom the side (upper illustration), the ¢bers are aligned along the line of sight at rotation multiples of 60³ thus giving the appearanceof repeating striations. For an oblique lattice, the inter¢ber distance across each striated zone would vary. This variation might not beresolvable in the cryo-EM images if the obliquity is only slightly di¡erent from hexagonal, as seems to be the case here, based onSAXS results.



DOX-sulfate and DOX-citrate ¢ber bundles stemsdirectly from di¡erences in the counterions them-selves. Although the inter¢ber spacing for DOX-cit-rate aggregates was 30^35 Aî the separation noted forDOX-sulfate ¢bers was reported to be approx. 27 Aî

[12]. This tighter packing is consistent with the factthat sulfate is a smaller ion. Additionally, ¢ber £ex-ibility may also involve di¡erences in the multi-anion's binding a¤nity. Although we have no senseof sulfate binding we did ¢nd here from NMR ex-periments with 13C-citrate that the citrate-DOX in-teraction was quite dynamic with rapid exchange of`free' and `bound' citrate.

We have not detailed how DOX forms ¢bers butgiven its planar structure it seems likely that DOXmolecules stack. Because citrate did not a¡ect mono-mer to dimer conversion one might argue that theamino sugar moieties are separated far enough thatcitrate could not bridge/couple monomers. Interest-ingly, upon stacking/ aggregating porphyrins exhibita signi¢cantly increased scattering intensity at theabsorption wavelengths; this phenomenon has beentermed resonance light scattering (RLS) [26,27,40]. Inour hands, the DOX ¢brous structures did not ex-hibit RLS (data not shown). This result, though notconclusive, may indicate that RLS is a speci¢c phe-nomenon of the large porphyrin ring systems.

For DOX loaded into liposomes containing themonoanionic bu¡er LBA, its physical state was de-pendent upon the internal DOX concentration.

When loaded to only approx. 20 mM DOX, therewere no discernible structures inside LBA liposomesand the CD pattern was consistent with disaggre-gated DOX. It is quite interesting that this CD pat-tern was identical to that for dimeric and not mono-meric DOX because Gallois et al. [36] using CD havereported that DOX partitioned onto large unilamel-lar vesicles (LUVs) was monomeric and not dimeric.While this might suggest that no binding of DOX tothe membrane occurred we cannot rule out that forour liposomes DOX binds to the membrane surfaceas the dimeric species. In any case, no membraneinvagination was noted.

When loaded into LBA liposomes at high internalconcentrations DOX formed ¢bers. Again no mem-brane invagination was noted by cryo-EM. Unlikethe DOX in citrate containing liposomes, these ¢berswere disorganized and not condensed into bundles

(Fig. 8B). Apparently DOX stacking to form ¢bersoccurs above its solubility limit regardless if whetherthere is a multivalent counterion present or not.These results with LBA would also seem to con¢rmthat the ¢ber bundles in citrate containing liposomesarise because the citrate multianion facilitates inter-¢ber crosslinking.

Interestingly, this crosslinking of ¢bers by citratedid not dramatically slow the rate of DOX release asit was similar to that for DOX out of LBA liposomeswhere the DOX was disaggregated. From 13C-NMRexperiments it was apparent that `bound' DOX wasin rapid exchange with `free' DOX. Such a dynamicrelationship would be consistent with the unhinderedleak of DOX from the citrate containing liposomes.However, because liposome integrity/stability is amajor concern in vivo, ¢ber formation (and ¢berbundles in particular) may enhance the stability ofDOX loaded liposomes by providing a possible sterichindrance to vesicle rupture. It has been shown thatlipoprotein adsorption onto PC/cholesterol LUVssigni¢cantly decreased their ability to withstand os-motic stress [41]. Consequently, liposomes with un-complexed DOX may leak more signi¢cantly in vivo(work now underway).

Cullis and co-workers proposed that DOX is pre-dominantly bound to the inner monolayer [14] andthat this binding leads to an invagination of themembrane [15]. From the work presented here it isobvious that citrate produced organized aggregatedDOX-citrate structures inside liposomes. Also ob-vious from our cryo-EM images is that membranemorphology is una¡ected by DOX loading. The con-clusion by Cullis and co-workers that membrane in-vagination occurred was based upon less well re-solved cryo-EM images (DOX liposomes lookedlike `co¡ee beans') where the DOX-citrate ¢ber ag-gregates appeared as blurred lines inside the lipo-somes which could be mistaken for a membraneedge [15]. To discard any possibility that DOXfrom di¡erent vendors might yield a di¡erent inter-action, we examined three di¡erent DOX vendor ma-terials (one which included methyl paraben and twothat did not) and found identical structures insideliposomes (data not shown). Cullis and co-workersfound that DOX's 13C-NMR signal was signi¢cantlybroadened when internalized in citrate bearing lip-osomes [14]. While they interpreted this as due to



inner monolayer binding, our data clearly indicatethat the explanation for this molecular immobiliza-tion is caused by DOX complexation with citrate. Aswe have demonstrated here, packing of adjacentDOX ¢bers into bundles was facilitated by citratewhich lowered the solubility limit of DOX signi¢-cantly. Although we cannot rule out that someDOX may be partitioned into the inner monolayer,we believe that this DOX accounts for only a smallpercentage of the total internalized DOX.

Acknowledgements

This work was supported by the Liposome Com-pany, Inc. We thank Dr. Eric Mayhew for his assist-ance performing the confocal microscopy, Dr. Rob-ert Pasternack for a fruitful discussion of resonancelight scattering, and Rao Fu for help creating Fig. 9color graphic. X-Ray instrumentation developmentin SMG's lab is supported by the Dept. of Energy(DEFG02-97ER624434).

References

[1] J.G. Hardman, L.E. Limbird (Eds.-in-Chief), Goodman andGilman's The Pharmacological Basis of Therapeutics, 9thedn., McGraw-Hill, New York, 1996, p. 1265.

[2] J.A.E. Balazsovits, L.D. Mayer, M.B. Bally, P.R. Cullis, M.McDonell, R.S. Ginsberg, R.E. Falk, Cancer Chemother.Pharmacol. 23 (1989) 81^86.

[3] F. Olson, E. Mayhew, D. Maslow, Y. Rustum, F. Szoka,Eur. J. Cancer. Clin. Oncol. 18 (1982) 167^176.

[4] A. Gabizon, A. Meshorer, Y. Barenholz, J. Natl. CancerInst. 77 (1986) 459^469.

[5] P.M. Kanter, G.A. Bullard, R.A. Ginsberg, F.G. Pilkiewicz,L.D. Mayer, P.R. Cullis, Z.P. Pavelic, In Vivo 7 (1993) 17^26.

[6] G.A. Fonseca, V. Valero, A. Buzdar, R. Walters, J. Wiley,R. Benjamin, M. Ewer, B. Mackay, D. Gordon, G. Horto-bagyi, Proc. 31st Annual Meeting Am. Soc. Clin. Oncol., 20^23 May 1995, abstr. 95, p. 99.

[7] T.D. Madden, P.R. Harrigan, L.C.L. Tai, M.B. Bally, L.D.Mayer, T.E. Redelmeier, H.C. Loughrey, C.P.S. Tilcock,L.W. Reinish, P.R. Cullis, Chem. Phys. Lipids 53 (1990)37^46.

[8] L.D. Mayer, M.B. Bally, M.J. Hope, P.R. Cullis, Biochim.Biophys. Acta 816 (1985) 294^302.

[9] G. Haran, C. Rivka, L.K. Bar, Y. Barenholz, Biochim. Bio-phys. Acta 1151 (1993) 201^215.

[10] A. Vigevani, M.J. Williamson, in: K. Florey (Ed.), Analyt-ical Pro¢les of Drug Substances, vol. 9, Academic Press,New York, 1980, pp. 245^274.

[11] L.D. Mayer, L.C.L. Tai, M.B. Bally, G.N. Mitilenes, R.S.Ginsberg, P.R. Cullis, Biochim. Biophys. Acta 1025 (1990)143^151.

[12] D.D. Lasic, P.M. Frederik, M.C.A. Stuart, Y. Barenholz,T.J. McIntosh, FEBS Lett. 312 (1992) 255^258.

[13] D.D. Lasic, B. Ceh, M.C.A. Stuart, L. Guo, P.M. Frederik,Y. Barenholz, Biochim. Biophys. Acta 1239 (1995) 145^156.

[14] P.R. Harrigan, K.F. Wong, T.E. Redelmeier, J.J. Wheeler,P.R. Cullis, Biochim. Biophys. Acta 1149 (1993) 329^338.

[15] P.R. Cullis, M.J. Hope, M.B. Bally, T.D. Madden, L.D.Mayer, D.B. Fenske, Biochim. Biophys. Acta 1331 (1997)187^211.

[16] M.B. Bally, R. Nayar, D. Masin, M.J. Hope, P.R. Cullis,L.D. Mayer, Biochim. Biophys. Acta 1023 (1990) 133^139.

[17] F. Frezard, J. Liposome Res. 4 (1994) 1063^1073.[18] S.M. Gruner, R.P. Lenk, A.S. Jano¡, M.J. Ostro, Biochem-

istry 24 (1985) 2833^2842.[19] W.R. Perkins, S.R. Minchey, M.J. Ostro, T.F. Taraschi,

A.S. Jano¡, Biochim. Biophys. Acta 943 (1988) 103^107.[20] P.S. Chen Jr., T.Y. Toribara, H. Warner, Anal. Chem. 28

(1956) 1756^1758.[21] M.J.H. Janssen, D.J.A. Crommelin, G. Storm, A Hulsho¡,

Int. J. Pharm. 23 (1985) 1^11.[22] R.J. Mehlhorn, L. Packer, R. Macey, T. Alexandru, I. Dra-

gutan, Methods Enzymol. 127 (1986) 738^745.[23] W.R. Perkins, S.R. Minchey, P.L. Ahl, A.S. Jano¡, Chem.

Phys. Lipids 64 (1993) 197^217.[24] K. Anzai, M. Yoshida, Y. Kirino, Biochim. Biophys. Acta

1021 (1990) 21^26.[25] M.W. Tate, S.M. Gruner, E.F. Eikenberry, Rev. Sci. Instr.

68 (1997) 47^54.[26] R.F. Pasternack, C. Bustamante, P.J. Collings, A. Giannet-

to, E.J. Gibbs, J. Am. Chem. Soc. 115 (1993) 5393^5399.[27] J.C. de Paula, J.H. Robblee, R.F. Pasternack, Biophys. J. 68

(1995) 335^341.[28] L.D. Mayer, M.B. Bally, P.R. Cullis, Biochim. Biophys.

Acta 857 (1986) 123^126.[29] C.L. Shapiro, T. Ervin, N. Azamia, J. Keating, V. Suppers,

L. Ayash, D. Hayes, and the D99-Study Group, Programs/Proceedings of the 32nd Annual Meeting of the AmericanSociety of Clinical Oncology, Philadelphia, PA, May 18^211996, abstr. 115.

[30] E. Early, S. Shorter, A. Sugarman, G.K. Schwartz, J. Wood-ru¡, M.F. Brennan, E.S. Casper, Programs/Proceedings ofthe 32nd Annual Meeting of the American Society of Clin-ical Oncology, Philadelphia, PA, May 18^21 1996, abstr.1690.

[31] E.S. Casper, G.K. Schwartz, A. Sugarman, D. Leung, M.F.Brennan, J. Clin. Oncol. 15 (1997) 2111^2117.

[32] J.-H. Fuhrlop, J. Ko«nig, Membranes and Molecular Assem-blies: the Synkinetic Approach, Royal Society of Chemistry,Cambridge, 1994, pp. 126^139.



[33] H. Matsuki, R. Ichikawa, S. Kaneshina, H. Kamaya, I.Ueda, J. Colloid Interface Sci. 181 (1996) 362^369.

[34] H. Satake, H. Matsuki, S. Kaneshina, Coll. Surf. A Phys-icochem. Eng. Asp. 71 (1993) 135^140.

[35] A. Garnier-Suillerot, L. Gattegno, Biochim. Biophys. Acta936 (1988) 50^60.

[36] L. Gallois, M. Fiallo, A. Garnier-Suillerot, Biochim. Bio-phys. Acta 1370 (1998) 31^40.

[37] N. Henry-Toulme, B. Stefanska, E. Borowski, J. Bolard,Mol. Pharmacol. 33 (1988) 574^579.

[38] B.L.-S. Mui, P.R. Cullis, E.A. Evans, T.D. Madden, Bio-phys. J. 64 (1993) 443^453.

[39] D.D. Lasic, Nature 380 (1996) 561^562.[40] R.F. Pasternack, P.J. Collings, Science 269 (1995) 935^936.[41] B.L.-S. Mui, P.R. Cullis, P.H. Pritchard, T.D. Madden,

J. Biol. Chem. 269 (1994) 7364^7370.



Doxorubicin physical state in solution and inside ...bigbro.biophys.cornell.edu/publications/139 Li...

Documents

Transcript of Doxorubicin physical state in solution and inside ...bigbro.biophys.cornell.edu/publications/139 Li...