Sei sulla pagina 1di 8

pubs.acs.

org/Langmuir
2011 American Chemical Society

Ionization Behavior of Amino Lipids for siRNA Delivery: Determination


of Ionization Constants, SAR, and the Impact of Lipid pKa on Cationic
Lipid-Biomembrane Interactions
Jingtao Zhang,* Haihong Fan, Dorothy A. Levorse, and Louis S. Crocker
Department of Pharmaceutical Sciences, Merck Research Laboratories, Merck & Co., Inc., West Point,
Pennsylvania, United States

Downloaded by HARVARD UNIV on September 1, 2015 | http://pubs.acs.org


Publication Date (Web): January 20, 2011 | doi: 10.1021/la104590k

Received November 18, 2010


Ionizable amino lipids are being pursued as an important class of materials for delivering small interfering RNA
(siRNA) therapeutics, and research is being conducted to elucidate the structure-activity relationships (SAR) of these
lipids. The pKa of cationic lipid headgroups is one of the critical physiochemical properties of interest due to the strong
impact of lipid ionization on the assembly and performance of these lipids. This research focused on developing
approaches that permit the rapid determination of the relevant pKa of the ionizable amino lipids. Two distinct
approaches were investigated: (1) potentiometric titration of amino lipids dissolved in neutral surfactant micelles; and
(2) pH-dependent partitioning of a fluorescent dye to cationic liposomes formulated from amino lipids. Using the
approaches developed here, the pKa values of cationic lipids with distinct headgroups were measured and found to be
significantly lower than calculated values. It was also found that lipid-lipid interaction has a strong impact on the pKa
values of lipids. Lysis of model biomembranes by cationic lipids was used to evaluate the impact of lipid pKa on the
interaction between cationic lipids and cell membranes. It was found that cationic lipid-biomembrane interaction
depends strongly on lipid pKa and solution pH, and this interaction is much stronger when amino lipids are highly
charged. The presence of an optimal pKa range of ionizable amino lipids for siRNA delivery was suggested based on
these results. The pKa methods reported here can be used to support the SAR screen of cationic lipids for siRNA
delivery, and the information revealed through studying the impact of pKa on the interaction between cationic lipids and
cell membranes will contribute significantly to the design of more efficient siRNA delivery vehicles.

Introduction
Since the recent discovery of RNA interference as a powerful
mechanism for regulating gene expression,1,2 the potential to use
siRNA (small interfering RNA) as a novel therapeutic modality
has attracted significant attention and has been heavily exploited
in academia and the biotech and pharmaceutical industries.3,4
Although the use of naked siRNA for localized treatment in lung
and eye has progressed rapidly through preclinical and clinical
stages,4-6 significant challenges still remain for the systemic administration of siRNA for broader therapeutic indications. Among
them, one of the biggest hurdles is the lack of highly efficient
delivery vehicles to enable safe and efficacious delivery of siRNA
for systemic administration.7
Lipid-based delivery systems such as liposomes, lipoplexes, and
lipid nanoparticles have been broadly utilized to deliver plasmid
*Corresponding author. E-mail: jingtao_zhang@merck.com.
(1) Fire, A.; Xu, S.; Montgomery, M. K.; Kostas, S. A.; Driver, S. E.; Mello,
C. C. Nature 1998, 391, 80611.
(2) Elbashir, S. M.; Harborth, J.; Lendeckel, W.; Yalcin, A.; Weber, K.; Tuschl,
T. Nature 2001, 411, 494498.
(3) Bumcrot, D.; Manoharan, M.; Koteliansky, V.; Sah, D. W. Y. Nat. Chem.
Biol. 2006, 2, 711719.
(4) Castanotto, D.; Rossi, J. J. Nature 2009, 457, 426433.
(5) de Fougerolles, A.; Vornlocher, H. P.; Maraganore, J.; Lieberman, J. Nat.
Rev. Drug Discovery 2007, 6, 443453.
(6) Sepp-Lorenzino, L.; Ruddy, M. K. Clin. Pharmacol. Ther. 2008, 84, 628632.
(7) Whitehead, K. A.; Langer, R.; Anderson, D. G. Nat. Rev. Drug Discovery
2009, 8, 129138.
(8) Heyes, J.; Palmer, L.; Bremner, K.; MacLachlan, I. J. Controlled Release
2005, 107, 276287.
(9) Santel, A.; Aleku, M.; Keil, O.; Endruschat, J.; Esche, V.; Fisch, G.; Dames,
S.; Loffler, K.; Fechtner, M.; Arnold, W.; Giese, K.; Klippel, A.; Kaufmann, J.
Gene Ther. 2006, 13, 12221234.
(10) Akinc, A.; Zumbuehl, A.; Goldberg, M.; Leshchiner, E. S.; Busini, V.;
Hossain, N.; et al. Nat. Biotechnol. 2008, 26, 561569.

Langmuir 2011, 27(5), 19071914

DNA, and more recently to deliver siRNA.8-12 In general,


cationic lipids as well as various helper lipids are used to promote
the encapsulation of DNA or siRNA and the formation of these
lipid delivery systems with desired physicochemical and pharmacokinetic properties.13 Currently, lipid nanoparticles consisting of
amino lipids with ionizable amine headgroups are being pursued
as a promising approach to siRNA delivery, and research is being
conducted to elucidate the structure-activity relationships (SAR)
of these amino lipids.10,12,14 Among the physicochemical properties of these ionizable amino lipids, the ionization constant (Ka) of
the lipid headgroups is one of the most important attributes, since
it directly determines the charge interaction behavior of the lipids,
as well as the surface charge properties of the assembled lipid
nanoparticles. Because the interaction between cationic lipids and
nucleic acids is strongly influenced by charge-charge interaction,15 lipid pKa is expected to play a very important role in
determining not only the encapsulation of nucleic acids in the
assembled lipid particles but also the release of nucleic acids from
the particles during intracellular delivery. In addition to its effect
on lipid-nucleic acid interaction, the pKa of ionizable amino
lipids also affects the surface charge behavior of assembled lipid
nanoparticles, since surface charge densities of lipid nanoparticles
(11) Abrams, M. T.; Koser, M. L.; Seitzer, J.; Williams, S. C.; DiPietro, M. A.;
Wang, W. M.; Shaw, A. W.; Mao, X. Z.; Jadhav, V.; Davide, J. P.; Burke, P. A.;
Sachs, A. B.; Stirdivant, S. M.; Sepp-Lorenzino, L. Mol. Ther. 2010, 18, 171180.
(12) Semple, S. C.; Akinc, A.; Chen, J.; Sandhu, A. P.; Mui, B. L.; Cho, C. K.;
et al. Nat. Biotechnol. 2010, 28, 1726.
(13) Li, W. J.; Szoka, F. C. Pharm. Res. 2007, 24, 438449.
(14) Spelios, M.; Nedd, S.; Matsunaga, N.; Savva, M. Biophys. Chem. 2007, 129,
137147.
(15) Kennedy, M. T.; Pozharski, E. V.; Rakhmanova, V. A.; MacDonald, R. C.
Biophys. J. 2000, 78, 16201633.

Published on Web 01/20/2011

DOI: 10.1021/la104590k

1907

Downloaded by HARVARD UNIV on September 1, 2015 | http://pubs.acs.org


Publication Date (Web): January 20, 2011 | doi: 10.1021/la104590k

Article

depend not only on the ratio of the amino lipids to neutral lipids,
but also on the ionization state of the amine headgroup. It is
widely known that physicochemical properties of lipid nanoparticles, such as particle size, morphology, and especially surface
charge, have a dramatic impact on the performance of delivery
vehicles.13,16-20 For example, it was demonstrated in DNA
delivery that surface charges of lipid-DNA nanoparticles under
physiological conditions played a critical role in determining the
interaction of lipid nanoparticles with cell membranes and ultimately determined the internalization of the nanoparticles by cells
and the toxicity induced during delivery.19-23 Later, it was also
suggested that the surface charge densities of lipid-DNA nanoparticles in acidic environments such as those found in cellular
endosomes determined the rate of membrane fusion between
lamellar lipid vehicles and endosomes, which was one of the ratelimiting steps for transfection processes.17,24,25 Therefore, understanding the charge state of ionizable amino lipids by measuring
pKa could prove to be crucial to the understanding of the charge
behavior of lipids and lipid nanoparticles during assembly, systemic circulation, internalization, and intracellular trafficking,
and could contribute significantly to the design and development
of lipid nanoparticles for siRNA delivery.
Most conventional approaches to determining the pKa of
pharmaceutical compounds,26-28 such as potentiometric titration, spectrophotometric titration, capillary electrophoresis, or
chromatographic measurements, cannot be directly applied to
determine the relevant pKa value of ionizable amino lipids used
for siRNA delivery. For example, potentiometric titration, the
best option in determining ionization constants, requires the
titrated compounds to have sufficient aqueous solubility and be
fully dissolved in solution; unprotonated amino lipids generally
possess extremely low aqueous solubility, and they form colloidal
aggregates after protonation due to their amphiphilic nature. In
addition, it is widely known that the value of pKa is strongly affected by environmental conditions such as dielectric constant, ionic
strength, and the presence of neighboring charges.29,30 For instance, the pKa of undecyl-hydroxycoumarin shifted significantly
higher in a hydrophobic environment compared with the value
when it was in aqueous solution; the measured pKa values also
changed dramatically depending on whether it was incorporated
in a cationic or anionic surfactant micelle.29 Consequently, in
order to gain a relevant understanding of the ionization status of
amino lipids, the determination of pKa should be conducted in an
(16) Barteau, B.; Chevre, R.; Letrou-Bonneval, E.; Labas, R.; Lambert, O.;
Pitard, B. Curr. Gene Ther. 2008, 8, 313323.
(17) Ewert, K. K.; Ahmad, A.; Evans, H. M.; Safinya, C. R. Expert Opin. Biol.
Ther. 2005, 5, 3353.
(18) Resina, S.; Prevot, P.; Thierry, A. R. PLoS One 2009, 4, 11.
(19) Ahmad, A.; Evans, H. M.; Ewert, K.; George, C. X.; Samuel, C. E.; Safinya,
C. R. J. Gene Med. 2005, 7, 739748.
(20) Caracciolo, G.; Pozzi, D.; Caminiti, R.; Marchini, C.; Montani, M.; Amici,
A.; Amenitsch, H. J. Phys. Chem. B 2008, 112, 1129811304.
(21) Lv, H. T.; Zhang, S. B.; Wang, B.; Cui, S. H.; Yan, J. J. Controlled Release
2006, 114, 100109.
(22) Mislick, K. A.; Baldeschwieler, J. D. Proc. Natl. Acad. Sci. U.S.A. 1996, 93,
1234912354.
(23) Mounkes, L. C.; Zhong, W.; Cipres-Palacin, G.; Heath, T. D.; Debs, R. J. J.
Biol. Chem. 1998, 273, 2616426170.
(24) Budker, V.; Gurevich, V.; Hagstrom, J. E.; Bortzov, F.; Wolff, J. A. Nat.
Biotechnol. 1996, 14, 760764.
(25) Lin, A. J.; Slack, N. L.; Ahmad, A.; George, C. X.; Samuel, C. E.; Safinya,
C. R. Biophys. J. 2003, 84, 33073316.
(26) Albert, A.; Serjeant, E. P. Ionization Constants of Acids and Bases: A
Laboratory Manual; Methuen: London, 1962.
(27) Allen, R. I.; Box, K. J.; Comer, J. E. A.; Peake, C.; Tam, K. Y. J. Pharm.
Biomed. Anal. 1998, 17, 699712.
(28) Poole, S. K.; Patel, S.; Dehring, K.; Workman, H.; Poole, C. F. J.
Chromatogr., A 2004, 1037, 445454.
(29) Fernandez, M. S.; Fromherz, P. J. Phys. Chem. 1977, 81, 17551761.
(30) Fromherz, P. Biochim. Biophys. Acta 1973, 323, 326334.

1908 DOI: 10.1021/la104590k

Zhang et al.

environment similar to what lipids will encounter in delivery


vehicles.
Heyes et al. reported a fluorescence-based approach to assessing the ionization of amino lipids by taking advantage of the pHdependent partitioning of a fluorescent dye to lipid nanoparticles
consisting of amino lipids.8 This represented an important advancement toward determining the pKa values of amino lipids in a
relevant environment and could serve as an indirect approach to
determining the ionization status of lipid nanoparticles. However,
the determination of pKa by this method requires the assembly of
lipids and nucleic acids into lipid nanoparticles, and the value of
pKa could depend on the dye and the particle assembly process
employed. Therefore, it is challenging to apply it to a large-scale
lipid screening where the pKa values of individual lipids are
required prior to the assembly of lipid nanoparticles.
To support the structure-activity relationship (SAR) screening of cationic lipids for siRNA delivery, our research sought to
develop more straightforward approaches to determine the pKa of
amino lipids bearing ionizable amine headgroups. Here, we report
our use of two distinct approaches to determining the pKa of
amino lipids for siRNA delivery: (1) potentiometric titration of
amino lipids dissolved in neutral surfactant micelles; and (2)
pH-dependent partitioning of a fluorescent dye into cationic
liposomes formulated from amino lipids. Using the approaches
described here, the pKa values of amino lipids with distinct
headgroups are measured and compared to calculated values.
To determine the potential impact of lipid pKa on the interaction
between cationic lipids and cell membranes, pH-dependent lysis
of model membranes caused by amino lipids is evaluated and
found to be highly dependent on the pKa values of amino lipids.
The pKa methods reported here can be used to support the SAR
screen of cationic lipids for siRNA delivery, and the information
revealed through studying the impact of pKa on the interaction
between cationic lipids and cell membranes will contribute significantly to the design of more efficient siRNA delivery vehicles.

Materials and Methods


Materials. All amino lipids and the PEGylated lipid (PEGDMG) used in the study were synthesized in Merck Research
Laboratories and purities by HPLC using a charged aerosol detection (CAD) detector were determined to be greater than 90%.31
n-Octyl--D-glucopyranoside and reduced Triton X-100 were obtained from Sigma-Aldrich (Milwaukee, WI). TNS (2-(p-toluidino)6-naphthalene sulfonic acid) and carboxyfluorescein were obtained
from Molecular Probes (Eugene, OR). DOPC, DOPE, DOPS, and
cholesterol were obtained from Avanti Polar Lipids (Alabaster, AL)
and used without further purification. All buffers used in the study
were prepared from the acid or sodium salt form of the components.
Preparation of Amino Lipids in Surfactant Micelles and
Potentiometric Titration. Surfactant micelle solutions used for
potentiometric titrations (120 mM) were prepared by dissolving
the appropriate amount of n-Octyl--D-glucopyranoside in a
0.15 M NaCl solution. To aid the protonation of amino lipids
in micelles, solution pH was adjusted to 2.0 by the addition of 1 M
hydrochloric acid. Appropriate amounts of amino lipids were
then weighed and directly dissolved in the micelle solutions. The
ratio of amino lipids to surfactants was generally kept at 1:100
unless otherwise noted to ensure that at most one lipid was
incorporated in each micelle. The resulting lipid solutions were
then vigorously vortexed to permit the distribution of lipids in
surfactant micelles and were equilibrated at 4 C for one to two
days. Potentiometric titration of the micelles was performed on a
(31) Chen, T.; Vargeese, C.; Vagle, K.; Wang, W.; Zhang, Y. Lipid Nanoparticle
Based Compositions and Methods for the Delivery of Biologically Active
Molecules. U.S. Patent 7,641,915 B2, 2010.

Langmuir 2011, 27(5), 19071914

Zhang et al.

Article

Sirius GLpKa/D-PAS instrument (Sirius Analytical Instruments,


East Sussex, UK) using a double junction electrode. The sample
was titrated with 0.5 M KOH at 25 C from pH 2.0 to pH 11.0,
and the data were analyzed using Refinement Pro v 2.2 software or
processed manually using buffer capacity analysis. Duplicate
measurements of pKa using this method generally differed from
one another by no more than 0.2.

Formulation of Cationic Lipids into Cationic Liposomes.


Cationic lipids were formulated into cationic liposomes as described.11 Briefly, cationic lipids (30 mM) were directly suspended
in a pH 4.0 citrate buffer (25 mM citrate and 100 mM NaCl)
containing 3 mM PEG-DMG (PEG 2000-dimyristoyl glycerol).
The resulting lipid suspension was vigorously vortexed and sonicated for 30 min to form a homogeneous solution.

Downloaded by HARVARD UNIV on September 1, 2015 | http://pubs.acs.org


Publication Date (Web): January 20, 2011 | doi: 10.1021/la104590k

TNS Fluorescence Assay for the Determination of Lipids


pKa. Except as noted otherwise, 30 mM cationic lipid liposome

solution was diluted to 75 M cationic lipid in a series of buffers


with pH ranging between 3 and 12 (buffers are composed of
10 mM citrate, 10 mM phosphate, 10 mM borate, and 150 mM
NaCl). A stock solution of TNS dissolved in DMSO was then
added to the above buffer solution to make a 6 M TNS solution.
All procedures were performed using a Tecan Evo liquid handling
robot. The fluorescence of the resulting solution was read on a
SpectraMax M5 fluorescence plate reader (Molecular Devices,
Sunnyvale, CA) with the excitation wavelength (ex) set at 325 nm
and the emission wavelength (em) set at 435 nm. pH of the
samples was confirmed using an automated Tiamo pH measurement robot. The fluorescence of TNS was plotted against pH and
fitted using a three-parameter-sigmoid function (see below). It is
assumed that in the presence of amino lipids TNS fluorescence
reaches a maximum when 100% of the amino lipids are ionized,
while TNS has little fluorescence when the amino lipids are in the
un-ionized state.8 The pH values at which half of the maximum
fluorescence is reached in the TNS fluorescence assay are reported
as the apparent pKa values of amino lipids.
Fitting function : FL

a

!
pH - pKa
1 exp
b

where FL is TNS fluorescence, pKa is apparent pKa of amino lipid,


a is maximum TNS fluorescence in the presence of amino lipids,
b is a measure of the cooperativeness of the protonation process of
amino lipids; and pKa-b is the pH at which fluorescence reaches
73% of maximum fluorescence.

Liposome Lysis for Determining the Impact of Lipid pKa


on Cationic Lipid-Biomembrane Interactions. Liposomes
mimicking cell membranes were prepared from phospholipid
mixtures containing DOPC, DOPE, DOPS, and cholesterol at a
weight ratio of 45:20:20:15.32 Briefly, 10 mg of lipid mixture
dissolved in chloroform was evaporated on a rotary evaporator
to remove the bulk solvent and further dried overnight under
vacuum. The resulting lipid film was hydrated with 1 mL of carboxyfluorescein solution (100 mM carboxyfluorescein, 20 mM
phosphate, 20 mM citrate, pH 8.0) and sonicated for 60 min to
form 100 nm membrane-mimicking liposomes as determined by
dynamic light scattering (Zetasizer Nano, Malvern Instruments).
Unencapsulated carboxyfluorescein was separated from liposomes through size exclusion chromatography with a NAP-25
column (GE Healthcare) using a pH 8.0 buffer (20 mM phosphate, 20 mM citrate, and 100 mM NaCl) as eluant.
To evaluate the pH-dependent lysis of carboxyfluoresceinliposomes by cationic lipid liposomes, purified carboxyfluorescein-liposomes were further diluted 200-fold in pH 7.5, 6.0, and
5.4 buffers (20 mM phosphate, 20 mM citrate, and 100 mM NaCl)
and the final lipid concentration was 25 M; varying amounts of
(32) Koynova, R.; Wang, L.; MacDonald, R. C. Proc. Natl. Acad. Sci. U.S.A.
2006, 103, 1437314378.

Langmuir 2011, 27(5), 19071914

cationic lipid liposomes were then added to the above liposome


solution to reach final cationic lipid concentrations of 50 M. All
procedures were performed using a Tecan Evo liquid handling
robot. Fluorescence intensities of carboxyfluorescein (ex = 480
nm, em = 520 nm) were measured after 30 min on a fluorescence
plate reader. Fluorescence intensities at maximum (I100) or minimum (I0) lysis were determined in the presence of 0.5% reduced
Triton X-100 or a blank phosphate-citrate buffer at the same pH.
Lysis activities caused by the cationic lipids were then calculated
as % liposome lysis = (I - I0)/(I100 - I0)  100, and data were
reported as the average of triplicate measurements.

Results and Discussion


Potentiometric Titration of Amino Lipids Prepared in
Surfactant Micelles. Although the pKa of an acid or base in
aqueous solution can be conveniently determined through potentiometric titration by strong base or acid, the low solubility of
amino lipids in aqueous solutions prohibits the direct measurement of pKa using this approach. In addition, the pKa values of
amino lipids determined in an aqueous environment generally do
not represent the true protonation and deprotonation behavior
when amino lipids are incorporated in lipid nanoparticles due to
the change in the chemical environment that these lipids will
encounter. For example, amino lipids in lipid nanoparticles will be
surrounded predominately by other lipids with low dielectric
constants, while molecules in aqueous solutions encounter a high
dielectric constant solvent-water. This large difference in dielectric constants will shift the acid-base equilibrium toward
noncharged species and effectively decrease the pKa of an amine.29
Dissolving amino lipids in neutral surfactant micelles is an
effective approach to solubilize them and could be used to determine the apparent pKa of amino lipids in a hydrophobic environment similar to the one in lipid nanoparticles.24 Due to the
amphiphilic nature of neutral surfactants, they self-assemble in
aqueous solutions to form micelles with hydrophobic tails facing
in and hydrophilic headgroup facing out (i.e., water) when the
concentration is above the critical micelle concentration (CMC).
Because the amino lipid is also amphiphilic when protonated, it
can be easily solubilized in surfactant micelles and will orient itself
in a micelle with the lipid tail buried inside the micelle. This
configuration closely resembles the conditions that the lipid will
experience in a lipid nanoparticle and also permits potentiometric
titrations in a conventional manner.
To facilitate the titration of amino lipids in neutral surfactant
micelles, 1.2 mM amino lipid in 120 mM n-octyl--D-glucopyranoside solution was prepared. The ratio of surfactants to amino
lipids was chosen to be greater than the micelle aggregate number
to minimize potential charge interactions between multiple amino
lipids (see section II for more discussion). Figure 1 shows a
representative titration curve and corresponding buffer capacity
analysis obtained from the titration of lipid 1 in 0.15 M NaCl
solution. Table 1 lists the pKa value obtained from the titration
curve of lipid 1 as well as the measured and calculated pKa values
of other amino lipids and model compounds with diverse chemical structures (the structure of lipids is shown below and in
Supporting Information Figure S1).

Lipid drug delivery formulations routinely utilize sterols as


functional excipients to adjust biomembrane fluidity to achieve
DOI: 10.1021/la104590k

1909

Article

Zhang et al.

Figure 1. (a) Potentiometric titration of lipid 1 (1.2 mM) prepared in neutral surfactant micelles (120 mM); surfactant micelles and titration

media were in 0.15 M ionic strength adjusted water. (b) Buffer capacity analysis (d(OH-)/d pH) of the potentiometric titration in part (a); the
squares (0) correspond to experimental data and the line came from the fitting of the data. The peak in maximum buffer capacity
corresponded to the pKa of the titrated lipid.
Table 1. Measured (Micelle Method and TNS Fluorescence Method) and Calculated pKa Values of Amino Lipids and Model Compounds at 25 Ca
Downloaded by HARVARD UNIV on September 1, 2015 | http://pubs.acs.org
Publication Date (Web): January 20, 2011 | doi: 10.1021/la104590k

ID

lipid headgroup

lipid tail L1

lipid tail L2

measured pKa (micelle)

measured pKa (TNS)

calculated pKab

1
dimethyl amine
linoleyl
octyl cholesteryl ether
7.9
7.7
8.6
2
dimethyl amine
linoleyl
butyl cholesteryl ether
8.2
8.1
8.6
3
morpholine
linoleyl
butyl cholesteryl ether
5.7
6.1
6.8
4
pyrrolidine
linoleyl
butyl cholesteryl ether
8.4
8.0
9.6
5
dimethyl amine
oleyl
hexyl cholesteryl ether
8.1
8.2
8.6
6
diethyl amine
oleyl
hexyl cholesteryl ether
7.9
7.3
9.6
7
1-methyl pyrrolidine
oleyl
hexyl cholesteryl ether
7.6
7.3
9.7
8
3-fluoropiperidine
oleyl
hexyl cholesteryl ether
6.6
6.1
6.6
9
pyrrolidine
oleyl
hexyl cholesteryl ether
8.1
7.8
9.6
10
piperidine
oleyl
hexyl cholesteryl ether
7.4
7.1
8.5
11
morpholine
oleyl
hexyl cholesteryl ether
5.3
4.8
6.8
12
imidazole
oleyl
hexyl cholesteryl ether
5.6
5.6
6.9
13
benzyl amine
oleyl
hexyl cholesteryl ether
5.2
4.1
7.7
14
methyl piperazine
oleyl
hexyl cholesteryl ether
2.8/7.9
7.9
3.3/7.6
9.1
8.8
3-(dimethylamino)-1,2-propanediolc
c
9.7
8.8
3-piperidino-1,2-propanediol
a
Measured pKa values for amino lipids are determined either by potentiometric titration in the presence of neutral surfactants (micelle method) or by a
TNS fluorescence approach. b Calculated pKa values are obtained from ACD Laboratories v 11.0 software. c pKa values of model compound are
determined directly in the absence of micelles.

desired pharmacokinetic properties. Recently, strategies to incorporate sterols as part of lipid structures were shown to further
improve biomembrane properties, and formulations based on
these sterol-modified lipids were successfully applied in drug
delivery and siRNA delivery.11,31,33 For instance, a new class of
asymmetric cationic lipids consisting of a cholesteryl ether tail and
an aliphatic tail (see the above generic lipid structure) was found
effective in mediating the delivery of siRNA in vitro and in
vivo.11,31 To understand the relevant ionization behavior of these
lipids, we decided to develop appropriate methods to evaluate the
pKa of a series of these ionizable amino lipids. Headgroups of
these ionizable amino lipids (Table 1) were selected to span across
a broad range of calculated pKa values to evaluate the SAR of
lipids. A subset of the lipids was chosen to have identical tails to
allow for a simple assessment of the impact of headgroup structures on pKa.
From Figure 1a, it is clear that potentiometric titration of lipids
dissolved in micelles resembled those expected from the titration
of fully soluble molecules. This shows that the micelle titration
method can be used to understand the ionization behavior of
cationic lipids. Buffer capacity analysis in Figure 1b demonstrates
that the pKa of cationic lipid can be easily obtained from the
potentiometric titration data. Comparison between the pKa values of amino lipids prepared in micelles and the pKa values of
(33) Huang, Z.; Szoka, F. C., Jr. J. Am. Chem. Soc. 2008, 130, 1570212.

1910 DOI: 10.1021/la104590k

model compounds consisting of lipid headgroups alone (i.e., lipid


1, 2, 5, and 10 vs model compounds) reveals that the pKa values of
these ionizable amine headgroups are lower by more than 1 after
lipids are incorporated in micelles. Similar decreases are observed
between the measured pKa values and those calculated through
modeling. This decrease in the pKa values of lipids assembled in
micelles is consistent with the increase in the hydrophobicity
around lipid headgroups after they are incorporated in micelles
and suggests that the pKa of amino lipids in lipid nanoparticles
could also be lower than those calculated and those of the model
compounds. It is noted that, although majority of lipid headgroups appear to have decreased measured pKa values compared
with calculated values, the extent of this decrease appears to
depend on headgroup structures. For example, certain lipid headgroups (e.g., benzyl amine: lipid 13) can have a greater difference
between measured and calculated pKa values compared with
other lipid headgroups (e.g., 3-fluoropiperidine: lipid 8). This
might be due to the differences in the hydrophobicities of lipid
headgroups and suggests that it will be difficult to find a universal
correction factor between measured and calculated pKa values.
Additionally, these results also show that the use of only calculated pKa values to guide the SAR evaluation of cationic lipid
headgroups might be insufficient to obtain the correct understanding on the impact of lipid pKa on siRNA delivery. However,
it is further noted that three subsets of lipids with the same headgroups and different tail structures (i.e., dimethyl amine: lipids 1,
Langmuir 2011, 27(5), 19071914

Zhang et al.

Article

Table 2. Impact of Multiple Amino Lipids on the Apparent pKa of


Amino Lipidsa
amino lipid
in surfactant (mol %)

estimated average numbers of


measured apparent
amino lipid molecules
pKa of amino lipid
per surfactant micelleb

Downloaded by HARVARD UNIV on September 1, 2015 | http://pubs.acs.org


Publication Date (Web): January 20, 2011 | doi: 10.1021/la104590k

0.5
0.4
8.2
1
0.8
8.2
1.25
1
8.2
1.5
1.2
8.1
2
1.6
8.0
2.5
2
7.9
5
4
7.7
10
8
7.4
a
pKa values for amino lipids are determined by potentiometric
titration in the presence of neutral surfactants (micelle method).
b
Micelle aggregate number of 80 is used in the calculation.34

2, and 5; morpholine: lipids 3 and 11; pyrrolidine: lipids 4 and 9)


show similar but measurable difference in measured pKa values.
This suggests that measured pKa values are mostly determined by
the chemical structures of headgroups while tail structures will
also contribute to the results, likely by influencing the conformation of cationic lipids within micelles. Therefore, it might be
possible to estimate the pKa range of other lipids with similar
headgroups using the data reported here. Finally, although only
lipids with asymmetric tail and glycerol backbone structures were
tested here, it was observed that this micelle titration method
could also be applied to lipids with symmetric tails or different
backbone structures (data not shown). This suggests that the
micelle preparation method to determine lipid pKa is a convenient
approach that could be broadly applicable to other classes of ionizable amino lipids.
Potentiometric Titration of Multiple Amino Lipids Incorporated in a Micelle. The use of surfactant micelles to
dissolve amino lipids allows for the assessment of the impact of
charge interactions between lipid headgroups on the value of pKa.
Most pKa values reported previously in the literature are determined under conditions where amines or acids are soluble in
solution and will not interact with each other during ionization.
However, in a self-assembled lipid delivery vehicle, cationic lipids
are packed within a 100 nm particle and headgroups of lipids are
in close contact with each other. Therefore, due to charge repulsion, the charge on a protonated amine headgroup will inhibit the
acceptance of another proton by the neighboring amine headgroup. To assess the impact of charge interaction and lipid packing on the value of pKa, surfactant micelles were prepared with
varying ratios of surfactants to lipid (lipid 2). Table 2 lists the
measured pKa values of lipid 2 in surfactant micelles as a function
of the number of amino lipid molecules encapsulated per micelle.
When cationic lipid concentration is less than 1.25%, the ratio of
surfactant to amino lipid is greater than the micelle aggregate
number (80)34 and there is less than one cationic lipid per
micelle. The pKa at these concentrations corresponds to lipid pKa
without any contribution from the charge interaction between
cationic lipids, and the measured values are independent of the
lipid concentration in the solution. Although measurement of pKa
at a low lipid concentration is desirable for obtaining the intrinsic
value of lipid pKa, ionization of the dissolved carbonate in solution will interfere with the measurement at low lipid concentrations. Therefore, intrinsic pKa values of amino lipids are generally
measured at a lipid concentration of 1% (1 mM amino lipid). At
cationic lipid concentrations between 1.5% and 10%, there will be
more than one lipid within each micelle. It is obvious that the
(34) Lorber, B.; Bishop, J. B.; Delucas, L. J. Biochim. Biophys. Acta 1990, 1023,
254265.

Langmuir 2011, 27(5), 19071914

presence of multiple lipids within micelles significantly lowers the


pKa values of lipids and the pKa values decreased systematically
with the increasing amount of amino lipids within micelles. The
decrease in pKa values is likely due to the increasing charge repulsion between protonated amine headgroups. It is noted that the
highest level of tested cationic lipid concentration in a micelle is
10% due to the decreased solubility of amino lipids in micelles.
However, cationic lipid concentration in lipid nanoparticles could
be as high as 50%. This suggests that the pKa values of cationic
lipids in lipid nanoparticles where lipids are in close proximity
with each other will likely be even lower than the pKa values determined here.
TNS Fluorescence Approach to Determining the Apparent pKa of Amino Lipids. An alternative nonpotentiometric
approach for determining the pKa values of amino lipids is to
assemble amino lipids into liposomes and utilize the chargedependent binding property of TNS (2-(p-toluidino)-6-naphthalene sulfonic acid) to indirectly measure the apparent pKa.35 TNS
is a negatively charged fluorescent dye, and the partition of TNS is
stronger into positively charged lipid membrane due to electrostatic interactions. After sequestration of TNS in lipid membranes, the fluorescence of TNS is significantly increased due to
the removal of water, which acts as a fluorescence quencher, from
the environment of the TNS molecule. Therefore, the increase of
TNS fluorescence in the presence of ionized amino lipids could be
used to estimate the surface charges of lipid membranes and
provides an approach to measure the apparent pKa of amino
lipids. To determine the pKa of amino lipids using TNS fluorescence, cationic lipids were formulated into cationic lipid liposomes by suspending cationic lipids (30 mM) with a PEG-DMG
lipid solution (3 mM) and further homogenized using sonication.
The resulting liposomes resembled unilamellar vesicles by CryoTEM (data not shown) and were generally in the range 100500 nm. To determine the optimum lipid concentration for TNS
fluorescence, varying amounts of lipid 2 were combined with a
fixed concentration of TNS (6 M) in buffers at three different
levels of pH. The fluorescence results are shown in Figure 2. As
shown, TNS fluorescence shows a strong dependence on both
buffer pH and lipid concentration. For example, TNS fluorescence at pH 5.0 increased linearly as a function of lipid concentration until the fluorescence plateaus when lipid concentration is
above 75 M. The increase in TNS fluorescence in the presence of
lipid suggests that TNS partitions into the hydrophobic lipid
environment. However, the concentration dependence behavior
of TNS fluorescence suggests that at low lipid concentrations the
binding of TNS to cationic lipids is limited by the amount of
binding sites in lipids, while at high lipid concentrations, all TNS
molecules bind to cationic lipids and fluorescence values correspond to the maximum TNS fluorescence when they are all in a
hydrophobic environment. It is noted that the TNS/lipid ratio at
the transition point should correspond to the maximum TNS
concentration in cationic lipids (i.e., solubility of TNS in lipid
membranes). This value (1/12.5) is significantly lower than 1,
which suggests that the binding of TNS to lipid is significantly
hindered by the accessibility of lipids to TNS. To maximize
fluorescence signal and minimize lipid usage, 75 M lipid concentration was used in all later experiments. Binding of TNS to
amino lipids at higher pH (8 and 11) showed similar concentration-dependent profiles although with significantly reduced fluorescence values. The decrease of TNS fluorescence as the pH
increases is consistent with the hypothesis that less TNS binds to
surfaces as the charges of amino lipids are neutralized at high pH.
(35) Bailey, A. L.; Cullis, P. R. Biochemistry 1994, 33, 1257312580.

DOI: 10.1021/la104590k

1911

Article

Downloaded by HARVARD UNIV on September 1, 2015 | http://pubs.acs.org


Publication Date (Web): January 20, 2011 | doi: 10.1021/la104590k

Figure 2. TNS fluorescence as a function of lipid 2 concentration


at three levels of pH; all solutions contain the same concentration
of TNS (6 M). Fluorescence values and error bars are the results
from triplicate measurements.

This suggests that TNS can be used as a reporter molecule for the
charge state of amino lipids.
To further determine the usefulness of TNS to measure pKa of
amino lipids, TNS fluorescence was evaluated at a broader pH
range for multiple lipids. Figure 3 shows the representative plot of
TNS fluorescence as a function of pH in the presence of lipid 2 and
five other cationic lipids. It is clear that TNS fluorescence is
significantly affected by the charge states of the lipids, and the pHdependent fluorescence data follow a simple sigmoid function.
The profiles shown here are consistent with the TNS binding
data on other lipids previously reported in the literature.8,14 The
resemblance of these curves to the sigmoid function of an ionization event suggests that TNS can be used to measure the pKa of
amino lipids. A fit of the data using a sigmoid function reveals the
pH at which fluorescence values are half of the maximum fluorescence values. These pH values correspond to apparent pKa
values of the amino lipids. Apparent pKa values of amino lipids
from TNS fluorescence assays are reported in Table 1 and compared with the pKa values measured from micelle titration method
and calculation. It was shown that the apparent pKa values
determined from TNS method are generally in good agreement
with the values determined from the micelle titration method,
while both values are significantly lower than those estimated
from modeling. This shows again that the determination of pKa of
amino lipids in a relevant hydrophobic environment is critical in
understanding the ionization behavior of amino lipids.
It is noted that the TNS fluorescence assay for pKa determination relies on the binding of TNS to the positive charges of the
lipid headgroups, which have a relatively smaller size compared
with TNS molecule. As mentioned above, binding of TNS to
lipids is limited by the accessibility of lipids to TNS. Therefore, if
amino lipids bearing multiple amines are used, it might be
challenging for multiple TNS molecules to bind around one lipid
headgroup due to even greater steric hindrance, and therefore, no
increases in TNS fluorescence will be observed even if more than
one amine on a single lipid headgroup are ionized. In fact,
evaluation of the TNS fluorescence of lipid 14 (dibasic amine
with measured pKa around 2.8 and 7.9 (based on the micelle
approach)) at pH 2 showed the same fluorescence value as pH 5,
while the cationic lipid liposome showed much greater surface
charge at pH 2 compared with pH 5. A further evaluation of the
lipid 14 concentration-dependent fluorescence profiles at pH 1
and 5 showed overlapping fluorescence values at all concentration
1912 DOI: 10.1021/la104590k

Zhang et al.

Figure 3. Representative plots showing TNS fluorescence as a


function of pH in the presence of cationic lipids (75 M); all
solutions contain the same concentration of TNS (6 M). Fluorescence values shown here are normalized by the maximum TNS
fluorescence values of each lipid. The apparent pKa values of the
amino lipids are the pH values at which TNS fluorescence is half of
the maximum fluorescence and are obtained after fitting the data
using a sigmoid function (discussed in detail in the Methods and
Materials section).

ranges (data not shown). This suggests that TNS does not respond
to two charges on a single lipid molecule, and it is unlikely that the
TNS fluorescence approach will be able to differentiate among the
individual pKa values of the multiple amines on the same lipid
headgroup.
Impact of Lipid pKa on the Interaction between Cationic
Lipids and Model Cell Membranes. The charge state of the
amine headgroups on cationic lipids directly determines surface
charge properties of lipid nanoparticles, which are mostly composed of cationic lipids and neutral lipids and can have a dramatic
impact on the efficacy and toxicity of these lipid nanoparticles as
siRNA delivery vehicles. For example, the delivery of siRNA
using lipid nanoparticles generally needs to overcome several
critical extracellular and intracellular barriers such as blood protein adsorption, biodistribution, cell uptake, endosomal escape,
dissociation of siRNA from delivery vehicles, and binding of
siRNA with RNA interference silencing complex (RISC).36 A
common feature in many of these steps is the involvement of
negatively charged biological components including peptides,
proteins, and most importantly the external and internal cell
membranes. It was hypothesized that the charge state of cationic
lipids mostly affects its biological performance by influencing the
charge-charge interaction between cationic lipids and negatively
charged cell membranes and blood proteins.21-23,37-39 In addition to affecting the interaction with external cell membranes
which happens at physiological pH (7), the charge state of
amino lipids at acidic pH significantly affects the ability of
cationic lipids to interact with internal cell membranes in an
acidified environment such as those in the endosomes (pH 5-6)
and therefore plays an important role in determining the efficiency of endosomal escape of lipid nanoparticles, a key ratelimiting step in siRNA delivery.24,25
(36) Juliano, R.; Bauman, J.; Kang, H.; Ming, X. Mol. Pharmaceutics 2009, 6,
686695.
(37) Xu, Y.; Szoka, F. C., Jr. Biochemistry 1996, 35, 561623.
(38) Zelphati, O.; Szoka, F. C., Jr. Proc. Natl. Acad. Sci. U.S.A. 1996, 93,
114938.
(39) Caracciolo, G.; Pozzi, D.; Amenitsch, H.; Caminiti, R. Langmuir 2007, 23,
87138717.

Langmuir 2011, 27(5), 19071914

Zhang et al.

Article

Downloaded by HARVARD UNIV on September 1, 2015 | http://pubs.acs.org


Publication Date (Web): January 20, 2011 | doi: 10.1021/la104590k

Figure 4. Lysis of membrane-mimicking liposomes by ionizable amino lipids with diverse amine headgroups. Amino lipids tested here all
have the same tail structure of one oleyl and one hexyl cholesteryl ether tail. The pKa values of amino lipids are determined using the micelle
method (see Table 1 for detail). Liposome lysis is conducted in buffer solutions at pH 5.4, 6.0, and 7.5 using carboxyfluorescein as a reporter
for lysis. Extent of liposome lysis is reported as an average of triplicate measurements of lysis activities at 50 M cationic lipids concentration.

To evaluate the impact of the ionization of lipid headgroups on


the interaction between cationic lipids and biomembranes, cationic lipids with pKa values ranging between 5 and 8 are formulated
into cationic lipid liposomes, and their ability to lyse a model cell
membrane over a range of physiologically relevant pH conditions
(pH 7.5, 6.0, and 5.4) was used to assess their tendency to interact
with biomembranes. A phospholipid mixture of DOPC/DOPE/
DOPS/cholesterol (45:20:20:15, w/w) was used to construct liposomes to mimic cellular membranes,32 with carboxyfluorescein
encapsulated as a reporter for membrane lysis.40 Due to the selfquenching of carboxyfluorescein at high concentrations when
encapsulated inside liposomes mimicking biomembranes, little or
no fluorescence should be detected. Upon incubation with cationic lipid liposomes, interactions between cationic lipids and model
biomembranes will cause fusion and breakup of liposome membranes and the subsequent release of carboxyfluorescein into
solution. An increased fluorescence signal can be detected due to
the dilution of carboxyfluorescein in solution and the reduced
quenching effect at low carboxyfluorescein concentrations. The
extent of liposome lysis is quantified based on the relative amount
of fluorescence signal increase, normalized to the fluorescence increase from complete lysis caused by the addition of surfactants.
The liposome lysis activities of five ionizable amino lipids with
different headgroups (with pKa spanning across the physiologically relevant pH range), and otherwise, the same tail structures
are compared in Figure 4. Tail structures of lipids were kept the
same to ensure that lysis differences observed only resulted from
the differences in the lipid headgroups (i.e., ionization states of
amine headgroups). As shown, lysis activities of cationic lipids
depended strongly on the pKa values of the amino lipids. For
example, amino lipids with high pKa headgroups (pyrrolidine,
dimethyl amine, and piperidine) had strong lysis activities at pH
5.5, while cationic lipids with much lower pKa (e.g., morpholine
headgroup) demonstrated little or no lysis activities at all three pH
conditions. The lack of lysis activity for the amino lipid with
morpholine headgroup is probably due to the lack of ionization in
the lipid headgroup as its pKa is lower than the lowest pH
conditions tested (5.4). This suggests that cationic charges in the
lipid headgroup play an important role in determining the
interaction between amino lipid and cell membranes. Consistent
(40) Ralston, E.; Hjelmeland, L. M.; Klausner, R. D.; Weinstein, J. N.;
Blumenthal, R. Biochim. Biophys. Acta 1981, 649, 133137.

Langmuir 2011, 27(5), 19071914

with this result, closer inspection of the lysis activities of cationic


lipids with high pKa headgroups (pyrrolidine, dimethyl amine,
piperidine, and fluoropiperidine) shows that their lysis activities
have a strong pH dependency, with a much higher lysis activity
under acidic pH conditions and almost no lysis activity at pH 7.5.
This is related to the increased ionization of lipids at low pH, as
amino lipids will be more positively charged at acidic conditions
and are expected to interact more strongly with the negatively
charged biomembranes. In contrast, amino lipids with pKa close
to or below 7.5 will have little tendency to interact with membranes at physiological pH since most of these amino lipids will
remain neutral at this pH. Further inspection of the data shows
that the greatest differences in pH-dependent lysis activities were
achieved with the cationic lipids with pKa values close to physiological pH (pH 7.5). For example, the cationic lipid bearing a
piperidine headgroup showed a difference of 37% between the
lysis activities at pH 7.5 and 5.4, while the cationic lipid with
a 3-fluoropiperidine group (pKa of 6.6) showed only a difference
of 12%.
Results in Figure 4 clearly show that the interaction between
amino lipid and cell membranes depends heavily on the ionization
state of the amino lipids and can be tuned systematically by adjusting the pKa of the lipid headgroup. Importantly, these data
suggest that the use of a lipid with an ionizable lipid headgroup
could be beneficial for siRNA delivery, since they can mediate a
pH-dependent membrane interaction. A pH-dependent ability of
cationic lipids to interact with cell membranes is desirable because
a lower level of interaction with biomembranes at physiological
pH can help to avoid toxicity during systemic circulation, while
stronger interactions with endosomal membranes under acidic
environments can cause higher membrane fusion rates and could
potentially lead to improved endosomal escape.13,41 From a pure
charge-charge membrane interaction perspective, there will be
an optimal pKa value for amino lipids to achieve the optimal pHdependent membrane interaction behavior. For siRNA intracellular delivery, we speculate that the optimal pKa value of the
amino lipids will be in the range 5-8, because the headgroup of
amino lipids needs to be protonated within the acidified endosomes (pH 5-6.5) to interact with the membranes while the cationic
(41) Semple, S. C.; Klimuk, S. K.; Harasym, T. O.; Dos Santos, N.; Ansell,
S. M.; Wong, K. F.; Maurer, N.; Stark, H.; Cullis, P. R.; Hope, M. J.; Scherrer, P.
Biochim. Biophys. Acta 2001, 1510, 152166.

DOI: 10.1021/la104590k

1913

Article

Zhang et al.

Downloaded by HARVARD UNIV on September 1, 2015 | http://pubs.acs.org


Publication Date (Web): January 20, 2011 | doi: 10.1021/la104590k

charge of the headgroup need to be masked during systemic circulation (pH 7.4). It will be interesting to evaluate this hypothesis in
vitro or in vivo and identify the optimal lipid pKa value for siRNA
delivery. However, it is noted that the optimal pKa value could
depend heavily on the extent of acidification in endosomes and
thus could be different when delivering to different cell types or
going through different intracellular pathways. It is further noted
that successful intracellular siRNA delivery depends on not only
the endosomal escape of delivery vehicle, but also the final release
of siRNA from the delivery vehicle, a step which could also be
affected by the pKa of amino lipids and the interaction strength
between amino lipid and siRNA.19,20,39,42 Therefore, it might be
difficult to find a universal optimal lipid pKa value for siRNA
delivery. Finally, it is expected that coupling of the lipid pKa
determination methods described here with methods that can
permit quantitative understanding of endosomal escape efficiency
or membrane interaction mechanism of amino lipids could greatly
help to advance the rational design and use of amino lipid for
siRNA delivery.

Summary and Perspectives


Ionizable amino lipids represent an important class of materials
for siRNA therapeutics delivery. Rapid synthesis and structural
variation on ionizable amino lipids are being carried out to
optimize the lipids for delivery, and demand approaches for the
physicochemical properties of amino lipids to inform structureactivity relationships (SAR). pKa of lipids, an important parameter in SAR study, is critical for understanding the phase and
self-assembly property of the ionizable amino lipids. It also plays
a very important role in determining the interaction between
amino lipids and cell membranes as well as blood proteins, which
ultimately determines the delivery efficacy and toxicity of the
assembled lipid carrier. We described two distinct approaches
that permit the rapid determination of pKa of ionizable amino
(42) Tarahovsky, Y. S.; Koynova, R.; MacDonald, R. C. Biophys. J. 2004, 87,
10541064.

1914 DOI: 10.1021/la104590k

lipids. The first method determines the pKa of lipids using


potentiometric titration by suspending amino lipids in neutral
surfactant micelles, while the second method takes advantage of
the charge-dependent partition of TNS to cationic lipids. Both
methods are straightforward, can be easily automated, and are
well-suited for supporting the pharmaceutical development of
cationic lipids for siRNA delivery. However, due to the ease of
sample preparation and the ability to measure headgroups with
multiple charges, the micelle method appears to be more broadly
applicable to determine the pKa of ionizable amino lipids. Using
the methods developed here, pKa values of a set of different amino
lipids are measured and observed to be significantly lower than
calculated values. These values are also found to be highly dependent on the amount of cationic lipids incorporated per micelle.
The impact of pKa of lipid headgroups on cationic lipidbiomembrane interaction is assessed using lipids with pKa values
ranging between 5 and 8. It is found that cationic lipidbiomembrane interaction is strongly affected by the ionization
state of amino lipid headgroup and is strongest when amino lipids
are highly charged. The use of lipid headgroups with pKa in the
range 6-8 provides pH-dependent membrane interaction within
physiological pH range, which suggests the presence of an optimal
amino lipid pKa for siRNA delivery. Approaches developed here
can be used to support the SAR evaluation of cationic lipids for
siRNA delivery and help to advance the development of siRNA
as a novel therapeutic modality.
Acknowledgment. We would like to thank Ye Zhang for the
discussions on liposome lysis and Andrew Shaw for the methods
of preparing cationic lipid liposomes. Review and comments on
the manuscript by Merck colleagues are gratefully acknowledged.
Provision of cationic lipids and PEGylated lipids by Merck
medicinal chemists is gratefully acknowledged.
Supporting Information Available: Additional figure as
described in text. This material is available free of charge via
the Internet at http://pubs.acs.org.

Langmuir 2011, 27(5), 19071914

Potrebbero piacerti anche