Lipid-modified oligonucleotide conjugates: Insights into gene silencing, interaction with model membranes and cellular uptake mechanisms

Article abstract of DOI:10.1016/j.bmc.2016.10.024

The ability of oligonucleotides to silence specific genes or inhibit the biological activity of specific proteins has generated great interest in their use as research tools and therapeutic agents. Unfortunately, their biological applications meet the limitation of their poor cellular accessibility. Developing an appropriate delivery system for oligonucleotides is essential to achieve their efficient cellular uptake. In the present work a series of phosphorothioate lipid–oligonucleotide hybrids were synthesized introducing covalently single or double lipid tails at both 3′- and 5′-termini of an antisense oligonucleotide. Gene transfections in cultured cells showed antisense luciferase inhibition without the use of a transfecting agent for conjugates modified with the double-lipid tail at 5′-termini. The effect of the double lipid-tailed modification was further studied in detail in several model membrane systems as well as in cellular uptake experiments. During these studies the spontaneous formation of self-assembled microstructures is clearly observed. Lipidation allowed the efficient incorporation of the oligonucleotide in HeLa cells by a macropinocytosis mechanism without causing cytotoxicity in cells or altering the binding properties of the oligonucleotide conjugates. In addition, both single- and double-tailed compounds showed a similar behavior in lipid model membranes, making them useful in nucleotide-based technologies.

Full text of DOI:10.1016/j.bmc.2016.10.024

Accepted Manuscript
Lipid-Modified Oligonucleotide Conjugates: Insights into Gene Silencing, In-
teraction with Model Membranes and Cellular Uptake Mechanisms
Begoña Ugarte-Uribe, Santiago Grijalvo, Samuel Nú ñez Pertí ñez, Jon V.
Busto, César Martín, Adele Alagia, Félix M. Goñi, Ramón Eritja, Itziar Alkorta
PII:
S0968-0896(16)31034-3
DOI:
http://dx.doi.org/10.1016/j.bmc.2016.10.024
BMC 13351
Reference:
Bioorganic & Medicinal Chemistry
To appear in:
Received Date:
Revised Date:
Accepted Date:
22 August 2016
18 October 2016
19 October 2016
Please cite this article as: Ugarte-Uribe, B., Grijalvo, S., Nú ñez Pertí ñez, S., Busto, J.V., Martín, C., Alagia, A.,
Goñi, F.M., Eritja, R., Alkorta, I., Lipid-Modified Oligonucleotide Conjugates: Insights into Gene Silencing,
Interaction with Model Membranes and Cellular Uptake Mechanisms, Bioorganic & Medicinal Chemistry (2016),
doi: http://dx.doi.org/10.1016/j.bmc.2016.10.024
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Lipid-Modified Oligonucleotide Conjugates: Insights into Gene
Silencing, Interaction with Model Membranes and Cellular Uptake
Mechanisms
Begoña Ugarte-Uribe1, Santiago Grijalvo2,3, Samuel Núñez Pertíñez2, Jon V. Busto1,
César Martín1, Adele Alagia2, Félix M. Goñi1, Ramón Eritja2,3, Itziar Alkorta1*
1
Instituto Biofisika (UPV/EHU, CSIC), and Departamento de Bioquímica, Universidad del País
Vasco, P.O. 644, Bilbao 48080, Spain
Institute of Advanced Chemistry of Catalonia (IQAC-CSIC), Jordi Girona 18-26, E-08034 Barcelona,
2
Spain
3
Networking Center on Bioengineering, Biomaterials and Nanomedicine (CIBER-BBN)
* Correspondence: Email address: itzi.alkorta@ehu.es (I. Alkorta); Tel: +34-94-6012568; fax: +34-94-
6013500

Keywords: Antisense technology, Cellular uptake, Gene delivery, Lipid-
oligonucleotide conjugates, Model membrane systems

Abstract
The ability of oligonucleotides to silence specific genes or inhibit the biological activity
of specific proteins has generated great interest in their use as research tools and
therapeutic agents. Unfortunately, their biological applications meet the limitation of
their poor cellular accessibility. Developing an appropriate delivery system for
oligonucleotides is essential to achieve their efficient cellular uptake. In the present
work a series of phosphorothioate lipid-oligonucleotide hybrids were synthesized
introducing covalently single or double lipid tails at both 3’- and 5’-termini of an
antisense oligonucleotide. Gene transfections in cultured cells showed antisense
luciferase inhibition without the use of a transfecting agent for conjugates modified with
the double-lipid tail at 5’-termini. The effect of the double lipid-tailed modification was
further studied in detail in several model membrane systems as well as in cellular
uptake experiments. During these studies the spontaneous formation of self-assembled
microstructures is clearly observed. Lipidation allowed the efficient incorporation of the
oligonucleotide in HeLa cells by a macropinocytosis mechanism without causing
cytotoxicity in cells or altering the binding properties of the oligonucleotide conjugates.
In addition, both single- and double-tailed compounds showed a similar behavior in
lipid model membranes, making them useful in nucleotide-based technologies.

1. Introduction
Nucleic acid-based therapy is acknowledged as a promising approach to treat a
large number of diseases.1 Antisense, small interfering RNA (siRNA), aptamers and
exon-skipping are the most used mechanisms to generate novel drugs based on nucleic
acids. For this reason, nucleic acid therapy has been considered as an alternative to
small molecule drugs.2 However, serious drawbacks regarding stability and cellular
uptake need to be still improved. The use of non-viral vectors in combination with
nucleic acids has remarkably increased in the last decades due to its simple large scale
production and low host immunogenicity. 3
There are two major strategies for associating non-viral vectors with nucleic acids.
The formulation approach is often one of the most commonly-used strategies since the
combination between vector and cargo is carried out electrostatically.4 The second
approach involves the use of a covalent strategy that allows the non-viral vector to be
directly linked to nucleic acids. The latter strategy leads to stable conjugates that
normally improve their biological properties in terms of stability, pharmacokinetic
profiles and internalization without altering the properties of natural nucleic acids.5 This
approach has allowed the development of synthetic strategies to obtain covalently
polymers,6 peptides,7 lipids,8 among others9 conjugated with nucleic acids.
Lipids are commonly used for modifying oligonucleotides.10 The importance of
such hybrid conjugates has been demonstrated in areas like nanobiotechnology,11 cell
biology12 and medicine.5 From a therapeutic point of view, lipid-oligonucleotide
conjugates (LOCs) have emerged as an alternative to cationic particles in order to
reduce toxicity and improve biodistribution.13 Nonetheless, a few studies deal with their
biophysical and biochemical properties and their interactions with cellular
components.14-17 Oligonucleotide entry into cells normally takes place through receptor-

mediated endocytosis by binding with high affinity to cell surface receptors.18 However,
this uptake efficiency could change when modified oligonucleotides are internalized
into the cell.
In this context, a number of issues remain unsolved such as the LOCs interaction
with the plasma membrane, the mechanism by which these kinds of conjugates are
internalized and their final destination within cells.19 These pathways may vary with the
membrane coat composition, the transport vesicle size and the fate of the internalized
materials. Additionally, understanding of how cellular uptake and subcellular trafficking
of lipid oligonucleotide conjugates is an essential starting point to gain insight into
pharmacological effectiveness and gene silencing activities of LOCs. Many synthetic
strategies have been reported in recent years for obtaining oligonucleotides covalently
linked to a variety of lipophilic moieties.20,21 The conjugation of these units can occur at
either the nucleoside or the nucleotide level, giving rise to the corresponding
amphiphilic-based compounds.20 Other strategies have incorporated lipid moieties
through a phosphodiester bond in order to obtain LOC conjugates with better
pharmacokinetic behaviors and improved cellular uptake. In general, the
oligonucleotide ends (3’- and 5’-termini) are the most accessible conjugation sites for
lipophilic units.21
The primary aim of the current study is to synthesize a small library made up of
phosphorothioate LOCs which contains a series of neutral lipids that were covalently
introduced at both the 3’- and 5’-termini. This library was initially designed to evaluate
the efficiency for knocking down the mRNA Renilla luciferase gene through antisense
technology. We also evaluated its ability to provide suitable cellular uptake and inhibit
luciferase gene expression in the absence of a transfection agent. After selecting the
optimal LOC, we explored its physicochemical and biological properties in model

membranes with different lipid compositions and its internalization into cells. Our
results constitute a starting point for the development of novel lipid modifications and
more effective amphiphilic oligonucleotides for antisense, aptamer, exon-skipping and
single- or double-stranded siRNA technologies.22

2. Materials and Methods
All chemicals, reagents and solvents were purchased from Sigma-Aldrich (St Louis,
MO) or Acros Organics (Geel, Belgium) and were used as received. All the standard
phosphoramidites (A, C, G and T) and ancillary reagents used for oligonucleotide
synthesis were purchased from Applied Biosystems (Foster City, CA) or Link
Technologies (Glasgow, UK). Phosphorothioate oligonucleotide sequences: 5’-CGT
TTC CTT TGT TCT GGA-3’ (Luc, 19, Table 1) and 5’-CTG TCT GAC GTT CTT
TGT-3’ (Scr, Table 1) were purchased from Sigma-Aldrich (St Louis, MO). The
sequences 5’-TCC AGA ACA AAG GAA ACG-3’ (Table 1), 5’-CTC TCG CAC CCA
TCT CTC TCC TTC T-3’ (GEM91, Table 1) and 5’-A GAA GGA GAG AGA TGG
GTG CGA GAG-3’ (22, Table S1 in Supporting Information) were synthesized in
house (phosphate form). Additional nuclease-free water was also prepared by using
0.1% (v/v) of diethylpyrocarbonate (DEPC) to ensure the removal of RNase
contamination. Luciferase assay kits were purchased from Promega (Fitchburg, WI).
Qiagen Giga plasmid purification kit was purchased from Qiagen (Hilden, Germany).
1,2-dioleoyl-L-α-phosphatidylcholine (DOPC), egg sphingomyelin (eSM) and
cholesterol (Chol) were from Avanti Polar Lipids (Alabaster, AL). Mouse monoclonal
antibody to mannose-6-phosphate receptor (M6PR), rabbit polyclonal antibody to
clathrin heavy chain (HC) and rabbit polyclonal antibody to caveolin-1 were from
Abcam (Cambridge, UK) and HRP-linked Goat anti-rabbit IgG antibody was supplied
by Cell Signaling Technology (Danvers, MA). Lipofectamine 2000, Oligofectamine,
OptiMEM, Alexa Fluor 546 phalloidin, Alexa Fluor 546 Donkey anti-mouse IgG (H+L)
antibody, Alexa Fluor 633 Goat anti-rabbit IgG (H+L) antibody, Hoechst 33258, DNA
stain, DiD dye, SYBR® Gold nucleic acid gel stain and 1 Kb Plus DNA Ladder were
from Life Technologies (Carlsbad, CA). Caveolin-1 siRNA, clathrin HC siRNA and

rabbit polyclonal antibody to glyceraldehyde-3-phosphate dehydrogenase (GAPDH)
were purchased from Santa Cruz Biotechnology (Santa Cruz, CA, USA). 5-(N,N-
Dimethyl)amiloride hydrochloride (DMA), chlorpromazine hydrochloride (CPZ),
cytochalasin D (Cyt D), genistein, nocodazole, nystatin and the VenorGEM
Mycoplasma Detection Kit were supplied from Sigma (St Louis, MO).
2.1 Organic synthesis
All reactions were carried out under an inert atmosphere of dry nitrogen. Reaction
and mixture of compounds were monitored by TLC on silica gel (F254 Merck plates).
TLC plates were visualized with phosphomolybdic acid in ethanol (20%). Flash
1
chromatography was performed using SDS 0.063-0.2 mm/70-230 mesh silica gel. H
and 13C NMR spectra were recorded at 25ºC on a Varian Mercury 400 MHz
spectrometer (Unitat de RMN, IQAC-CSIC) using deuterated solvents.
1
Tetramethylsilane (TMS) was used as an internal reference (0 ppm) for H spectra and
13
the residual signal of the solvent (77.16 ppm) for C spectra, both recorded in CDCl3.
For CD3OD the residual signal of the solvent was used as a reference. Chemical shifts
are reported in part per million (ppm) in the d scale, coupling constants (J) in Hz and
multiplicity as follows: s (singlet), d (doublet), t (triplet), q (quadruplet), quint
(quintuplet), sext (sextuplet), m (multiplet) and br (broad signal). Electrospray
ionization mass spectra (ESI-MS) were recorded on a Micromass ZQ instrument with
single quadrupole detector coupled to an HPLC, and high-resolution (HR) ESI-MS on
an Agilent 1100 LC/MS-TOF instrument (Servei d’Espectrometría de Masses,
Universitat de Barcelona). Absorbance measurements and melting temperature studies
were recorded on a Jasco FP-6200 spectrofluorometer equipped with a thermoregulated
cell holder as well.

2.1.1 Synthesis of the double-tailed lipid derivatives 12 and 13
Scheme 1 shows the main steps of the synthesis of the double-lipid derivatives. Further
information regarding experimental procedures can be found in Supporting Information.
2.1.2 Synthesis of the phosphorothioate LOCs
Antisense LOCs (LOC-14, -15, -16, -17) (Table 1 and Supporting Information)
were synthesized on a DNA/RNA synthesizer (Applied Biosystems) on a 1 µmol scale
using the corresponding lipid functionalized CPG-solid supports (CPG-1, -2, -3) and the
lipid phosphoramidite-4 (Figure 1). Coupling yields were >97%. The resulting DMTon-
oligonucleotides (LOC-14, -15, -16) and LOC-17 were deprotected and cleaved from
the solid supports with aqueous concentrated ammonia (32%) at 55ºC overnight. The
mixtures were filtered and ammonia solutions were concentrated to dryness. DMTon-
crudes and LOC-17 were purified by semi-preparative HPLC. Solvent A: 5% ACN in
100 mM triethylammonium acetate (TEAA) (pH = 7) and solvent B: 70% ACN in 100
mM TEAA (pH = 7). Column: XBridgeTM OST C18 semipreparative column (10x50
mm, 2.5 mm). Flow rate: 3 mL/min. Conditions: 15 min of linear gradient from 0 to
85% of B. The pure DMTon-peak was concentrated and dissolved in AcOH (80%) in
order to remove the last DMT protecting group. After 30 min, the deprotected antisense
oligonucleotides were washed with ethylic ether and finally desalted with Sephadex-
G25 (NAP-10 column). The isolated oligonucleotides LOC-14, -15, -16, -17 were
finally analysed by HPLC (Column: XBridgeTM OST C18 column; (4.6x50 mm, 2.5
mm); Flow rate: 1 mL/min) and analysed by MALDI-TOF mass spectrometry (Table 1
and Supporting Information).

The oligonucleotide conjugate T10-18 was isolated following the same experimental
conditions described above (Table 1 and Supporting Information).
2.1.3 Synthesis of fluorescently labelled DNA sequence 24
For Alexa Fluor 488 labeling, a commercially available 5’-amino-modifier-C6-Tfa
phosphoramidite (Glen Research) was introduced into the sequence 5’-A GAA GGA
GAG AGA TGG GTG CGA GAG-3’ (22, Table S1 and Supporting Information)
obtaining the respective 5’-protected amino-derivative (23, Table S1 and Supporting
Information). The CPG solid support was deprotected according to experimental
conditions described above. Finally, oligonucleotide amino derivative 23 was analyzed
by HPLC and confirmed by MALDI-TOF mass spectrometry (Table S1 and Supporting
Information). 23 was used without further purification in the next step.
A solution of the deprotected oligonucleotide 23 was passed through a column
filled with Dowex 50x2, sodium form to remove the interfering ammonium ions coming
from the cleavage step. It was dissolved with 50 µL of an aqueous solution of 0.2 M
NaHCO3 (pH 9). Then, 1.1 equiv of Alexa Fluor 488 tetrafluorophenyl ester dissolved
in 30 µL DMF was added to the solution and left to react overnight at room
temperature. The mixture was concentrated to dryness, and the residue was dissolved in
H2O (1 mL). The green solution was desalted with Sephadex G-25 (NAP-10 Column)
and the anticipated fluorescently labelled oligonucleotide 24 (Table S1 and Supporting
Information) was purified by semi-preparative HPLC and confirmed by MALDI-TOF
mass spectrometry.
2.2 Thermal denaturation studies
The thermal melting curves for duplexes formed by the phosphorothioate LOCs
(LOC-14, -15, -16, -17), or the unmodified oligonucleotide 19, and the DNA

complementary strand in its phosphate form (mixed in the appropriate buffer to a 1:1
molar ratio) were carried out recording the absorbance changes at 260 nm with a linear
temperature ramp of 0.5ºC/min (Figure S1 in Supporting Information). The
measurements were recorded on a JASCO V-650 spectrophotometer equipped with a
Peltier temperature control. Sample concentrations were 1 µM in 100 mM phosphate
buffer supplemented with 100 mM NaCl, pH 7.02. Measurements were performed in
triplicate.
2.3 Formation of DNA duplexes
Double-stranded LOC hybrids were prepared by mixing LOC-20 or LOC-21 with
the sequence complementary strand 22 at equimolar ratio. For fluorescence experiments
one of the duplex strands was labelled with Alexa 488. The annealing procedure was
performed by incubating samples dissolved in phosphate-buffered saline (1 PBS)
solution at 95ºC for 2 min, followed by a cooling step at 1ºC /min to room temperature
in a temperature-controlled water circulating bath Julabo F10-MH (Julabo
Labortechnik, Seelbach, Germany). The formation of the DNA duplexes was checked
by either non-denaturing polyacrylamide or agarose gel electrophoresis (Figure S2 in
Supporting Information). DNA duplexes were stored at -20ºC.
2.4 Transfection studies carried out with or without lipofectamine
The luciferase inhibition assay was carried out on Mycoplasma free-HeLa (human
cervical carcinoma) cell cultures that were maintained in exponential growth (Figure S3
in Supporting Information). Cells were seeded one day prior to the experiment in a 24-
well plate at a density of 105 cells/well. Transfection of the antisense LOCs (LOC-14, -

15, -16, -17) were carried out in the absence or presence of Lipofectamine 2000 as
described by the manufacturer for adherent cell lines.
For Lipofectamine-treated cells, pRL-TK (0.1 µg) and pGL3 (1.0 µg) plasmids,
which were used as a reporter and control, together with the phosphorothioate LOCs
(LOC-14, -15, -16, -17) (60 nM), were formulated into liposomes and added to each
well with a final volume of 600 µL. After an incubation period of 20 h, cells were
washed once with PBS and were lysed with passive lysis according to the
manufacturer’s instructions.
To evaluate the ability of the LOC conjugates to silence mRNA Renilla luciferase,
DMEM either supplemented with 10% FBS or without FBS, was used to carried out
transfection experiments in the absence of Lipofectamine 2000. Firstly, pRL-TK (0.1
µg) and pGL3 (1.0 µg) plasmids were transfected by using commercially available
cationic lipids. After 5 h incubation, cells were thoroughly washed with PBS (2 x 500
µL) and fed with fresh medium (either supplemented with 10% FBS, or without FBS,
depending on the type of the transfection experiment). Phosphorothioate LOCs (LOC-
14, -15, -16, or -17) were added at a particular concentration and incubated for 24 h.
After 24 h cells were lysed with passive lysis buffer according to the instructions of the
Dual-Luciferase Reporter Assay System. An unmodified antisense oligonucleotide 19
and a scramble oligonucleotide were used as controls. The luciferase activities of the
samples were measured in a luminometer using the Luciferase Assay Reagent II and
Stop and Glo Reagent. Lumminiscence was measured by using SpectraMax M5
spectrophotometer. The inhibitory effects generated were expressed as normalized ratios
between the activities of the reporter (Renilla) luciferase gene and the control (firefly)
luciferase gene.

2.5 Surface tension measurements
The studies were carried out at 22ºC using a µTrough-S equipment (Kibron Inc.,
Finland) consisting of a small (2-cm) rounded multi-well plate that allowed for 1 mL
subphase measurements. For possible LOC-21 interaction with lipids, DOPC:eSM:Chol
(2:2:1) monolayers were formed at the air-buffer interface at increasing initial pressures
by spreading a small amount of lipid (about 2 nmol) in chloroform:methanol (2:1, v/v)
solution over 1 PBS. After allowing for solvent evaporation, LOCs were injected into
the subphase with constant stirring and the changes in surface pressure were recorded.
The plateau values of surface pressure increase () after LOC insertion were plotted as
a function of initial  value (i), and line fitted. Average values  SD (n=2).
2.6 SPB and GUV preparation
SPB preparation was performed as described previously.23 Sucrose-enclosing
GUVs were prepared using the electroformation method as described previously.23
DOPC, DOPC/eSM/Chol (2:2:1) and eSM:Chol (2:1) stock solutions were prepared
containing 0.4 mol% DiD (Invitrogen, Waltham, Massachusetts, USA). Vesicles were
generated in a 300 mM sucrose solution at 60ºC, a temperature rendering bilayers in a
fluid phase state. 30 min after electroformation and vesicle equilibration at room
temperature, 50 L vesicles were pipetted onto 250 L of an equiosmolar 1 PBS
buffer solution in BSA-pre-treated Lab-Tek Chamber slides (Thermo Fisher Scientific,
NY, USA).
2.7 Analysis of LOC incorporation into GUVs and SPBs by confocal microscopy
Images were obtained in an inverted fluorescence confocal microscope (Leica
TCSSP5, Leica Microsystems, Wetzlar, Germany). Briefly, BSA-pre-treated 8 well-

Lab-Tek Chamber slides containing GUVs or Biocell-liquid-cell containing SPBs were
placed on the microscope followed by image acquisition with a 40 LD objective using
an excitation wavelength of 488 for Alexa 488 and 633 nm for DiD, and the emission
was recovered between 505 and 540 nm for Alexa 488 and 655 and 795 nm for DiD.
Binding of Alexa 488-labelled LOC-21 to fluorescently labelled GUVs or SPBs was
observed at 22ºC under the microscope in sequential mode imaging. Image treatment
was performed using Leica Application Suite software.
2.8 Cellular binding/uptake studies by flow cytometry
HeLa cells seeded in 24-well plates (1.25105 cell/well) were pretreated with
endocytic inhibitors for 30 min at 37ºC in OptiMEM followed by incubation with 500
nM Alexa 488-labeled LOC-21 for 2 h at 37ºC.Selected inhibitors: 0.3 M sucrose, 10
µg/mL CPZ, 5 µg/mL filipin, 200 µM genistein, 100 µM DMA, 10 µg/mL Cyt D, 20
µM nocodazole were tested (Figure S4 in Supporting Information). After incubation,
cells were completely dissociated with 0.5% Trypsin-EDTA, washed by centrifugation
at 520  g for 10 min and resuspended into ice-cold 1 PBS, followed by a flow
cytometry analysis (FACSCalibur, Beckton Dickinson) analyzing 104 cells per sample.
The fluorescence of Alexa 488-labeled LOC-21 remaining at the cellular surface was
also quenched with 0.2% Trypan Blue (final concentration) to measure only the
fluorescence corresponding to the internalized LOC-21. Alexa 488 fluorescence was
measured using an air-cooled 488 nm argon-ion laser, collecting emission signals by
FL1 detector. GMFI (Geometric Mean Fluorescence Intensity) data of histograms were
used for analysis comparison using WinMDI 2.9 and Cell Quest (BD Biosciences)
softwares.

2.9 siRNA-mediated silencing protein expression
The expression of clathrin heavy chain and caveolin-1 proteins were silenced by
siRNA treatment. siRNA transfection protocols were performed following the
manufacturer’s instructions. Cells seeded in 24-well plates (1.25105 cell/well) were
treated with siRNA-oligofectamine mixtures in OptiMEM. After siRNA incubation for
4-5 h, 1 mL opti-MEM containing 20% FBS was added to the wells without removal of
the transfection mixture. The cells were incubated further for 24 h, the medium was
replaced by appropriate growth medium, and cells were incubated for another 24 h.
Cells were scrapped, counted and used for cellular uptake assays and Western blotting
experiments in order to test siRNA silencing efficiency.
2.10 Western blotting
After siRNA treatment, HeLa cells were harvested and lysed in ice-cold
homogenization buffer. Proteins were electrophoretically separated on a 12% SDS-
polyacrylamide gel and transferred to nitrocellulose membranes. The membranes were
then blocked for 1 h with 5% skim milk in Tris-buffered saline (TBS) containing 0.01%
NaN3 and 0.1% Tween 20, then incubated overnight with the primary antibody (1:1,000
dilution) in TBS-0.1% Tween 20 at 4ºC. After three washes with TBS-0.1% Tween 20,
membranes were incubated with horseradish peroxidase-conjugated secondary antibody
at 1:4,000 dilution for 1 h. Equal loading of protein was monitored using a specific
antibody against GAPDH protein. Proteins were detected using the enhanced
chemiluminiscence detection system (ECL®, Amersham Biosciences). The Quantity
One® Image Analyzer software program (Bio-Rad, Hercules, CA) was used for
quantitative densitometric analysis.

2.11 Cellular localization assays by confocal microscopy
Cells seeded in Poly-L-lysine pre-treated 8 well-Lab-Tek Chamber slides (5104
cell/well) were pre-incubated with endocytic inhibitors (100 M DMA, 0.3 M sucrose
or 5 g/mL filipin) for 2 h at 37ºC, followed by incubation with 100 nM Alexa 488-
labeled LOC-21 for 30 min at 37ºC in OptiMEM. After incubation, cells were washed
three times, fixed with 4% (w/v) paraformaldehyde for 15 min at room temperature
(RT) and permeabilized with 0.1% (w/v) Triton X-100 dissolved in 1 PBS for 10 min
at RT. Next, samples were blocked with 1% (w/v) BSA dissolved in 1 PBS at RT for
30 min. After blocking samples were incubated with the appropriate primary antibodies
(1:200 dilution) at RT for 1 h followed by incubation with the appropriate Alexa-
conjugated secondary antibodies (1:200 dilution) at RT for 1 h. Then samples were
stained with nuclear marker Hoechst 33258 at RT for 10 min and visualized using
Olympus IX 81 inverted fluorescence confocal microscope. Image acquisition was
performed with a digital camera (Axiocam NRc5, Zeiss). Samples were placed on top of
the microscope followed by image acquisition with a 60X oil immersion objective,
using an excitation wavelength of 405, 488, 543 and 633 nm for Hoechst 33258, Alexa
488, Alexa 546 and Alexa 633, respectively. The emission was recovered between 430-
460 nm for Hoechst 33258, 505-525 nm for Alexa 488, 560-600 nm for Alexa 546, 660-
IF nm for Alexa 633. Image treatment was performed using Fluoview v.50 software
(Olympus Life Science, Pennsylvania, USA).
2.12 Statistical analysis
All measurements were performed at least 3 times, and results are presented as
mean ± s.d. Levels of significance were determined by a two-tailed Student's t-test, and

a confidence level (*p < 0.05, **p < 0.01, ***p < 0.001) was used to establish statistical
significance.

3. Results and discussion
3.1. Synthesis of phosphorothioate lipid-oligonucleotide conjugates
The strategy followed for introducing lipophilic residues at the 3’- and 5’-termini of an
antisense oligonucleotide was based on obtaining the appropriate modified CPG-solid
supports21 (CPG-1, -2, -3) and the lipid phosphoramidite-4,21 respectively (Figure 1).
Following this strategy, small series of siRNA conjugates containing neutral and
cationic residues were obtained and their RNAi silencing activities were previously
evaluated in cell culture. We found that a modified siRNA conjugate containing a lipid
double-tailed pendant group at the 5’-termini showed the best transfection efficiency
compared to other lipophilic conjugates also synthesized.21 Inspired by these results, the
development of a synthetic strategy to introduce the same double lipid-tailed
modification at the 3’-termini of a phosphorothioate oligonucleotide was found
particularly interesting.
The synthesis of the solid support 3 (CPG-3) containing the double lipid-tailed
modification with two aliphatic chains of fourteen carbon atoms in length is depicted in
Scheme 1. First, the double alkylation of our starting material 5 in the presence of a
phase-transfer catalyst and 50% NaOH solution in toluene gave the corresponding
alkylated compound 6 in 52% yield.24 Next, Boc-deprotection in acid conditions (30%
trifluoroacetic acid in dichloromethane) produced the anticipated amine in its
trifluoroacetate form that was converted to the anticipated free amine derivative 7 by
using a cationic ion exchange resin in carbonate form.25
INSERT FIGURE 1. Lipid functionalized CPG-solid supports (CPG-1, CPG-2, CPG-3) used for the
synthesis of antisense LOCs at the 3’-termini (A) and the lipid-phosphoramidite derivative

(phosphoramidite-4) used for modifying the 5’-termini of an antisense phosphorothioate (PS) and DNA
phosphate (PO) oligonucleotides (B).
A
X
R
X = OMe; CPG-1
X
R
X = OMe; CPG-2
O
O
O
O
O
NH
O
X = H;
CPG-3
O
O
NH
B
N
NC
P
O
O
O
O
phosphoramidite-4
We then studied both dimethoxytrityl and trityl as protecting groups to selectively
block the primary alcohol of the triol 8. This reaction generated the anticipated
protected diols 9 and 10 in 40 and 70% yield, respectively. In order to combine the
amino lipid 7 and the diols 9 and 10,26,27 we thought about obtaining the matching
carbamate as the best approach for getting the modified lipid derivative with a trityl
residue. Although the introduction of p-nitrophenyl chloroformate into the amino lipid 7
worked well in our hands and quantitatively yielded the carbamate 11, several attempts
to displace the leaving group with the appropriate alcoxide (generated from the diols 9
or 10 in the presence of sodium hydride) were not as efficient as we originally thought
they would be. The anticipated lipophilic compounds 12 and 13 were achieved although
none of the reactions exceeded a 15% yield. Finally, with the tritylated lipid derivative
13 in our hands, we were able to functionalize CPG solid support according with well-
established methods28 yielding the expected modified solid support CPG-3 (L = 44
µmol/gr).

INSERT SCHEME 1. Synthesis of the Controlled Pore Glass (CPG)-solid support 3 containing the
double-tailed lipid modification used for the conjugation of an antisense oligonucleotide at 3’-termini. (a)
C14H29Br, Bu4NHSO4, 50% NaOH, Toluene, 60ºC, overnight; (b) i. 30% TFA in CH2Cl2, RT, 1 h; ii.
Carbonate resin, AcOEt, RT, 1 h; (c) 4-nitrophenyl chloroformate, CH2Cl2:THF (1:1), DIPEA, RT, 3 h;
(d) NaH (60%), 9 or 10, DIPEA, DMF, overnight at RT or 60ºC, for 9 and 10, respectively; (e) i. 13,
succinic anhydride, DMAP, CH2Cl2, RT, overnight; ii. CPG functionalization; (f) DMTr-Cl or Tr-Cl,
DMAP, pyridine, 6 h at RT or 60ºC for DMTr-Cl and Tr-Cl, respectively.
a
b
yield = 53%
yield = 100%
7
6
5
c
d
yield = 100%
11
12; X= OMe (yield = 15%)
13; X= H (yield = 10%)
e
13
f
9; X= OMe (yield = 40%)
10; X= H (yield = 70%)
8
3
Two sequences modified with lipids (Luc and GEM91, Table 1) were selected for the
present work. The first one corresponds to a 18-mer phosphorothioate oligonucleotide
(Luc: 5’-CGT TTC CTT TGT TCT GGA-3’) that is complementary to the coding
region of the mRNA Renilla luciferase gene.29 The second one corresponds to a
phosphate 25-mer oligonucleotide (GEM91: 5’-CTC TCG CAC CCA TCT CTC TCC
TTC T-3’) which is complementary to the gag initiation site of HIV.30 The covalent
introduction of neutral lipid residues into sequence 19 (Table 1) at 3’-termini was
carried out from the CPG-solid supports (CPG-1, -2, -3). In addition, the lipid
phosphoramidite-421 was also incorporated at the 5’-termini of the same antisense
oligonucleotide. Once the sequences were completed, detachment from the solid support

and full deprotection of the nucleobases protecting groups were achieved using an
overnight treatment with ammonia (32%) at 55ºC (see Materials and Methods section
for further experimental details). Purification of the antisense LOCs (LOC-14, -15 and -
16) was carried out according to DMTon-based protocols, except for the LOC-17 DMTr
protecting group, which was deprotected at the end of the DNA synthesis). All LOCs
were analyzed by HPLC and finally characterized by MALDI-TOF mass spectrometry
(Table 1 and Supporting Information), affording the expected MS except for LOC-16
which gave a mass that corresponded to the loss of the double lipid moiety. To confirm
this result, we coupled the same solid-support CPG-3 on a decathymidine sequence at
the 3’-termini which afforded the conjugate T10-18. Although the MS analysis by
MALDI-TOF showed the same loss of the lipid residue observed earlier, ESI
experiments confirmed the exact mass of the conjugate T10-18 (Table 1 and Supporting
Information).
The effect of the terminal lipids at the 3’- and 5’-end on the hybridization properties of
the LOC conjugates (LOC-14, -15, -16, -17) was measured by thermal denaturation
analysis of the corresponding duplexes. Results are displayed on Table 1 and Figure S1
(Supporting Information). These data showed similar thermal stabilities in all cases
(50.3, 51.5, 51.5 and 50.9ºC for LOC-14, -15, -16 and -17, respectively) with respect to
the unmodified sequence 19 (50.8ºC). These results indicated that introducing lipid
pendants at the ends of a phosphorothioate oligonucleotide single strand did not cause
any significant destabilizing effect on the duplex structure. Similar results were reported
in literature when hydrocarbonated lipids were introduced at the end of the siRNA
passenger strand.21, 31,32
INSERT TABLE 1. MALDI-TOF mass spectrometry and melting temperatures for lipid-
oligonucleotide conjugates shown in this work

Lipid
Oligonucleotide
Conjugate
(LOC)
Mass
[calcd]
Mass
[found]
Name1
Modification2
Backbone3
Tm / ºC (ΔTm)4
Luc
Luc
Luc
3’-C14 (CPG-1)
3’-C18 (CPG-2)
3’-C28 (CPG-3)
PS
PS
PS
LOC-14
LOC-15
LOC-16
6097
6131
6410
6095
6136
60127
50.3 (-0.6)
51.5 (0.6)
51.5 (0.6)
5’-C28
(phosphoramidite-4)
Luc
PS
LOC-17
6276
6279
50.9 (0.1)
T10
3’-C28 (CPG-3)
PO
PS
T10-185
19
3643
nd
3643
Nd
nd
Luc
unmodified
50.8
Scr
unmodified
PS
PO
PO
Scr
nd
Nd
nd
nd
nd
5’-C28
(phosphoramidite-4)
GEM91
GEM91
LOC-206
LOC-216
8147
8629
8149
8631
5’-C28
(phosphoramidite-4)
1 Luciferase (Luc): 5’-CGT TTC CTT TGT TCT GGA-3’; T10: Decathymidine; Scramble (Scr): 5’-CTG TCT GAC
GTT CTT TGT-3’; GEM91: 5’-CTC TCG CAC CCA TCT CTC TCC TTC T-3’
2 3’-lipid at the 3’-termini or 5’-lipid at the 5’-termini and reagents used for the incorporation of the modification
3 PS: phosphorothioate; PO: phosphodiester
4 ΔTm corresponds to the difference between the melting temperatures of the corresponding phosphorothioate
LOCs/antisense strand (phosphate form) hybrids minus the melting temperature of the unmodified phosphorothioate
oligonucleotide (19)/antisense strand duplex (phosphate form; 5’-TCC AGA ACA AAG GAA ACG-3’)
5 see Supplementary Information for additional information: MALDI-TOFF mass spectrometry and HPLC
chromatogram
6 for more information see ref. 25
7 main mass peak corresponds to the loss of the two -C14 aliphatic chains
nd: not determined
3.2 In vitro gene silencing studies
The silencing activity of the amphiphilic phosphorothioate LOCs (LOC-14, -15, -
16, -17) and the effect of the lipophilic moieties in antisense technology were studied in
mammalian cells using the unmodified oligonucleotide (19) as a control. First, the
silencing efficiency of the lipophilic LOCs on Renilla mRNA knockdown was
evaluated in serum-free conditions at 60 nM in the absence of a transfecting agent in
order to avoid unwanted side effects of FBS on the transfection process (Figure 2A). All
amphiphilic conjugates were able to silence luciferase activity (46%, 23%, 26% and

49% for LOC-14, LOC-15, LOC-16 and LOC-17, respectively). The highest antisense
silencing efficiencies were shown by LOC-14 and LOC-17 (46% and 49%,
respectively) which improved the silencing activity compared to the unmodified
phosphorothioate 19 (22%) (*p < 0.01 and ***p < 0.001 for LOC-14 and LOC-17,
respectively). The silencing efficiency for LOC-15 and LOC-16 showed that the
presence of the alkenyl and the saturated double-lipid residues at the 3’-termini was not
as effective as LOC-14 and LOC-17 (23 and 26%, versus 46 and 49%, respectively).
Although it has been described that both the lipophilicity degree and the position of
the hydrocarbonated chain may affect the cellular uptake process,33 we did not expect
big differences between LOC-16 and LOC-17. A potential source for these differences
(***p < 0.001) could be hypothesized due to differences on their degradability
capabilities. Thus, in the case of LOC-17, the 5’-phosphate hydrolysis between the
double-lipid alkyl chain at the 5’-end and the antisense oligonucleotide might keep
intact the antisense oligonucleotide and might facilitate endosomal escape enhancement.
Once released, it would be ready to recognize and block a particular mRNA. As a result,
it would improve the corresponding silencing efficiency.
Unexpectedly, this luciferase inhibition efficiency was not found in the case of
LOC-16 as well as for other 3’-LOCs, in our hands. This difference might be due to the
presence of threoninol-based lipid modification at the 3’-termini which might slow
down the hydrolysis process. Thus, the oligonucleotide-3’-lipid would remain longer in
the endosomal compartment. Interestingly, during our studies on model membranes
explained below, we observed that the oligonucleotide-lipid carrying the double lipid at
the 5’-end (LOC-21) had a tendency to self-assemble into larger structures and these
process may facilitate cellular uptake process. The linker needed to attach the double
linker at the 3’-end may disrupt this self-assembling process, disturbing this

phenomenon. Moreover, we found that LOC-21 showed and enhanced internalization
via receptor-mediated endocytosis in β2 integrin (CR3) expressing cells23 but nothing is
known for oligonucleotide-3’-lipid. Hence, this clearly indicates that the modification at
the 5´-end enables the binding and interaction between the oligonucleotide and its
interacting partners, such as β2 integrin receptor.
LOC-14 and LOC-17, which afforded the best results in Figure 2A, were selected
for further transfection studies. We evaluated the LOCs silencing activities in the
presence of a commercially available transfecting agent and 10% FBS at 60 nM of LOC
in order to study whether the effect of lipid modifications together with their position in
the oligonucleotide (3’- or 5’-termini) could disrupt the antisense machinery (Figure
2B). We also included LOC-16 because was modified with the same double lipid
residue than LOC-17 but at the 3’-termini. Although all conjugates inhibited luciferase
expression (61%, 46% and 10% for LOC-14, LOC-16 and LOC-17, respectively), these
inhibition results were lower than the inhibition showed by the unmodified
phosphorothioate oligonucleotide 19, which displayed the best inhibition profile (82%)
of the series. Specifically, the silencing activities diminished gradually when lipid
modifications containing one or two hydrocarbon branches were introduced at the 3’-
termini (61% and 46% for LOC-14 and LOC-16, respectively) (*p < 0.05). Regarding
the hydrocarbon chain position (either 3’- or 5’- termini), the inhibition activity of
Renilla luciferase in the presence of lipofectamine was practically abolished in our
hands when the double lipid modification was introduced at the 5’- termini (10% versus
61% and 46% for LOC-17, LOC-14 and LOC-16, respectively). Interestingly, a
significant reduction of luciferase activity in the case of LOC-14 was achieved when
compared to the double-lipid modification introduced at the 3’-termini (*p < 0.05). If

only compared the effect induced by the double-lipid tail, we found statistically
differences between LOC-16 and LOC-17, respectively (***p < 0.001).
Based on these preliminary inhibition results, it is suggested that the kind of lipid
modification (saturated versus unsaturated with one or two branches) together with their
position in the antisense oligonucleotide tract (3’- or 5’- termini) could consider two
important parameters in gene silencing process. Furthermore, the low effectiveness
observed for the conjugate LOC-17 might point out not only to the unpacking of the
LOC-17/lipofectamine lipoplexes but also to the inefficient interaction of LOC-17 with
its interacting partners. This suggests that this modification could somehow disrupt the
binding properties of the oligonucleotide.
INSERT FIGURE 2. Analysis of Renilla mRNA knockdown in HeLa cells with different
phosphorothioate LOCs (LOC-14, -15, 16, -17). (A) Normalized luciferase activity in serum-free
conditions with a LOCs concentration in cell culture of 60 nM. (B) Normalized luciferase activity with
phosphorothioate LOCs (LOC-14, -16 and -17) in the presence of lipofectamine at 60 nM. (C) Dose-
response experiment using phosphorothioate LOC-14 and LOC-17 at several concentrations (60, 150 and
300 nM). A phosphorothioate scramble (Scr) (without any modification) showed the specificity of the
experiment at 60 nM. Phosphorothioate 19 (without any modification) was used as a positive control.
Results are expressed as mean ± SD for an average of three to six independent replicates. A regular two-
way ANOVA variance analysis combined with Bonferroni post-test for multiple comparisons was used.
***p < 0.001, **p < 0.01, *p < 0.05 when the comparison was established between conditions indicated
in (A), (B) or (C) to establish statistical significance.

In an attempt to emulate the physiological conditions, we further investigated the
ability of the phosphorothioate LOC-14 and LOC-17 to impart cellular uptake at several
concentrations (60, 150 and 300 nM) in the absence of lipofectamine but in the presence
of serum proteins. As shown in Figure 2C, the antisense silencing activity of LOC-14
was drastically reduced in the presence of 10% FBS (v/v), indicating that LOC-14 was
not able to silence Renilla luciferase gene at the three concentrations tested. Conversely,
LOC-17 could inhibit the Renilla luciferase expression in a dose-response manner.
While antisense activity for LOC-17 was barely detected at 60 and 150 nM, luciferase
expression was inhibited to 22% when a higher concentration was tested (300 nM) (*p
< 0.05). Similarly, this result improved the silencing activities achieved for the
unmodified oligonucleotide 19 and LOC-14 at the same concentration (300 nM; **p <
0.01). In view of these results, we could hypothesize, among other factors, that
transfection efficiencies mediated by LOC-14 and -17 might be compromised due to the
dependence and/or association of these two amphiphilic conjugates with serum
proteins.33
In order to understand the effect induced by this double-lipid modification, we
characterized in detail the insertion of a DNA sequence modified with the same double
lipid-tail at 5’-termini into model membranes (giant unilamellar vesicles, GUVs, and
supported planar bilayers, SPBs) as well as its mechanism of entry in cultured cells.
3.3 Insertion into model membranes
We selected the GEM91 (gene-expression modulator 91) sequence (5’-CTC TCG
CAC CCA TCT CTC TCC TTC TX-3’; where X = 3-amino modifier C7-CPG) as a
model DNA sequence to incorporate the lipid phosphoramidite-4 at the 5’-termini
(LOC-20) since GEM91 oligonucleotide has been shown to be a potent inhibitor of

HIV-1 replication in cell culture by interacting with the conserved gag initiation site of
HIV-1.30 Furthermore, to carry out membrane binding studies and cellular uptake
experiments, fluorescent dye Alexa Fluor 488 was introduced into the LOC-20 model
conjugate at 3’-termini (LOC-21) taking advantage of the presence of the free primary
amine group, as previously described.23 Although model membranes cannot mimic the
complex cellular environment, their use would allow the study of certain biophysical
aspects of oligonucleotide delivery under controlled experimental conditions and well-
defined lipid compositions.34,35
We could clearly observe that this LOC could be efficiently inserted into
DOPC/eSM/Chol (2:2:1) monolayers at initial  above 35 mN/m (Figure S5A in
Supporting Information), indicating that its incorporation into different domains of cell
membranes would be expected since biological membranes are considered to support a
lateral pressure  30 mN/m, as previously observed in several cell lines,23 albeit with
large fluctuations around this average value.36 These results were in agreement with
what was previously observed in GUVs and SPBs composed of the same ternary
mixture,23 where the insertion of LOC showed a homogeneous incorporation into both
liquid-disordered (Ld) and liquid-ordered (Lo) domains. In fact, this LOC showed first a
rapid interaction with the Ld phase of the bilayer, followed by its accumulation in the
boundary between the Ld and Lo domains and finally leading to its incorporation to Lo
domains and the formation of Alexa 488-enriched rounded structures (Figure S5B in
Supporting Information).
Next, GUVs and SPBs model membranes presenting pure Ld (DOPC) and Lo
[eSM/Chol (2:1)] phase states were also used in order to study LOC-21 insertion into
each pure phase and to analyze the possible dependence of the Lo-Ld interface on the
insertion. Here, a rapid and remarkable interaction was observed with both model

membranes, independently of the lipid composition, which also lead to the formation of
Alexa 488-enriched structures in all SPBs tested (Figure S6 in Supporting Information).
These findings helped us discard a possible influence of domain boundaries on the LOC
insertion, suggesting that this LOC could interact and insert into either Ld or Lo phases
within a cell membrane.
Finally, we investigated whether the bilayer-binding properties of this LOC in
model membranes in a DNA duplex based conjugate. Two duplexes were formed to
carry out this study: LOC-21 was hybridized with its complementary strand (22) in
order to form a double-stranded hybrid that contained an Alexa 488 fluorophore in the
3’-termini (dsLOC-1) or LOC-20 was hybridized with the fluorescently labelled
complementary sequence 24 bearing the Alexa 488 dye at 5’-termini of the
complementary strand (dsLOC-2) (Figure S2 in Supporting Information).
Oligonucleotide 24 covalently labelled with Alexa 488 was obtained from the 5’-amino-
oligonucleotide 23 following the experimental procedures described (see Materials and
methods for further information). Interestingly, both double-stranded LOCs (dsLOC-1
and dsLOC-2) behave similarly to LOC-21. They could easily insert into both Ld and Lo
domains of GUVs and composed of DOPC, DOPC/eSM/Chol (2:2:1) and eSM/Chol
(2:1) (Figure 3A). In addition, Alexa 488-enriched intense spots (bigger than 1 m in
diameter) were also observed in SPBs composed of DOPC/eSM/Chol (2:2:1),
suggesting that these reproducible microstructures are at least in part formed by these
compounds (Figure 3B).
INSERT FIGURE 3. Double-stranded LOC binding to lipid bilayers. (A) Confocal microscopy
images of DiD-labelled DOPC (upper panels), DOPC/eSM/Chol (2:2:1) (middle panels) and eSM/Chol
(2:1) (bottom panels) GUVs (equatorial sections) incubated for 2 h with 100 nM (50 nM Alexa 488-
labelled) double-stranded LOCs, carrying Alexa 488 fluorophore bound to the 3’-end of the LOC
(dsLOC-1, right panels) or to the 5’-end of the complementary strand (dsLOC-2, left panels). (B)