Degradable self-assembling dendrons for gene delivery: Experimental and theoretical insights into the barriers to cellular uptake

Article abstract of DOI:10.1021/ja2070736

This paper uses a combined experimental and theoretical approach to gain unique insight into gene delivery. We report the synthesis and investigation of a new family of second-generation dendrons with four triamine surface ligands capable of binding to DN

Full text of DOI:10.1021/ja2070736

ARTICLE
pubs.acs.org/JACS
Degradable Self-Assembling Dendrons for Gene Delivery:
Experimental and Theoretical Insights into the Barriers to
Cellular Uptake
Anna Barnard,† Paola Posocco,‡ Sabrina Pricl,‡ Marcelo Calderon,§ Rainer Haag,§ Mark E. Hwang,||
Victor W. T. Shum,|| Daniel W. Pack,|| and David K. Smith*,†
†Department of Chemistry, University of York, Heslington, York, YO10 5DD, U.K.
‡Molecular Simulation Engineering (MOSE) Laboratory, Department of Industrial Engineering and Information Technology (DI3),
University of Trieste, 34127 Trieste, Italy
§Institut f€ur Chemie und Biochemie, Freie Universit€at Berlin, Takustrasse 3, D-14195 Berlin, Germany
Department of Chemical and Biomolecular Engineering, University of Illinois at UrbanaÀChampaign, Urbana, Illinois 61801,
United States
S Supporting Information
b
ABSTRACT: This paper uses a combined experimental and
theoretical approach to gain unique insight into gene delivery.
We report the synthesis and investigation of a new family of
second-generation dendrons with four triamine surface ligands
capable of binding to DNA, degradable aliphatic-ester dendritic
scaffolds, and hydrophobic units at their focal points. Dendron
self-assembly significantly enhances DNA binding as monitored
by a range of experimental methods and confirmed by multi-
scale modeling. Cellular uptake studies indicate that some of these dendrons are highly effective at transporting DNA into cells
(ca. 10 times better than poly(ethyleneimine), PEI). However, levels of transgene expression are relatively low (ca. 10% of PEI).
This indicates that these dendrons cannot navigate all of the intracellular barriers to gene delivery. The addition of chloroquine
indicates that endosomal escape is not the limiting factor in this case, and it is shown, both experimentally and theoretically, that gene
delivery can be correlated with the ability of the dendron assemblies to release DNA. Mass spectrometric assays demonstrate that the
dendrons, as intended, do degrade under biologically relevant conditions over a period of hours. Multiscale modeling of degraded
dendron structures suggests that complete dendron degradation would be required for DNA release. Importantly, in the presence of
the lower pH associated with endosomes, or when bound to DNA, complete degradation of these dendrons becomes ineffective on
the transfection time scale—we propose this explains the poor transfection performance of these dendrons. As such, this paper
demonstrates that taking this kind of multidisciplinary approach can yield a fundamental insight into the way in which
dendrons can navigate barriers to cellular uptake. Lessons learned from this work will inform future dendron design for
enhanced gene delivery.
’ INTRODUCTION
dendron constitute an optimum multivalent array for displaying
bioactive ligands—such arrays significantly enhance the binding
affinity for key biological targets as a consequence of the entropic
benefits of ligand organization.4 It is also possible to make the
dendritic branched scaffold degradable, such that over time, or in the
presence of specific biological triggers, it breaks down in a con-
trollable and predictable way into smaller subunits.5 This can en-
hance the biocompatibility of the dendron, lower its toxicity, and
limit its persistence in cells. Dendron degradation also disassembles
the multivalent array and therefore acts as an effective way of
“switching off” the multivalent binding effect, significantly decreas-
ing the affinity of the system for the biological target.6,7
Dendrons have been suggested as appropriate nanoscale car-
rier molecules for the delivery of bioactive materials into cells.1
These branched, well-defined, wedge-like molecular structures
are highly tunable using simple organic synthesis—for example,
it is possible to modify the functional groups present at the focal
point, those within the branching, and also those displayed on the
dendritic surface. This means that a huge variety of potentially
bioactive dendrons can be accessed with relative ease.1 Each part
of the dendron can play a distinctive role in controlling the biological
behavior. Choice of an appropriate hydrophobic group at the
dendron focal point can give rise to controlled self-assembly into
larger nanoscale aggregates,2 with the relative size of the hydro-
phobic and hydrophilic groups controlling the architecture of the
resulting self-assembled structure.3 The multiple surface groups of a
Received: July 28, 2011
Published: October 31, 2011
r
2011 American Chemical Society
20288
dx.doi.org/10.1021/ja2070736 J. Am. Chem. Soc. 2011, 133, 20288–20300
|

Journal of the American Chemical Society
ARTICLE
has enabled us to develop a detailed understanding of structureÀ
activity relationships and multivalency effects in DNA binding
and gene delivery.24 In recent years, to improve biocompatibility,
the attention of Kostiainen and ourselves has turned to biode-
gradable dendron frameworks,6,7 and in 2009, we reported a simple
system based on Frꢀechet-type aliphatic ester dendrons25
modified with spermine surface groups which exhibited multi-
valent binding and degraded such that DNA binding became
disfavored in vitro.7
In this paper, we modified the focal point of these dendrons,
replacing the Z-protecting group with hydrophobic units (Figure 2),
promoting dendron self-assembly. We have also changed the
surface groups, replacing them with a triamine, which we have
previously reported lowers the toxicity of the dendron constructs.24b
We have then experimentally measured the effect of these mod-
ifications on DNA binding and cellular gene delivery as well as
monitored the degradation of the dendrons using an in vitro mass
spectrometric assay. We have then modeled the behavior of both
the intact and the degraded dendrons using a mesoscale approach
and monitored their ability to self-assemble and bind DNA.
Using this powerful combination of experimental and theoretical
approaches, we gain a unique insight into the gene delivery
process and a detailed understanding of the way in which these
systems negotiate each of the barriers to gene delivery.
Figure 1. Schematic of gene delivery process, highlighting the barriers
which have to be overcome by effective vectors.
Dendrons (and spherical dendrimers) have been of particular
interest as a consequence of their ability to deliver genetic material
into cells.8 Since the pioneering work from the groups of Tomalia9
and Szoka10 using poly(amidoamine) spherical dendrimers, a
wide range of different dendritic architectures have been ex-
plored for their gene delivery potential. In most cases, large poly-
cationic dendrimers based on polyamines such as dendritic poly-
L-lysine have been employed.11 Such systems can be considered
to belong to the cationic polymer class of gene delivery vehicle.12
Although such systems can perform highly effective gene deliv-
ery, they also exhibit problematic toxicity profiles13 and can cause
difficulties as a consequence of their persistence in cells after
gene delivery has taken place.14 An alternative approach has therefore
used dendron architectures in which a hydrophobic group at the
focal point encourages self-assembly of the dendrons into a larger
“pseudodendrimer”. Such systems, reported by the groups of
Diederich,15 Florence,16 and Hammond,17 are also capable of
binding DNA and delivering it into cells. Dendrons of this type
are somewhat closer in nature to the cationic lipid class of gene
delivery vehicle.18
For effective gene delivery into cells, it is necessary to over-
come a number of extra- and intracellular barriers.19 To achieve
effective gene delivery, a synthetic vector must: (1) complex and
protect the DNA in the extracellular environment, (2) enable
transit of the cell membrane—usually via endosomal uptake me-
chanisms, (3) allow escape from the endosomal vesicles within
the cell after uptake, (4) enable transit to and entry into the cell
nucleus, (5) release the DNA so that it can be effectively translated
into the desired protein, and (6) (not shown) degrade such that the
vector does not persist and cause problems within the cell once
delivery has taken place (Figure 1).14 Gene delivery is therefore a
challenging target, and as of yet, although synthetic vectors are used
commercially for in vitro applications, there has only been limited
success using synthetic gene delivery vehicles for in vivo work.20
Since 2005, we have been reporting on a family of dendrons
designed to show ultrahigh-affinity DNA binding21 and explored
their ability to deliver genetic material into cells.22 By modifying
the dendrons with different hydrophobic groups at the focal point,
we have demonstrated that self-assembly of the dendrons takes
place, with their gene delivery ability correlating with the nanoscale
architecture into which the dendrons assemble.3,23 This approach
’ RESULTS AND DISCUSSION
Dendron Synthesis and Characterization. The synthesis
and characterization data for all compounds are provided in the
Supporting Information (SI). In summary, second-generation
dendrons with amine groups on the surface and an alkyne
functional group at the focal point were synthesized using a
combination of the methodology previously reported by
Frꢀechet,25 Sharpless,26 and ourselves.27 We then employed
“click' chemistry methods to conjugate azide-functionalized
hydrophobic units to the focal point of the alkyne-modified
dendron (see SI for full synthetic schemes). Appropriate
azides were prepared according to the methods outlined in the
SI. After coupling of the azides to the alkyne group, compound
purification yielded the modified dendrons with hydrophobic
groups connected at the focal point via a triazole linkage. The Boc
protecting groups were finally removed from the amine surface
ligands using HCl gas. This gave rise to the target compounds,
Chol-G2, Chol2-G2, C12-G2, C16-G2, and C22-G2 (Figure 2).
As a control, we also synthesized Z-G2 (Figure 2) in which a
simple benzyl ester protecting group is present at the focal point
of the dendron. All intermediate compounds were fully char-
acterized, and all target compounds were synthesized in good
overall yields and fully characterized (see SI).
Self-Assembly of Dendrons. Initially, we monitored whether
the dendrons aggregated in aqueous solution, using solubiliza-
tion experiments with the hydrophobic dye Nile Red.28 If self-
assembled architectures are present, Nile Red is solubilized into
the hydrophobic domain of the aggregates. Varying the concen-
tration of the self-assembling peptide will then lead to increasing
levels of dye solubilization. Plotting the fluorescence emission
intensity of Nile Red at 635 nm versus log10[dendron] allows the
determination of critical micelle concentrations (CMCs) (see SI
for data). Table 1 reports the CMC values for the six dendrons
investigated in this paper. As can be seen from the data, the
functional group at the dendron focal point has a profound effect
on the dendron aggregation process. Control compound Z-G2
20289
dx.doi.org/10.1021/ja2070736 |J. Am. Chem. Soc. 2011, 133, 20288–20300

Journal of the American Chemical Society
ARTICLE
Figure 2. Dendrons investigated in this paper.
Table 1. Self-Assembly Data for Dendrons Investigated in
This Paper
Table 2. Modeling of Dendron Self-Assembly by Mesoscale
Methods
CMC (μM)a
zeta (mV)b
diameter (nm)b
compound
Dm
Rcb
Nagg
σm
ΔGmic
CMCf
a
c
d
e
compound
Z-G2
N/A
+34.1 ( 1.8
+42.6 ( 2.0
+45.0 ( 1.9
+56.1 ( 3.2
+56.3 ( 4.4
+55.4 ( 2.2
does not fit DLS model
does not fit DLS model
5.93 ( 0.65
C12-G2
C16-G2
C22-G2
Chol-G2
Chol2-G2
2.9 ( 0.1
3.1 ( 0.1
3.3 ( 0.2
3.9 ( 0.2
5.1 ( 0.1
0.8
0.9
1.0
1.1
1.5
6
7
1.77
1.89
2.09
2.68
2.15
À63.6
À75.7
À92.3
À96.4
À99.6
2.7
C12-G2
C16-G2
C22-G2
Chol-G2
Chol2-G2
208 ( 56
37 ( 6
0.23
9
0.080
0.035
0.018
2.0 ( 0.1
4.9 ( 0.6
4.9 ( 0.6
6.96 ( 0.18
16
22
7.55 ( 0.04
15.04 ( 0.65
a Dm (diameter of the micelle in nm). b Rc (radius of the micellar core
c
a The CMC is the critical micelle concentration as determined from the
Nile Red assay. b Zeta potential and average diameter assessed by zeta
sizing and intensity distribution in DLS measurements, respectively.
in nm). Nagg (aggregation number, i.e., the number of dendrons in a
micelle). d σm (surface charge density of the micelle in e nmÀ2). e ΔGmic
(predicted free energy of micellization in kJ molÀ1). f CMC (predicted
critical micelle concentration in μM).
did not show any evidence of aggregation under the conditions
assayed (up to 1 mM). For the alkyl chain modified dendrons,
there is an inverse relationship between the length of the
hydrophobic chain and the CMC value. As the hydrophobic
chain increases in length from C12-G2 to C16-G2 to C22-G2, the
CMC value drops from 208 to 2 μM, a consequence of more
effective packing of the longer hydrophobic chains resulting in
better self-assembly. Both cholesterol-functionalized dendrons
exhibited similar CMC values of ca. 5 μM. Clearly, cholesterol is
quite effective in encouraging the self-organization of these dendrons.
The aggregates were also investigated by zeta sizing methods
to estimate the sizes and surface charges of the self-assembled
“pseudodendrimers” being formed by these compounds, with the
diameters reported being based on the volume contribution
(Table 1). As can be seen, the data were in general agreement
with the Nile Red assay, in showing that the dendrons with the
most extensive hydrophobic functionalization at the focal point
are best able to self-assemble into structured aggregates. Speci-
fically, compounds Z-G2 and C12-G2 only fitted poorly to the
dynamic light scattering (DLS) model, exhibiting relatively low
zeta potentials indicative of low positive charge densities. This is
consistent with the results of the Nile Red assay which indicated
that these compounds have low self-assembly potential. Con-
versely, compounds Chol-G2, Chol2-G2, and C22-G2, all of
which have significant hydrophobic groups attached at the focal
point and formed aggregates with CMCs < 10 μM according to
the Nile Red assay, appeared to form well-defined aggregates via
the zeta sizing method. Furthermore, those systems with the
greatest potential to self-assemble achieved the highest surface
charge potential, of ca. +56 mV. This would seem to indicate that
favorable energetic terms associated with the self-assembly of the
hydrophobic functional groups are able to drive the close-
packing of positively charged groups on the surfaces of the
nanoscale assembled architectures. Interestingly, as the size of
the hydrophobic unit increases, so does the apparent diameter of
the micellar aggregates, from ca. 5.9 nm for C16-G2 to 7.0 nm for
C22-G2 and 7.5 nm for Chol-G2. Intriguingly, the assemblies
formed by Chol-G2 were significantly different from those for-
med by Chol2-G2, and although they exhibited roughly the same
zeta potential values, the diameters of the observed complexes
were ca. 7.5 and 15.0 nm, respectively.
We employed mesoscale modeling methods to better under-
stand the eventual self-assembly of all modified dendrons.
Mesoscale methods are ideal for modeling nanoscale self-assem-
bly processes, as they are able to cope with spatial inhomogene-
ities over the nanometer length scale, as well as variable phenomena
over longer time scales (up to milliseconds). We chose to use
dissipative particle dynamics (DPD) to model these self-
assemblies.29 In DPD, a group of atoms is coarse-grained into
a bead, thereby substantially reducing the number of particles to be
simulated. Further, rather than interact through Lennard-Jones
forces, the beads feel a simple soft pairwise conservative potential
which embodies the essential chemistry of the system. This mod-
eling was carried out in a biorelevant aqueous solvent environ-
ment, in the presence of NaCl (150 mM).
As indicated by the data in Table 2, with the exception of Z-
G2, all the remaining dendrons self-assembled into nanoscale
20290
dx.doi.org/10.1021/ja2070736 |J. Am. Chem. Soc. 2011, 133, 20288–20300

Journal of the American Chemical Society
ARTICLE
Table 3. DNA Binding Data for Dendrons Investigated in
This Paper
compound
CE50 (EthBr assay)a
N:P (gel assay)b
Z-G2
N/A
4.61
2.38
2.30
2.20
2.09
1.82
C12-G2
C16-G2
C22-G2
Chol-G2
Chol2-G2
N/A
1.42 ( 0.80
0.84 ( 0.34
0.66 ( 0.13
0.57 ( 0.03
Figure 3. Mesoscale modeling of the hydrophobically modified Z-G2
(left), Chol-G2 (center), and C22-G2 (right). The white sticks represent
the hydrophobic units. Different colored sticks (purple, light blue, and
green) are adopted to represent the various hydrophilic regions. A
transparent gray field is used to portray water.
a The CE50 value is the charge excess required for 50% displacement of
ethidium bromide in the fluorescence displacement assay. b The N:P
value is the charge excess required for retention of 1 μg of plasmid DNA
in the well in the gel electrophoresis assay.
objects under mesoscale modeling conditions. Israelachvili re-
ported that the packing parameter, P (an effective ratio of hydro-
phobic/hydrophilic size), could be used to predict the shape of
nanoscale aggregates for ionic amphiphiles.30 Adopting the same
concept in the present case, for all modified dendrons, this value
is <0.33 and would lead us to predict that all dendrons but Z-G2
would self-assemble into spherical micelles, as observed in the
modeling (Figure 3 and Figure S8.3, SI).
Importantly, the predicted CMC values follow the correct,
increasing trend as the hydrophobic character of the dendron
substituents decreases. For example, on going from C22-G2 to
C16-G2 to C12-G2, as the hydrophobic chain becomes shorter,
the CMC increases. This is in general agreement with the results
of the Nile Red assay. Furthermore, the micellar surface charge
density (σm) increases as the hydrophobic chain lengthens and
becomes better able to pack—in agreement with the measure-
ments of zeta potentials (Table 1). Modeling indicates that
Chol2-G2 should have a lower CMC than Chol-G2, although
this was not observed experimentally—but it is well-known that
the use of dyes such as Nile Red can lead to perturbations in the
apparent aggregation. Notably, even if these CMC values esti-
mated in silico are taken with due caution, they are below the
experimental concentrations employed in the transfection ex-
periments, and the presence of micelles as nanovectors is there-
fore supported by the modeling.
It was predicted from modeling (Table 2) that the average
micellar diameter increases as the hydrocarbon content of the
hydrophobic chain increases—in agreement with the trends
found in the experimental study (Table 1). Interestingly, the dif-
ferent architectures of the hydrophobic portion ultimately result
in differently sized micelles and/or a different number of
dendrons per micelle (the aggregation number Nagg) and, hence,
a different micellar surface charge density σm. Comparing Chol-
G2 and Chol2-G2 with the other modified dendrons, we see that
these compounds assemble into micelles of bigger diameters
than the other counterparts. This is likely due to the highly hy-
drophobic nature and flat, rigid cholesterol moieties leading to
highly effective packing within the micellar interior. Also, com-
paring the two cholesterol-bearing molecules, modeling predicts
the formation of larger micelles for Chol2-G2 than Chol-G2.
These larger aggregates are a consequence of the fact that a larger
number of Chol2-G2 are incorporated into each micelle com-
pared with Chol-G2 (22 molecules per micelle rather than 16)
and due to the two cholesterol units at the focal point providing a
more favorable free energy of micellization (by ca. 3 kJ molÀ1). This
modeling is in general agreement with the results of zeta sizing,
which indicated that Chol2-G2 did indeed form the largest aggre-
gates, followed by Chol-G2, C22-G2, and C16-G2, respectively.
Although it is predicted that the micelles formed by Chol-G2
and Chol2-G2 are somewhat different in terms of their dimen-
sions, it is noteworthy that the surface charges (σm) of the two
self-assembled dendron micelles are fairly comparable, with
Chol-G2 having a slightly higher surface charge. Once again,
this is in good agreement with the results of zeta sizing, which
indicated the Chol-G2 and Chol2-G2 aggregates had surface
charges of +56.3 and +55.4 mV, respectively. Consequently, the
combined experimental and theoretical approach convinced us
that we had a good understanding of the dendron self-assembly
process.
DNA Binding Studies. We then went on to assay the ability of
these dendrons to bind DNA. Initially we bound calf thymus
DNA using the standard ethidium bromide (EthBr) displace-
ment fluorescence spectroscopy assay which we have employed
in our previous research.7,21,31 This method uses competition
between the DNA binder and EthBr to assess the concentration
at which the DNA binder becomes effective. This can be ex-
pressed as the concentration of DNA binder required for half of
the EthBr to be displaced from binding to DNA. This concen-
tration can usefully be expressed as a charge excess (CE50) value.
To calculate this CE50 value, it is assumed that each amine in the
DNA binder is protonated, and each phosphate in the DNA is
deprotonated. As such, the CE50 value is equivalent to a N:P
ratio. Lower CE50 values represent more effective binding, as a
smaller amount of positive charge is required to effectively bind
the negative charge associated with the DNA. This assay there-
fore provides an excellent comparative method for considering
the DNA affinities of a family of compounds such as this, in which
each member has the same amine-derived DNA binding motif
and, as such, is ideally suited for the development of structureÀ
activity relationships. For this study, we compared all of our DNA
binders under physiologically relevant salt concentrations (150 mM
NaCl).
As indicated in Table 3, compounds Z-G2 and C12-G2 were
unable to effectively displace ethidium bromide from the DNA
double helix under the conditions investigated. These com-
pounds are the less effective dendrons in terms of self-assembly
and should therefore not have such effectively organized multi-
valent arrays of DNA binding ligands. Each individual dendron
only has a net positive charge of 8+, clearly insufficient to displace
more than 50% of EthBr from its intercalation sites. Those
dendrons which self-assemble more effectively, however, did
achieve EthBr displacement, and CE50 values could be deter-
mined. As anticipated, C22-G2 was a more effective DNA binder
than C16-G2, presumably because of its greater self-assembly
20291
dx.doi.org/10.1021/ja2070736 |J. Am. Chem. Soc. 2011, 133, 20288–20300

Journal of the American Chemical Society
ARTICLE
Table 4. Selected Zeta Sizing Data for Short Strand DNA in
the Presence of Dendrons
compound
N:P ratio
zeta (mV)
diameter (nm)a:
DNA
-----
5
À49.8 ( 1.8
+6.2 ( 0.3
---------
Z-G2
548.1 ( 25.2
507.2 ( 59.4
297.2 ( 16.1
219.1 ( 1.21
106.6 ( 0.4
79.7 ( 1.2
C12-G2
C16-G2
C22-G2
Chol-G2
Chol2G2
5
+5.85 ( 0.1
+18.5 ( 0.3
+45.5 ( 1.4
+48.8 ( 1.1
+45.0 ( 1.0
5
5
5
5
a: Average diameter assessed by Z-average in DLS measurements.
potential. The most effective DNA binders, however, were those
with cholesterol groups at their focal points, with Chol2-G2
being slightly more effective than Chol-G2. Pleasingly, the ex-
perimentally verified CE50 values directly correlate with the surface
charge density, σm, values estimated from simulation, indicating
that the micelles characterized by higher values of σm (i.e., Chol2-
G2, Chol-G2, and C22-G2) are tighter DNA binders than their
counterparts with lower σm values (see Table 2).
Figure 4. Mesoscale modeling of the hydrophobically modified den-
drons synthesized in this work in the presence of DNA: Z-G2 (top left);
Chol-G2 (top right); C22-G2 (bottom left), PEI (bottom right). In all
pictures, white sticks represent hydrophobic units. Different colored
sticks (purple, light blue, and green) are adopted to represent the various
hydrophilic regions. DNA chains are depicted as red sticks. A transparent
gray field is used to represent water.
The plasmid DNA binding affinity of the dendrons was then
studied using gel electrophoresis with increasing amounts of
dendron added to 1 μg of plasmid DNA. We used the dendron:
DNA weight ratio necessary to fully retard the DNA in the
loading well as an effective indicator of the DNA binding affinity
of the vector. Table 3 reports the N:P ratio at which the DNA is
fully retarded in the well (equivalent to a charge excess, CE,
value). From the data, we can see that the cholesterol-function-
alized dendrons were able to bind DNA at the lowest N:P ratio
indicative of the most effective binding. Once again, for the aliphatic
chain functionalized dendrons there was a correlation between
aliphatic chain length and binding strength, with the shortest
chain compound C12-G2 binding DNA with the lowest affinity
and C22-G2 binding DNA much more effectively. Dendron Z-
G2, which has no ability to self-assemble under the experimental
conditions, binds to DNA with the lowest affinity of all the den-
drons. It therefore appears that those compounds most effective
at self-assembly are the strongest binders. It should be noted that
the N:P values from gel electrophoresis are consistently higher
than the CE50 values from the EthBr assay—this is because in
electrophoresis we are looking for complete binding of plasmid,
whereas in the EthBr assay, we are only looking for 50% dis-
placement of the fluorescent dye. Gel electrophoresis images can
be found in the Supporting Information.
We then carried out zeta sizing experiments of the complexes
formed between the different dendrons and short strand DNA
(see Supporting Information for full details). This allowed us to
gain insight into the sizes of the complexes formed in the binding
process and, as a consequence, the ability of the dendrons to
condense plasmid DNA—an important step in the transfection
pathway. Table 4 summarizes some of the key data which indicate
that those dendrons which exhibit the highest affinity DNA
binding as revealed by EthBr displacement and gel electrophor-
esis (i.e., Chol-G2, Chol2-G2, and C22-G2) are also those which
condense the plasmid DNA most effectively into complexes
which are typically ca. 100 nm in diameter (for full data see
Supporting Information). The less effective DNA binders form
aggregates with DNA which are ca. 500 nm in size—indicative of
ineffective plasmid condensation and far too large for effective
transfection. Furthermore, the dendrons which are better able to
self-assemble form complexes with DNA which have a higher net
positive charge, indicative of their better ability to neutralize the
net anionic charge of the DNA.
We then carried out a mesoscale modeling study of the inter-
actions between DNA and the dendron micelles. Conversion of
the atomistic level structures into coarse-grained form and mapping
the energetic information onto the appropriate value of the DPD
interaction parameters aij allowed the mesoscale optimization of
the interaction between self-assembled dendrons and extended
DNA chains. The morphology of the self-assembled micelles and
DNA helices is visualized in Figure 4 (see also Figure S8.4). As can
be seen, the DNA molecules are loosely packed without a well-
defined interhelical pattern or distances. The micelles appear to
undergo a small degree of deformation in the complexes, with a
tendency to elongate along the DNA longitudinal direction pre-
sumably to enhance the adhesion with the rather planar, extended
surface of DNA (e.g., see Chol-G2 in Figure 4). In these cases, the
DNA molecules seem to comply with the well-known “bead-on-
a-string” model,32 according to which some regions of the nucleic
acid are engulfed by the micelles while some others are no longer
surrounded by them. Thus, the DNA helices are partially em-
bedded within the micellar organization and partially exposed to
the solution environment, where the counterions originating
from the electrolyte provide the charge neutralization required
for eventual condensation. Figure 4 (d) shows the results obtained
from the mesoscopic simulations performed on the PEI/DNA
system. These mesoscale results offer a sensible explanation for
the high transfection capacity of PEI—the DNA is homoge-
neously intertwined between the PEI chains, so that it is quite
efficiently protected from the outer environment. At the same
time, since no particular structuring (e.g., beads on a necklace) is
attained in solution, there are some areas where DNA bundles
20292
dx.doi.org/10.1021/ja2070736 |J. Am. Chem. Soc. 2011, 133, 20288–20300

Journal of the American Chemical Society
ARTICLE
Figure 5. Atomistic MD simulations of a Chol-G2 (pink, left) and a
Chol2-G2 (light blue, right) micelle in complex with DNA (gold and
orange ribbons). Na+ and ClÀ ions are depicted as magenta and green
spheres, respectively. Water is not shown for clarity.
Table 5. Binding Affinities between DNA and the Self-
Assembled Dendron Micelles from Atomistic Level Modeling
Figure 6. Fluorescence data for the uptake of fluorescently tagged DNA
into HEK293 cells demonstrating effective cellular uptake by vectors
Chol-G2, Chol2-G2, C16-G2, and C22-G2 which are best able to self-
assemble into micellar aggregates. The data are corrected by subtraction
of the uptake observed for plasmid DNA in the absence of any vector.
a
Compound
ΔGbind
ΔGbind/Nb
C12-G2
C16-G2
C22-G2
Chol-G2
Chol2-G2
PEI
À16.8 ( 1.3
À22.4 ( 1.7
À33.8 ( 2.8
À71.2 ( 3.1
À119.2 ( 3.1
À269.0 ( 2.6
À0.35 ( 0.03
À0.40 ( 0.03
À0.47 ( 0.04
À0.56 ( 0.02
À0.68 ( 0.02
À0.63 ( 0.01
than, the experimental observations (Table 3). This is perhaps
surprising given that modeling indicated that the Chol2-G2 self-
assembled structures had lower surface charge densities than
those resulting from Chol-G2 (Table 2) and suggests that the
larger micelle size of Chol2-G2 (Table 2) may play an important
role in modulating the ability of the charges to bind to the DNA
double helix without overcrowding at the micellar surface. Since
PEI is a current standard vector system for DNA cell trasfection
experiments, MD atomistic simulations were also run on a PEI/
DNA for comparison. Interestingly, the estimated binding affi-
nity of PEI toward DNA is, in terms of ΔGbind/N, somewhat
intermediate between Chol-G2 and Chol2-G2. This finding
is indicative of the fact that both cholesterol bearing, self-
assembling dendrons possess excellent DNA binding properties
which could be potentially exploited with success in gene delivery
processes.
Gene Delivery into Cells. The ability of the dendronÀDNA
complexes to cross cell membranes was then investigated using
flow cytometry. Human embryonic kidney cells (HEK293) were
incubated with the vector mixed with fluorescently tagged
plasmid DNA for 2 h at 37 °C. The cell uptake was measured
using flow cytometry, and the fluorescence of any cells with char-
acteristic dimensions for live cells was then recorded. The fluo-
rescence is normalized to the value obtained for uptake of DNA
alone, and the results are shown in Figure 6.
From the data, it can be clearly seen that the most hydrophobic
dendrons, Chol-G2, Chol2-G2, C16-G2, and C22-G2, demon-
strated in the previous section to be capable of effective self-
assembly and DNA binding are all capable of carrying DNA
across cell membranes. Interestingly, they all do so even more
effectively than positive control PEI. Indeed, compound C22-G2
is ca. 12 times more effective than PEI at getting DNA into live
cells. To check that these results did not simply reflect the
polyplexes sticking to cell surfaces, this experiment was repeated
at 4 °C, and cell uptake was depressed to <20% of the levels
observed at 37 °C for the two most effective delivery vehicles,
C22-G2 and Chol2-G2 (see Supporting Information). In agree-
ment with the self-assembly and DNA binding studies, Z-G2 and
C12-G2 show relatively poor cellular uptake, with activities only
ca. 25% of that of PEI. The results of this study clearly indicate
that cellular delivery can be directly correlated with the ability of
a ΔGbind represents the free energy of binding between micelles and
DNA in kcal molÀ1. b ΔGbind/N is the normalized free energy per charge
on the micelle.
appear and some free space which, ultimately, might yield better
performance during release of this system.
The average number of DNA base pairs covered by each den-
dron micelle can be estimated to a first approximation by using
the following, simple relationship: nbp = (Dm/3.4) Â 10, where
Dm is the spherical micelle diameter, 3.4 is the DNA duplex pitch
(in nm), and 10 is the number of base pairs per duplex pitch.
Inserting the estimated values of Dm listed in Table 2 in the above
formula, we get nbp = 15, 11.5, 9.7, 9.1, and 8.5 for Chol2-G2,
Chol-G2, C22-G2, C16-G2, and C12-G2, respectively—in agree-
ment with the observed DNA binding affinities (Table 3). The
corresponding DNA phosphate groups per micelle are, however,
not all bound directly to the micelle, considering the mismatch
between the surface curvature of DNA and that of the micelle,
even though, as discussed above, the dendritic micelles may elongate
slightly along the DNA chain axis to maximize the favorable electro-
static interactions.
Quantitative modeling of micelle/DNA interactions at a fully
atomistic level was used to rank the affinity of each type of
modified dendron micelle toward DNA, ΔGbind. Interactions be-
tween the ligands on the dendron and the grooves on the DNA
double helix were clearly visible (Figure 5). As shown in Table 5,
the MD-based MM/PBSA calculations indicated that Chol2-G2
is by far the most effective DNA binder, followed by Chol-G2,
C22-G2, C16-G2, and C12-G2, in agreement with the corre-
sponding experimental evidence (Table 3). By considering the
per charge normalized values of ΔGbind on the micelle (ΔGbind
/
N), it is clear that Chol2-G2 is better able to utilize each charge in
binding to the DNA double helix than Chol-G2 and all other self-
assembled systems. Focusing on the two tighter DNA binders,
this observation, that Chol2-G2 binds DNA more strongly than
Chol-G2, is consistent with, although somewhat larger in magnitude
20293
dx.doi.org/10.1021/ja2070736 |J. Am. Chem. Soc. 2011, 133, 20288–20300

Journal of the American Chemical Society
ARTICLE
Figure 7. Transfection data for dendritic vectors in HEK 293 cells.
Luciferase expression was normalized by total cellular protein; i.e., data
were calculated in RLU/mg of protein and are then quoted as
percentages of the transfection efficiency of PEI (N = 6, error bars
represent standard deviation).
the dendrons to self-assemble, with hydrophobically modified
dendrons showing a significant improvement on PEI. This
demonstrates that these dendrons are well able to navigate
barriers 1 and 2 to gene delivery (Figure 1). These flow cyto-
metry results were also supported by confocal microscopy (see
SI) which indicated cellular uptake of green-fluorescence tagged
DNA after incubation for 1 h. The fluorescence associated with
the DNA was spread within the cell and did not indicate localization
in any specific organelle.
We then went on to investigate the ability of these dendrons to
transfect cells. This experiment measures the ability of the dendrons
to navigate all of the first five barriers on the pathway to gene
delivery (as shown in Figure 1). The ability of the ester dendrons
to transfect cells was studied using a luciferase expression assay.
HEK293 cells were incubated for 4 h in serum-free media with
mixtures of each dendron and plasmid (pGL3) DNA at different
weight ratios. The transfection was then measured via lumines-
cence assay after a further 20 h incubation in the presence of
serum containing media. The transfection activity of the most ef-
fective dendrons as a percentage of the activity of polyethylenei-
mine (25k, branched) as a positive control33 is shown in Figure 7.
From the transfection data, we can see that Chol-G2 was the
most effective vector, with an activity of around 10% of that of
PEI. Chol2-G2 showed some transfection at ca. 4% PEI positive
control. C22-G2 only showed negligible levels, while Z-G2, C12-
G2, and C16-G2 exhibited no measurable transfection (data not
shown in the figure).
Interestingly enough, the value of one of the micellar key
parameters, the surface charge density σm, as estimated by our
modeling procedures, correlates directly with the cellular trans-
fection efficiency (TE) data discussed above. Indeed, the higher
the value of σm, the higher the transfection efficiency of the cor-
responding system. This finding is not only qualitatively but also
quantitatively in agreement with what is observed for cationic
lipid/DNA-mediated delivery systems for which—depending on
the morphological structure of the overall assembly—high values
of σm are beneficial for achieving high TEs. Overall, however,
none of the new dendrons perform as well as PEI, and this indicates
that somewhere in steps 3À5 on the transfection pathway one of the
barriers must be preventing overall transfection activity.
Figure 8. Percentage cell viability of HEK293 cells in the presence of:
(A) Z-G2, Chol-G2, and Chol2-G2 and (B) C12-G2, C16-G2, and C22-
G2, at a range of different concentrations compared to that of 25 kDa
PEI. Lines are drawn to guide the eye.
this basic drug has been used to enhance transfection levels by
facilitating the endosomal escape process.34 The transfection
studies described above were therefore repeated in the presence
of three different concentrations of chloroquine, 20, 50, and 100 μM.
No enhancement in transfection levels was observed in any case,
and this leads us to conclude that endosomal escape (barrier 3) is
unlikely to be the rate-limiting step in gene delivery for these
vectors. This is supported by the fact that no accumulation/
localization of fluorescence within endosomes/lysosomes was
detected by confocal microscopy (see SI). As such, we can con-
clude that barriers 4 and 5 are likely to be preventing effective
gene transfection in this case.
We were concerned that one reason for ineffective gene trans-
fection might be associated with vector toxicity, with the cells
dying during the 24 h period of the assay. We therefore monitored
toxicity using a commercial Cell Titer 96 assay (Promega). HEK293
cells were treated with the dendrons in serum free media and
incubated for 4 h. Then the media was replaced with serum
containing media, and the cells were incubated for a further 20 h.
The cholesterol-functionalized dendrons, Chol-G2 and Chol2-
G2, are on their own completely nontoxic at the full range of
concentrations measured (up to 60 μg/mL) (Figure 8A). Den-
dron Z-G2, on the other hand, has a toxicity similar to the PEI
control, which is regarded as too toxic for clinical use.13 The
hydrocarbon chain-functionalized dendrons, however, showed
very high toxicity. Indeed, the compounds with longer aliphatic
chains, C16-G2 and C22-G2, were significantly more toxic than
25 kDa PEI (Figure 8B). For this reason, although C22-G2
appeared optimum in terms of flow cytometry for its ability to
cross cell membranes, it is clearly not appropriate for use in gene
delivery. It is unclear why changing the hydrophobic group from
a cholesterol moiety to a hydrocarbon chain has such a dramatic
effect on the toxicity of these compounds, but it is possible that
To consider barrier 3, endosomal escape, in more detail, we
performed transfection in the presence of chloroquine. Previously,
20294
dx.doi.org/10.1021/ja2070736 |J. Am. Chem. Soc. 2011, 133, 20288–20300

Journal of the American Chemical Society
ARTICLE
Figure 10. Atomistic MD simulations of a Chol-G2 (pink, left) and a
Chol2-G2 (light blue, right) in complex with heparin sulfate (atom-
colored CPK spheres: red, O; blue, N; gray, C; white, H). Na+ and ClÀ
ions are depicted as magenta and green spheres, respectively. Water is
not shown for clarity.
Table 6. Atomistic MD Modeling of the Binding to Heparin
Sulfate of the Self-Assembled Dendron Micelles
a
b
ΔGbind
ΔΔGbind
C12-G2
À12.0 ( 0.7
À12.2 ( 1.1
À22.4 ( 1.9
À67.6 ( 1.5
À103.9 ( 2.6
4.8 ( 1.5
10.2 ( 2.0
11.4 ( 3.4
3.6 ( 2.8
C16-G2
Figure 9. Gel electrophoresis images of pGL3 plasmid DNA (left-hand
lane). The remaining lanes are dendron-bound plasmid DNA (5:1 wt/
wt ratio) in the presence of increasing quantities of heparin sulfate (left
to right)—lanes 2À15 (0:1 (heparin:DNA), 1:1, 2:1, 3:1, 4:1, 5:1, 7.5:1,
10:1, 15:1, 20:1, 25:1, 20:1, 35:1, 40:1). Image A, dendron = Z-G2; B,
Chol-G2; C, Chol2-G2.
C22-G2
Chol-G2
Chol2-G2
15.3 ( 4.0
a ΔGbind represents the free energy of binding between micelles and
DNA in kcal molÀ1
.
b ΔΔGbind = ΔGbind(Heparin sulfate) À ΔGbind
-
(DNA) (see text for more details).
the more flexible C22 hydrophobic unit is able to insert itself
directly into cell membranes, disrupting normal cell behavior. In
the presence of DNA, Chol-G2 also showed significant levels of
toxicity (50% cell viability at ca. 20 μg/mL). This may be a
consequence of the good cellular uptake shown by this polyplex,
followed by an inability to disassemble intracellularly (see below).
Interestingly, Chol2-G2 showed lower toxicity when complexed
to DNA.
DNA Release Studies. Given the conclusions from the gene
delivery studies described above, we reasoned that although our
dendrons self-assemble into highly effective multivalent DNA
binders perhaps this binding is in fact too strong to facilitate
intracellular DNA release. The ability of the dendrons to effectively
release DNA was therefore investigated using a heparin sulfate
displacement assay. Heparin sulfate is an anionic polymer, which
can compete with DNA for binding to the cationic dendrons.
The assay was performed via gel electrophoresis, using a fixed
weight ratio of 5:1 vector:DNA in the wells and with increasing
amounts of heparin sulfate being added to subsequent wells. The
amount of heparin sulfate required to fully displace the DNA
from the vector is a good indicator of how effectively the vector
can release the DNA from its complex intracellularly.
Of all the vectors, it was interesting to note that Chol-G2
shows the most effective DNA release (Figure 9), even though it
is also one of the more effective DNA binders under these con-
ditions. This dendron achieves full DNA release at a ratio of 20:1
(heparin:DNA), although very little DNA is released at lower
weight ratios. This result correlates well with the transfection,
where Chol-G2 was found to be the most effective gene delivery
agent. Dendrons Z-G2 and C12-G2 (data not shown for C12-G2)
were the next most effective in terms of DNA release, with ratios
of 35:1 heparin sulfate:dendron required for complete release. In
the case of the remaining dendrons, C16-G2, C22-G2 (data not
shown), and Chol2-G2, there was still incomplete release even at
ratios of up to 40:1. This suggests that, particularly for Chol2-G2
which showed excellent DNA binding profiles and cellular uptake
ability, ineffective DNA release (barrier 5 in the transfection
pathway, Figure 2) is likely to be the cause of the low transfection
levels observed. It is interesting to speculate that Chol2-G2
achieves more effective DNA binding as indicated both experi-
mentally and by modeling methods, at the expense of effective
DNA release, as the complex simply becomes too highly opti-
mized and cannot be broken apart. Chol-G2, on the other hand,
forms smaller micelles and slightly weaker complexes with DNA,
facilitating intracellular DNA release. We noted that in this
experiment PEI could be relatively easily displaced from the
DNA by the addition of heparin (data not shown, ca. 15:1
heparin:DNA). This may agree with the morphology of the PEI/
DNA aggregates as shown in Figure 4, in which there are some
areas where DNA bundles appear between the PEI polymer
chains and some free space which should cooperate in a better
ability to release DNA from the overall complex.
The experimental heparin sulfate competition assays were paral-
leled by atomistic molecular dynamics simulation (Figure 10) of
direct binding ΔGbind of each dendron micelle and a chain of
heparin sulfate. As shown in Table 6, MM/PBSA ranked Chol2-
G2 as having the highest affinity toward heparin sulfate, followed
by Chol-G2, C22-G2, C16-G2, and C12-G2, respectively, in agree-
ment with the affinities exhibited by these self-assemblies for DNA.
More enlightening information is however obtained by con-
sidering the last column in Table 6, presenting the difference in
binding affinity of each dendron micelle toward heparin sulfate
and DNA, respectively. Since here ΔΔGbind is defined as
ΔΔGbind = ΔGbind(heparin sulfate) À ΔGbind(DNA), the more
positive the value of ΔΔGbind for a given dendron micelle, the
higher the affinity of that micelle toward DNA rather than to
20295
dx.doi.org/10.1021/ja2070736 |J. Am. Chem. Soc. 2011, 133, 20288–20300

Journal of the American Chemical Society
ARTICLE
heparin sulfate. The data indicate that compound Chol2-G2
binds DNA much more effectively than heparin, while for Chol-
G2 the binding of heparin is much more competitive with DNA
binding. This modeling study therefore verifies the experimental
results provided by gel electrophoresis, in which Chol-G2 was
best able to release the DNA and transfection studies in which
Chol-G2 was the most effective delivery vehicle. This provides
further support to the concept that, although beneficial for com-
plexation, compaction, protection transport, and cellular entry, if
DNA binding is too tight in comparison with other anionic
species it is detrimental to the fundamental step of DNA release
inside the target cell.
We propose that the better relative ability of Chol-G2 to bind
heparin and release DNA may be a consequence of the higher
surface charge density of Chol-G2 being better matched to the
very high anionic surface charge density of heparin. We therefore
reason that the cationic charge density of the self-assembled
nanostructures may play an important role in enabling heparin-
mediated DNA release. This would also explain why PEI, which
has very high charge density, achieves such effective DNA release
under these conditions.
Figure 11. Electrospray mass spectrometry of Chol-G2 from buffer at
pH 7.4 at time zero (above) and after standing for 10 h (below).
Degradation of the dendron framework can be clearly observed, as
labeled.
Dendron Degradation and DNA Release. Our dendrons
were designed to incorporate ester groups into the branched
framework, providing them with the capacity to degrade under
biological conditions of pH.25 It had been our intention that over
biologically relevant time scales dendron degradation would
occur, turning a multivalent ligand array into smaller units which
were not capable of such effective DNA binding.6,7 We had hoped
that this would actually facilitate intracellular DNA release, helping
overcome barrier 5 to gene transfection, as well as breaking down
the dendron into smaller nontoxic fragments (assisting with bar-
rier 6 in the development of effective in vivo transfection agents).
We therefore designed an electrospray mass spectrometric (ES-
MS) assay to probe the pathways of dendron degradation which
were actually taking place. A sample of each dendron was
dissolved in aqueous buffer at pH 7.5 and stirred at 37 °C.
Aliquots were removed from the bulk solution at regular time
intervals and combined with dipeptide (glycine-alanine) at a fixed
concentration as an internal standard. The samples were then
analyzed by ES-MS, and the peak height of the molecular ion was
monitored in relation to that of the standard. This allows the rate
of degradation to be determined. It has been previously shown
that similar dendrons fragment under ES conditions,35 so the
compound stability was set to 30% in the spectrometer to ensure
that the peaks in the spectra were the result of degradation alone
and did not correspond to spectrometer-induced fragmentation.
Figure 11 shows the mass spectra for Chol-G2 taken at time
zero and after 10 h stirring. In this way, and by considering the
intermediate spectra, we could identify some of the products of
degradation from the peaks which appear over time. Initially one
of the ester bonds is cleaved, resulting in a carboxylic acid frag-
ment and a larger alcohol-containing fragment of varying weight.
After this initial ester hydrolysis, the carboxylic acid undergoes
decarboxylation. Subsequently, the second branch becomes
disconnected via ester hydrolysis. All of the dendrons effectively
degrade via the same pathway (see Supporting Information).
By monitoring the ratio between the height of the molecular
ion peaks and the peaks corresponding to the dipeptide, the rate
of ester-hydrolysis-mediated degradation can be estimated. Plots
of dendron concentration against time are shown in Figure 12. In
all cases, the starting concentration of the dendron for the assay
was 100 μM, and the concentration of the standard was constant
Figure 12. Time course of degradation for Chol-G2 extracted from
mass spectrometric data and determined at two different pH values (7.5
and 5.0), demonstrating that the dendron degrades at pH 7.5 but
remains stable at pH 5.0.
at 500 μM. The degradation only monitors the loss of starting
material and does not represent subsequent degradation steps. It
exhibits zero-order kinetics with respect to dendron concentra-
tion as the concentration change is linear with respect to time.
This demonstrates that there are no intermolecular interactions
affecting the rate of degradation. We suggest it is possible that the
amine groups on the dendron periphery may play an intramo-
lecular role in catalyzing dendron degradation, which would
explain why these dendrons degrade more rapidly under these
conditions than those previously reported by Frꢀechet and co-
workers.25
Data for all dendrons are tabulated in Table 7. As can be seen,
dendron self-assembly does appear to have some effect on ester
degradation. The non-self-assembling dendron Z-G2 has an
initial degradation rate of ca. 10.1 mmol dm3 hÀ1. The cholesterol
functionalized dendrons have slightly lower rates of degradation,
which might appear to indicate that self-assembly and effective pack-
ing of the cholesterol groups somewhat lower the degradation rate.
However, for dendrons Cn-G2 as the hydrophobic chain becomes
20296
dx.doi.org/10.1021/ja2070736 |J. Am. Chem. Soc. 2011, 133, 20288–20300

Journal of the American Chemical Society
ARTICLE
Table 7. Rates of Dendron Degradation in μmol dmÀ3 hÀ1
Based on the Breakdown of the Initial Dendron, As Deter-
mined by the Mass Spectrometric Assay
,
Table 8. Atomistic Modeling of the Binding to DNA of the
Self-Assembled Dendron Micelles with Sequential Ligand
Detachment
rate of initial degradation/μmol dmÀ3 hÀ1
dendron
number of ligands
ΔGbind
ΔGbind/Nb
a
dendron
Z-G2
10.1
9.9
Chol-G2
4
3
2
1
0
4
3
2
1
0
À71.2 ( 2.4
À47.7 ( 1.3
À21.1 ( 1.1
À10.0 ( 0.4
À1.3 ( 0.1
À119.2 ( 3.1
À82.6 ( 2.2
À49.7 ( 1.6
À14.0 ( 0.3
À1.8 ( 0.1
À0.56 ( 0.02
À0.50 ( 0.01
À0.33 ( 0.02
À0.31 ( 0.01
C12-G2
C16-G2
C22-G2
Chol-G2
Chol2-G2
12.2
15.0
9.6
9.1
Chol2-G2
À0.68 ( 0.02
À0.63 ( 0.02
À0.56 ( 0.02
À0.32 ( 0.02
a ΔGbind represents free energy of binding between micelles and DNA in
kcal molÀ1. b ΔGbind/N is the normalized free energy per charge on the
micelle.
At first sight, therefore, the MS assay would indicate that
dendron degradation should occur under cellular conditions and
encourage DNA release. However, we also found that dendron
degradation does not occur at pH 5.0, relevant to the interior of
endosomes. It is known that ester hydrolysis in this type of
dendron relies on base-catalyzed hydrolysis, and therefore, we
argue that this degradation mode is switched off at pH 5.0.25 We
had thought that the carabamates may hydrolyze under this pH
regime, in analogy to the results of Frechet and co-workers.25
Clearly no carbamate hydrolysis takes place in our case. We
assign the absence of carbamate hydrolysis either at pH 5.0 or pH
7.5 to the presence of protonated amines on the ligands proximal
to the carbamate group, which would be expected to hinder the
required protonation of the carbamate CdO group. It is there-
fore possible that during trafficking into the cell degradation of
the dendron is limited as it experiences the more acidic endoso-
mal environment and that this is responsible for the poor DNA
release and transfection profile.
To confirm that dendron degradation should, in principle, lead
to DNA release, we performed atomistic MD simulations. For
this study, we detached surface ligands from the dendron by in
silico degradation, and monitored the thermodynamics of bind-
ing by MM/PBSA calculations. The modeling is represented
graphically in Figure 13, which illustrates how the micelle loses its
ability to bind to DNA as surface ligands are detached in this
manner. Table 8 collects the quantitative data which demonstrate
that, as expected, the degraded products are less able to bind
DNA than the intact dendrons when assembled into micellar
form. Interestingly, the loss of DNA binding is more marked for
Chol-G2 than for Chol2-G2. For Chol-G2, when half of the
ligands have been lost (i.e., degradation of one of the ester bonds
as evidenced by mass spectrometry), the binding strength for
DNA has dropped by ca. 70%. However, for Chol2-G2, loss of
half the binding ligands only leads to a decrease in binding affinity
of 58%. Only on further degradation does the free energy of DNA
Figure 13. Images from the modeling of the degradation of Chol-G2.
(A) represents intact Chol-G2 binding DNA. (B) represents Chol-G2
with half of the surface ligands detached. (C) represents Chol-G2 with
all of the surface ligands detached.
binding for Chol2-G2 drop significantly below 50 kcal molÀ1
.
We suggest that the two hydrophobic cholesterol units in Chol2-
G2 are better able to maintain the self-assembled nanostructure
during degradation, enabling the maintenance of a higher micellar
surface positive charge density and allowing the micelles to hold
onto DNA more effectively. These modeling observations would
imply that if intracellular degradation is indeed occurring then
longer, the rate of degradation increases, which would indicate that
in this case self-assembly appears to promote more effective degra-
dation. Nonetheless, in general terms, all of the dendrons degrade,
with none of the initial dendron (100 μM) remaining after a period
of 6À10 h—a transfection relevant time scale.
20297
dx.doi.org/10.1021/ja2070736 |J. Am. Chem. Soc. 2011, 133, 20288–20300

Journal of the American Chemical Society
ARTICLE
outperform PEI in terms of delivering DNA across cell mem-
branes. However, we found that intracellular release of the DNA
from the self-assembled vector was problematic.
The degradation of the dendrons was explored in detail, and it
was found that although ester hydrolysis of one of the branches
occurs on a transfection-relevant time scale (hours) this initial
degraded system can still self-assemble and bind DNA quite
effectively. Multilevel modeling was used to demonstrate that
complete dendron degradation would be necessary for effective
DNA release. Furthermore, when in the presence of lower pH
associated with endosomes, or when bound to DNA, the degra-
dation of these dendrons becomes ineffective on the transfection
time scale—we propose that this explains the poor transfection
performance of these dendrons.
As such, this integrated experimental and theoretical approach
has provided a unique insight into the way in which gene delivery
vectors can approach the multiple different extra- and intracel-
lular barriers to gene delivery. Our primary target is now the
development of high-affinity DNA binding systems with the potential
to degrade, while still bound to DNA, under biologically useful
time scales and at biologically relevant pH values. Clearly, future
work on these vectors will therefore focus on enhancing the
degradation pathways and enabling degradation-induced intra-
cellular disassembly of the nanostructures. This kind of enhanced
degradation process will trigger the loss of multivalency, improv-
ing DNA release, and ultimately enabling effective gene delivery.
Figure 14. Fluorescence data demonstrating that on incubating a
mixture of Chol-G2, calf thymus DNA, and ethidium bromide for extended
periods of time the EthBr fluorescence remains quenched, indicating that
the dendron does not release the DNA and allow reintercalation of the
EthBr.
Chol-G2 would be the better dendron for achieving rapid DNA
release, in agreement with our experimental observations of gene
delivery.
To gain further insight into the process of DNA release and
dendron degradation under biologically relevant conditions, we
then designed an assay to monitor the DNA release capability of
the self-assembling dendrons. This was achieved using a variant
of the ethidium bromide (EthBr) assay in which calf thymus
DNA, EthBr, and the DNA binder (either Chol-G2 or Chol2-
G2) were mixed and incubated at 37 °C. This leads to a decrease
in fluorescence as the EthBr is displaced from the DNA double
helix by the presence of the dendron.
We reasoned that as the dendron degraded under these con-
ditions it should release the DNA, and the EthBr would reintercalate
into the DNA, switching its fluorescence back on. However, even
on standing the solution for a period of 1 week, the EthBr fluo-
rescence intensity remained essentially unchanged (Figure 14).
This is in direct contrast to our previous studies, in which the
dendron was degraded alone in buffer for a similar period of time
and then shown to be completely incapable of binding DNA and
displacing EthBr.7 We therefore propose that complete degrada-
tion of the dendron does not occur effectively when it is complexed to
the DNA. It is possible that the presence of the DNA prevents
the base-catalyzed hydrolysis of ester bonds in a similar way to
lowering pH or that once the amine groups are bound to the
DNA they are no longer able to catalyze intramolecular break-
down of the dendron framework. As a consequence, we therefore
propose that ineffective dendron degradation when bound to
DNA is a primary cause of incomplete transfection and is a key
problem associated with barrier 5 on the transfection pathway in
this case.
’ ASSOCIATED CONTENT
S
Supporting Information. Full synthetic methods and
b
characterization data for all new compounds, experimental methods
for all DNA binding, gene delivery, toxicity and other assays, full
details of mesoscale modeling methods, and additional data. This
material is available free of charge via the Internet at http://pubs.
acs.org.
’ AUTHOR INFORMATION
Corresponding Author
david.smith@york.ac.uk
’ ACKNOWLEDGMENT
We acknowledge The University of York and BBSRC for
supporting this research through the award of a DTA studentship
as well as the EU for assisting networking activities through the
COST mechanism as part of TD0802 (Dendrimers in Bio-
medicine). Part of this work was carried out in the York Centre
of Excellence in Mass Spectrometry, which was created thanks to
a major capital investment through Science City York, supported
by Yorkshire Forward with funds from the Northern Way Initiative.
P.P. and S.P. wish to acknowledge the financial support from
ESTECO through the project DDOS. D.W.P. acknowledges
financial support from National Institutes of Health grants
GM085222 and DK083875 (to M.E.H.).
’ CONCLUSIONS AND FUTURE WORK
We have reported a novel series of dendrons, in which it is
demonstrated that self-assembly, DNA binding ability, and gene
delivery potential can all be controlled in predictable ways by the
nature of the functional group at the focal point. Multiscale
modeling can be used to rationalize the DNA binding ability of all
self-assembled nanostructures. Using a combination of experi-
mental methods, it was then demonstrated that the dendrons
could navigate barriers 1À4 on the transfection pathway. Pleas-
ingly, it was noted that some of these vectors could significantly
’ REFERENCES
(1) (a) Rolland, O.; Turrin, C.-O.; Caminade, A.-M.; Marjoral, J.-P.
New J. Chem. 2009, 33, 1809–1824. (b) Boas, U.; Christensen, J. B.;
Heegaard, P. M. H. Dendrimers in Medicine and Biotechnology; Royal
Society of Chemistry: Cambridge, 2006. (c) Klajnert, B.; Bryszewska, M.
Dendrimers in Medicine; Nova Science: UK, 2007. (d) Majoros, I. J.;
20298
dx.doi.org/10.1021/ja2070736 |J. Am. Chem. Soc. 2011, 133, 20288–20300

Journal of the American Chemical Society
ARTICLE
Baker, J. R. Dendrimer-Based Nanomedicine; Pan Stanford Publishing:
Singapore, 2008.
Wagner, B.; Kansy, M.; Heinzelmann, H.; Pugin, R.; Diederich, F.;
Gallani, J. L. Langmuir 2007, 23, 737–746.
(2) (a) Smith, D. K.; Hirst, A. R.; Love, C. S.; Hardy, J. G.; Brignell,
S. V.; Huang, B. Prog. Polym. Sci. 2005, 30, 220–293. (b) Rosen, B. M.;
Wilson, C. J.; Wilson, D. A.; Peterca, M.; Imam, M. R.; Percec, V. Chem.
Rev. 2009, 109, 6275–6570.
(3) Posocco, P.; Pricl, S.; Jones, S. P.; Barnard, A.; Smith, D. K. Chem.
Sci. 2010, 1, 393–404.
(16) (a) Toth, I.; Sakthivel, T.; Wilderspin, A. F.; Bayele, H.;
O’Donnell, M.; Perry, D. J.; Pasi, K. J.; Lee, C. A.; Florence, A. T. STP
Pharm. Sci. 1999, 9, 93–99. (b) Shah, D. S.; Sakthivel, T.; Toth, I.;
Florence, A. T.; Wilderspin, A. F. Int. J. Pharm. 2000, 208, 41–48. (c) Al-
Jamal, K. T.; Ramaswamy, C.; Singh, B.; Florence, A. T. J. Drug Delivery
Sci. Technol. 2005, 15, 11–18. (d) Bayele, H. K.; Sakthivel, T.; O’Donell,
M.; Pasi, K. J.; Wilderspin, A. F.; Lee, C. A.; Toth, I.; Florence, A. T.
J. Pharm. Sci. 2005, 94, 446–457. (e) Bayele, H. K.; Ramaswamy, C.;
Wilderspin, A. F.; Srai, K. S.; Toth, I.; Florence, A. T. J. Pharm. Sci. 2006,
95, 1227–1237.
(17) (a) Wood, K. C.; Little, S. R.; Langer, R.; Hammond, P. T.
Angew. Chem., Int. Ed. 2005, 44, 6704–6708. (b) Wood, K. C.; Azarin,
S. M.; Arap, W.; Pasqualini, R.; Langer, R.; Hammond, P. T. Bioconjugate
Chem. 2008, 19, 403–405. (c) Poon, Z.; Lee, J. A.; Huang, S.; Prevost,
R. J.; Hammond, P. T. Nanomedicine 2011, 7, 201–209.
(18) (a) Karmali, P. P.; Chaudhuri, A. Med. Res. Rev. 2007, 27, 696–722.
(b) Bhattacharya, S.; Bajaj, A. Chem. Commun. 2009, 4632–4656.
(19) (a) Kodama, K.; Katayama, Y.; Shoji, Y.; Nakashima, H. Curr.
Med. Chem. 2006, 13, 2155–2161. (b) Mintzer, M. A.; Simanek, E. E.
Chem. Rev. 2009, 109, 259–302.
(20) Edelstein, M. L.; Abedi, M. R.; Wixon, J. J. Gene Med. 2007,
9, 833–842.
(21) Kostiainen, M. A.; Hardy, J. G.; Smith, D. K. Angew. Chem., Int.
Ed. 2005, 44, 2556–2559.
(22) (a) Hardy, J. G.; Kostiainen, M. A.; Smith, D. K.; Gabrielson,
N. P.; Pack, D. W. Bioconjugate Chem. 2006, 17, 172–178. (b) Jones,
S. P.; Gabrielson, N. P.; Pack, D. W.; Smith, D. K. Chem. Commun.
2008, 4700–4702.
(23) Jones, S. P.; Gabrielson, N. P.; Wong, C.-H.; Chow, H.-F.; Pack,
D. W.; Posocco, P.; Fermeglia, M.; Pricl, S.; Smith, D. K. Mol. Pharm.
2011, 8, 416–429.
(24) (a) Pavan, G. M.; Danani, A.; Pricl, S.; Smith, D. K. J. Am. Chem.
Soc. 2009, 131, 9686–9694. (b) Jones, S. P.; Pavan, G. M.; Danani, A.;
Pricl, S.; Smith, D. K. Chem.—Eur. J. 2010, 16, 4519–4532.
(25) (a) Ihre, H.; Hult, A.; Soederlind, E. J. Am. Chem. Soc. 1996,
118, 6388–6395. (b) Ihre, H.; Hult, A.; Frꢀechet, J. M. J.; Gitsov, I.
Macromolecules 1998, 31, 4061–4068. (c) Ihre, H.; Padilla de Jesus,
O. L.; Frꢀechet, J. M. J. J. Am. Chem. Soc. 2001, 123, 5908. (d) Gillies,
E. R.; Frꢀechet, J. M. J. J. Am. Chem. Soc. 2002, 124, 14137–14146. (e)
Gillies, E. R.; Dy, E.; Frꢀechet, J. M. J.; Szoka, F. C. Mol. Pharmaceutics
2005, 2, 129–138. (f) Lee, C. C.; Gillies, E. R.; Fox, M. E.; Guillaudeu,
S. J.; Frꢀechet, J. M. J.; Dy, E. E.; Szoka, F. C. Proc. Natl. Acad. Sci. U.S.A.
2006, 103, 16649–16654.
(26) Wu, P.; Malkoch, M.; Hunt, J. N.; Vestberg, R.; Kaltgrad, E.;
Finn, M. J.; Fokin, V. V.; Sharpless, K. B.; Hawker, C. J. Chem. Commun.
2005, 5775–5777.
(27) Rodrigo, A. C.; Barnard, A.; Cooper, J.; Smith, D. K. Angew.
Chem., Int. Ed. 2011, 50, 4675–4679.
(4) (a) Mammen, M.; Choi, S. K.; Whitesides, G. M. Angew. Chem.,
Int. Ed. 1998, 37, 2754–2794. (b) Mulder, A.; Huskens, J.; Reinhoudt,
D. N. Org. Biomol. Chem. 2004, 2, 3409–3424. (c) Badjic, J. D.; Nelson,
A.; Cantrill, S. J.; Turnbull, W. B.; Stoddart, J. F. Acc. Chem. Res. 2005,
38, 723–732. (d) Huskens, J. Curr. Opin. Chem. Biol. 2006, 10, 537–543.
(e) Martos, V.; Castreno, P.; Valero, J.; de Mendoza, J. Curr. Opin. Chem.
Biol. 2008, 12, 698–706. (f) Welsh, D. J.; Smith, D. K. Org. Biomol. Chem.
2011, 9, 4795–4801.
(5) (a) Amir, R. J.; Pessah, N.; Shamis, M.; Shabat, D. Angew. Chem.,
Int. Ed. 2003, 42, 4494–4499. (b) Shamis, M.; Lode, H. N.; Shabat, D.
J. Am. Chem. Soc. 2004, 126, 1726–1731. (c) Perry, R.; Amir, R. J.;
Shabat, D. New J. Chem. 2007, 31, 1307–1312. (d) Sagi, A.; Segal, E.;
Satchi-Fainaro, R.; Shabat, D. Bioorg. Med. Chem. 2007, 15, 3720–3727.
(e) Sella, E.; Shabat, D. Chem. Commun. 2008, 5701–5703. (f) Szalai,
M. L.; Kevwitch, R. M.; McGrath, D. V. J. Am. Chem. Soc. 2003,
125, 15688–15689. (g) Kevwitch, R. M.; McGrath, D. V. New J. Chem.
2007, 31, 1332–1336. (h) de Groot, F. M. H.; Albrecht, C.; Koekkoek,
R.; Beusker, P. H.; Scheeren, H. W. Angew. Chem., Int. Ed. 2003, 42,
4490–4494.
(6) (a) Kostiainen, M. A.; Smith, D. K.; Ikkala, O. Angew. Chem., Int.
Ed. 2007, 46, 7600–7604. (b) Kostiainen, M. A.; Rosilo, H. Chem.—Eur.
J. 2009, 15, 5656–5660. (c) Pavan, G. M.; Kostiainen, M. A.; Danani, A.
J. Phys. Chem. B 2010, 114, 5686–5693. (d) Kostiainen, M. A.; Kotimaa,
J.; Laukkanen, M. L.; Pavan, G. M. Chem.—Eur. J. 2010, 16, 6912–6918.
(e) Kostiainen, M. A.; Kasyutich, O.; Cornelissen, J. J. L. M.; Nolte,
R. J. M. Nature Chem. 2010, 2, 394–399.
(7) Welsh, D. J.; Jones, S. P.; Smith, D. K. Angew. Chem., Int. Ed.
2009, 48, 4047–4051.
(8) (a) Dufes, C.; Uchegbu, I. F.; Schatzlein, A. G. Adv. Drug Delivery
Rev. 2005, 57, 2177–2202. (b) Guillot-Nieckowski, M.; Eisler, S.;
Diederich, F. New J. Chem. 2007, 31, 1111–1127. (c) Smith, D. K. Curr.
Top. Med. Chem. 2008, 8, 1187–1203.
(9) (a) Kukowska-Latallo, J.-F.; Bielinska, A. U.;Johnson, J.;Spindler,
R.; Tomalia, D. A.; Baker, J. R. Proc. Natl. Acad. Sci. U.S.A. 1996, 93,
4897–4902. (b) Bielinska, A.; Kukowska-Latallo, J. F.; Johnson, J.;
Tomalia, D. A.; Baker, J. R. Nucleic Acids Res. 1996, 24, 2176–2182.
(10) (a) Haensler, J.; Szoka, F. C. Bioconjugate Chem. 1993,
4, 372–379. (b) Tang, M. X.; Redemann, C. T.; Szoka, F. C. Bioconjugate
Chem. 1996, 7, 703–714. (c) Turunen, M. P.; Hiltunen, M. O.; Ruponen,
M.; Virkamaki, L.; Szoka, F. C.; Urtti, A.; Yla-Herttuala, S. Gene Ther.
1999, 6, 6–11.
(11) (a) Choi, J. S.; Lee, E. J.; Choi, Y. H.; Jeong, Y. J.; Park, J. S.
Bioconjugate Chem. 1999, 10, 62–65. (b) Ohsaki, M.; Okuda, T.; Wada,
A.; Hirayama, T.; Niidome, T.; Aoyagi, H. Bioconjugate Chem. 2002,
13, 510–517.
(12) (a) Wong, S. Y.; Pelet, J. M.; Putnam, D. Prog. Polym. Sci. 2007,
32, 799–837. (b) Tiera, M. J.; Winnik, F. M.; Fernandes, J. C. Curr. Gene
Ther. 2006, 6, 59–71. (c) Midoux, P.; Breuzard, G.; Gomez, J. P.; Pichon,
C. Curr. Gene Ther. 2008, 8, 335–352.
(13) (a) Lv, H.; Zhang, S.; Wang, B.; Cui, S.; Yan, J. J. Controlled
Release 2006, 114, 100–109. (b) Duncan, R.; Izzo, L. Adv. Drug Delivery
Rev. 2005, 57, 2215–2237.
(14) (a) Kuo, J.-h. S.; Lin, Y. L. J. Biotechnol. 2007, 129, 383–390.
(b) Kuo, J.-h. S.; Liou, M.-j.; Chiu, H.-c. Mol. Pharmaceutics 2010, 7,
805–814.
(15) (a) Joester, D.; Losson, M.; Pugin, R.; Heinzelmann, H.;
Walter, E.; Merkle, H. P.; Diederich, F. Angew. Chem., Int. Ed. 2003,
42, 1486–1490. (b) Guillot, M.; Eisler, S.; Weller, K.; Merkle, H. P.;
Gallani, J. L.; Diederich, F. Org. Biomol. Chem. 2006, 4, 766–769. (c)
Guillot-Nieckowski, M.; Joester, D.; Stohr, M.; Losson, M.; Adrian, M.;
(28) Stuart, M. C. A.; van de Pas, J. C.; Engberts, J. B. F. N. J. Phys.
Org. Chem. 2005, 18, 929–934.
(29) (a) Hoogerbrugge, P. J.; Koelman, J. Europhys. Lett. 1992, 19,
155–160. (b) Espa~nol, P.; Warren, P. Europhys. Lett. 1995, 30, 191–196.
(c) Groot, R. D.; Warren, P. B. J. Chem. Phys. 1997, 107, 4423–4435.
(d) Groot, R. D.; Rabone, K. L. Biophys. J. 2001, 81, 725–736.
(30) (a) Israelachvili, J. N.; Mitchell, D. J.; Ninham, B. W. J. Chem.
Soc., Faraday Trans. 2 1976, 72, 1525–1568. (b) Israelachvili, J. N.;
Mitchell, D. J.; Ninham, B. W. Biochim. Biophys. Acta 1977, 470, 185–201.
(31) (a) Cain, B. F.; Baguley, B. C.; Denny, W. A. J. Med. Chem.
1978, 21, 658–668. (b) Gershon, H.; Ghirlando, R.; Guttman, S. B.;
Minsky, A. Biochemistry 1993, 32, 7143–7151.
(32) (a) Wong, H.; Victor, J. M.; Mozziconacci, J. PLos One 2007, 2
(9), e877. (b) Abdelhady, H. G.; Allen, S.; Davies, M. C.; Roberts, C. J.;
Tendler, S. J. B.; Williams, P. M. Nucleic Acid Res. 2003, 31, 4001–4005.
(c) Dorigo, B.; Schalch, T.; Kulangara, A.; Duda, S.; Schroeder, R.;
Richmond, T. J. Science 2004, 306, 1571–1573. (d) Olins, A. L.; Olins,
D. E. Science 1974, 183, 330–332.
20299
dx.doi.org/10.1021/ja2070736 |J. Am. Chem. Soc. 2011, 133, 20288–20300

Journal of the American Chemical Society
ARTICLE
(33) (a) Thomas, M.; Klibanov, A. M. Proc. Natl. Acad. Sci. U.S.A.
2002, 99, 14640–14645. (b) Forrest, M. L.; Koerber, J. T.; Pack, D. W.
Bioconjugate Chem. 2003, 14, 934–940. (c) Gabrielson, N. P.; Pack,
D. W. Biomacromolecules 2006, 7, 2427–2435. (d) Gabrielson, N. P.;
Pack, D. W. J. Controlled Release 2009, 136, 54–61.
(34) (a) Guy, J.; Drabek, D.; Antoniou, M. Mol. Biotechnol. 1995,
3, 237–248. (b) Pless, D. D.; Wellner, R. B. J. Cell. Biochem. 1996,
62, 27–39. (c) Erbacher, P.; Roche, A. C.; Monsigny, M.; Midoux, P. Exp.
Cell Res. 1996, 225, 186–194.
(35) Baytekin, B.; Baytekin, H. T.; Hahn, U.; Reckien, W.; Kirchner,
B.; Schalley, C. A. Chem.—Eur. J. 2009, 15, 7139–7149.
20300
dx.doi.org/10.1021/ja2070736 |J. Am. Chem. Soc. 2011, 133, 20288–20300