Activation of Cell-Penetrating Peptides with Ionpair-π Interactions and Fluorophiles

Article abstract of DOI:10.1021/jacs.6b06253

In this report, we elaborate on two new concepts to activate arginine-rich cell-penetrating peptides (CPPs). Early on, we have argued that repulsion-driven ion-pairing interactions with anionic lipids account for their ability to move across hydrophobic c

Full text of DOI:10.1021/jacs.6b06253

Article
pubs.acs.org/JACS
Activation of Cell-Penetrating Peptides with Ionpair−π Interactions
and Fluorophiles
Nicolas Chuard,† Kaori Fujisawa,†,§ Paola Morelli,† Jacques Saarbach,† Nicolas Winssinger,†
Pierangelo Metrangolo,‡ Giuseppe Resnati,‡ Naomi Sakai,† and Stefan Matile*,†
†Department of Organic Chemistry, University of Geneva, CH-1211 Geneva, Switzerland
‡NFMLab, Department of Chemistry, Materials and Chemical Engineering “Giulio Natta”, Politecnico di Milano, via Mancinelli 7,
I-20131 Milan, Italy
S
* Supporting Information
ABSTRACT: In this report, we elaborate on two new
concepts to activate arginine-rich cell-penetrating peptides
(CPPs). Early on, we have argued that repulsion-driven ion-
pairing interactions with anionic lipids account for their ability
to move across hydrophobic cell membranes and that
hydrophobic anions such as pyrenebutyrate can accelerate
this process to kinetically outcompete endosomal capture. The
original explanation that the high activity of pyrenebutyrate
might originate from ionpair−π interactions between CPP and
activator implied that replacement of the π-basic pyrene with
polarized push−pull aromatics should afford more powerful CPP activators. To elaborate on this hypothesis, we prepared a small
collection of anionic amphiphiles that could recognize cations by ionpair−π interactions. Consistent with theoretical predictions,
we find that parallel but not antiparallel ionpair−π interactions afford operational CPP activators in model membranes and cells.
The alternative suggestion that the high activity of pyrenebutyrate might originate from self-assembly in membranes was explored
with perfluorinated fatty acids. Their fluorophilicity was expected to promote self-assembly in membranes, while their high acidity
should prevent charge neutralization in response to self-assembly, i.e., generate repulsion-driven ion-pairing interactions.
Consistent with these expectations, we find that perfluorinated fatty acids are powerful CPP activators in HeLa cells but not in
model membranes. These findings support parallel ionpair−π interactions and repulsion-driven ion pairing with self-assembled
fluorophiles as innovative concepts to activate CPPs. These results also add much corroborative support for counterion-mediated
uptake as the productive mode of action of arginine-rich CPPs.
proximity effect, counterions are bound tightly, but their
exchange remains fast. In other words, repulsion-driven ion-
pairing interactions ensure that CPPs will never be alone, but
the nature of their counterions can change easily. For cellular
uptake, CPPs will thus arrive at lipid bilayer membranes
together with hydrophilic counterions such as phosphates or
chlorides. Counterion exchange with anionic lipids will bind the
CPPs to the cell membrane (Figure 1D). CPPs with sufficient
activity will cross the membrane through micellar pores (Figure
1E).2,5 These transient pores are dynamic enough to let large
substrates pass and to heal once the CPPs have passed through.
In the cytosol, CPPs detach from the membrane by counterion
exchange with intracellular polyanions (Figure 1F). If ion-
pairing mediated translocation is slow (kt, Figure 1D), then
kinetically competing endocytosis (ke) dominates. Also in such
cases, i.e., ke > kt, repulsion-driven ion-pairing translocation
contributes to endosomal escape. This mode of action applies
to arginine-rich CPPs and their models, particularly oligoargi-
nines. Their activity is dependent on size, cargo, concentration,
INTRODUCTION
■
Since their discovery in the late 1980s, the question how
arginine-rich cell-penetrating peptides (CPPs) and their mimics
move across cell membranes has caused intense debates in
chemistry and biology.1,2 After all, oligomers such as the
fluorescently labeled octaarginine 1 used in this study are
hydrophilic polycations (Figure 1A). They could not be
expected to pass a hydrophobic membrane barrier as if it
would not be there. Intrigued by a puzzling cation selectivity
observed with arginine-rich synthetic pores,3 we proposed early
on that repulsion-driven ion-pairing interactions account for the
productive mode of action of CPPs.4 These most important
interactions originate from the poor acidity of the guanidinium
cation.2 When brought into close proximity, the more acidic
ammonium cations can eliminate charge repulsion by release of
protons, i.e., reduction of the pKa of all proximal ammonium
cations except for one (Figure 1B). This proximity effect is
operational throughout biology (e.g., class I aldolases).2 With
the less acidic guanidinium cations, proton release is essentially
excluded in neutral water. The only remaining solution to
minimize intramolecular charge repulsion in guanidinium-rich
oligomers are counterions (Figure 1C). As a result of this
Received: June 17, 2016
© XXXX American Chemical Society
A
DOI: 10.1021/jacs.6b06253
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
accelerate direct translocation across lipid bilayer membranes
and thus kinetically outcompete endocytosis. Toward this end,
many hydrophobic anions have been characterized as CPP
activators in bulk (U tube, identifying counterion-activated
CPPs as anion carriers4) and model membranes (dye-loaded
vesicles).2,4,10−12 Activities have been found to depend on the
nature of both the CPPs and the CPP activators. Particularly
attractive are the perfectly planar carboxylate-guanidinium
pairs,2,4 naturally including assistance from preorganized
hydrogen bonding.10
Until recently, pyrenebutyrate 2 was the only counterion
known to activate CPPs in a broad variety of cells (Figure 2).13
Figure 2. (A) The originally proposed structure of CPPs bound to
pyrenebutyrate 2. (B) Activators 3 could operate with parallel (p) and
antiparallel (a) ionpair−π interactions on push−pull surfaces. They are
defined by the direction of the push−pull dipoles (blue-to-red arrows,
with partial charges) relative to orientation of the ion pair on their π
surface (red and blue circles, small arrows).
Figure 1. (A) Structure of the fluorescently labeled Cy5-CPP 1 used in
this study. (B, C) Repulsion-driven ion-pairing interactions: The more
acidic ammonium cations minimize charge repulsion between proximal
lysine residues by deprotonation (B), and the less acidic guanidinium
cations have to minimize charge repulsion between proximal arginine
residues with repulsion-driven ion-pairing interactions (C). (D−F)
The productive mode of action of CPPs: Counterion exchange with
anionic lipids (or CPP activators, red) to bind to the outer membrane
surface (D), formation of and translocation through transient dynamic
micellar pores (E), and release by counterion exchange with internal
hydrophilic polyions (F); kinetic competition from endocytosis (ke)
occurs if repulsion-driven ion-pairing translocation is slow (kt).
The mode of activation has originally been explained with
repulsion-driven ion-pairing on the electron-rich pyrene surface
(Figure 2A).4 The directionality of the resulting complex could
possibly orient the CPPs more toward the interior of the
membrane and thus facilitate the formation of transient micellar
pores (Figure 1E). The originally proposed structure of the
complex between CPP and activator 2 has recently been
recognized as one of the rare examples of operational
ionpair−π interactions in biology (Figure 2A).14 However,
the introduction of ionpair−π interactions also suggested that
pyrenebutyrate activators should be far from perfect. Electron-
rich π surfaces such as the one of pyrene are π basic,
characterized by a negative quadrupole moment Qzz in the z
direction perpendicular to the plane and inward in-plane
dipoles from donating substituents. This situation is suited to
attract cations to the π surface.15 For anion−π interactions, π-
acidic surfaces with Qzz > 0 and outward in-plane dipoles from
withdrawing substituents are required.16 To accommodate both
an anion and a cation on the same aromatic surface, polarized
push−pull surfaces with both π-acidic and π-basic side should
be ideal (Figure 2B).14
and cell type.1,2 Their cytotoxicity usually increases with
concentration and number of guanidinium cations per oligo-/
polymer.1,2,6 Cell-penetrating poly(disulfide)s (CPDs) have
been introduced recently to address this important problem
with CPPs.2,6
The concept of counterion-mediated function with repul-
sion-driven ion-pairing interactions2,4 has been not only
introduced to explain the productive mode of action of
arginine-rich CPPs but also applied to explain the voltage gating
of potassium channels7 and to create conceptually new
sensors2,8 and other responsive organic materials.9
In ionpair−π interactions, the orientation of the ion pair can
be parallel or antiparallel with regard to the dipole of the push−
pull system (Figure 2B). Recent insights from covalent systems
indicated that parallel ionpair−π interactions are stronger than
antiparallel ionpair−π interactions.12,14 These findings sug-
For cellular uptake, the concept of repulsion-driven ion-
pairing interactions implied that counterions other than the
intrinsic anionic lipids could be used to activate CPPs.2,4
Reminiscent of catalysts, CPP activators have been conceived to
B
DOI: 10.1021/jacs.6b06253
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
gested that push−pull aminonaphthalimides (ANIs) 3p
designed for parallel ionpair−π interactions should be better
CPP activators than antiparallel activators 3a and should also be
better than pyrenebutyrate 2. Very recent preliminary results in
model vesicles were consistent with these admittedly high
expectations.12
Synthesis. The Cy5-labeled octaarginine 1 was synthesized
by automated solid-phase synthesis (Figure 1A, Schemes S1
and S2). Activators 2 and 13−16 were commercially available,
and all other activators were easily accessible in a few steps
following the procedures developed previously for the original
ionpair−π activators 3 (Figure 3, Schemes 1 and S3−S10). The
CPP activators other than pyrenebutyrate 2 that work in cells
have been reported very recently.11 Namely, several ordinary
fatty acids, essentially inactive at physiological conditions, have
been shown to activate CPPs under more basic conditions. This
finding suggested that fatty acid activators are not sufficiently
deprotonated under neutral conditions, which in turn implies
that they self-assemble in the membrane (see below).
a
Scheme 1. Synthesis of Ionpair−π Activators 7a and 7p
Stimulated by these recent findings, we decided to more
systematically explore the possibility to activate CPPs with
parallel ionpair−π interactions on push−pull surfaces on the
one hand and self-assembled activators on the other. In the
following, we report the design, synthesis, and evaluation of a
small collection of new activators. Experimental support is
provided that parallel but not antiparallel ionpair−π inter-
actions on push−pull surfaces might contribute to CPP
activation, and perfluorinated fatty acids are introduced as
remarkably powerful supramolecular CPP activators.
RESULTS AND DISCUSSION
■
Design. The design of ionpair−π activators 4−16 was based
on pyrenebutyrate 2 and parallel and antiparallel ionpair−π
activators 3a and 3p (Figure 3).12 The shortcomings of
a
(a) EtOH, 90 °C, 15 h, 34%; (b) DMSO, TEA, 90 °C, 3 d, 72%; (c)
TFA, CH2Cl2, rt, 2 h, 50%; (d) TEA, DMF, 2 h, μW, 52%; (e) 90 °C,
15 h, 79%; (f) TFA, CH2Cl2, rt, 2 h, 79%.
synthesis of one of the most interesting pair of activators, i.e., 7,
is outlined in Scheme 1. All details on synthesis and compound
characterization can be found in the Supporting Information.19
CPP Activation in Vesicles. Activities in the model
membranes were determined with the CF assay.4 For these
experiments, large unilamellar vesicles (LUVs) composed of
egg yolk phosphatidylcholine (EYPC) were loaded with
carboxyfluorescein (CF). Intravesicular CF concentrations
were selected high enough to ensure self-quenching. CF export
then results in fluorescence recovery due to local dilution.
In a typical experiment, CF emission was followed during the
addition of first the activator, then polyarginine (pR) as
simplest possible CPP mimic, and finally a detergent to destroy
the vesicles for calibration (Figure 4A). Significant increases in
emission before CPP addition identify activators that act as
detergents. This can be observed for most activators4 at
sufficiently high concentrations; studies on CPP activation
naturally have to focus on sublytic concentrations. Fluorescence
kinetics for CPP activation were repeated with different
concentrations, and the results were analyzed in dose response
curves (Figure 4B). The activator characteristics obtained from
Hill analysis are the maximal activity Ymax and the EC50, which
is the activator concentration needed to observe 50% of Ymax
(Table 1, Figure 5).
Figure 3. Structure of amphiphiles tested as CPP activators.
ionpair−π activators 3 are (i) poor push−pull polarization,
because of a weak, twisted tertiary amine donor and (ii)
inconsistent attachment of alkyl chains, making the amphiphi-
licity of the two constitutional isomers incomparable. To
overcome these deficiencies, the collection of ionpair−π
activators 4−11 was designed based on the central motif a
and p arranged for antiparallel and parallel ionpair−π
interactions, respectively. Activators 12 were of interest as
controls with spacers between carboxylate on the π surface that
are too short to fold into correct Leonard turns.17
Fluorophiles18 13−16 were added to the collection to elaborate
on activator self-assembly in lipid bilayer membranes.
The Ymax obtained for different activators varied broadly. The
Ymax = 97% of the parallel ionpair−π activator 7p, for example,
revealed excellent activity (Figures 5A, 4B, red, ○, Table 1,
entry 5). On the other hand, the Ymax = 14% observed for
antiparallel ionpair−π activator 7a (Figures 5A, 4B, blue ●,
Table 1, entry 15) and Ymax = 21% for perfluorinated fatty acid
14 (Figures 5A, 4B, green △, Table 1, entry 22) indicated that
C
DOI: 10.1021/jacs.6b06253
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
Figure 4. (A) Changes in CF emission I (λex 492 nm, λem 517 nm) in
CF-loaded EYPC LUVs (62.5 μM EYPC) as a function of time during
the addition of activators 7p (red, solid) or 14 (green, dashed) at t =
50 s (12.5 μM), pR (250 nM) at t = 100 s (I = 0%) and excess triton
X-100 at t = 500 s for final calibration (I = 100%). (B) Dependence of
final activity Y of pR (right before lysis in A) on the concentration of
○
●
△
activators 7p (red ), 7a (blue ), and 14 (green , with curve fit to
Hill equation).
Table 1. Characteristic Parameters for CPP Activators in
a
Vesicles Determined by Hill Analyses
b
c
d
e
activators
Ymax (%)
EC50 (μM)
n
f
1
2
3p
61
9
4
1.05 0.05
inactive
2.1 0.2
4p
3
5p
87
66
97
24
27
59
3
2
5
3
1
3
6.2 0.2
3.8 0.6
2.4 0.3
1.8 0.5
1.3 0.2
1.9 0.2
4.4 0.7
1.8 0.7
4
6p
1.18 0.07
0.54 0.09
5
7p
6
8p
6
1
7
9p
2.6 0.2
30
8
10p
11p
12p
1
9
70 10
0.31 0.08
1.2 0.2
3.6 0.2
inactive
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
68
75
3
4
4
1
f
g
3a
1.8 0.2
Figure 5. (A) Maximal activity Ymax and (B) effective concentration
EC50 in vesicles as a function of the calculated log P, i.e., the
hydrophobicity of the CPP activators estimated from their structure.
From Hill analyses of dose response curves (Figure 4B). (C) Maxima
in absorption spectra of ANI activators in EYPC LUVs. ANI activators
a series: red circles; p series: blue squares; fluorinated fatty acids: green
triangles; pyrenebutyrate: black diamonds; irregular or control
compounds: open symbols. Isomers a and p are connected by gray
dashed lines. Trend lines are added for activators 4−10.
4a
g
5a
58
36
14
19
33
12
25
32
21
21
7
19
15
1
3
3.0 0.6
1.9 0.7
g
6a
7a
inactive
8a
1
9
3.5 0.1
2.8 0.2
9a
30
8
4
3
10a
11a
12a
13
14
15
16
inactive
inactive
g
28
61
1
8
2.3 0.2
1
1
5
8
4
2
hydrophobicity of the activator (Figure 5A). Ymax ≥ 50% was
defined as range of meaningful activity (Figure 5A, gray line).
In addition, EC50 = 10 μM was considered as an appropriate
limit of meaningful activities (Figure 5B, gray line). Activators
4−11 feature a flexible Leonard turn17 to position the
carboxylate on the aromatic surface (Figure 6). In the series
p, the carboxylate is above the imide acceptor to support CPP
binding by parallel ionpair−π interactions. In constitutional
5.1 0.3
inactive
inactive
14
a
Determined using eq S2.19 Hill analyses were not applied to those
b
with low Ymax values. Errors are of curve fit. Activators, see Figures 2
and 3. Maximal activity. Effective concentration to reach 50% activity
of Ymax. Hill coefficient. From ref 12. Maximum observed Y before
the onset of the detergent effect. Hill analyses were made assuming
Ymax = 100%.
c
d
e
f
g
these CPP activators are essentially inactive in vesicles. The
apparent EC50 = 5.1 0.3 μM obtained for 14 was obviously
nearly meaningless considering this inacceptable Ymax = 21%
(Figures 5B, 4B, green △, Table 1, entry 22), whereas the
excellent Ymax = 97% identified the EC50 = 0.54
0.09 μM
measured for 7p as a meaningful value (Figures 5B, 4B, red ○,
Table 1, entry 5).
All new CPP activators were characterized in vesicles
following this procedure (Table 1). The found activities Ymax
were ordered as a function of the calculated log P, that is the
Figure 6. CPP activators 4−11 (m = 1) and 12 (m = 0) that can
possibly operate with parallel (left) and antiparallel (right) ionpair−π
interactions.
D
DOI: 10.1021/jacs.6b06253
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
isomers a, the carboxylate is above the amine donor for
antiparallel ionpair−π interactions with CPPs. In the series 4−
8, these parallel and antiparallel isomers were equipped with
linear alkyl tails of increasing length from C4 to C18 (Figure 3).
This collection was complemented with unsaturated oleyl tails
in 9 because of their importance in the delivery of siRNA2 and a
cycloalkyl tail in 10.
In CF-loaded vesicles, all parallel ionpair−π activators
showed excellent Ymax except for the hydrophilic and hydro-
phobic extremes, i.e., 4p, 8p and 9p (Figure 5A, red circles,
Table 1, entries 1−10). In sharp contrast, most antiparallel
ionpair−π activators were essentially inactive, except for the
original 3a and 5a (Figure 5A, blue squares, Table 1, entries
11−20).
increasing hydrophobicity of the ionpair−π activators, the
absorption maxima of both isomers shifted steadily to the blue,
from almost 460 nm down to almost 430 nm. Considering the
positive solvatochromism of the push−pull chromophore,12 this
trend indicated that with increasing overall hydrophobicity of
the activators, the ANI core penetrates deeper in the
hydrophobic core of the membrane. Similar location of parallel
and antiparallel ionpair−π activators in the membrane (Figure
5C) suggested that the observed dramatic differences in CPP
activation (Figure 5A, B) do not originate from differences in
partitioning. This conclusion was in excellent agreement with
operational parallel and dysfunctional antiparallel ionpair−π
interactions in CPP activators that are otherwise almost
identical (Figure 6).
The trends found with Ymax were overall nicely reproduced
with the EC50. Except for extreme log P, parallel ionpair−π
activators p were all very active (EC50 < 10 μM, Figure 5B, red
circles), whereas their antiparallel constitutional isomers a
showed much less convincing activities (Figure 5B, blue
squares).
The original pyrenebutyrate 2 was special in the sense that
high Ymax = 77% coincided with an EC50 = 43 4 μM (Figures
5A and B, black diamonds). This unusually poor effective
concentration was consistent with weak ionpair−π interactions
on the π-basic pyrene surface (Figure 2).
The only significant exception from the general blue shift of
ionpair−π activators with increasing hydrophobicity concerned
3p and 11p (Figure 5C). Their absorption in vesicles appeared
about 20 nm more hypsochromic than expected from their
hydrophobicity. These blue shifts probably originate from the
twisted push−pull system in 3p and self-assembly of fluorous
11p, rather than particularly deep partitioning into the
membrane.
Interestingly, polyfluorinated activators 13−16 were essen-
tially inactive in vesicles, independent of the length of their tails
and the basicity of their carboxylates (Figure 5A, B, green
triangles). Hybrids 11 behaved more like ionpair−π rather than
fluorous activators, i.e., parallel 11p was active, while
antiparallel 11a was not. Polyfluorinated activator 11p was
slightly more active than expected from the general trend
(Figure 5B, red lines), its EC50 = 310 80 nM was the best of
the entire collection (Table 1, entry 9).
CPP Activation in Cells. The uptake of fluorescently
labeled octaarginine 1 into HeLa Kyoto cells in the absence and
the presence of ionpair−π activators and fluorophiles was
examined using confocal laser scanning microscopy (CLSM).
The cells were incubated first with the activators for 5 min and
then with 1 μM Cy5-CPP 1 for another 15 min at 37 °C in the
standard Leibovitz medium (commercially available for cell
culture in CO2-free atmosphere). Then the cells were washed
with PBS containing heparin (20 U/mL)21 and Leibovitz
medium to remove reversibly bound CPPs. Activators were
mostly used at 10 μM, cell death could usually be observed
above 20 μM, increasing with increasing concentration,
dependent on the nature of the activator.
In the absence of activators, HeLa cells incubated with Cy5-
CPP 1 showed the punctate fluorescence characteristic for
endosomal capture (Figure 7A). Moreover, fluorescence
intensities were overall relatively weak. Nearly identical images
were obtained with antiparallel ionpair−π activator 7a (Figure
7B). In clear contrast, some of the cells preincubated with
activator 7p showed exceptionally bright fluorescence after
incubation with Cy5-CPP 1, staining all over within the cells
(Figure 7C). Interestingly, two extreme populations of cells
could be observed, together with examples in-between. A
significant minority of highly fluorescent cells (Figure 7C)
alternated with a majority of others that showed weak emission
from endosomes only, as for uptake of Cy5-CPP 1 in the
absence of activators 7p (Figure 7A). The overall appearance of
the cells was similar to the original findings with pyrenebuty-
rate.13
Controls 12 with a shortened turn were characterized by less
distinct behavior. The EC50 = 1.2
0.2 μM of the most
relevant 12p was slightly less than expected from the
hydrophobicity (Figure 5B). This slight underperformance
was consistent with less than perfect turns to place the
carboxylates on the push−pull surface (Figure 6). Previous
results with shortened turns in anion−π catalysis confirmed
that it would be unreasonable to expect larger changes from the
removal of one carbon in the turn.17
Interestingly, the original 3a with weakened push−pull
system and disturbed amphiphilicity was the best among the
otherwise consistently inactive antiparallel ionpair−π activators
a (Figure 5A, B). In all controls with less than perfect
architecture, i.e., 3 and 12, the difference between EC50’s for
parallel and antiparallel ionpair−π interactions was less
pronounced compared to the general trends (Figure 5A, B,
red and blue lines).
Hill coefficients should be interpreted carefully because they
increase not only with the stoichiometry but also with the
instability of the formed complexes.20 With all activators, Hill
coefficients n were much smaller than expected for stoichio-
metric ion pairing (Table 1). These observations support that
CPP−activator complexes are as stable as expected and that
stoichiometric ion pairing with all the 33−99 guanidinium
groups in pR is presumably not necessary to mediate its passage
across the bilayer membrane. Within meaningful Ymax ≥ 50%,
highest n were found for too hydrophilic activators (5p, 6p,
10p) and too short spacers (12p, Table 1, Figure 5A). These
observations were consistent with less stable ionpair−π
complexes formed in more polar environments and with
incorrect turns.
With correct Leonard turns17 and meaningfully balanced
intermediate hydrophobicity, the difference in activity between
parallel and antiparallel ionpair−π activators was overall much
larger and more consistent than expected (Figure 5A, B, red vs
blue lines). Nearly identical absorption maxima found for the
constitutional isomers in vesicles suggested that their local-
ization in membranes is similar (Figure 5C). Namely, with
The observed trends were consistent with results in CF-
loaded vesicles (Figure 5). They supported that parallel
ionpair−π activators interact better with the positively charged
E
DOI: 10.1021/jacs.6b06253
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
Figure 7. CLSM (left, 5% laser power), differential interference
contrast (DIC, middle), and merged images (right) of HeLa cells after
15 min incubation with Cy5-CPP 1 (1 μM) without activators (A) and
after preincubation for 5 min with 10 μM activator 7a (B) and 7p (C),
all in Leibovitz medium at 37 °C. Scale bar: 10 μm.
CPPs, resulting in better uptake in cells, whereas antiparallel
ionpair−π activators are much less active and give cellular
distributions as for CPP 1 without activators (Figure 6).
Control experiments with activators only confirmed that the
observed fluorescence originates from the Cy5 attached to CPP
1 and not from the much weaker ANI fluorophore (Figure S4).
Similar trends were observed with activators 11 with
perfluorinated tails. CPP uptake with parallel ionpair−π
interactions in activator 11p was clearly better than with
antiparallel ionpair−π interactions in activators 11a (Figure
S5). Moreover, fluorous ionpair−π activators 11p enabled CPP
uptake in larger populations of cells in comparison to
nonfluorinated activators 7p. This increased apparent activity
of ionpair−π activators 11 with fluorous tails could possibly
originate from fluorophilic self-assembly18 in cell membranes.
To elaborate on this hypothesis, CPP activation by pure
fluorophiles 13 and 14 was explored. Under the conditions
developed for ionpair−π activators, uptake experiments with
CPP 1 in the presence of perfluorinated fatty acids 13 and 14
gave highly fluorescent HeLa Kyoto cells (Figure 8A, B).
Compared to results with ionpair−π activators, perfluorinated
fatty acids 13 and 14 delivered CPP 1 more evenly to most
cells. Fluorophile-activated CPPs appeared to localize prefer-
entially in intracellular membranes. Significant diffuse overall
fluorescence indicated CPP release into the cytosol, whereas
the nuclei were much less fluorescent (Figures 8A, B).
Polyfluorinated fatty acids 15 and 16 did not activate CPP 1
for uptake into HeLa cells. The CLSM images showed weak
fluorescence with a distribution similar to Cy5-CPP 1 in the
absence of any activator (Figure 8C, D). This inactivity of
controls 15 and 16 provided direct experimental support that
activators 13 and 14 operate by repulsion-driven ion-pairing
interactions with self-assembled fluorophiles (Figure 9).
Namely, the fluorous self-assembly of activators in membranes
will bring their negative charges into close proximity. The
resulting proximity effects should be exactly complementary to
those that account for the productive mode of action of CPPs
(Figure 1C). Strongly basic bundles of anions will overcome
Figure 8. CLSM (left, 1% laser power for A, B, 5% for C, D), DIC
(middle), and merged images (right) of HeLa cells after 5 min of
incubation with activator 13 (A), 14 (B), 15 (C), and 16 (D, 10 μM
each) and then 15 min with Cy5-CPP 1 (1 μM) in Leibovitz medium
at 37 °C. Scale bar: 10 μm.
Figure 9. Fluorophilic CPP activators that operate with repulsion-
driven ion-pairing interactions: (A) More basic carboxylates minimize
charge repulsion between self-assembled activators by protonation.
(B) Less basic carboxylates minimize charge repulsion between self-
assembled activators by repulsion-driven ion pairing, i.e., the very
stable yet sufficiently labile binding of CPPs (compare Figure 1).
charge repulsion by protonation (Figure 9A). “Unprotonat-
able,” weakly basic bundles of anions will strongly bind
counterions (Figure 9B). Examples for the protonation of
proximal carboxylates at pH 7 include acid catalysis in
glycosidases or proteases,2 self-assembled fatty acids from
premicellar aggregates to soap bubbles,22 and so on. Repulsion-
driven ion-pairing interactions with weakly basic, unprotonat-
F
DOI: 10.1021/jacs.6b06253
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
able oligophosphodiesters (DNA, RNA, etc.) afford cation
transporters and sensors in bulk and lipid bilayer membranes2
and are the basis of all approaches to DNA/RNA delivery that
operate with lipoplexes.2
π acidity of push−pull ANIs and the nearly identical
spectroscopic properties of parallel and antiparallel ionpair−π
activators in membranes (Figure 5C) did not support this
conceivable alternative interpretation.
In amphiphiles 15 and 16, the absence of fluorines on the
proximal carbons increases the basicity of the carboxylates.
Fluorophilic self-assembly18 in cell membranes thus fails to
increase their activity because of proximity-induced protonation
(Figure 9A). The recent observation that ordinary fatty acids
can activate CPPs under basic conditions,11 not under neutral,
was consistent with this interpretation, suggesting that these
fatty acids self-assemble in the membrane and minimize charge
repulsion by protonation at pH = 722 and by CPP binding only
at pH > 7.
Under physiological conditions, the fluorous fatty acids 13
and 14 offer the perfect combination of fluorophilicity and
“unprotonatable” carboxylates. Complementary to guanidinium
cations in CPPs (Figure 1C), these weaker bases generate
operational repulsion-driven ion-pairing interactions in re-
sponse to fluorophilic self-assembly18 in cell membranes
(Figure 9B). The result is most efficient CPP uptake (Figure
8). The cumulation of repulsion-driven ion-pairing interactions
from self-assembled perfluorinated fatty acids on the one side
and CPPs on the other side could also explain the apparent
accumulation in intracellular membranes: Release into the
cytosol by exchange with hydrophilic polyanions could become
increasingly difficult (Figures 9B, 1F).
Interestingly, the most powerful activation of CPPs by
repulsion-driven ion pairing with self-assembled perfluorinated
fatty acids in HeLa cells was not predictable from transport
experiments in vesicles. In these model systems, polyfluorinated
fatty acids 13−16 were all similarly inactive, independent of the
basicity of the carboxylate (Figure 5A). This was in clear
contrast to ionpair−π activators including perfluorinated fatty
acid 11, which showed clear preference for parallel over
antiparallel activators in vesicles and cells. This important
difference was, however, understandable and meaningful
considering the different processes detected in vesicles and
cells. In CF-loaded vesicles, the export of intravesicular CF is
monitored. Detailed studies, including positive controls in bulk
membranes and anionic lipids as activators in bilayer
membranes, provided experimental support that the export of
anionic CF by pR−activator complexes occurs via counterion
exchange.2,4 In competition with ionpair−π activators, this ion
exchange on the CPP scaffold is as unproblematic as with all
other anionic activators tested previously. With the cumulation
of repulsion-driven ion-pairing interactions in both self-
assembled fluorophiles and CPPs, the temporary substitution
of one or more of the tightly bound activators by CF becomes
more difficult. As a result, CF export from CF-loaded vesicles
does not occur. In clear contrast, occurrence and detection of
cellular uptake does not require exchange of one or more
activators in CPP−activator complexes with an anionic
fluorophore. This mechanistic difference thus explains why
with conventional activator−CPP complexes, uptake is
predictable from vesicles, and with stronger activator−CPP
complexes, cellular uptake is possible and detectable, whereas
CF transport in vesicles is not.
CONCLUSIONS
■
The concept of CPP activators has been introduced as early as
2003.4 Reminiscent of catalysts, activators were expected to
accelerate direct translocation of CPPs across lipid bilayer
membranes and thus overcome endosomal capture. However,
significant activation of CPPs to enter cells at pH 7 without
endosomal capture has been limited so far to one example, i.e.,
pyrenebutyrate.13 This report introduces more CPP activators
that work in cells. Most importantly, their design is rational and
conceptually innovative, aiming to apply lessons from supra-
molecular chemistry to cellular uptake.
Repulsion-driven ion-pairing interactions,4 stable and labile
at the same time, account for the counterion hopping that
allows CPPs to move across the hydrophobic membrane
barriers.2,4 Perfluorinated fatty acids, the most powerful
activators discovered in this study, offer repulsion-driven ion-
pairing interactions also from the activator side. They combine
the fluorophilicity18 to self-assemble in the membrane with the
poor basicity needed to attract CPPs with repulsion-driven ion-
pairing interactions. The best fluorophilic activators, as simple
as perfluorinated lauric acid, are shown to enable most efficient
delivery of fluorescently labeled octaarginines into HeLa cells
without endosomal capture. The excellent performance of
fluorophilic activators in cells is not reproduced in model
vesicles, presumably because assays in CF-loaded vesicles report
on CPP-mediated anion export rather than CPP uptake. New
assays reporting directly on CPP uptake in model membranes
would be desirable.
Parallel ionpair−π interactions12 are found to contribute to
CPP activation, whereas antiparallel ionpair−π interactions
inhibit CPP activators in model vesicles and in cell. Parallel
ionpair−π activators cause exceptionally efficient CPP uptake
into some cells, which become highly fluorescent, whereas
other cells remain untouched. Although consistent with
theoretical predictions and pertinent controls, including
binding studies to membranes and CPPs, it could not be
expected that the difference between parallel and antiparallel
ionpair−π interactions would be so dramatic. Despite
remarkably consistent results, it is thus important to reiterate
that direct evidence for operational ionpair−π interactions for
CPP activation is probably inaccessible. Contributions from
other effects should never be fully excluded.
In summary, these results introduce repulsion-driven ion-
pairing and parallel ionpair−π interactions as innovative
concepts to elaborate on CPP activators. Naturally, these
results also support counterion-mediated uptake with repul-
sion-driven ion-pairing interactions as the productive mode of
action of arginine-rich CPPs.2,4 Application of the lessons
learned should enable rapid progress with the development of
CPP activators for general use in practice.
ASSOCIATED CONTENT
■
The mode of action of fluorous CPP activators implied that
differences in carboxylate basicity could possibly account also
for the differences in activity between parallel and antiparallel
ionpair−π activators. Anion−π interactions have been found to
cause basicity changes up to ΔpKa = 5.5.23 However, the weak
S
* Supporting Information
The Supporting Information is available free of charge on the
ACS Publications website at DOI: 10.1021/jacs.6b06253.
Experimental details (PDF)
G
DOI: 10.1021/jacs.6b06253
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
(8) Whitney, M.; Savariar, E. N.; Friedman, B.; Levin, R. A.; Crisp, J.
L.; Glasgow, H. L.; Lefkowitz, R.; Adams, S. R.; Steinbach, P.; Nashi,
N.; Nguyen, Q. T.; Tsien, R. Y. Angew. Chem., Int. Ed. 2013, 52, 325−
330.
(9) Okuro, K.; Sasaki, S.; Aida, T. J. Am. Chem. Soc. 2016, 138, 5527−
5530.
AUTHOR INFORMATION
■
Corresponding Author
*stefan.matile@unige.ch
Present Address
§Aichi Institute of Technology, Aichi 470-0392, Japan
(10) Rothbard, J. B.; Jessop, T. C.; Lewis, R. S.; Murray, B. A.;
Wender, P. A. J. Am. Chem. Soc. 2004, 126, 9506−9507.
(11) Herce, H. D.; Garcia, A. E.; Cardoso, M. C. J. Am. Chem. Soc.
2014, 136, 17459−17467.
(12) Fujisawa, K.; Humbert-Droz, M.; Letrun, R.; Vauthey, E.;
Wesolowski, T. A.; Sakai, N.; Matile, S. J. Am. Chem. Soc. 2015, 137,
11047−11056.
(13) (a) Takeuchi, T.; Kosuge, M.; Tadokoro, A.; Sugiura, Y.; Nishi,
M.; Kawata, M.; Sakai, N.; Matile, S.; Futaki, S. ACS Chem. Biol. 2006,
1, 299−303. (b) Candan, G.; Michiue, H.; Ishikawa, S.; Fujimura, A.;
Hayashi, K.; Uneda, A.; Mori, A.; Ohmori, I.; Nishiki, T.; Matsui, H.;
Tomizawa, K. Biomaterials 2012, 33, 6468−6475.
(14) Fujisawa, K.; Beuchat, C.; Humbert-Droz, M.; Wilson, A.;
Wesolowski, T. A.; Mareda, J.; Sakai, N.; Matile, S. Angew. Chem., Int.
Ed. 2014, 53, 11266−11269.
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
We thank the NMR, the Sciences Mass Spectrometry (SMS)
and the imaging platforms for services, A. Roux and G.
Gasparini for access to and assistance with cell culture and the
University of Geneva, the European Research Council (ERC
Advanced Investigator), the Swiss National Centre of
Competence in Research (NCCR) Chemical Biology, the
NCCR Molecular Systems Engineering and the Swiss NSF for
financial support.
(15) (a) Dougherty, D. A. Acc. Chem. Res. 2013, 46, 885−893.
(b) Mahadevi, A. S.; Sastry, G. N. Chem. Rev. 2013, 113, 2100−2138.
(16) (a) Giese, M.; Albrecht, M.; Rissanen, K. Chem. Commun. 2016,
52, 1778−1795. (b) Chifotides, H. T.; Dunbar, K. R. Acc. Chem. Res.
2013, 46, 894−906. (c) Ballester, P. Acc. Chem. Res. 2013, 46, 874−
884. (d) Frontera, A.; Gamez, P.; Mascal, M.; Mooibroek, T. J.;
Reedijk, J. Angew. Chem., Int. Ed. 2011, 50, 9564−9583. (e) Wang, D.-
X.; Wang, M.-X. Chimia 2011, 65, 939−943.
REFERENCES
■
(1) (a) Bartolami, E.; Bouillon, C.; Dumy, P.; Ulrich, S. Chem.
Commun. 2016, 52, 4257−4273. (b) Priegue, J. M.; Crisan, D. N.;
Martínez-Costas, J.; Granja, J. R.; Fernandez-Trillo, F.; Montenegro, J.
Angew. Chem., Int. Ed. 2016, 55, 7492−7495. (c) McKinlay, C. J.;
Waymouth, R. M.; Wender, P. A. J. Am. Chem. Soc. 2016, 138, 3510−
3517. (d) Li, M.; Schlesiger, S.; Knauer, S. K.; Schmuck, C. Angew.
Chem., Int. Ed. 2015, 54, 2941−2944. (e) DeRonde, B. M.; Birke, A.;
Tew, G. N. Chem. - Eur. J. 2015, 21, 3013−3019. (f) Douat, C.;
Aisenbrey, C.; Antunes, S.; Decossas, M.; Lambert, O.; Bechinger, B.;
Kichler, A.; Guichard, G. Angew. Chem., Int. Ed. 2015, 54, 11133−
11137. (g) Brock, R. Bioconjugate Chem. 2014, 25, 863−868.
(h) Eggimann, G. A.; Blattes, E.; Buschor, S.; Biswas, R.; Kammer,
S. M.; Darbre, T.; Reymond, J. L. Chem. Commun. 2014, 50, 7254−
7257. (i) Bechara, C.; Sagan, S. FEBS Lett. 2013, 587, 1693−1702.
(n) Inoue, M.; Tong, W.; Esko, J. D.; Tor, Y. ACS Chem. Biol. 2013, 8,
1383−1388. (j) Tang, H.; Yin, L.; Kim, K. H.; Cheng, J. Chem. Sci.
(17) Zhao, Y.; Beuchat, C.; Mareda, J.; Domoto, Y.; Gajewy, J.;
Wilson, A.; Sakai, N.; Matile, S. J. Am. Chem. Soc. 2014, 136, 2101−
2111.
(18) (a) Marsh, E. N. G. Acc. Chem. Res. 2014, 47, 2878−2886.
(b) Krafft, M. P.; Riess, J. G. Chem. Rev. 2009, 109, 1714−1792.
(c) Zhang, W.; Curran, D. P. Tetrahedron 2006, 62, 11837−11865.
(19) See SI.
(20) Bhosale, S.; Matile, S. Chirality 2006, 18, 849−856.
(21) Sakai, N.; Takeuchi, T.; Futaki, S.; Matile, S. ChemBioChem
2005, 6, 114−122.
(22) Kanicky, J. R.; Shah, D. O. Langmuir 2003, 19, 2034−2038.
(23) Miros, F. N.; Zhao, Y.; Sargsyan, G.; Pupier, M.; Besnard, C.;
Beuchat, C.; Mareda, J.; Sakai, N.; Matile, S. Chem. - Eur. J. 2016, 22,
2648−2657.
2013, 4, 3839−3844. (k) Nakase, I.; Akita, H.; Kogure, K.; Graslund,
̈
̈
A.; Langel, U.; Harashima, H.; Futaki, S. Acc. Chem. Res. 2012, 45,
1132−1139. (l) Ornelas-Megiatto, C.; Wich, P. R.; Frechet, J. M. J.
́
Am. Chem. Soc. 2012, 134, 1902−1905. (m) Appelbaum, J. S.;
LaRochelle, J. R.; Smith, B. A.; Balkin, D. M.; Holub, J. M.; Schepartz,
A. Chem. Biol. 2012, 19, 819−830. (n) Piest, M.; Engbersen, J. F. J. J.
Controlled Release 2011, 155, 331−340. (o) Kolonko, E. M.; Pontrello,
J. K.; Mangold, S. L.; Kiessling, L. L. J. Am. Chem. Soc. 2009, 131,
7327−7333. (p) McNaughton, B. R.; Cronican, J. J.; Thompson, D. B.;
Liu, D. R. Proc. Natl. Acad. Sci. U. S. A. 2009, 106, 6111−6116.
(q) Torchilin, V. P. Adv. Drug Delivery Rev. 2008, 60, 548−558.
(r) Gump, J. M.; Dowdy, S. F. Trends Mol. Med. 2007, 13, 443−448.
(s) Fernandez-Carneado, J.; Van Gool, M.; Martos, V.; Castel, S.;
Prados, P.; De Mendoza, J.; Giralt, E. J. Am. Chem. Soc. 2005, 127,
869−874. (t) Fillon, Y.; Anderson, J.; Chmielewski, J. J. Am. Chem. Soc.
2005, 127, 11798−11803. (u) Zhou, P.; Wang, M.; Du, L.; Fisher, G.
W.; Waggoner, A.; Ly, D. H. J. Am. Chem. Soc. 2003, 125, 6878−6879.
(2) Gasparini, G.; Bang, E.-K.; Montenegro, J.; Matile, S. Chem.
Commun. 2015, 51, 10389−10402.
́
(3) Sakai, N.; Sorde, N.; Das, G.; Perrottet, P.; Gerard, D.; Matile, S.
Org. Biomol. Chem. 2003, 1, 1226−1231.
(4) Sakai, N.; Matile, S. J. Am. Chem. Soc. 2003, 125, 14348−14356.
(5) Herce, H. D.; Garcia, A. E.; Litt, J.; Kane, R. S.; Martin, P.;
Enrique, N.; Rebolledo, A.; Miles, V. Biophys. J. 2009, 97, 1917−1925.
(6) (a) Gasparini, G.; Bang, E.-K.; Molinard, G.; Tulumello, D. V.;
Ward, S.; Kelley, S. O.; Roux, A.; Sakai, N.; Matile, S. J. Am. Chem. Soc.
2014, 136, 6069−6074. (b) Fu, J.; Yu, C.; Li, L.; Yao, S. Q. J. Am.
Chem. Soc. 2015, 137, 12153−12160.
(7) Schmidt, D.; Jiang, Q. X.; MacKinnon, R. Nature 2006, 444,
775−779.
H
DOI: 10.1021/jacs.6b06253
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX