Functional determinants of a synthetic vesicle fusion system

Article abstract of DOI:10.1021/ja711184u

Selective membrane mergers may be driven by small-molecule recognition between synthetic surface-displayed fusogens which bear vancomycin glycopeptide and its native binding target, D-Ala-D-Ala dipeptide. These recognition motifs are membrane anchored by

Full text of DOI:10.1021/ja711184u

Published on Web 04/11/2008
Functional Determinants of a Synthetic Vesicle Fusion
System
Yun Gong, Mingming Ma, Yumei Luo, and Dennis Bong*
Department of Chemistry, The Ohio State UniVersity, Columbus, Ohio 43210
Received December 17, 2007; E-mail: bong@chem.osu.edu
Abstract: Selective membrane mergers may be driven by small-molecule recognition between synthetic
surface-displayed fusogens which bear vancomycin glycopeptide and its native binding target, D-Ala-D-Ala
dipeptide. These recognition motifs are membrane anchored by antimicrobial peptide magainin II and a
phosphatidylethanolamine lipid derivative, respectively. We report herein characterization of this synthetic
membrane fusion reaction with regard to the following: effects of fusogen concentration, lipid composition,
and membrane charge. Our findings indicate that these parameters are determinants of fusion rate, vesicle
stability, peptide binding, catalytic fusion and membrane disruption during fusion. Notably, these data indicate
the importance of coupling between molecular recognition and insertion for bilayer activation as well as
the critical role of membrane subdomain formation for membrane fusion reactivity. These phenomena are
general to lipid membrane chemistry, and therefore these findings provide a guideline for understanding
more complex biomembrane systems.
Introduction
mechanically deform the membrane, lowering the activation
energy for lipid mixing and fusion.1 Notably, enveloped viruses
Experimental and theoretical studies1,2 concur that membrane
fusion proceeds through at least two steps: membrane docking
and actual fusion, resulting in the mixing of membrane lipids
and membrane-bound contents (Scheme 1). Fusion may occur
upon close (1–2 nm)2 docking of target membranes, driven by
the binding of surface groups. Docking “strains” the surfaces,
allowing lipids from the two membranes to mix and ultimately
form a fusion pore connecting the two compartments. Insertion
of a hydrophobic anchor into the lipid matrix can frustrate
efficient lipid packing and activate the membrane toward
noncovalent reactions such as lysis and fusion; these reactions
are essentially lipidic3 and are precisely controlled in Nature
by molecular recognition events. Much of experimental data
on selective membrane fusion has been gathered in studies of
synaptic4 vesicle fusion machinery as well as in viral fusion
machinery.5–7 Physical membrane deformation or insertion of
a hydrophobic fusion peptide allows the formation of a high-
energy intermediate nonbilayer lipid surface8 that fuses with
its target membrane when drawn into apposition by surface
binding. Native membrane recognition elements are proteins,
which in class I viral fusion5 and synaptic vesicle fusion are
coiled-coils. Helical bundle formation draws the membranes into
apposition; this binding is thought to locally dehydrate and
such as HIV6 and influenza7 similarly employ coiled-coil
recognition to guide fusion with the host membrane. That all
known native membrane fusion is driven by protein recognition
(and often with coiled coils), raises the question of whether
recognition strategies between small molecules would also be
fusion competent. Such a minimal fusogenic molecular system
would be useful to determine the fundamental requirements for
membrane fusion catalysis. A simple synthetic model system
of specific fusion via molecular recognition would allow
physical organic methods to be applied to rigorously probe the
scope and limitations of controlled lipid membrane fusion. All
known molecular recognition systems that induce selective
fusion are derived from native fusogenic proteins that must be
expressed; further, their fusogenic behavior is complex and
modulated by other species in vivo.9 Thus, detailed study of
membrane fusion phenomena using native fusogenic systems
as probes is complicated by limited control over chemical
content and the participation of multiple proteins in the native
system. Though there have been reports of small molecule
fusogenic systems,10,11 these systems lack controlled valency,
defined partner selection and are not thoroughly characterized.
(9) (a) Stein, A.; Radhakrishnan, A.; Riedel, D.; Fasshauer, D.; Jahn, R.
Nat. Struct. Mol. Biol. 2007, 14, 904–911. (b) Martens, S.; Kozlov,
M. M.; McMahon, H. T. Science 2007, 316, 1205–1208. (c) Dennison,
S. M.; Bowen, M. E.; Brunger, A. T.; Lentz, B. R. Biophys. J. 2006,
90, 1661–1675. (d) Scales, S. J.; Finley, M. F. A.; Scheller, R. H.
Science 2001, 294, 1015–1016.
(1) Jahn, R.; Lang, T.; Sudhof, T. C. Cell 2003, 112, 519–533.
(2) Blumenthal, R.; Clague, M. J.; Durell, S. R.; Epand, R. M. Chem.
ReV. 2003, 103, 53–69.
(3) Lentz, B. R.; Malinin, V.; Haque, M. E.; Evans, K. Curr. Opin. Struct.
Biol. 2000, 10, 607–615.
(10) (a) Marchi-Artzner, V.; Gulik-Krzywicki, T.; Guedeau-Boudeville,
M.-A.; Gosse, C.; Sanderson, J. M.; Dedieu, J.-C.; Lehn, J.-M. Chem-
PhysChem 2001, 2, 367–376. (b) Marchi-Artzner, V.; Jullien, L.;
Gulik-Krzywicki, T.; Lehn, J.-M. Chem. Commun. 1997, 117–118.
(c) Paleos, C. M.; Tsiourvas, D. J. Mol. Rec. 2006, 19, 60–67.
(11) Richard, A.; Marchi-Artzner, V.; Lalloz, M.-N.; Brienne, M.-J.;
Artzner, F.; Gulik-Krzywicki, T.; Guedeau-Boudeville, M.-A.; Lehn,
J.-M. Proc. Natl. Acad. Sci. U.S.A. 2004, 101, 15279–15284.
(4) Lin, R. C.; Scheller, R. H. Annu. ReV. Cell DeV. Biol. 2000, 16, 19–
49.
(5) Sollner, T. H. Curr. Opin. Cell Biol. 2004, 16, 429–435.
(6) Chan, D. C.; Kim, P. S. Cell 1998, 93, 681–684.
(7) Carr, C. M.; Chaudhry, C.; Kim, P. S. Proc. Natl. Acad. Sci. U.S.A.
1997, 94, 14306–13.
(8) Yang, L.; Ding, L.; Huang, H. W. Biochemistry 2003, 42, 6631–6635.
9
6196 J. AM. CHEM. SOC. 2008, 130, 6196–6205
10.1021/ja711184u CCC: $40.75
2008 American Chemical Society

Determinants of a Synthetic Vesicle Fusion System
A R T I C L E S
Scheme 1. Model of Molecular Recognition Guided Vesicle Fusion
Scheme 2. Selective Vesicle Fusion Guided by Vancomycin/ D-Ala-D-Ala Recognition
We have previously reported a well-defined chemical model
system which has allowed us to examine the molecular
determinants of membrane fusion.12 We report here our detailed
examination of the parameters that define membrane fusion
behavior. Our designed fusion system incorporates the two
functions of recognition and disruption as compactly as possible
(Scheme 2).
We chose vancomycin glycopeptide and D-Ala-D-Ala dipep-
tide as a recognition pair to guide fusion and an antimicrobial
peptide, magainin II, fused to vancomycin served both as a
membrane anchor and a membrane disrupting module, while
the D-Ala-D-Ala dipeptide was anchored using a POPE lipid
derivative. Reaction of fusogen-functionalized vesicles as shown
in Scheme 2 results in rapid vesicle aggregation and fusion;
this can be completely suppressed by addition of underivatized
vancomycin, which blocks all available surface D-Ala-D-Ala sites
and therefore, membrane apposition. Vancomycin is an antibiotic
of last resort that inhibits the transpeptidation step of pepti-
doglycan synthesis by binding to D-Ala-D-Ala dipeptide on lipid
II, a key biosynthetic intermediate of the peptidoglycan cell wall
of bacteria.13 Recognition occurs via the formation of five
hydrogen bonds between the two peptide backbones and the
free C-terminus of the dipeptide (Figure 1). Due to its therapeutic
importance, this micromolar dissociation binding event is very
well studied.13 Though only loosely structured, membrane ligand
presentation results in increased binding avidity,14 which can
overcome surface repulsion energies and result in membrane
apposition. The importance of vancomycin has led to a large
body of literature on how it interacts with D-Ala-D-Ala as well
as synthetic methodology for modification of the commercially
available drug; we therefore judged the vancomycin-D-Ala-D-
Ala recognition pair to be an ideal starting point for exploration
of small-molecule triggered fusion.
Frog skin-derived antimicrobial peptide magainin II,15,16 was
used to anchor vancomycin to the membrane for reasons similar
to our choice of recognition pair: magainin is a very well-studied
peptide whose membrane binding mode is known, and it is also
(12) Gong, Y.; Luo, Y.; Bong, D. J. Am. Chem. Soc. 2006, 128, 14430–
14431.
(13) Kahne, D.; Leimkuhler, C.; Lu, W.; Walsh, C. Chem. ReV. 2005, 105,
425–448.
(14) (a) Lahiri, J.; Isaacs, L.; Tien, J.; Whitesides, G. M. Anal. Chem. 1999,
71, 777–790. (b) Mammen, M.; Chio, S.-K.; Whitesides, G. M. Angew.
Chem., Int. Ed. 1998, 37, 2755–2794.
(15) Zasloff, M. Nature 2002, 415, 389–395.
(16) Zasloff, M. Proc. Natl. Acad. Sci. U.S.A. 1987, 84, 5449–5453.
(17) Gennis, R. B. Biomembranes: Molecular Structure and Function;
Springer-Verlag: New York, 1989.
(18) Brown, W. J.; Chambers, K.; Doody, A. Traffic 2003, 4, 214–221.
(19) Epand, R. M.; Shai, Y.; Segrest, J. P.; Anantharamaiah, G. M.
Biopolymers 1995, 37, 319–38.
Figure 1. Vancomycin (top) binds to Lys-D-Ala-D-Ala (bottom) via five
hydrogen bonds.
(20) Oren, Z.; Shai, Y. Biopolymers 1998, 47, 451–463.
9
J. AM. CHEM. SOC. VOL. 130, NO. 19, 2008 6197

A R T I C L E S
Gong et al.
Figure 2. Synthetic fusogens 1 (vancomycin-magainin conjugate) and 2 (Lys-D-Ala-D-Ala phospholipid).
Scheme 3. Sulfhydryl Functionalization of Vancomycin
known to destabilize (perturb) membranes in a concentration
dependent manner. Notably, experiments using vancomycin
coupled to the relatively nonperturbative POPE lipid anchor
instead of the magainin anchor resulted in liposome aggregation
without fusion, suggesting that a more disruptive anchor is
required to trigger lipid mixing and fusion. Lipid packing
determines membrane strain,17 thus disruption of hydrophobic
packing actiVates membranes for noncovalent reactions such
as fusion. While phospholipases can increase the rate of fusion
by producing single-chain lipids that are hydrophobically
mismatched with two-chain lipids,18 noncoValent modification
by peptide insertion can activate the bilayer for reaction in much
the same way. Shallow insertion of a hydrophobic anchor such
as a peptide into the top monolayer of a bilayer (Figure 2) can
create negative membrane curvature,19–21 which destabilizes the
lamellar phase. Many antimicrobial (AMPs)15,20,22 and viral fusion
peptides23 insert in this way and activate permeation and fusion
by lowering the energy barrier to non-bilayer lipid phases.
Permeation and lysis are dose dependent, and in low concentration
AMPs may bind to membranes without lysis or fusion and with
minimal permeability increases. Thus, natural product AMPs and
viral fusion peptides are possible perturbative components of
designed fusogenic systems if coupled to molecular recognition
motifs. Magainin II binds selectively into the hydrophobic matrix
of negatively charged membranes and perturbs lipid packing
without vesicle fusion in the micromolar peptide concentration
regime; there are possibly many other natural and synthetic peptides
that could serve this role.24 Thus, we have previously prepared
synthetic fusogens 1 and 2 (Figure 2) and verified that they can
indeed selectively induce fusion of vesicular membranes via specific
molecular recognition of surface-displayed vancomycin and Kaa
peptides.12 We present herein detailed study of the parameters
affecting membrane fusion in this system.
Results and Discussion
Synthesis of Vancomycin-Magainin Conjugates. Sulfhydryl
functionality was installed on the vancomycin skeleton either
at the N-terminus or at the C-terminus (Scheme 3) to allow
Michael addition to maleimide-functionalized membrane an-
chors, which are readily prepared by solid-phase peptide
(21) Plank, C.; Zauner, W.; Wagner, E. AdV. Drug DeliV. ReV. 1998, 34,
21–35.
(22) (a) Matsuzaki, K. Pept. Sci. 2005, 41st, 87–88. (b) Tossi, A.; Sandri,
L.; Giangaspero, A. Biopolymers 2000, 55, 4–30.
(23) Epand, R. M. Biochim. Biophys. Acta 2003, 1614, 116–121.
(24) (a) Patch, J. A.; Barron, A. E. Curr. Opin. Chem. Biol. 2002, 6, 872–
877. (b) Porter, E. A.; Wang, X.; Lee, H.-S.; Weisblum, B.; Gellman,
S. H. Nature 2000, 404, 565. (c) Porter, E. A.; Weisblum, B.; Gellman,
S. H. J. Am. Chem. Soc. 2002, 124, 7324–7330. (d) Tew, G. N.; Liu,
D.; Chen, B.; Doerksen, R. J.; Kaplan, J.; Carroll, P. J.; Klein, M. L.;
DeGrado, W. F. Proc. Natl. Acad. Sci. U.S.A. 2002, 99, 5110–5114.
9
6198 J. AM. CHEM. SOC. VOL. 130, NO. 19, 2008

Determinants of a Synthetic Vesicle Fusion System
A R T I C L E S
Scheme 4. Preparation of Linkers for Phospholipid Fusogen 2a
a (a) MsCl, Et3N, THF; (b) NaN3, MeOH; (c) CrO3, acetone; (d) HS(CH2)2SH, Et3N, MeOH; (e) Boc2O, KOH, THF, H2O; (f) TBTU, DIEA, DMF; (g)
LiOH, MeOH, H2O; (h) TFA; (i) N-methoxy-carbonylmaleimide, Et3N.
synthesis.25,26 Reductive alkylation of vancomycin with 1,4-
dithiane-2,5-diol yielded the tertiary N-terminal aminoethane-
thiol 5. Alkylation of the vancosamine sugar was not observed;
this was confirmed by acidic cleavage of the aminoglycoside
to yield a single product by HPLC with mass of aglycon plus
the thioethyl fragment. Glycopeptide was also derivatized by
carboxyamidation at the C-terminus27 with aminoethanedisul-
fide. Interestingly, the thioethyl derivatives of vancomycin 4
and 5 were highly membrane active against synthetic phospho-
choline membranes, inducing rapid and irreversible lipid precipita-
tion upon mixing. We hypothesized that the neutral adducts may
be increasing the membrane interaction of vancomycin and thus
chose to couple cysteine to the glycopeptide C-terminus by
carboxamidation with cystine dimethylester. Saponification and
disulfide reduction yielded a thiolated vancomycin derivative 3 that
retains the native negative charge at the C-terminus. Gratifyingly,
this derivative did not detectably induce vesicle aggregation, as
judged by light-scattering measurements, and was successfully
coupled to a synthetic magainin II derivative bearing an N-terminal
maleimide to yield magainin–vancomycin conjugate 1.
Synthesis of Lys-D-Ala-D-Ala-Phospholipid Derivatives. The
binding partner to 1 was similarly prepared by coupling a
synthetic mercaptopropionamide-capped tripeptide sequence
from the lipid II peptide, Lys-D-Ala-D-Ala (Kaa) to a palmitoyl-
oleoyl phosphatidyl ethanolamine (POPE) lipid derivative.12 The
POPE lipid anchor was functionalized with a PEG linker25,
bearing a terminal maleimide (Scheme 4). Both the magainin
and POPE anchors bear an acetamidobenzamide (ABA) moiety
as a UV label (ε270 ) 18,000 M-1cm-1) to determine concen-
tration. While Lehn and co-workers have shown that long PEG
linkers are necessary to induce fusion in LUVs upon metal
complexation with tethered ligands,11 we have used relatively
short linkers to minimize the entropic cost of surface binding
and to avoid possible complications from PEG-induced mem-
brane activity.
With these maleimide-derivatized anchors in hand, prepara-
tion of peptide and lipid fusogens 1 and 2 was straightforward
(Scheme 5). Mixing vancomycin-cysteinamide with the ma-
gainin anchor in water in a 1:1 ratio resulted in clean formation
of the adduct upon mixing, which was purified on HPLC.
Similarly, phospholipid derivative was reacted with the Kaa-
mercaptopropionate in methanol with diisopropylethylamine to
yield fusogen 2, which was purified by HPLC.
Magainin-Vancomycin Conjugate Binding to Large
Unilamellar Vesicles (LUVs). The binding of magainin to
negatively charged membranes is well established,2829 as is its
inability to bind to neutral membranes. Our magainin-vancomy-
cin conjugate is designed to bind to both negatively charged
membranes and any membrane bearing Lys-D-Ala-D-Ala
(Kaa) lipid fusogen 2. To better understand the role of the
peptide anchor in fusion, we set out to determine the binding
constants of our magainin-vancomycin fusogen 1 to both
negatively charged and Kaa-displaying vesicular membranes.
Fusogen 1 is random coil (unfolded) in aqueous solution,
but rapidly folds into an R-helix as judged by circular
dichroism (CD) upon interaction with LUVs with 10% and
20% negative charge from palmitoyl-oleoyl phosphoglycerol
(POPG) lipids.19,20 While 1 did not fold with neutral
phosphocholine (PC) LUVs, POPG LUVs induced a lipid-
dependent coil-helix transition. There is clearly a strong
electrostatic component for surface binding as the CD signal
increases more intensely and rapidly with increasing POPG
(25) Frisch, B.; Boeckler, C.; Schuber, F. Bioconjugate Chem. 1996, 7,
180–186.
(26) Rich, D. H.; Gesellchen, P. D.; Tong, A.; Cheung, A.; Buckner, C. K.
J. Med. Chem. 1975, 18, 1004–1010.
(28) (a) Jeong, S. W.; O’Brien, D. F. J. Org. Chem. 2001, 66, 4799–802.
(b) Nicolaou, K. C.; Mitchell, H. J.; Jain, N. F.; Bando, T.; Hughes,
R.; Winssinger, N.; Natarajan, S.; Koumbis, A. E. Chem.-Eur. J. 1999,
5, 2648–2667. (c) Schoenmakers Ronald, G.; van de Wetering, P.;
Elbert Donald, L.; Hubbell Jeffrey, A. J. Controlled Release 2004,
95, 291–300.
(27) (a) Metallo, S. J.; Kane, R. S.; Holmlin, R. E.; Whitesides, G. M.
J. Am. Chem. Soc. 2003, 125, 4534–4540. (b) Sharman, G. J.; Try,
A. C.; Dancer, R. J.; Cho, Y. R.; Staroske, T.; Bardsley, B.; Maguire,
A. J.; Cooper, M. A.; O’Brien, D. P.; Williams, D. H. J. Am. Chem.
Soc. 1997, 119, 12041–12047. (c) Try, A. C.; Sharman, G. J.; Dancer,
R. J.; Bardsley, B.; Entress, R. M. H.; Williams, D. H. J. Chem. Soc.,
Perkin Trans. 1997, 2911–2917. (d) Sundram, U. N.; Griffin, J. H. J.
Org. Chem. 1995, 60, 1102–1103.
(29) (a) Wenk, M. R.; Seelig, J. Biochemistry 1998, 37, 3909–3916. (b)
Wieprecht, T.; Apostolov, O.; Seelig, J. Biophys. Chem. 2000, 85,
187–198. (c) Wieprecht, T.; Beyermann, M.; Seelig, J. Biochemistry
1999, 38, 10377–10387.
9
J. AM. CHEM. SOC. VOL. 130, NO. 19, 2008 6199

A R T I C L E S
Gong et al.
Scheme 5. Fragment Coupling to Produce Fusogens 1 and 2
content in the LUV titration. The lipid dependence of helicity
was analyzed using a surface-partitioning model,30 in which
the partitioning constant Kp is defined as:
1, saturating at ∼3% surface concentration. Furthermore,
fractional binding to the POPG surface is consistently higher
than the Kaa surface at the same equilibrium free peptide
(fusogen 1) concentration. Though scatter noise from high lipid
concentration in the CD titration prevents exploration of lower
peptide-lipid ratios, the trend clearly indicates that partitioning
into the 20% POPG membrane is favored over binding to the
1% Kaa membrane. For example, when the equilibrium
concentration of peptide fusogen 1 is 1 µM in solution, the
apparent partitioning constant the membrane is 7000 M-1 as
compared to 18,000 M-1 for the POPG membrane. The difference
in partitioning constants should widen further at lower free peptide
concentrations. It is clear from this analysis that under the conditions
of the fusion reaction, there is effectively no free peptide in solution;
all is membrane-bound, and mostly on the POPG membrane. A
1% loading of fusogen 1 correlates to a peptide-lipid ratio of 0.01
on 20% POPG LUVs, which is completely bound (Xb ) 0.01),
whereas the same loading on a 1% Kaa membrane is only 70%
bound (Xb ) 0.007) (Figure 3). Importantly, although electrostatics
drives peptide fusogen 1 partitioning more so than specific lipid
recognition, the partitioning constants are still within the same order
of magnitude, meaning that fusogenic peptide has significant
affinity to both POPG and Kaa membranes; this is in line with
expectations for a peptide that acts to drive apposition and fusion
of these two surfaces.
Cb ⁄ Cl
Kp )
Cp,free
where Cb, Cl, and Cp, free are concentrations of surface-bound
peptide, total lipid, and free peptide in solution, respectively.
Thus, when surface concentration Xb (Cb/Cl) is plotted as a
function of Cp,free, the slope yields the partitioning constant Kp:
Xb ) Kp × Cp,free
This model assumes that the partitioning constant does not
change with bound peptide concentration, an assumption which
breaks down when binding is driven by electrostatic interactions
or specific lipid binding.29,31 In these latter cases, the propensity
for membrane partitioning decreases the more peptide is bound,
due to either decreased surface potential from the positively
charged peptide, or as a result of saturation of the surface
binding sites. Thus, simple partitioning yields a linear Xb vs
Cp,free plot, while situations with both hydrophobic partitioning
and a secondary interaction (electrostatics or specific recogni-
tion) exhibit nonlinear saturation behavior. In the nonlinear plot,
the apparent partitioning constant may be calculated at each
peptide concentration. We performed this analysis for the
binding of magainin-vancomycin to vesicles bearing a 10%
and 20% negative charge as well as neutral vesicles bearing
2% D-Ala-D-Ala lipid, using CD and isothermal titration
calorimetry (ITC) to follow binding, respectively (Figure 3).
The 10% POPG membrane exhibits a roughly linear fractional
binding plot, consistent with weak electrostatic binding, with
an apparent Kp (slope) of 1300 M-1 (data not shown). As charge
content increases to 20% POPG, the Xb plot becomes signifi-
cantly nonlinear, indicative of a partitioning constant that
decreases as more cationic peptide is bound. Notably, binding
to the 2% Kaa membrane saturates at a surface concentration
Xb of ∼0.01 (1%), indicating two points: (1) the peptide fusogen
1 can only bind the Kaa on the outer monolayer of the vesicle,
which has roughly half the total Kaa, and (2) binding is
completely driven by specific Kaa lipid recognition by 1. The
20% POPG surface has a higher binding capacity for fusogen
Fusion Is Catalytic in Peptide Fusogen 1. We followed fusion
using the same fusion assay as in previous studies;12 NBD and
rhodamine functionalized lipids were incorporated into LUVs
with Kaa lipid 2 and this suspension was treated with POPG
LUVs and peptide fusogen 1; increase in NBD (donor)
fluorescence was taken as a reporter of loss of FRET, membrane
dilution and therefore fusion. In typical fusion reactions, we
found that the rate of change in NBD fluorescence decreased
and leveled off with time, suggesting that fusion also slows and
then stops.12 The observation that charge content greatly affects
fusogen binding suggested that fusion rate should depend on
surface charge differential between the two fusing LUVs. This
is likely the reason why fusion slows and stops; the charge
gradient between the two liposome populations erodes as fusion
proceeds, leading to product inhibition and eventual halting of
fusion as fusogen 1 binds preferentially to the product (fused)
membrane which has both negative charge and Kaa lipid,
forming a “cis” complex. Thus, we reasoned that a fusion system
(30) Beschiaschvili, G.; Seelig, J. Biochemistry 1990, 29, 10995–11000.
(31) Seelig, J. Biochim. Biophys. Acta 1997, 1331, 103–116.
9
6200 J. AM. CHEM. SOC. VOL. 130, NO. 19, 2008

Determinants of a Synthetic Vesicle Fusion System
A R T I C L E S
Figure 3. (A) Mean residue ellipticity of fusogen 1 with 20% POPG LUVs. CD scans decrease in ellipticity at 222 nm when total lipid/peptide mole ratio
(L/P) varies from (green O) zero to (red O ) 62; traces in between (O) represent L/P of 10, 20, 30, 42 and 50, respectively. (B) ITC trace of fusogen 1 into
LUVs with 2% Kaa lipid 2. (C) Fractional binding vs free peptide plots from peptide-lipid titration of fusogen 1 and 20% POPG LUVs followed by CD
(red O), and with 1% Kaa LUVs followed by ITC (blue O). (D) As in (C), plotted as a function of peptide:lipid ratio. Smoothed curves are visual guides
only.
Scheme 6. Charge Gradients Render Fusion Catalytic in Peptide Fusogen 1
catalytic in 1 would be possible. One round of fusion between
neutral and 10% POPG LUVs results in product liposomes with
5% POPG and unreacted 10% POPG LUVs. Addition of 20%
POPG LUVs should competitively bind all 1 in the system as
the partitioning constant is 1–2 orders of magnitude higher than
to 10% POPG, permitting the formation of new “trans”
complexes and a new round of fusion (Scheme 6).
system with 1 equiv of 20% POPG LUVs, without additional
fusogen 1. This resulted in a second burst of fluorescence
change, suggesting that new fusion active LUVs may be formed
if the charge gradient is sufficient. The system showed a
response to multiple injections of naked 20% POPG LUVs, with
diminishing response as expected with the eroding charge
gradient between product and unreacted LUVs. Thus, sequential
rounds of fusion may be accomplished with this strategy; these
FRET dilution experiments32 confirm that this indeed happens
(Figure 4). We allowed a fusion reaction with a 10% negative
charge differential come to equilibrium and then charged the
(32) Struck, D. K.; Hoekstra, D.; Pagano, R. E. Biochemistry 1981, 20,
4093–4099.
9
J. AM. CHEM. SOC. VOL. 130, NO. 19, 2008 6201

A R T I C L E S
Gong et al.
Figure 4. Fusion followed by FRET dilution assay. 4:1 donor (1/10%
POPG) to acceptor (2% 2 in ePC) LUVs. One equivalent LUVs (20%
POPG) added at (a) 168 min, (b) 289 min, and (c) 409 min.
characteristics indicate an artificial fusogenic system whose
behavior is not only robust, but predictable.
Effects of Altering Charge Content and Ligand Concentra-
tion on Fusion Rate. To determine the effects of varying the
relative affinity of peptide fusogen to either reacting membrane,
fusion rate was measured as a function of charge content in the
membrane as well as surface concentration of fusogen Kaa lipid
2. Interestingly, analysis of initial rates indicates that there is a
minimum concentration of ∼0.5% Kaa lipid required before
fusion will occur; after this “turn on” concentration, fusion rate
increases steadily and ultimately levels off at higher Kaa lipid
concentration (Figure 5). The dependence on surface fusogen
concentration suggests there must be a minimum level of
fusogen partitioning into the Kaa membrane before the mem-
branes are activated enough to fuse. A similar experiment was
undertaken by varying charge content in the apposing mem-
brane. In this case, 10% POPG LUVs fused more slowly than
20% POPG, but above 20% negative charge, fusion rates
decreased sharply with a significant change in rate profile (Figure
5). We speculate that more negatively charged membranes bind
the magainin anchor of the peptide fusogen too tightly to allow
the pendant glycopeptide to interact with its Kaa binding partner.
Unlike vancomycin/Kaa lipid binding, magainin-bilayer inter-
actions involve the entire length of the peptide anchor, and
tightening this binding by increased electrostatic interactions
may decrease steric accessibility to the Kaa reactant membrane.
This does not appear to be an issue when Kaa lipid concentration
is increased, possibly because the magainin-vancomycin
conjugate 1 is only anchored in the membrane at a single point
by glycopeptide-lipid recognition, with the magainin anchor
likely partitioning weakly in the neutral PC membrane. These
differences in the way fusogen 1 anchors to the two reacting
membranes may have important functional consequences, as
discussed below.
Peptide-Membrane Binding and Membrane Activation.
Vancomycin binding facilitates insertion of the magainin anchor
into a neutral egg PC membrane, as judged by the Kaa LUV-
dependent CD spectrum, which shows peptide helicity increasing
with lipid concentration (Figure 6). The amphipathic sequence
of many AMPs favors insertion in the lipid matrix33 although
not all of these peptides (such as magainin) insert spontaneously,
requiring assistance from electrostatic interactions. Lipid rec-
ognition could similarly provide the binding energy to drive
spontaneous partitioning. These two different insertion pathways
have strikingly different effects on membrane function.
Figure 5. Fusion rates are represented by relative change in NBD
fluorescence. (A) Fusion as a function of percentage Kaa lipid fusogen 2
in membrane. (B) Initial rates for fusion as function of Kaa surface
concentration. (C) Fusion as a function of percentage of POPG in membrane.
We studied contents release upon fusogen 1 binding with
vesicles that encapsulated dye at self-quenching concentration;34,35
we noted that peptide fusogen 1 induces rapid dequenching
(release) of contents from vesicles with 1% Kaa lipid fusogen
9
6202 J. AM. CHEM. SOC. VOL. 130, NO. 19, 2008

Determinants of a Synthetic Vesicle Fusion System
A R T I C L E S
Figure 6. (A) CD monitoring of lipid to magainin-vancomycin (MV) titration. Two mole percent Kaa lipid in eggPC liposomes are titrated into the
magainin-vancomycin conjugate. (red b) lipid/peptide mole ratio ) 0; (blue 2) lipid/peptide mole ratio ) 30; (O) lipid/peptide mole ratios in between 0
and 30, all traces decrease at 222 nm with ratio ) 30 the lowest. (B) Dye release from liposomes upon treatment with MV at a peptide/lipid ratio of 1:139
and concentrations of 0.625 µM and 86.9 µM, respectively. Green trace ) 2% Kaa lipid in eggPC, lower traces ) 20% POPG in eggPC and 100% eggPC.
(C) Model for how lipid headgroup binding can alter peptide insertion angle and result in more disruptive insertion.
Scheme 7. Possible Domain Interaction Leading to Enhanced Membrane Activity
2, whereas 1 binding to 20% POPG LUVs under similar
conditions elicits no detectable contents release.36 This indicates
that peptide insertion driven by specific lipid headgroup
recognition is much more disruptive than insertion of an
amphipathic peptide into the membrane. Clearly, the physical
parameters of insertion are different. While electrostatically
driven binding results in roughly uniform insertion of the
hydrophobic face of helical magainin, specific binding interac-
tion with a lipid headgroup dictates that the binding terminus
must remain out of the membrane while the other end is inserted
(Figure 6). The latter mode would involve only partial insertion
which could make it more difficult for the lipid matrix to
compensate for lipid packing defects. This is similar in concept
to the greater membrane activity of obliquely inserting peptides
versus transmembrane inserted peptides. Transmembrane inser-
tion merely compresses lipid packing, whereas oblique or partial
insertion frustrates packing in the outer lipid monolayer only.
This study supports the notion that specific lipid recognition37
can dramatically alter the functional and physical consequences
of peptide-membrane insertion by controlling peptide insertion;
presumably, insertion angle and depth are important parameters.
This finding lays the groundwork for future designs of mem-
brane activators, including selective fusogenic and pore-forming
systems.
Effect of Lipid Composition on Contents Leakage, Domain
Formation, and Fusion Rate. One function of cholesterol in
biomembranes is to fill in packing defects and stabilize
unsaturated lipid membranes.38 We hypothesized that peptide
insertion was causing lipid matrix packing defects that resulted
in permeability increases, and thus we incorporated cholesterol
in reacting membranes in an effort to decrease leakage during
fusion. Using the Cbf release assay as above, we discovered
that cholesterol additives resulted in a minor decrease in leakage,
but also slowed the fusion reaction. The latter does make sense,
as the more rigid the membrane, the less reactive it should be,
although the reason why cholesterol does not significantly
decrease leakage is unclear. However, when dipalmitoyl phos-
phatidyl choline (DPPC) was added to the egg PC membrane,
leakage decreased and fusion rates increased, in conjunction
with increasing DPPC content (Figure 7). We interpret these
results in the following way: the increased DPPC content causes
the membrane to rigidify and thus decrease leakage, while
increased fusion rate reflects domain formation in the membrane.
It has been reported that mixtures of saturated and unsaturated
lipids such as DPPC and POPC or POPG will phase separate,39
resulting in separate fluid and gel-phase membrane domains.
(33) Hessa, T.; Kim, H.; Bihlmaier, K.; Lundin, C.; Boekel, J.; Andersson,
H.; Nilsson, I.; White, S. H.; von Heijne, G. Nature 2005, 433, 377–
381.
(34) New, R. R. C., Liposomes: A Practical Approach; Oxford: NY, 1990.
(35) Blumenthal, R.; Weinstein, J. N.; Sharrow, S. O.; Henkart, P. Proc.
Natl. Acad. Sci. U.S.A. 1977, 74, 5603–5607.
(36) (a) Previous experiments using an ANTS/DPX vesicle contents-mixing
assay had indicated nonleaky fusion in our system, but a simple
carboxyfluorscein (Cbf) dequenching assay indicated that contents
release was rapid from both reacting vesicles. This may indicate that
both leakage and transfer occur during fusion. In the ANTS/DPX assay,
ANTS leakage results in an increase in fluorescence through dilution
dequenching while transfer results in quenching by DPX (decrease).
We found a nominal fluorescence decrease upon fusion, possibly a
sum of two effects. (b) Ellens, H.; Bentz, J.; Szoka, F. C. Biochemistry
1985, 24, 3099–3106.
(38) Silvius, J. R. Biochemistry 1992, 31, 3398–408.
(39) (a) Beattie, M. E.; Veatch, S. L.; Stottrup, B. L.; Keller, S. L. Biophys.
J. 2005, 89, 1760–1768. (b) Gopal, A.; Lee, K. Y. C. J. Phys. Chem.
B 2001, 105, 10348–10354. (c) Bringezu, F.; Ding, J.; Brezesinski,
G.; Zasadzinski, J. A. Langmuir 2001, 17, 4641–4648.
(37) Chatterjee, C.; Paul, M.; Xie, L.; Van der Donk, W. A. Chem. ReV.
2005, 105, 633–683.
9
J. AM. CHEM. SOC. VOL. 130, NO. 19, 2008 6203

A R T I C L E S
Gong et al.
reactivity. These findings indicate that membrane activation for
fusion depends strongly on the way the fusogen anchors into
the lipid matrix. “Single-point” attachment via specific lipid
recognition by a peptide results in much more disruptive binding
than “multipoint,” as when the entire helical face is buried by
electrostatically driven binding. This suggests that insertion
depth and angle of peptide helices can determine membrane
activity and further indicates that specific lipid recognition may
generally enhance the membrane activity of all antimicrobial
peptides. The presence of gel-phase lipids was found to
significantly increase membrane fusion rate and decrease
leakage, which provides support for the notion that membrane
subdomains formed by lipid mismatch may serve to cluster lipids
and fusogens in the bilayer, enhancing binding and function.
While the functional role of such lipid rafts in biology remains
contentious,40 their physical existence in synthetic membranes
is well established,41 and this study suggests how lipid subdo-
mains may play a functional role in a synthetic fusion system.
Examination of this synthetic model system has yielded insights
into both fusion and permeation processes, which are general
for the lipidic phase of membrane function, and thus has
relevance for these same processes in biological membranes.
Experimental Section
General Methods and Instrumentation. All synthetic trans-
formations were carried out at room temperature and all measure-
ments and fusion assays were thermostatted at 25 °C. Stock
solutions of fusogens 1 and 2 were prepared and concentrations
determined using acetamidobenzamide absorbance (ε270 ) 18,000
M-1cm-1). Lipid fusogen 2 was dispersed in lipid films at 1.0–2.0
mol % while peptide fusogen 2 was added to preformed LUVs at
1 mol % concentration. Lipid film hydration was used to prepare
a polydisperse population of multilamellar vesicles that were
extruded through a 100-nm polycarbonate filter to obtain large
unilamellar vesicles34 (LUVs) with average diameter of 150 nm
as judged by light-scattering measurements. LUVs were prepared
using mixtures of neutral egg phosphocholine lipids (egg PC) and
negatively charged phosphoglycerol (POPG) lipids. Fluorescence
measurements were performed in a Perkin-Elmer LS-50B using a
3-mL cuvette or in 96-well plate format with a Spectramax M5
plate reader (Molecular Devices). Circular dichroism measurements
were made with an Aviv CD spectrometer. Titration calorimetry
experiments were run using a VP-ITC microcalorimeter. Peptides
were synthesized on an Advanced Chemtech (Apex 396) automated
peptide synthesizer using standard Fmoc chemistry and purified to
homogeneity on reverse phase HPLC. Purified peptides and
vancomycin derivatives were identified by ESI-MS or MALDI-
MS, as were purified lipid derivatives, prepared as previously
described.
Materials. Fluorescent lipids NBD-PE and Rh-DHPE were
purchased from Invitrogen and used as provided by the manufac-
turer. Phospholipids POPG, DPPC, and egg PC were purchased
from Avanti Polar Lipids. Amino acid derivatives were purchased
from Advanced Chemtech and EMD Biosciences, vancomycin
hydrochloride and other fine chemicals were purchased from Aldrich
and used as provided.
Synthesis of Vancomycin-Cysteinamide 3. Vancomycin hy-
drochloride hydrate (75 mg, 50 µmol) was dissolved in 1 mL of
DMSO and activated for 10 min with HATU (1 equiv, 19 mg) and
DIEA (2 equiv, 13 mg) and then treated with cystine dimethyl ester
dihydrochloride (4 equiv, 68 mg) and DIEA (8 equiv, 52 mg). The
Figure 7. Fusogenic LUVs are mixed at t ) 0. Donor LUVs preincubated
with magainin-vancomycin conjugate and all have 20% POPG and vary
in ePC/DPPC ratio: O ) 80% ePC; O (red) ) 60% ePC, 20% DPPC; O
(green) ) 40% ePC, 40% DPPC; O (blue) ) 20% ePC, 60% DPPC. (A)
Fusion rate: acceptor LUVs composed of 2% Kaa lipid, 1.5% NBD-PE,
1.5% Rh-lipid, 40% DPPC in ePC. (B) Dye release: acceptor LUVs
composed of ePC/2% Kaa encapsulating Cbf at self-quenching concentration.
Cholesterol has the opposite effect, causing a mixing of domains
that results in one liquid crystalline domain that is both strong
and fluid, which is critical for biomembranes.38 These biophysi-
cal properties of DPPC and cholesterol-containing membranes
may be the underlying cause of the observed chemistry.
Domain formation in a POPG/DPPC/eggPC membrane would
segregate the fluid (POPG, egg PC) from the gel (DPPC) and
also would segregate the negative charge (POPG) from the gel.
Therefore, binding of the magainin-vancomycin fusogen would
occur on the negatively charged fluid phase island only (Scheme
7). The POPG domain would have higher charge density, and
the membrane as a whole would have peptide binding directed
to a surface subdomain instead of being uniformly distributed.
This would increase both the density of ligands and degree of
activation, thereby increasing the avidity and productivity of
initial membrane apposition, resulting in faster fusion.
(40) (a) Cottingham, K. Anal. Chem. 2004, 76, 403A–406A. (b) Munro,
S. Cell 2003, 115, 377–388.
Conclusions
(41) (a) Binder, W. H.; Barragan, V.; Menger, F. M. Angew. Chem., Int.
Ed. 2003, 42, 5802–5827. (b) Kahya, N.; Scherfeld, D.; Bacia, K.;
Schwille, P. J. Struct. Biol. 2004, 147, 77–89.
We have presented herein studies which identify the aspects
of our designed membrane fusion system that are critical for
9
6204 J. AM. CHEM. SOC. VOL. 130, NO. 19, 2008

Determinants of a Synthetic Vesicle Fusion System
A R T I C L E S
mixture was stirred for 6 h and then diluted with water and
neutralized with 1 M HCl and purified on a C12 semiprep HPLC
column using a gradient of 15–50% solvent B/solvent A over 40
min (solvent A ) 1% acetonitrile, 0.01% TFA in water, solvent B
) 90% acetonitrile, 0.007% TFA in water). The purified glyco-
peptide was lyophilized to give the cystine adduct in 66% yield
based on vancomycin (MALDI-MS, calculated: 1699, found: 1722
(M + Na+). Vancomycin-cystinedimethyl ester (17 mg, 10 µmol)
was dissolved in water (1 mL), and lithium hydroxide (10 equiv,
2.4 mg) was added. The mixture was stirred for 30 min after
nitrogen sparge then neutralized by 1 M HCl. Tris-carboxyethyl
phosphine (6 mg, 2 equiv) was added, and pH was adjusted to be
7–8 with NaHCO3. After stirring for 10 min, the reaction mixture
was purified on a C-12 semiprep column with the same gradient
used for the cystine adduct. Lyophilization yielded compound 3 in
85% yield. Molecular weight was verified by MALDI (calculated:
1551, found: 1574 (M + Na)).
Addition of Kaa Thiol to POPE Lipid Anchor. S-Trityl and
N-Boc protected peptide (TrtS(CH2)2CO2-Lys(Boc)-D-Ala-D-Ala-
OH, 28 mg, 38 µmol) was dissolved in a minimum amount of TFA
(200 µL), yielding a bright-yellow solution which was diluted with
2 mL of ethyl ether after 2 min. The suspension was centrifuged
to obtain a white pellet. The supernatant was removed, and the
pellet was resuspended in fresh ether and spun down twice more
to remove TFA. The white precipitate was dissolved in degassed
methanol and added directly to a methanol solution of POPE anchor
6 (30 mg, 25 µmol) with 6 equiv DIEA under argon. Reaction was
followed by HPLC and purified to homogeneity on a C4 column
with a gradient of 70% solvent B to 100% solvent B in solvent A
over 40 min (solvent A ) 4.99% methanol, 95% H2O, 0.1% formic
acid; solvent B ) 4.99% H2O, 45% methanol, 50% isopropanol,
0.1% formic acid). Product mass was confirmed by electrospray
mass spectroscopy (calculated: 1539, found: 1540.2).
Membrane Binding Experiments. All peptide-lipid titrations
were performed using 10 mM Tris buffer, pH 7.4, 100 mM NaCl.
Changes in signal upon interaction with lipid (ellipticity, heat) were
normalized by the maximum change in signal for each experiment
after background subtraction. The resulting normalized fractional
change in signal at each step of the titration was taken as fractional
binding. Each addition of titrant was followed by an equilibration
time; for ITC this was taken as the time required for heat flow to
return to baseline, and for CD equilibration time was 3 min.
Fusion and Release Experiments. All experiments were run
in 10 mM Tris buffer, pH 7.4, 100 mM NaCl. A 5:1 ratio of POPG
LUVs to egg PC LUVs containing NBD-PE and Rh-PE at 1.5 mole
percent was used unless otherwise noted. Total lipid concentrations
were typically to 312.5 µM (POPG LUVs) to 62.5 µM (ePC LUVs).
In the charge dependence study (Figure 5C), a 9:1 ratio of POPG
LUVs (562.5 µM) to ePC LUVs (62.5 µM) was used. Studies of
the effects of DPPC on dye release during fusion (Figure 7B) were
performed at a 1:1 LUV ratio (62.5 µM each).
Acknowledgment. We thank The Ohio State University for
start-up funds to support this research.
JA711184U
9
J. AM. CHEM. SOC. VOL. 130, NO. 19, 2008 6205