Large molecular assembly of amphotericin B formed in ergosterol-containing membrane evidenced by solid-state NMR of intramolecular bridged derivative

Article abstract of DOI:10.1021/ja063433w

Amphotericin B (AmB 1) is known to assemble and form an ion channel across biomembranes. We have recently reported that conformation-restricted derivatives of AmB 2-4 show different ergosterol preferences in ion-channel assays, which suggested that the orientation of the mycosamine strongly affects the sterol selectivity of AmB. The data allowed us to assume that compound 3 showing the highest selectivity would reflect the active conformation of AmB in the channel assembly. In this study, to gain further insight into the active conformation of AmB, we prepared a new intramolecular-bridged derivative 5, where the linker encompassed a hydrophilic glycine moiety. The derivative has almost equivalent ion-channel activity to those of AmB and 3. The antifungal activity of 5 compared with 3 improves significantly, possibly because the increasing hydrophilicity in the linker enhances the penetrability through the fungal cell wall. Conformation of 5 was well converged and very similar to that of 3, thus further supporting the notion that the conformations of these derivatives reproduce the active structure of AmB in the channel complex. Then we used the derivative to probe the mobility of AmB in the membrane by solid-state NMR. To measure dipolar couplings and chemical shift anisotropies, we incorporated [1-¹³C,¹⁵N]glycine into the linker. The results indicate that 5 is mostly immobilized in ergosterol-containing DMPC bilayers, implying formation of large aggregates of 5. Meanwhile some fraction of 5 remains mobile in sterol-free DMPC bilayers, suggesting promotion of Amb aggregation by ergosterol.

Full text of DOI:10.1021/ja063433w

Published on Web 08/17/2006
Large Molecular Assembly of Amphotericin B Formed in
Ergosterol-Containing Membrane Evidenced by Solid-State
NMR of Intramolecular Bridged Derivative
Nobuaki Matsumori,* Yuri Sawada, and Michio Murata*
Contribution from the Department of Chemistry, Graduate School of Science, Osaka UniVersity,
Toyonaka, Osaka 560-0043, Japan
Received May 17, 2006; E-mail: murata@ch.wani.osaka-u.ac.jp; matumori@ch.wani.osaka-u.ac.jp
Abstract: Amphotericin B (AmB 1) is known to assemble and form an ion channel across biomembranes.
We have recently reported that conformation-restricted derivatives of AmB 2-4 show different ergosterol
preferences in ion-channel assays, which suggested that the orientation of the mycosamine strongly affects
the sterol selectivity of AmB. The data allowed us to assume that compound 3 showing the highest selectivity
would reflect the active conformation of AmB in the channel assembly. In this study, to gain further insight
into the active conformation of AmB, we prepared a new intramolecular-bridged derivative 5, where the
linker encompassed a hydrophilic glycine moiety. The derivative has almost equivalent ion-channel activity
to those of AmB and 3. The antifungal activity of 5 compared with 3 improves significantly, possibly because
the increasing hydrophilicity in the linker enhances the penetrability through the fungal cell wall. Conformation
of 5 was well converged and very similar to that of 3, thus further supporting the notion that the conformations
of these derivatives reproduce the active structure of AmB in the channel complex. Then we used the
derivative to probe the mobility of AmB in the membrane by solid-state NMR. To measure dipolar couplings
and chemical shift anisotropies, we incorporated [1-13C,15N]glycine into the linker. The results indicate that
5 is mostly immobilized in ergosterol-containing DMPC bilayers, implying formation of large aggregates of
5. Meanwhile some fraction of 5 remains mobile in sterol-free DMPC bilayers, suggesting promotion of
AmB aggregation by ergosterol.
Introduction
branes are more sensitive to AmB than cholesterol-containing
mammalian membranes (Chart 1).8,10,11
For over 40 years, amphotericin B (AmB, 1) has been a
standard drug for treatment of deep-seated systemic fungal
infections.1-3 Due to lack of better alternatives as well as the
scarce occurrence of resistant strains,4 the clinical importance
of AmB has been unchanged. It is now widely accepted that
AmB in a phospholipid membrane associates with sterols to
form barrel-stave type pores, with their polyhydroxy side
pointing inward and their lipophilic heptaene part directing
outward.5,6 The occurrences of such molecular assemblies in a
fungal cell membrane increases ion permeability and alters
membrane potentials, ultimately leading to the cell death.6-9
The pharmacological action of AmB is based on its selective
toxicity against eukaryotic microbes over mammalians. It has
been repeatedly shown that ergosterol-containing fungal mem-
The structure of AmB comprises two rigid fragments, a
macrolide ring and a mycosamine sugar moiety, which are
linked by a rotatable â-glycosidic bond. Their relative position
is thought to control AmB-AmB and AmB-sterol interactions,
where inter- and intramolecular hydrogen bonds play crucial
roles.12,13 To reveal the relationship between the orientation of
aminosugar and activities, we have recently prepared intramo-
lecular bridged derivatives of AmB 2-4 (Chart 2), where
carboxyl and amino groups are bridged with different lengths
of alkyl chains.14 Consequently, 2-4 showed different ergosterol
preferences in membrane-permeabilizing activity, probably due
to the alteration of the amino-sugar orientation induced by the
alkyl bridge. In particular, the ergosterol selectivity of 3, of
which amino-sugar orientation is immobilozed by the bridge,
markedly exceeds that of AmB. These results allowed us to
assume the conformation of 3 somewhat reflects the active
structure of AmB in the channel complex. The bridged deriva-
(1) Gallis, H. A.; Drew, R. H.; Pickard, W. W. ReV. Infect. Dis. 1990, 12,
308-329.
(2) Hartsel, S. C.; Bolard, J. Trends Pharmacol. Sci. 1996, 12, 445-449.
(3) Ablordeppey, S. Y.; Fan, P. C.; Ablordeppey, J. H.; Mardenborough, L.
Curr. Med. Chem. 1999, 6, 1151-1195.
(4) Vanden-Bossche, H.; Dromer, F.; Improvisi, I.; LozanoChiu, M.; Rex, J.
H.; Sanglard, D. Med. Mycol. 1998, 36, 119-128.
(10) Milhaud, J.; Hartman, M. A.; Bolard, J. Biochimie 1989, 71, 49-56.
(11) Vertut-Qroquin, A.; Bolard, J.; Gary-Bobo, C. M. Biochem. Biophys. Res.
Commun. 1984, 125, 360-366.
(5) Andreoli, T. E. Ann. N.Y. Acad. Sci. 1974, 235, 448-468.
(6) De-Kruijff, B.; Demel, R. A. Biochim. Biophys. Acta 1974, 339, 57-70.
(7) Cohen, B. E. Biochim. Biophys. Acta 1986, 857, 117-122.
(8) Vertut-Croquin, A.; Bolard, J.; Chabbert, M.; Gary-Bobo, C. Biochemistry
1983, 22, 2939-2944.
(12) Baginski, M.; Gariboldi, P.; Bruni, P.; Borowski, E. Biophys. Chem. 1997,
65, 91-100.
(13) Baginski, M.; Resat, H.; Borowski, E. Biochim. Biophys. Acta 2002, 1567,
63-78.
(9) Van-Hoogevest, P.; De-Kruijff, B. Biochim. Biophys. Acta 1978, 511, 397-
407.
(14) Matsumori, N.; Sawada, Y.; Murata, M. J. Am. Chem. Soc. 2005, 127,
10667-10675.
9
10.1021/ja063433w CCC: $33.50 © 2006 American Chemical Society
J. AM. CHEM. SOC. 2006, 128, 11977-11984
11977

A R T I C L E S
Matsumori et al.
Chart 1. Structures of AmB 1, Ergosterol, and Cholesterol
the case with 3, derivative 5 was supposed to adopt a rigid
conformation due to the short bridge. The synthetic route of
derivative 5 is represented in Scheme 1. Fmoc-glycine was
coupled with aminopropanol to give 7, which was then
converted to aldehyde 8 by Dess-Martin oxidation. The
aldehyde was coupled with the amino group of AmB by the
reductive amino-alkylation reaction with use of NaBH3CN.
Deprotection of the Fmoc group followed by intramolecular
amide formation in the presence of PyBOP as a condensation
reagent afforded bridged derivative 5. [1-13C,15N]glycine was
used for synthesis of double-labeled 5′.
K+ Flux and Biological Activities. Then we assessed the
K+ flux activity of these derivatives.15,16 The liposomes used
in this assay have transmembrane gradients of pH (pH 5.5 inside
and pH 7.5 outside) as well as K+ concentration (higher on the
outside). Once ion channels form, K+ influx into the liposomes
induces H+/K+ exchange in the presence of proton-transporter
FCCP, leading to a rise in inner pH. This pH change was
monitored by chemical shift changes of phosphate anions in
31P NMR. Details on the spectral features of 31P NMR observed
in this assay are described in our recent publications.17-22 The
molar ratio of drug/lipid at which 50% of the liposomes were
permeabilized was determined as EC50. Table 1 lists the EC50
values of AmB, 3 and 5 in sterol-free, cholesterol-containing
and ergosterol-containing liposomes. As is clear from the table,
compound 5 shows comparable or even higher K+-influx activity
than those of AmB and 3. In addition, the effect of cholesterol
on the activity is only marginal for 5 as is the case with AmB
and 3, further supporting the idea that cholesterol is not involved
in formation of the channel complex.19,23-26 It is noteworthy
that compound 5 also retains the ergosterol selectivity compa-
rable with AmB, though less selective than 3. Compound 5 has
higher channel activity than the other two compounds in sterol-
free and cholesterol-containing membranes, which partly con-
tributes to the reduction in the ergosterol selectivity of 5
compared with that of 3. Taken together, the hydrophilic bridge
of 5, possibly due to the formation of additional intermolecular
hydrogen bonds, might stabilize the assembly even in the
absence of ergosterol, thus resulting in the elevation of the
channel activity without ergosterol.
Chart 2. Structures of Intramolecular Bridged Derivatives 2-4
and a New One 5
tives including 3, however, are less active toward fungi
compared with AmB, possibly because polysaccharides of the
fungal cell wall, which form a strong diffusion barrier against
hydrophobic molecules, reduce the penetrability through the cell
wall due to the hydrophobic bridge part of the derivatives.
In this report, to gain further insight into the active conforma-
tion of AmB in the channel complex, we designed and prepared
a new derivative 5, to which a glycine-containing linker was
incorporated instead of an alkyl chain to enhance the hydro-
philicity of the bridge portion. We examined the conformation
and activities of the derivative, which further supports our idea
on the active conformation of AmB. We then investigated the
dynamics of membrane-bound AmB using 5 in DMPC bilayers
by solid-state NMR techniques. To facilitate the measurements,
[1-13C,15N]glycine was introduced into the bridge. Our data
implied the less mobile feature of the AmB assembly in DMPC
bilayers, which suggested the formation of large aggregates in
membranes.
We then evaluated the membrane permeabilizing activity of
5 in biological systems (Table 1). It has already been reported
that the hemolytic activities (EC50) of AmB and derivative 3
are 4.8 and 2.2 µM, respectively.14 Derivative 5 also revealed
potent hemolytic action at 3.0 µM. It should be noted that the
antifungal activity of 5 is significantly improved in comparison
(15) Herve, M.; Cybulska, B.; Gary-Bobo, C. M. Eur. Biophys. J. 1985, 12,
121-128.
(16) Herve, M.; Debouzy, J. C.; Borowski, E.; Cybulska, B.; Gary-Bobo, C.
M. Biochim. Biophys. Acta 1989, 980, 261-272.
Results and Discussion
(17) Matsumori, N.; Yamaji, N.; Matsuoka, S.; Oishi, T.; Murata, M. J. Am.
Chem. Soc. 2002, 124, 4180-4181.
(18) Yamaji, N.; Matsumori, N.; Matsuoka, S.; Oishi, T.; Murata, M. Org. Lett.
2002, 4, 2087-2089.
Preparation of Intramolecular Bridged Derivative 5. In
the previous report,14 we have prepared three conformationally
constrained derivatives of AmB (2-4), where their carboxyl
and amino groups are intramolecularly linked with C4, C6, and
C8 alkyl chains, respectively. Among them, derivative 3 showed
the most potent ion-channel activity and highest ergosterol
selectivity. In this study we prepared a new derivative 5, in
which a glycine-containing bridge was introduced instead of
alkyl chains to enhance the hydrophilicity. The chain length of
derivative 5 was designed to be almost equal to that of 3. As is
(19) Matsuoka, S.; Murata, M. Biochim. Biophys. Acta 2002, 1564, 429-434.
(20) Matsuoka, S.; Murata, M. Biochim. Biophys. Acta 2003, 1617, 109-115.
(21) Matsuoka, S.; Matsumori, N.; Murata, M. Org. Biomol. Chem. 2003, 1,
3882-3884.
(22) Matsumori, N.; Eiraku, N.; Matsuoka, S.; Oishi, T.; Murata, M.; Aoki, T.;
Ide, T. Chem. Biol. 2004, 11, 673-679.
(23) Cotero, B. V.; Rebolledo-AntUnez, S.; Ortega-Blake, I. Biochim. Biophys.
Acta 1998, 1375, 43-51.
(24) Dufourc, E. J.; Smith, I. C. P.; Jarrell, H. C. Biochim. Biophys. Acta 1984,
776, 317-329.
(25) Ruckwardt, T.; Scott, A.; Scott, J.; Mikulecky, P.; Hartsel, S. C. Biochim.
Biophys. Acta 1998, 1372, 283-288.
(26) Cuntinho, A.; Prieto, M. Biophys. J. 1995, 69, 2541-2557.
9
11978 J. AM. CHEM. SOC. VOL. 128, NO. 36, 2006

Large Molecular Assembly of Amphotericin B
A R T I C L E S
Scheme 1. Preparation of 5 and 5′a
a Reagents: (a) Fmoc-OSu, pyridine, MeOH; (b) aminopropanol, PyBOP, HOBt, N-ethyldiisopropylamine, DMF; (c) Dess-Martin reagent, CH2Cl2; (d)
AmB, NaBH3CN, DMF-MeOH; (e) piperidine, DMF-MeOH; (f) PyBOP, HOBt, N-ethyldiisopropylamine, DMF (for product yield in each step, see
Experimental Section).
Table 1. K+ Influx Activity Using Liposomes and Biological Activities of AmB, 3, and 5
K+ influx activity EC50 (sample/lipid)
×
106
hemolytic
activitya
sterol-free
liposomes
cholesterol-containing
liposomes
ergosterol-containing
liposomes
antifungal
activity,b
µg
compound
EC50
,
µM
1 (AmB)
3
5
47
60
30
43
58
27
11
5.8
9.5
4.8
2.2
3.0
5
20
10
a Against 1% human erythrocytes. b The minimal amount of samples on a paper disk that shows inhibitory zone on the culture of Aspergillus niger.
with 3 (Table 1). In the previous paper we postulated that
hydrophilic polysaccharides of fungal cell walls form a stronger
diffusion barrier against 3, which is more hydrophobic than
AmB. Accordingly, the improvement of antifungal activity in
5 may be accounted for by the enhanced hydrophilicity in the
bridge portion.
conformation occurring in the channel complex of AmB. To
gain further insight into the geometry of the amino-sugar in
AmB, we performed conformational analysis of 5 in a similar
way. The chemical shifts in the macrolide portion of 5 agreed
well with those of AmB within 0.1 ppm except for the C16-
C21 and the amino-sugar portion where the change in space
arrangement of the sugar is likely to influence the chemical shifts
(Supporting Information). Spin-coupling patterns of the mac-
rolide rings were also similar between AmB and 5 (data not
shown). These observations indicate that the macrolide confor-
mation of 5 is virtually identical with that of AmB. Due to the
possible inflexibility of the conjugated double bonds, the
macrolide ring of 5 was treated as a motion-restricted part, in
which a (30° allowance from the crystal structure28 was given
to each C-C single bond upon calculation. Distance constraints
for 5 were derived from 2D NOESY spectra in DMSO-d6. The
total number of interproton distances derived from the NMR
data was 22. On the basis of these conformational constraints,
molecular modeling calculations were performed using the
Monte Carlo conformation search algorithm.29 In comparison
with 2-4, derivative 5 has additional amide and glycine protons,
which yielded well-resolved cross-peaks in NOESY spectra and,
therefore, allowed us to extract more distance constraints around
the bridged portion. Figure 1b shows superposition of 20 lowest
energy conformations around the sugar portion of 5. As in the
case of 3, derivative 5 has well converged conformers due to
the constraints of the short bridge, thus implying the rigid
Alternatively, it might be also possible to explain the
improved antifungal activity of 5 by formation of water-soluble
oligomers in aqueous conditions. It was reported that AmB
adopts three forms in water, monomers, water-soluble oligomers,
and nonsoluble aggregates, and that monomeric and water-
soluble oligomeric species are biologically active.27 To evaluate
the proportion of the three forms in water, we recorded the UV
and CD spectra of 5 µM AmB, 3, and 5 in PBS buffer (see
Supporting Information). Consequently, 3 mostly formed non-
soluble aggregates in water probably due to the low hydrophi-
licity. On the other hand, AmB and 5 formed biologically active
water-soluble oligomers in buffer to a similar extent, which may
be also attributed to the comparable antifungal efficacy between
AmB and 5. In both cases, the hydrophilicity must be an
important index for designing new AmB derivatives in terms
of amelioration of the antifungal activity.
Conformation of Intramolecular Bridged Derivative 5. We
have previously determined the conformation of 3 based on
distance and dihedral constraints obtained in a DMSO solution14
and found that the orientation of the amino sugar is restricted
due to the short bridge (Figure 1a). Since the ion channel
property of 3 is similar yet superior to that of AmB itself, we
assumed that the conformation of 3 reproduces the active
(28) Ganis, P.; Avitabile, G.; Mechlinski, W.; Schaffner, C. P. J. Am. Chem.
Soc. 1971, 93, 4560-4564.
(29) Mohamadi, F.; Richards, N. G.; Guida, W. C.; Liskamp, R.; Lipton, M.;
Canfield, C.; Chang, G.; Hendrickson, T.; Still, W. C. J. Comput. Chem.
1990, 11, 440-467.
(27) Legrand, P.; Romero, E. A.; Cohen, B. E.; Bolard, J. Antimicrob. Agents
Chemother. 1992, 36, 2518-2522.
9
J. AM. CHEM. SOC. VOL. 128, NO. 36, 2006 11979

A R T I C L E S
Matsumori et al.
Figure 2. Experimental 13C-15N REDOR dephasing values (∆S/S0) of 5′
in powder form (9), in membranes of ergosterol-containing DMPC (2),
and sterol-free DMPC ([). Dark solid, light solid, and dotted lines are the
best-fit curves for the square points, triangles, and diamonds, respectively.
amplitude of motion and, in favorable cases, the orientation of
the internuclear vector relative to the motional axis. Accordingly,
we measured 13C-15N dipolar couplings of 5′ in DMPC
membranes using REDOR experiments.29 Figure 2 shows the
13C-15N REDOR dephasings of 5′ in the powder form and in
DMPC membranes in the presence and absence of ergosterol
(5′/ergosterol/DMPC 1:1:9 or 1:0:10). 13C-15N dipolar coupling
values obtained from the best-fit dephasing curves are also
depicted in Figure 2. The rigid-limit C-N dipolar coupling of
5′ is calculated to be 198 Hz on the basis of the C-N
internuclear distance 2.49 Å. Bond vibration and imperfection
of π pulses in REDOR measurements may scale down the
dipolar coupling to 175 Hz even in the powder form.31,32
Surprisingly, the dipolar coupling of 5′ in the ergosterol-
containing DMPC membrane (170 Hz) is almost the same as
that in the powder form, while that in the sterol-free DMPC
membrane is significantly attenuated to be 145 Hz. These dipolar
couplings indicate that 5′ is almost immobilized in the ergosterol-
containing DMPC membrane, while a small amount of mobility
remains in the sterol-free DMPC membrane. The attenuated
dipolar coupling of 5′ in the sterol-free DMPC membrane can
be interpreted in two ways. One possibility is that both mobile
and immobile molecules coexist in the membrane, and the
mobile molecules contribute to the reduction of the dipolar
coupling. The other possibility is that all of 5′ molecules undergo
whole-body fast rotation in the sterol-free DMPC membrane.
If the rotation is uniaxial, the order parameter is represented by
(1 - 3 cos2 θ)/2, which yields the θ angle of ca. 20° between
the rotation axis and the C-N direction.
Figure 1. Ensembles of 20 lowest energy conformations for 3 (a) and 5
(b), generated by Monte Carlo algorithm restrained on the basis of distance
constrains from 2D-NOESY in DMSO-d6. Models are represented around
the polar head regions of the derivatives, and 20 lowest energy conformers
are superposed for a minimum rmsd. The æ (C18-C19-O-C1′) and ψ
(C19-O-C1′-C2′) angles of the most stable conformer for derivative 5
are determined to be -93.2° and 149.7°, respectively, which are almost
identical with those of 3 (-89.3° and 153.8°).
conformation. More importantly, the conformation of 5 agrees
well with what was reported for 314 (Figure 1). As described
above, we have reported that the conformation of 3 probably
reflects the active conformer of AmB in the channel assembly.14
Considering that the ion-channel and biological activities of 5
are comparable to those of AmB, the rigid conformation of 5
is also thought to reproduce the active structure of AmB in the
channel complex. The consistency of the conformations of 3
and 5 strongly supports this notion.
Dynamics of Intramolecular Bridged Derivative 5′ in
DMPC Liposomes. With a better derivative of AmB in hand,
we applied solid-state NMR techniques to obtain information
on the dynamics of AmB in the membrane. The motional
behavior of the derivative can provide information on the
molecular assembly formed in the membrane. In this study we
measured dipolar couplings and chemical shift anisotropies by
solid-state NMR to investigate the molecular motion of 5 in
membrane environments. For this purpose, we prepared doubly
isotope-labeled 5′ using [1-13C,15N]glycine (Scheme 1). Dipolar
couplings depend both on the internuclear distance and on
molecular motion. For a fixed internuclear distance such as 13C-
15N in 5′, reduction in the coupling strength from its rigid lattice
value is caused only by motional averaging. The scaling factor
between the measured coupling constant and the rigid-limit value
is the order parameter S, which provides information about the
Further information of the dynamics of 5′ was obtained from
13C chemical shift anisotropies. Changes in chemical shift
tensors also signify molecular motion. Figure 3 illustrates the
13C cross-polarization spectra of 5′ in the powder form and in
DMPC membranes in the presence and absence of ergosterol.
The spectra in DMPC membranes were acquired at slow MAS
of 1 kHz to generate spinning sidebands for sensitivity enhance-
ment. As is clear from Figure 3, the powder patterns expressed
by the sidebands are almost consistent in all spectra, indicating
identical chemical shift tensors even in membranes. This result
(30) Gullion, T.; Schaefer, J. J. Magn. Reson. 1989, 81, 196-200.
(31) Saito, H.; Tuzi, S.; Naito, A. Annu. Rep. NMR Spectrosc. 1998, 36, 79-
121.
(32) Naito, A.; Saito, H. Encycl. Nucl. Magn. Reson. 2002, 9, 283-291.
9
11980 J. AM. CHEM. SOC. VOL. 128, NO. 36, 2006

Large Molecular Assembly of Amphotericin B
A R T I C L E S
Figure 4. DD-MAS 13C spectra of 5′ in ergosterol-containing (a) and sterol-
free (b) DMPC bilayers. The signals at 171 ppm are due to the labeled
glycine moiety of 5′, while those at 173 ppm are from the ester groups of
DMPC bilayers, which are used for internal standards to compare the 13C
signal intensity of 5′.
not shown). The broad signals of 5′ further reveal that
monomeric or oligomeric forms, which were observed in the
aqueous phase as described earlier, hardly occurred under these
NMR conditions.
The above results totally suggested that 5′ is mostly im-
mobilized in ergosterol-containing DMPC membranes while a
considerable amount of movable molecules coexist in the sterol-
free DMPC membrane. Some membrane peptides are known
to form aggregates in the membrane and consequently show
low mobility.33 Therefore, the current results can also be
accounted for by the aggregate formation of 5′ in DMPC
membranes. Recently we have shown that the AmB molecule
can span the DMPC bilayer with a single molecular length by
measuring dipolar interactions between 13C labeled AmB and
31P of a phospholipid using the same membrane preparation.34
The relatively large 13C-31P dipolar interactions also indicated
the formation of a tight complex of AmBs and phospholipids.
Since the membrane permeabilizing activities of 5 are quite
similar to those of AmB, the previous results obtained for AmB
would hold true for derivative 5. Taken together, it can be
assumed that derivative 5 and boundary DMPC molecules form
an immobilized complex in the bilayer under the present
conditions. This notion is supported by earlier 2H NMR studies
which also suggest that DMPC is immobilized in the presence
of AmB.35
Here it should be emphasized that 5′ is immobilized on a
time scale of 5 ms or longer, which corresponds to the inverse
of the C-N dipolar interaction of ca. 200 Hz. Although chemical
shift anisotropy and quadrupolar coupling, which are large
interactions over 104 Hz, are frequently used for motional study
of membrane peptides in lipid bilayers, measurements of these
interactions can only provide information about the mobility
of membrane-bound molecules on time scales shorter than the
inverse of the interactions (µs orders). Accordingly, the advan-
tage in weak dipolar couplings used in this study is the detection
capability of the immobilization on millisecond scales.
Based on the rotational correlation time slower than 5 ms,
we can roughly estimate the size of the aggregate of 5′ in the
Figure 3. 13C powder patterns of 5′ in powder form (a), ergosterol-DMPC
(b), and sterol-free DMPC (c). The molar ratios of 5′/ergosterol/DMPC are
1:1:9 for ergosterol-containing liposome and 1:0:10 for sterol-free one.
Spectra b and c obtained in membranes were recorded under MAS at 1
kHz. Sharp signals at 173 ppm in spectra b and c are due to carbonyl carbon
atoms of esters in DMPC.
rules out the possibility of uniaxial fast molecular rotation of
5′ in the sterol-free DMPC membrane and further supports the
immobility of 5′ in the ergosterol-DMPC membrane.
However, whether both mobile and immobile molecules are
present in the sterol-free membrane was not clear from Figure
3. Hence, to gain further evidence on the presence of mobile
molecules in sterol-free DMPC membrane, we performed
dipolar-decoupling (DD)-MAS experiments.31 In DD-MAS
experiments, as opposed to the cross-polarization method,
molecules with high mobility yield stronger signals. The broader
13C signal at 171 ppm of 5′ in sterol-free DMPC was
significantly stronger that that in ergosterol-containing DMPC
(Figure 4), thus supporting the presence of a larger amount of
movable molecules in sterol-free DMPC bilayers. The signal
at 171 ppm in the ergosterol-containing membrane is thought
to be mostly due to immobilized molecules because the powder
form of 5′ also gave the signal in the DD-MAS experiment (data
(33) Buffy, J. J.; Waring, A. J.; Lehrer, R. I.; Hong M. Biochemistry 2003, 42,
13725-13734.
(34) Matsuoka, S.; Ikeuchi, H.; Matsumori, N.; Murata, M. Biochemistry 2005,
44, 704-710.
9
J. AM. CHEM. SOC. VOL. 128, NO. 36, 2006 11981

A R T I C L E S
Matsumori et al.
Figure 5. Models of aggregate formation of AmB derivative 5 in DMPC bilayers in the presence (top) and absence (bottom) of ergosterol. In the presence
of ergosterol, 5 aggregates with DMPC and possibly with ergosterol to form immobilized domains larger than 20 nm in diameter. Immobilized aggregates
of 5 are also formed even in the absence of ergosterol, although some fraction of 5 probably remains mobilized.
membrane since the rotational diffusion rate is inversely
proportional to the aggregate size. According to the two-
dimensional hydrodynamic model,36 a cylindrical particle with
a radius of a and thickness of h has a rotational diffusion
coefficient DR as follows:
hand, in the sterol-free DMPC membrane, although a large part
of the AmB molecules form immobilized aggregates with
phospholipids, mobile molecules, which are possibly sequestered
from the aggregates, are also present to a considerable extent.
In the aggregates, 5 is thought to traverse the membrane with
a single molecular length, as previously shown in AmB by 13C-
31P REDOR experiments.34 The higher aggregate formation in
the presence of ergosterol may be ascribable to the action of
ergosterol as a glue to stick the AmB derivatives, which further
shifts the equilibrium from monomer (or small oligomer) to
aggregates. Considering that the biological and ion-permeable
activities are almost equivalent between AmB and 5, it may be
possible to extrapolate this model to AmB itself.
Hing et al. have reported that the methyl-d3 amide derivative
of AmB undergoes fast rotation in the DPPC bilayer,43 which
is in contrast to the present data. This is probably due to the
different drug/lipid ratio; they adopted a lower ratio (ca 1 mol
%) compared with the current study (10 mol %). Earlier circular
dichroism (CD) studies by Perkins et al. also suggested that
AmB at less than 3 mol % of DMPC-DMPG was largely
monomeric.44 To observe the molecular assembly of AmB by
NMR, we are obliged to use higher drug/lipid ratios although
the drug exerts its effects at much lower doses, where a small
fraction of AmB molecules are thought to form ion channel
aggregates. Our preliminary results of solid-state NMR experi-
ments using deuterated AmB suggest that immobilized AmB
in large aggregates, which are predominant at 10 mol % AmB
in DMPC, partly exists even at 1 mol % AmB in DMPC (to be
published in due course). Since the complex formation is thought
to be necessary for the ion-channel activity of AmB, the
motionally restricted aggregates formed by bridged derivative
5 may be reasonably related to the molecular assembly of AmB
acting as an ion channel. In that context, these aggregates would
provide relevant information on the ion channel structure of
AmB as wall as intermolecular interactions in the channel
complex. Further structural studies of the aggregates using solid-
state NMR are currently underway.
kBT
4πη a2h
1
τR
DR )
)
where τR is the rotational correlation time, η is the viscosity of
the lipid bilayer, kB is the Boltzmann constant, and T is the
absolute temperature. Since we have shown that the AmB
molecule traverses the DMPC bilayer with a single molecular
length,34 which should be the case with 5, h can be estimated
as ca. 40 Å, corresponding to the thickness of the DMPC bilayer.
A conservative upper-limit estimate of the membrane viscosity
is 10 P.37 Using these values, a motional correlation time τR
longer than 5 ms yields an aggregate radius larger than 20 nm.
The pore size of the AmB ion channel is estimated as 0.6-0.8
nm,38-41 and the external diameter is reported to be around 1.7
nm.41 Therefore, it is unlikely that the radius of a single ion
channel complex exceeds 20 nm, even if boundary phospholipid
molecules are involved in the complex. Rather it is more likely
that ion channel complexes gather and form immobilized
domainlike aggregates with a radius larger than 20 nm. This
speculation is supported by the recent results of AFM observa-
tions of large AmB aggregates.42
Figure 5 summarizes our current model of aggregation of 5
in DMPC membranes. In the ergosterol-DMPC membrane, 5
together with DMPC and ergosterol forms immobilized ag-
gregates, which possess a diameter larger than 20 nm and are
phase-separated from liquid-crystalline bilayers. On the other
(35) Dufourc, E. J.; Smith, I. C. P.; Jarrell, H. C. Biochim. Biophys. Acta 1984,
778, 435-442.
(36) Saffman, P. G.; Delbruck, M. Proc. Natl. Acad. Sci. U.S.A. 1975, 72, 3111-
3113.
(37) Tamm, L. K. Biochim. Biophys. Acta 1991, 1071, 123-148.
(38) Scherrer, R.; Gerhardt, P. J. Bacteriol. 1971, 107, 718-735.
(39) Katsu, T. Biol. Pharm. Bull. 1999, 22, 978-980.
(40) Houdai, T.; Matsuoka, S.; Matsumori, N.; Murata, M. Biochim. Biophys.
Acta 2004, 1667, 91-100.
(43) Hing, A. W.; Schaefer, J.; Kobayashi, G. S. Biochim. Biophys. Acta 2000,
1463, 323-332.
(44) Perkins, W. R.; Minchey, S. R.; Boni, L. T.; Swenson, C. E.; Popescu, M.
C.; Pasternack, R. F.; Janoff, A. S. Biochim. Biophys. Acta 1992, 1107,
271-282.
(41) Gruszecki, W. I.; Gagos, M.; Kernen, P. FEBS Lett. 2002, 524, 92-96.
(42) Milhaud, J.; Ponsinet, V.; Takahashi, M.; Mchels, B. Biochim. Biophys.
Acta 2002, 1558, 95-108.
9
11982 J. AM. CHEM. SOC. VOL. 128, NO. 36, 2006

Large Molecular Assembly of Amphotericin B
A R T I C L E S
N-{N-(9-Fluorenylmethoxycarbonyl)-glycyl-3-aminopropyl}-
AmB (8). A solution of 7 (13 mg, 36 µmol) and AmB (50 mg, 54
µmol) in DMF-MeOH (4:3, 7 mL) was stirred for 3 h, and then NaBH3-
CN (14 mg, 216 µmol) was added to the solution. After being stirred
overnight, the solution was poured into diethyl ether (100 mL) to form
a yellow precipitate, which was filtered over Celite and washed with
diethyl ether. The precipitate and the Celite were extracted with CHCl3-
MeOH-H2O 10:6:1 and purified by column chromatography on SiO2
with the same solvent system to afford 8 (19 mg, 43%) as a yellow
solid. Rf 0.63 (CHCl3-MeOH-H2O ) 10:6:1. ESI-MS m/z 1260.4 (M
+ H)+.
Preparation of Derivatives 5 and 5′. To a solution of 8 (19 mg,
15 µmol) in DMF-MeOH (1:1, 4 mL) was added piperidine (500 µL,
5 mmol). After the solution stirred for 30 min, diethyl ether was added
to form a yellow precipitate. The precipitate was filtered over Celite
and washed with diethyl ether. The product was extracted with CHCl3-
MeOH-H2O to give crude 9 (18 mg), which was used for the next
reaction without further purification because of instability. To a solution
of crude 9 (18 mg) in DMF (10 mL) were added diisopropylethylamine
(6 µL, 35 µmol), 1-hydroxybenzotriazole (3.3 mg, 21 µmol), and
PyBOP (8.8 mg, 17 mmol) sequentially. After the mixture stirred for
20 h, it was poured into diethyl ether to form precipitates. The
precipitate was filtered over Celite, washed with diethyl ether, and
extracted with CHCl3-MeOH-H2O. The extract was purified by HPLC
to afford 2 (1.7 mg, 11% for two steps). HPLC conditions: column,
COSMOSIL 5C18-AR-II Φ 4.6 mm × 150 mm; flow rate, 0.5 mL/
min; mobile phase, MeOH-5 mM ammonium acetate (pH 5.3)
changing linearly from 70:30 to 100:0 for 30 min; retention time, 20.0
min. ESI-MS m/z 1020.5 (M + H)+. 1H NMR (500 MHz, DMSO-d6),
see Supporting Information. 5′ was prepared in the same way as 5 by
using [1-13C,15N]glycine as a starting material.
Conclusion
We have designed and prepared a new intramolecular bridged
derivative 5 and showed that 5 excellently retained membrane
permeabilizing activity and selective toxicity characteristic of
AmB. The conformation of 5, restricted with a short-chain
bridge, was similar to that of derivative 3 which has the highest
ergosterol selectivity among homologues with different lengths
of bridges. This indicates that the orientation of aminosugars
in 3 and 5 is critical to the activity of AmB, thus further
confirming the idea that the conformation of the bridged
derivatives reflects the active structure of AmB in the channel
complex.
Isotope-labeled derivative 5′ was then subjected to solid-state
NMR measurements in DMPC membrane preparations, which
demonstrated that 5 forms immobilized aggregates in the
membranes. It was also observed that the aggregate formation
is promoted in the presence of ergosterol, which might be related
to the higher activity of AmB in ergosterol-containing mem-
branes. Because derivative 5 has biological activity and ion-
channel properties similar to those of AmB, it is assumed that
motional properties of 5 somewhat reflect the behaviors of AmB
upon ion-channel formation.
Experimental Section
Materials and Methods. Amphotericin B (AmB), egg phosphati-
dylcholine, amino-alcohols, cholesterol, and ergosterol were from
Nakarai Tesque. 1,2-Dimyristoyl-sn-glycero-3-phosphocholine (DMPC)
was purchased from Avanti Polar Lipid Inc. (Alabaster, AL). [1-13C,15N]-
glycine was from Cambridge Isotope Laboratory (Cambridge, MA).
Deuterated solvents were purchased from Merck. All other chemicals
were obtained from standard venders and used without further purifica-
tion. NMR spectra were recorded on a JEOL GSX-500 spectrometer.
ESI-MS spectra were measured on an LCQ-deca (Thermo Finnigan).
HPLC was performed on a Shimadzu LC-10ADvp with an SPD-
M10Avp photodiode array detector. Thin-layer chromatography (TLC)
was performed on a glass plate precoated with silica gel (E. Merck
Kieselgel 60 F254). Column chromatography was performed with silica
gel 60 (E. Merck, particle size 0.063-0.200 mm, 60-230 mesh).
N-(9-Fluorenylmethoxycarbonyl)-glycyl-3-aminopropanol (6).
Fmoc-glycine was prepared with a standard method. To a stirred
solution of Fmoc-glycine (120 mg, 0.30 mmol) and 3-aminopropanol
(46µL, 40 µmol, 0.60 mmol) in DMF (12 mL) were added diisopro-
pylethylamine (138 µL, 0.80 mmol), 1-hydroxybenzotriazole (74 mg,
0.48 mmol), and PyBOP (208 mg, 0.40 mmol) sequentially. After being
stirred at 23 °C for 16 h, the mixture was diluted with water (100 mL)
and extracted with CHCl3. The organic layer was dried over MgSO4,
concentrated, and purified by column chromatography on SiO2 with
CHCl3-MeOH to give 6 as a white amorphous powder (89 mg, 63%).
Liquid NMR Measurements. The antibiotic solutions were prepared
under dry argon in DMSO-d6 at 5 mM concentration. All liquid NMR
spectra were recorded at 25 °C on a JEOL GSX500 spectrometer (1H
500 MHz). Spectra were processed using Alice2 V.4.1 (JEOL DATUM)
software. Homonuclear two-dimensional spectra COSY, TOCSY (HO-
HAHA), and NOESY were recorded with a 1.5 s recovery delay in the
phase-sensitive mode using the States method as data matrices of 512
(F1) × 1024 (F2) complex data points. Mixing times of 80 ms for
TOCSY and 300 ms for NOESY spectra were used. The spectral width
in both dimensions was 5000 Hz. The data were apodized with shifted
square sine-bell window functions in both F1 and F2 dimensions after
zero-filling in the F2 dimension to obtain a final matrix of 512 (F1) ×
2048 (F2) real data points. Chemical shifts were referenced to the
solvent chemical shift (DMSO-d5(1H), 2.49 ppm.).
Conformation Analysis of 5. All the interproton-distance restraints
between non-J-coupled protons are derived from the two-dimensional
homonuclear NOESY experiments. Interproton restraints were classified
into three categories. Upper bounds were fixed at 2.8, 3.4, and 4.0 Å
for strong, medium, and weak correlations, respectively. A lower bound
was fixed at 1.8 Å, which corresponds to the sum of the hydrogen van
der Waals radii. Pseudo atom corrections of the upper bounds were
applied for distance restraints involving the unresolved methylene and
methyl protons (+1 Å). For stereospecifically assigned diastereotopic
methylene protons, the interproton distances were applied to each proton
according to the NOE peak intensities. When possible, H-C-C-H
dihedral angles were restrained to dihedral domains according to the
1
Rf 0.41 (CHCl3-MeOH ) 5:1). H NMR (500 MHz, CDCl3) δ 7.74
(d, 2H, J ) 7.0 Hz), 7.57 (d, 2H, J ) 7.0 Hz), 7.39 (t, 2H, J ) 7.5
Hz), 7.30 (t, 2H, J ) 7.5 Hz), 6.43 (brs, 1H), 5.38 (brs, 1H), 4.44 (d,
2H, J ) 7.0 Hz), 4.20 (t, 1H, J ) 7.0 Hz), 3.83 (d, 2H), 3.62 (t, 2H),
3.42 (m, 2H), 1.67 (m, 2H).
N-(9-Fluorenylmethoxycarbonyl)-glycyl-3-aminopropanal (7). To
a CH2Cl2 solution (1 mL) of 6 (150 mg, 0.48 mmol) was added Dess-
Martin reagent (220 mg, 0.52 mmol) at 23 °C. After being stirred for
3 h, the solution was quenched with 10 mL of saturated Na2S2O3
solution and extracted with CHCl3. The organic layer was dried over
MgSO4 and concentrated in vacuo, yielding crude 7 as a white
amorphous powder (69 mg, 72%). Rf 0.41 (CHCl3-MeOH ) 5:1). 1H
NMR (500 MHz, CDCl3) δ 9.77 (s, 1H), 7.74 (d, 2H, J ) 7.5 Hz),
7.57 (d, 2H, J ) 7.5 Hz), 7.38 (t, 2H, J ) 7.5 Hz), 7.30 (t, 2H, J ) 7.5
Hz), 6.39 (brs, 1H), 5.38 (brs, 1H), 4.41 (d, 2H, J ) 7.0 Hz), 4.20 (t,
1H, J ) 7.0 Hz), 3.80 (d, 2H), 3.54 (m, 2H), 2.73 (m, 2H).
3
different JHH coupling constants measured using Karplus dihedral
relations with a (30° allowance. Conformation was calculated using
the MacroModel software version 8.529 installed on RedHat Linux 8
OS. Initial atomic coordinates and structure files were generated step-
by-step from the crystal data of N-iodoacetyl AmB.28 The macrolide
ring was treated as a semirigid group, in which a (30° allowance from
the crystal structure28 was given to each C-C single bond upon
calculation. The sampling of the conformational space was performed
9
J. AM. CHEM. SOC. VOL. 128, NO. 36, 2006 11983

A R T I C L E S
Matsumori et al.
following a Monte Carlo Multiple Minimum method.45 The MMFF
force field46 implemented in the MacroModel was used. The nonbonded
van der Waals interactions were represented by a simple repulsive
quadratic term. The experimental distance restraints were represented
as a soft asymptotic potential, and electrostatic interactions were
ignored. 5000 steps of MCMM were carried out.
DMSO were spotted on paper disks 8 mm in diameter. As a control, a
disk containing only DMSO was also prepared. These paper disks were
then placed on an agar plate containing A. niger mycelia. After
cultivating at 25 °C for 2 days, the diameter of the inhibitory zone on
each paper disk was measured.
Solid-State NMR Measurements. For membrane preparation, a
combination of 2.6 mg of 5′ and 16.7 mg of DMPC (molar ratio 1:10)
or that of 2.6 mg of 5′, 15.1 mg DMPC, and 1.0 mg of ergosterol
(1:9:1) was mixed in CHCl3-MeOH (6:1), and the solvent was removed
in vacuo for 8 h. The dried membrane was hydrated with 19.3 µL of
10 mM HEPES buffer (pH 7.0) under Ar and then diluted with 1 mL
of H2O. After a few minutes of sonication, the membrane was freeze-
thawed 3 times and vortexed to make multilamellar vesicles. The
suspension was lyophilized, rehydrated with 19.3 µL of D2O, and
transferred into a glass insert (Wilmad) for a φ 5-mm MAS rotor, which
was sealed with epoxy glue. Then the glass insert was set to the φ
5-mm MAS rotor. For the powder sample of 5′, 5 mg of 5′ was mixed
with Celite and packed into the φ 5-mm MAS rotor. Solid-state NMR
experiments were carried out on a CMX300 (Varian/Chemagnetics)
spectrometer operating at a resonance frequency of 299.494 MHz for
1H, 75.320 MHz for 13C, and 30.351 MHz for 15N. Cross-polarization
spectra of 13C were obtained under no spinning for the powder sample
and 1 kHz MAS for membrane samples. 13C{15N} REDOR spectra30
were acquired with the MAS frequency of 5000 ( 2 Hz. DD-MAS
spectra were obtained with single-pulse excitation of 13C under 5 kHz
MAS. In all measurements, the rotor temperature was kept at 30 ( 1
°C by a temperature controller, and the spectral width was 30 kHz.
Typically a 90° pulse width for 1H was 2.5 µs, and 180° pulse widths
for 13C and 15N were 7 µs and 14 µs, respectively. The contact time
for cross polarization transfer was set to be 2 ms. Spectra were acquired
K+ Flux Assays Using 31P NMR. Egg phosphatidylcholine (52.0
µmol) or egg phosphatidylcholine-sterol (46.8 and 7.2 µmol) was
dissolved in CHCl3 (2 mL), and the mixture was evaporated to a thin
film in a round-bottom flask. After dried in vacuo for over 8 h, pH 7.5
buffer (0.726 mL) containing KH2PO4 (0.4 mM) and EDTA (1 mM)
was added to the flask. The lipid mixture was suspended in the buffer
by vortexing and sonication. The resultant suspension was frozen at
-20 °C and thawed at 50 °C 3 times. The liposome solution thus
obtained was passed repeatedly through a membrane filter (pore size,
0.2 µm, 19 times) with a Liposofast apparatus (AVESTIN) and then
diluted with K2SO4 (0.4 M, 3.63 mL) to give a 12 mM LUV solution.
The LUV suspension was adjusted to pH 7.5 with 10 M KOH. A 750
µL aliquot of the LUV suspension was transferred to a 1.5 mL
Eppendorf tube and mixed with 2 µL of 10 mM carbonyl cyanide-p-
trifluoromethoxyphenyl hydrazone (FCCP) ethanol solution and 2 µL
of a DMSO solution of drugs. After incubation for 6 h at room
temperature, a 550 µL portion was transferred to a 5-mm NMR glass
tube and mixed with 4.4 µL of 100 mM MnCl2. 31P NMR spectra were
recorded at 23 °C on a JEOL GSX-500 spectrometer (31P at 202.35
MHz) with 1H-broad band decoupling. The assay was generally repeated
more than 3 times for each sample with good reproducibility. The
activity was evaluated by the decrease in peak area at δ 1.2. From the
concentration-activity relationship, EC50 values were determined.
Biological Activities. For hemolytic assays, freshly collected human
blood was centrifuged for 5 min at 1000 g, and separated erythrocytes
were washed twice by suspending in PBS buffer (pH 7.4). Packed
erythrocytes were then resuspended with PBS to a 100-fold volume of
the original blood. Gradually diluted drug solutions (DMSO solution,
2 µL) were added to the erythrocyte suspensions (198 µL), and the
suspensions were incubated with gentle shaking at 37 °C. After 18 h,
the suspensions were spun and the absorbance of the supernatant was
determined at 490 nm by a microplate reader (Molecular Devices). For
a positive control, water was used instead of a PBS buffer. As a negative
control, 2 µL of DMSO were added instead of the sample solution.
From dose-response curves, the dosage that led to 50% hemolysis
(EC50) was determined.
1
with a recycle delay of 5 s and a H decoupling field strength of 75
kHz. REDOR spectra were measured at dephasing times of 1.6, 3.2,
4.8, and 6.4 ms, using xy-8 phase cycling for 15N irradiations.47
Acknowledgment. We are grateful to Prof. Tohru Oishi and
Dr. Shigeru Matsuoka in our laboratory for invaluable discus-
sions. This work was supported by Grants-In-Aids for Scientific
Research A) (No. 15201048) and for Young Scientists (B) (No.
14780458) from MEXT, Japan.
Supporting Information Available: NMR spectra and NOE
correlations for 5; UV and CD spectra of 1, 3, and 5 in PBS
buffer. This material is available free of charge via the Internet
at http://pubs.acs.org.
For antifungal assays, Aspergillus niger was cultured in a GP liquid
medium (2% glucose, 0.2% yeast extract, 0.5% polypeptone, 0.05%
MgSO4, and 0.1% KH2PO4) at 25 °C for 2 days. An aliquot of the
broth was then spread onto a GP agar plate. The drugs dissolved in
JA063433W
(45) Kolossvary, I.; Guida, W. C. J. Comput. Chem. 1999, 20, 1671-1684.
(46) Halgren, T. A. J. Comput. Chem. 1999, 20, 730-748.
(47) Gullion, T.; Baker, D. B.; Conradi, M. S. J. Magn. Reson. 1990, 89, 479-
484.
9
11984 J. AM. CHEM. SOC. VOL. 128, NO. 36, 2006