Synthesis and biophysical studies on 35-Deoxy amphotericin b methyl ester

Article abstract of DOI:10.1002/chem.200900231

The use of molecular editing in the elucidation of the mechanism of action of amphotericin B is presented. A modular strategy for the synthesis of amphotericin B and its designed analogues is developed, which relies on an efficient gram-scale synthesis of various subunits of amphotericin B. A novel method for the coupling of the mycosa-mine to the aglycone was identified. The implementation of the approach has enabled the preparation of 35-deoxy amphotericin B methyl ester. Investigation of the antifungal activity and efflux-inducing ability of this amphotericin B congener provided new clues to the role of the 35-hydroxy group and is consistent with the involvement of double barrel ion channels in causing electrolyte efflux. 2009 Wiley-VCH Verlag GmbH & Co. KGaA.

Full text of DOI:10.1002/chem.200900231

FULL PAPER
DOI: 10.1002/chem.200900231
Synthesis and Biophysical Studies on 35-Deoxy Amphotericin B Methyl Ester
Alex M. Szpilman, Damiano M. Cereghetti, Jeffrey M. Manthorpe,
Nicholas R. Wurtz, and Erick M. Carreira*[a]
Abstract: The use of molecular editing
in the elucidation of the mechanism of
action of amphotericin B is presented.
A modular strategy for the synthesis of
amphotericin B and its designed ana-
logues is developed, which relies on an
efficient gram-scale synthesis of various
subunits of amphotericin B. A novel
method for the coupling of the mycosa-
mine to the aglycone was identified.
The implementation of the approach
has enabled the preparation of 35-
deoxy amphotericin B methyl ester. In-
vestigation of the antifungal activity
and efflux-inducing ability of this am-
photericin B congener provided new
clues to the role of the 35-hydroxy
group and is consistent with the in-
volvement of double barrel ion chan-
nels in causing electrolyte efflux.
Keywords: amphotericin B
· anti-
fungal agents · natural products ·
structure–activity relationship
total synthesis
·
Introduction
Amphotericin B (1), a prominent member of the mycosa-
mine family that also includes nystatin (2), candidin (3) and
rimocidin (4), remains the treatment of last resort for sys-
temic fungal infections.[1,2] Since amphotericin B (1) entered
the clinic in 1958, its use has been relegated to closely moni-
tored treatment in hospitals, due to a number of debilitating
side effects.[3] Its use has been extended to treat multi-drug
resistant types of the Leshmania parasite.[4] Leshmaniasis is
at present one of the most common parasite infections
worldwide, second only to malaria. As the first member of
the mycosamine family to have its structure fully elucidat-
ed,[5] amphotericin B (1) has attracted wide attention from
synthetic groups. Numerous synthetic approaches have been
reported since the early 1980s.[6] However, to date only one
total synthesis has been accomplished.[7] In addition, synthe-
ses of the aglycones of candidin (3) and rimocidin (4) have
been documented (Figure 1).[8]
The ability of amphotericin B (1) to cause efflux of intra-
cellular electrolytes was observed early on and linked to its
fungicidal activity.[9] Further studies led to the suggestion by
[a] Dr. A. M. Szpilman, Dr. D. M. Cereghetti, Dr. J. M. Manthorpe,
Dr. N. R. Wurtz, Prof. Dr. E. M. Carreira
Laboratorium fꢀr Organische Chemie der ETH-Zꢀrich
HCI H 335, Wolfgang Pauli Strasse 10
8093 Zurich (Switzerland)
E-mail: carreira@org.chem.ethz.ch
Supporting information for this article is available on the WWW
under http://dx.doi.org/10.1002/chem.200900231.
Figure 1. Structure of amphotericin B (1) and related mycosamine poly-
ketides.
Chem. Eur. J. 2009, 15, 7117 – 7128
ꢁ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
7117

Finkelstein et al. that electrolyte efflux could be accounted
for by the formation of barrel-like transmembrane ion chan-
nels.[10] These constructs are believed to arise from 4–12
molecules of amphotericin B (1) and are constituted in the
fungal membrane through a process of self-assembly. This
model is commonly referred to as the barrel-stave model.[1b]
The relatively strong electrolyte efflux observed for ergo-
sterol containing membranes in comparison to cholesterol
containing membranes was proposed to derive from the
active participation of sterols in the self-assembly process
and/or from the incorporation of the sterols as an integral
part of the ion channel construct.[11] An alternative mecha-
nism of action has also been proposed. According to this hy-
pothesis the polyene moiety participates in redox processes
that upset the intracellular redox balance.[12] In spite of nu-
merous studies a full understanding of the molecular mecha-
nism of action of amphotericin B (1) remains elusive.[1b] It
could very well be that multiple fungicidal mechanisms are
operative.
Figure 2. Barrel-stave models for ion channel formation in the fungal
membrane.
hydroxy group for biological activity.[18] We also disclose al-
ternative routes to the C1–C13 and C33–C37 moieties.
Our synthetic strategy is outlined in Scheme 1. The struc-
ture of amphotericin B (1) may be divided into three princi-
pal components: the mycosamine sugar 6 as well as the C1–
C20 and C21–C37 moieties that collectively comprise the
macrocycle.[19] Our approach is predicated on a strategy in
which each of these is prepared in a synthetically efficient
and flexible manner, thus enabling a practical strategy for
the molecular editing of amphotericin B (1).[20]
An intriguing discussion surrounding the barrel-stave
model, relates to the discrepancy between the overall length
of amphotericin B (1) and that of the typical fungal cell
membrane (21 and 40 ꢂ, respectively).[1b,5] This incongruity
has been addressed in models that suggest constriction of
the cell membrane around the site of the ion channel (Fig-
ure 2a) or the formation of ion channels consisting of two
AmB subunits oriented in
a tail-to-tail fashion (Fig-
ure 2b).[1b] Computational studies have pinpointed specific
structural elements of amphotericin B (1) as potential “hot
spots” of importance for antifungal activity (Figure 1).[13]
The C35-hydroxyl unit has been singled out and hypothe-
sized to be essential for the formation and stabilization of
these dimeric ion channels through hydrogen bonding. Ba-
ginski et al. noted that: “…Concerning the interactions be-
tween AmB molecules in the channel, we suggest eliminat-
ing these hydroxyl groups or substituting related (e.g., me-
thoxy or methyl) groups. However, we are aware that selec-
tive modification of the hydroxyl groups involves a very dif-
ficult chemical synthesis task. Other AmB hydroxyl groups
are very similar; probably only through total synthesis from
fragments can AmB derivatives be prepared with modified
hydroxyl groups in certain positions.”[13]
Recently the use of amphotericin B analogues as probes
for various aspects in the barrel-stave model has emerged as
a powerful tool for the study of its mechanism of action and
associated toxicity.[1b] Access to these analogues has been
achieved through semisynthesis as well as through genetic
modification of the amphotericin B producing polyketide
synthase of Streptomyces nodosus.[14–17] In line with the re-
marks of Baginski et al., we proposed that a fully synthetic
access to amphotericin B analogues would be a powerful
complement to these methods as they would in principal
render any structural part of the amphotericin B backbone
amenable to modification.
Scheme 1. Retrosynthetic analysis of amphotericin B (1).
Results and Discussion
Taking note of the inherent latent elements of symmetry in
the C1–C13 moiety we pursued a strategy that would allow
access to both the C1–C7 and C8–C13 subunits using identi-
cal tactics.[7] To this end, we developed two complementary
syntheses of the C1–C7 and C8–C13 fragments: One based
on catalytic asymmetric reactions developed in our labs and
In this paper, we deliver a full account of our synthetic
preparation of 35-deoxy amphotericin B methyl ester and
biophysical studies that revealed the significance of the C35-
7118
www.chemeurj.org
ꢁ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2009, 15, 7117 – 7128

Amphotericin B
FULL PAPER
one based on starting materials from the chiral pool
(Scheme 2). The former involves the copper-tol-BINAP (tol-
BINAP = 2,2’-bis(di-p-tolylphosphino)-1,1’-binaphthyl) cat-
alyzed enantioselective aldol addition of bis-enol ether 9 to
commercially available furylaldehyde.[21] In order to facili-
tate the gram-scale preparation of aldol adduct 10, we inves-
tigated the direct use of commercially available solvents.
Fortuitously, the reaction could be successfully scaled to
provide more than 160 g (0.67 mol, 99% yield) of aldol
product in a single batch, using only 2 mol% of the in situ
formed catalyst and HPLC grade THF (<0.02% water).
Refluxing the acetonide enol ether 10 in butanol converted
it into the keto n-butyl ester 11 in 92% yield. At this point
the 1,3-syn diol motif was introduced by application of the
boron-mediated Prasad reduction.[22] The hydroxyl ketone
was treated with in situ formed or premade methoxydiethyl-
boron to form a complex which was reduced with sodium
borohydride (3 equiv) at ꢀ788C. The resulting boron ester
was hydrolyzed using hydrogen peroxide in water/THF lead-
ing to liberation of the free syn-diol. The unpurified diol
was protected as the acetonide using pyridinium p-toluene-
sulfonate (PPTS; 0.9 mol%) as the catalyst and 2,3-dime-
thoxypropane as a co-solvent with DMF. Subsequently, the
carboxylic ester was converted into a primary alcohol by the
action of 2 equivalents of LiAlH4 at 08C (68% yield).
Scheme 2. a) 2 mol% CuF2, 2.2 mol% (R)-tol-BINAP, 4 mol% TBAT,
1 equiv furylaldehyde, THF, ꢀ788C, 99% yield, 94% ee; b) nBuOH,
1108C, 92%; c) 1.1 equiv Et2BOMe, 3 equiv NaBH4, MeOH, THF,
ꢀ788C; d) H2O2 water/THF; e) 0.9 mol% PPTS, (MeO)2CMe2, DMF,
72% over 3 steps; f) 2 equiv LiAlH4, THF, 08C, 68%; g) 1.55 equiv KH,
1.05 equiv BnBr, 12 mol% (nBu)4NI, THF/DMF 10:1, 95%; h) O3
CH2Cl2/MeOH 1:1, then 1.5 equiv Ph3P, then CH2N2, 66%; i) LiAlH4,
THF, 08C, 39%; j) 1 mol% TEMPO, 10 mol% KBr, NaOCl, pH 8.6
The synthetic plan required the C1 alcohol to be protect-
ed. This was achieved by deprotonating it with potassium
hydride (1.55 equiv) and alkylation by benzyl bromide
(1.04 equiv) in the presence of 12 mol% tetrabutylammoni-
um iodide as a catalyst. The furan was cleaved oxidatively
by ozone and the resulting ozonide was reduced in situ by
triphenylphosphine (1.5 equiv). Finally, the revealed carbox-
ylic acid was converted into a methyl ester by addition of di-
azomethane to the reaction vessel. The yield for this three-
step-one-pot transformation was 66%. The methyl ester was
dissolved in THF and reduced at 08C to a primary alcohol
using solid lithium aluminium hydride (1 equiv). Unfortu-
nately, the product proved difficult to extract from the het-
erogeneous reaction mixture. Addition of sodium sulfate
decahydrate, followed by filtration, concentration at reduced
pressure and flash chromatography led to the isolation of al-
cohol 13 in only 39% yield (see below). Subsequent oxida-
tion of the alcohol to an aldehyde can be achieved by reac-
tion with 1.2 equivalents of the Dess–Martin periodinane in
the presence of 10 equivalents of pyridine in dichlorome-
thane[23] in 80% yield. However, for large-scale applications,
the aldehyde was best prepared by oxidation of the alcohol
in dichloromethane by the action of buffered bleach
(1 equiv, phosphate buffer pH 8.6)) and TEMPO (1 mol%)/
potassium bromide (10 mol%) as catalysts at 08C.[24] This
method produces the aldehyde in quantitative yield and also
has the advantage that the air-sensitive aldehyde can be
used directly without need for time consuming chromato-
graphic purification. Furthermore, this protocol obviates the
need to produce the Dess–Martin reagent on large-scale.
Exposure of the unpurified aldehyde to a slight excess
(1.3 equiv) of the Ohira reagent under mild basic conditions
buffer, CH2Cl2, 0–78C; k) EtO2CC(N2)P(O)ACHTUNGTRENUNG(OEt)2, K2CO3, MeOH, 51%
over 3 steps; l) SOCl2, MeOH; m) BH3·SMe2, cat. NaBH4, THF, 08C;
n) TBSCl, imidazole, CH2Cl2, 0 to 238C; o) LDA, tert-butyl acetate, THF,
ꢀ78 to ꢀ108C, 55% over 4 steps; p) Bu3B, NaBH4, MeOH, THF ꢀ788C;
q) H2O2, water/THF; r) 1 mol% PPTS, 1 equiv 2-methoxypropene,
CH2Cl2, ꢀ358C to RT, 69% over 3 steps, d.r. 15:1; s) LiAlH4, THF,
ꢀ108C; t) 1.2 equiv NaH, 2 equiv BnBr, 1 equiv (C4H9)4NI, THF/DMF
10:1, 88% over 2 steps; u) excess HF/pyridine, THF, 08C. TEMPO=tet-
ramethylpiperidoxyl; TBAT = tetrabutylammonium triphenyldifluorosi-
licate; tol-BINAP = 2,2’-bis(di-p-tolylphosphino)-1,1’-binaphthyl.
(2 equiv of K2CO3/methanol) afforded the desired alkyne 14
in 51% yield over three steps.[25]
Alkyne 14 can also be accessed starting from commercial-
ly available, inexpensive (S)-malic acid (15). Esterification
as the dimethyl ester was effected using two equivalents of
thionyl chloride in methanol. The ester function proximal to
the hydroxy group could then be reduced selectively using
borane and 10 mol% sodium borohydride as a catalyst.[26]
The C6-primary alcohol of the resulting diol was then pro-
tected chemoselectively by TBS-Cl and two equivalents of
imidazole in dry dichloromethane. Earlier syntheses of
acetal protected syn-diol 18 had relied on the use of sturdy
but heavy TBDPS[27a] (MW =239) or TIPS[27b] (MW =157)
protecting groups. A synthesis of a large natural product
such as amphotericin B (1) must ultimately contend with the
issue of mass efficiency: introduction of a unduly heavy pro-
tecting group early on demands that larger quantities of
starting materials be carried through the early steps only to
be subject to large weight loss later in the synthesis. For this
reason we chose to rely on the lighter TBS (MW =115)
group. For comparison, the molecular weight of methyl-3,4-
Chem. Eur. J. 2009, 15, 7117 – 7128
ꢁ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
7119

E. M. Carreira et al.
dihydroxy butanoate is a mere 134 gmolꢀ1. As a conse-
quence of this decision, laborious operations such as large-
scale chromatographic separations were made considerably
easier, but, as a corollary, we had to overcome a rather curi-
ous problem in a subsequent acetonide protection step.
Prasad reduction of 17, this time using methoxydibutylboron
but otherwise identical conditions as above, followed by
treatment of the crude diol product with 10 mol% TsOH in
2,2-dimethoxypropane,[27] led to formation of a complex
mixture of products in which the TBS group had undergone
migration. Exposure of the diol to PPTS (10 mol%) and
excess 2-methoxypropene afforded the bis(methoxypropane)
acetal in poor yield. Only by using a combination of low
temperature (ꢀ358C to room temperature), low PPTS load-
ing (1 mol%) and one equivalent of 2-methoxypropane in
dichloromethane could TBS migration be avoided and high
yield ensured. Under these conditions the acetonide 18 was
isolated in 70% yield (two steps). At this point the C1 ester
group was reduced to the primary alcohol by lithium alumi-
num hydride in THF at ꢀ108C. Upon workup the crude al-
cohol was dried by distillation as an azeotrope with toluene
and immediately subjected to the following step. This en-
tailed benzylation of the alcohol by benzyl bromide
(2 equiv) and tetrabutylammonium iodide after deprotona-
tion by sodium hydride. The benzyl ether was isolated after
purification in a yield of 88% (2 steps). Key to obtaining
this result was the use of a commercial solution of lithium
aluminium hydride in the reduction step rather than the
solid form. As discussed above, the C8–C13 fragment 19/20
could be accessed using similar methods but starting from
either ent-12 or (R)-malic acid diethyl ester (Scheme 3).
We recognized this outcome as an opportunity to test the
asymmetric addition of zinc acetylides to aldehydes, dis-
closed from our laboratory.[30] In the situation at hand, the
chiral system is faced with a challenging mismatched scenar-
io, as evidenced from the result involving the organolithium
reagent. In the experiment, coupling of 14 and 20 using
1.1 equivalents of zinc triflate, 1.2 equivalents of N-methyle-
phedrine (NME) and 1.3 equivalents of triethylamine at
room temperature in toluene lead to the formation of the
desired R epimer 22 as the major stereoisomer in excellent
yield and in only 2.5 h (Table 1, entry 2). Similarly, union of
14 and 19 under these conditions affords the desired R prop-
argylic alcohol 21 (entry 3).
Table 1. Addition of alkyne 14 to aldehydes 19 and 20.
Entry
Alde-
hyde
Conditions
Yield [%] of
21 or 22
d.r.
C8-R/S
1
2
20
20
nBuLi, THF ꢀ788C
64 (22)
98 (22)
1:4
16:1
1.1 equiv Zn
(OTf)2,
1.2 equiv (ꢀ)-NME,[a]
1.3 equiv Et3N, toluene
3
19
1.1 equiv Zn
(OTf)2,
83 (21)
16:1
1.2 equiv (ꢀ)-NME,[a]
1.3 equiv Et3N, toluene
[a] NME=N-Methylephedrine.
Alkyne 22 underwent chemoselective reduction using H2,
10 w/w Pd/C as the catalyst in the presence of NaHCO3
(Scheme 4) in quantitative yield with no need for purifica-
tion. Next, the free C8-alcohol function was protected by
the action of excess TBS-Cl/imidazole in DMF. The TBS
ether 23 was isolated in quantitative yield after flash chro-
matography. Subsequently, the ester functionality was re-
duced to an alcohol using lithium aluminium hydride (solu-
tion) and without purification oxidized to an aldehyde. This
oxidation can be carried out with high yield using either the
Dess–Martin periodinane or the TEMPO/bleach protocols
described above. However, as discussed above TEMPO/
bleach oxidation proved superior for large-scale work as the
unpurified aldehyde could be used directly in the subse-
quent step. Accordingly, treatment of the crude aldehyde
with hydroxylamine hydrochloride (3 equiv) in pyridine
yielded the desired oxime 24 (84% over two steps). Only
two chromatographic purifications were needed in the
course of this five-step sequence, which greatly facilitated
the preparation of over 40 g of 24. In an alternative se-
quence, the alcohol function in 21 was protected as the TBS
ether using TBSOTf, 2,6-lutidine as the base at 08C in di-
chloromethane (90%) yield. Hydrogenation of the alkyne
was carried out as described above in 90% yield. Attempts
to chemoselectively cleave the primary TBS group using
CSA as the catalyst in methanol led to a complex mixture of
mono TBS-protected and unprotected alcohol products.
However, a yield of 89% of the desired alcohol 23 can be
achieved by carefully monitoring the reaction of the bis-
TBS-ether with excess HF/pyridine at 08C. Thus, the two
routes intersected at compound 23.
Scheme 3. a) LiAlH4, THF, 08C; b) 1.5 equiv TBSCl, 1.5 equiv imidazole,
CH2Cl2, 0 to 238C, 84% over 2 steps; c) O3, CH2Cl2/MeOH 1:1, then
1.5 equiv Ph3P, then CH2N2; d) LiAlH4, THF, 08C; e) 1.2 equiv DMP,
10 equiv pyridine, CH2Cl2, 80% yield; f) HF/pyridine, THF, 08C, 99%;
g) 1 mol% TEMPO, 10 mol% KBr, NaOCl, pH 8.6 buffer, CH2Cl2, 0–
78C, quantitative. DMP=Dess–Martin periodinane.
Initially, coupling of the C1–C7 and C8–C13 fragments
was attempted by lithiation of alkyne 14 with 1 equiv of bu-
tyllithium and subsequent addition to aldehyde 20. This re-
action proceeded at a surprisingly low rate, ultimately re-
quiring 14 h to reach completion. In addition, the propargyl-
ic alcohol 22 was formed with the undesired S epimer as the
major product as shown by Mosherꢃs ester analysis.[28,29]
Indeed this stereochemical result was expected based on a
polar Felkin–Anh model and is in line with the observations
of Masamune and Hanessian.[6a,7e]
7120
www.chemeurj.org
ꢁ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2009, 15, 7117 – 7128

Amphotericin B
FULL PAPER
Scheme 5. a) HF/pyridine, THF, 08C; b) 1.5 mol% TEMPO, 10 mol%
KBr, NaOCl, pH 8.6 buffer/CH2Cl2, 08C; c) 10 equiv NaClO2, tBuOH/2-
methyl-2-butene/2m NaH2PO4, 08C; d) TMSCHN2, MeOH/EtOAc, 63%
over 4 steps; e) 1.3 equiv LiOOH, water/THF; f) TMSCHN2, MeOH/
EtOAc, 63% over 2 steps; g) tert-butyltrichloroacetimidate, hexane, 15%
over 2 steps; h) see Table 2; i) 1.3 equiv LiOOH, water/dioxane;
j) TMSCHN2, MeOH/EtOAc, 67% over 2 steps. XN =Evansꢃ norephe-
drine derived auxiliary.
Scheme 4. a) See Table 1; R=COOtBu: b) H2, 3% w/w 10% Pd/C,
NaHCO3, MeOH; c) 2 equiv TBSCl, 4.2 equiv imidazole, DMF, 408C;
d) LiAlH4, THF, 08C 97% over three steps; R=CH2OTBS; b) 1.1 equiv
TBSOTf, 1.3 equiv 2,6-lutidine, CH2Cl2, 08C, 90%; c) H2, 10% w/w 10%
Pd/C, NaHCO3, MeOH, 90%; d) HF/pyridine, THF, 08C, 89%;
e) 1 mol% TEMPO, 10 mol% KBr, NaOCl, pH 8.6 buffer, CH2Cl2, 08C;
f) 3 equiv HONH2·HCl, pyridine, 84% over 2 steps.
Table 2. Synthesis of isoxazolines 29–31 (see Scheme 5).
A nitrile oxide cycloaddition was applied as a means to
append the C14–C19 fragment, in the form of alkene dipo-
larophiles 26, 27 and 28 to oxime 24 (Scheme 5, Table 2).
The homo-allylic alcohols 26–27 could be conveniently pre-
pared from the known Evans crotonate aldol product 25.[31]
First, the TBS-protecting group was removed using excess
HF/pyridine in THF at 08C. The free primary alcohol group
was oxidized to the aldehyde using the standard TEMPO/
bleach protocol[24] and then to the acid oxidation state using
the Pinnick modification of the Lindgren oxidation
(NaClO2, excess 2-methylbutene in tert-butanol).[32] The car-
boxylic acid was methylated using TMS-diazomethane. This
step was conveniently carried out by concentrating the com-
bined extracts (ethyl acetate) of the oxidation reaction and
using this solution, with the addition of about 25% metha-
nol (by volume), as the reaction medium for the alkylation
reaction. The acid was titrated with TMS-diazomethane
until a faint yellow color persisted. The methyl ester 26 was
purified by flash chromatography and isolated in 63% yield
(4 steps). Alternative coupling partners 27 and 28 were fash-
ioned by the extrusion of the Evansꢃ auxiliary using one
equivalent of lithium peroxide in water/dioxane.[33] Alkyla-
tion of the free carboxylic acid function could then be ach-
ieved by TMS-diazomethane to form the dimethyl ester 27
in 63% yield (2 steps) or by tert-butyl trichloroacetimidate
in hexane with a trace of acetic acid to afford the bis-tert-
butyl ester 28 in 30% yield (2 steps).
Entry Alkene Conditions
Yield [%] d.r.
1
2
3
4
26
26
27
28
24, iPrOH, iPrMgCl, tBuOCl,
ꢀ788C to RT
0
–
24, 0.55 equiv (Bu3Sn)2O, tBuOCl, 95 (29)
ꢀ308C ! RT
88:12
72:28
67:33
24, 0.55 equiv (Bu3Sn)2O, tBuOCl, 81 (30)
ꢀ40 C ! RT
24, 0.55 equiv (Bu3Sn)2O, tBuOCl, 50 (31)
ꢀ40 C ! RT
tributyltinoxide (0.55 equivalents) in dichloromethane prior
to oxidation by tert-butyl hyperchlorite (1 equiv) at ꢀ308C.
Upon slow warming to room temperature the incipient ni-
trile oxide undergoes cycloaddition to alkene 26 to form the
desired isoxazoline 29 in excellent yield and stereoselectivity
(95%, 88:12).[35] The tributyltin nitrile oxide adduct likely
acts as a reservoir that slowly releases the free nitrile oxide,
thus minimizing the unproductive dimerization side reac-
tion.
The significance of the nature of the C16-carbonyl sub-
stituent for the stereochemical outcome of the dipolar cyclo-
addition reaction was examined. The reaction of the dimeth-
yl ester 27 with 24 under the conditions described above led
to the formation of 29 but at the cost of a drop in yield and
stereoselectivity (entry 3). Similarly, reaction of 24 with the
di-tert-butyl ester 28 led to formation of the product 31 in
moderate yield and stereoselectivity (entry 4). Thus, the role
of the Evansꢃ auxiliary in the reaction is not simply that of a
bulky substituent. Stereoelectronic factors appear to be in-
volved in determining the stereoselectivity of the reaction.
Treatment of 29 with LiOOH led to efficient extrusion of
the Evansꢃ auxiliary.[33] The resulting free carboxylic acid
function was protected as the methyl ester by TMS-diazo-
methane as discussed above. Subsequent purification by
flash chromatography afforded the methyl ester 30 in 67%
yield (over 2 steps).
We have recently reported a stereospecific synthesis of
broad scope of isoxazolines via cycloaddition of magnesium
alkoxides of chiral allylic and homoallylic alcohols to in situ
formed nitrile oxides.[34] To our disappointment, no trace of
product could be isolated when the union of 24 and 26 was
attempted under the conditions shown in Table 2, entry 1.
Presumably, the rate of cycloaddition is outstripped by the
competing dimerization of the reactive nitrile oxide. An al-
ternative procedure involves stannylation of the oxime bis-
Chem. Eur. J. 2009, 15, 7117 – 7128
ꢁ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
7121

E. M. Carreira et al.
Finally, the synthesis of the C1–C20 moiety of the agly-
cone was completed by the sequence shown in Scheme 6.
pane 1:30 mixture. Using these conditions a yield of 75%
was achieved for the methyl ketal formation–silylation se-
quence.
ꢀ
Reduction of the N O bond by excess Raney-Nickel in the
presence of water led to formation of a hydroxyketone that
immediately cyclized to form hemiacetal 32. Unfortunately,
the product was isolated in only 34% yield. Better results
were achieved using molybdenum hexacarbonyl complex
(1.6 equiv) in refluxing wet acetonitrile for 4 h.[36] The yield
under these conditions was highly dependent on the workup
procedure. Thus, when the crude tar-like product formed on
removal of the reaction solvent was submitted directly to
column chromatography the product was formed in only
55% yield. In contrast, suspension of the crude product with
silica gel in ethyl acetate/hexane and filtration of the suspen-
sion through a short silica gel plug afforded the pure hemia-
cetal reproducibly in 80–86% yield. Thus, physical contact
with silica gel was essential for hydrolyzing the putative mo-
lybdenum product complex. However, simple evaporation
to dryness of the crude product on silica gel prior to flash
chromatography lead to a yield of only 28%.[37]
The keto-phosphonate function, an important handle for
the subsequent macrocyclization, was introduced by selec-
tive addition of (MeO)2P(O)CH2Li (1.8 equiv) to the less
sterically hindered methyl ester in 33 at 08C in THF. Opti-
mal yields were achieved by running this reaction to approx-
imately 70% conversion. Selectivity and thus overall yield
dropped if the reaction was allowed to proceed to higher
conversion. Subsequently, the benzyl ether was cleaved
under reductive conditions. The resulting free primary alco-
hol was converted into an aldehyde by Dess–Martin periodi-
nane[23] and finally to carboxylic acid 8 by a Pinnick/Lindg-
ren oxidation.[32]
The completion of C1–C20 fragment 8 entails a formal
total synthesis of amphotericin B (1).[7] Consequently, it is
important to take stock of the route. A total of 28 steps
were required to synthesize 8 from either bis-enol ether 9 or
malic acid 15. In both cases an overall yield of 4% was ach-
ieved, making these sequences the shortest and highest
yielding synthesis of 8 to date. Indeed more than 8 grams of
8 have been prepared to date. Importantly, we had fulfilled
the prerequisite of our synthetic strategy of having access to
all the subunits of amphotericin B through efficient and scal-
able syntheses (see above).[19] This, set the stage for the syn-
thesis of the first target, 35-deoxy amphotericin B (5).
The goal, to prepare 35-deoxy amphotericin B, dictated
that we first access the C21–C37 moiety of the amphoteri-
cin B aglycone in which the C35-hydroxy function would be
absent, namely 7b (Scheme 1). At the outset, a synthetic ap-
proach to 7b was by no means obvious as the exclusion of
the C35-hydroxyl function in the target precluded the use of
standard aldol methodology for its construction. We choose
(S)-3-hydroxybutanoic acid ethyl ester 34 as the starting
point and to rely on the use of substrate control in setting
the stereogenic center at C36. This was done by a Fratꢄr–
Seebach alkylation.[38] The enolate of 34 was formed in situ
using 1 equiv LDA, then 1.25 equiv methyl iodide in THF/
HMPA was added. The reaction was performed at ꢀ788C
and proceeded with excellent stereocontrol (>95:5 d.r.) and
in 92% yield (Scheme 7). Reduction of the ester group with
lithium aluminum hydride in Et2O at 08C afforded a diol in
84% yield. This diol could be selectively iodinated at the
primary alcohol position through an Appel reaction which
employed 1.5 equiv triphenylphosphine, 1.4 equiv of iodine
and 3.1 equiv of pyridine. The secondary alcohol was then
protected as the TES ether in 93% yield by triethylsilyl tri-
flate (1.04 equiv) and 2,6-lutidine (1.3 equiv) as the base in
dichloromethane (08C).
Scheme 6. a) 1.6 equiv
[Mo(CO)6],
MeCN/water,
808C,
86%;
b) 20 mol% 2-chloropyridine·CSA, (MeO)2CMe2, MeOH; c) 1.1 equiv
TBSOTf, 10 equiv 2,6-lutidine, CH2Cl2, 75% over 2 steps; d) 1.8 equiv
(MeO)2P(O)CH2Li, THF, ꢀ35–08C, 71% based on recovered starting
material; e) 10% w/w 10% Pd(OH)2/C, H2, EtOAc; f) DMP, pyridine,
CH2Cl2; g) 2 equiv NaClO2, tBuOH, 2-methyl-2-butene, 2m NaH2PO4,
99% (over 3 steps).
Hemi-acetal 32 was converted into the corresponding
methyl ketal and the remaining free hydroxyl group was si-
lylated using 1.1 equiv TBSOTf/2,6-lutidine (10 equiv) in di-
chloromethane at 08C to afford 33 (Scheme 6). Curiously,
methyl-ketal formation did not proceed using PPTS (0.1–
1 equiv) as the acid catalyst and acetic acid trimethylor-
thoester as the dehydrating reagent and cosolvent with
methanol.[7e] In contrast, the use of CSA (10 mol%) as the
catalyst led to the formation of a mixture of mono-acetonide
products. The need for a strongly acidic catalyst and condi-
tions that would minimize loss of acetonide protection
groups was reconciled by using chloro-pyridinum camphor
sulphonate (pKa ꢁ0.8) in a methanol/2,2-dimethoxy pro-
The powerful method developed by Myers,[39] which in-
volves alkylation of the enolate of 38, was used to set the
stereogenic center at C34 of fragment 7b. The enolate was
formed using LDA as the base in the presence of lithium
chloride at ꢀ788C. The iodide 35 was then added at 08C.
The reaction mixture was worked up by addition to a bipha-
sic system consisting of hexane and aqeous HCl. The hexane
7122
www.chemeurj.org
ꢁ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2009, 15, 7117 – 7128

Amphotericin B
FULL PAPER
ready described. The overall yield of the target aldehyde 7
for this three-step procedure was 62%.
Compound 7b was characterized by a strong yellow-
orange hue identical to that of the corresponding fragment
7a of amphotericin B.[7] In contrast to 7a, 7b proved insolu-
ble in all solvents examined with the exception of dichloro-
methane. This proved a serious impediment in optimizing
the subsequent coupling with carboxylic acid 8. Even more
problematic, 7b was found to be unreactive towards a large
variety of activated esters of 8. For example, the use of
DCC/DMAP[7c] or PYBOP, afforded none of the desired
ester. Esterification using standard two stage Yamaguchi
conditions[40] (i.e., 1.2 equiv of the Yamaguchi reagent and
1.2 equiv of triethylamine in THF followed by exchange of
the solvent for dichloromethane and addition of 7b) afford-
ed the ester 41 in only 12% yield. Exchanging strongly basic
triethylamine for the milder 2,6-lutidine increased the yield
somewhat although with poor reproducibility. Careful analy-
sis of the by-products led to identification of the b-elimina-
tion product of 8 by HRMS. This implied that the activated
esters of 8 in the presence of strong bases would undergo
elimination of acetone at a rate greater than reaction with
7b. Based on this observation a successful one pot protocol
was developed, that avoided the use of strong bases such as
DMAP (pKa 9.2). Accordingly, a mixture of acid 8 with
4.6 equivalents of pyridine (pKa 5.2) was added over 10 h to
a stirred mixture of Yamaguchiꢃs reagent (2.5 equiv), 7b
(3 equiv) and an additional 0.5 equivalent of pyridine.
Under these conditions ester 41 was produced in a reprodu-
cible yield of 48–52%. The Shiina reagent[41] was also exam-
ined, but did not offer any immediate advantage. Due to the
instability of ester 41 it was used immediately after purifica-
tion on silica gel. Excess alcohol 7b could be recovered and
recycled.
Scheme 7. a) 2.1 equiv LDA, 1.25 equiv MeI, HMPA/THF 1:10, ꢀ788C,
92%, d.r. 95:5; b) 2 equiv LiAlH4, Et2O, 0 8C, 84%; c) 1.5 equiv Ph3P,
1.4 equiv I2, 3.1 equiv pyridine, THF, 81%; d) 1.04 equiv TESOTf,
1.3 equiv 2,6-lutidine, CH2Cl2, 08C, 93%; e) 2.1 equiv 38, 4.3 equiv LDA,
LiCl, THF, ꢀ78 to 08C, 78%, brsm, d.r. 95:5; f) 3.9 equiv LDA, 3.9 equiv
BH3·NH3, THF, 08C to RT 97%; g) 1 mol% TEMPO, 10 mol% KBr,
NaOCl, pH 8.6 buffer/CH2Cl2, 08C, 98%; h) 1.25 equiv 39, 1.2 equiv
LDA, THF, ꢀ78 to 08C, 78%; i) 5 equiv DIBAL, CH2Cl2, ꢀ788C;
j) 10 equiv MnO2, CH2Cl2; k) 1.4 equiv 40, 1.2 equiv LDA, THF, ꢀ78 to
08C, 55% over 3 steps; l) HF/pyridine, THF, 08C; m) 6.3 equiv DIBAL,
CH2Cl2, ꢀ788C; n) 10 equiv MnO2, CH2Cl2, 62% over 3 steps. brsm =
based on recovered starting material.
cosolvent served to protect the sensitive TES ether from the
HCl while allowing slow quenching of the base. Excess 38
was crystallized out from a toluene solution of the crude
product. The product was isolated after chromatography in
60% yield along with 23% recovered iodide. Reductive
cleavage of the Myersꢃ auxiliary using lithium amidoborohy-
dride (LAB, 4 equiv) in THF reagent proceeded efficiently
in 97% yield even in the presence of the labile TES group.
Again special attention had to be given to the acidic
workup, which was carried out in a similar fashion to the al-
kylation step.
The alcohol was then oxidized to aldehyde 36 using the
TEMPO/bleach protocol[24] and the resulting aldehyde was
subjected to olefination using Horner–Wadsworth–Emmons
reagent 39, LDA in THF at ꢀ788C. The product triene 37
was obtained in 78% yield. The ester functionality was re-
duced to the primary alcohol using excess DIBAL in di-
chloromethane at ꢀ788C. The unpurified alcohol was dis-
solved in dichloromethane and then oxidized by MnO2
(10 equiv) to the aldehyde. Purification of the product was
performed by simply filtering off the excess reagent and
subjecting it to a Horner–Wadsworth–Emmons reaction this
time with phosphonate 40. Accordingly, using 1.4 equiva-
lents of 40, 1.2 equivalents of LDA as the base and THF sol-
vent, the desired hexane product was obtained in 55%
(over three steps). All that remained was to adjust the oxi-
dation stage of C20 to the aldehyde stage and liberate the
protected C37 alcohol. The latter step was tackled first using
excess HF/pyridine in THF at 08C. The crude product was
subjected to DIBAL reduction (6.3 equiv) in dichlorome-
thane and oxidation by MnO2 (10 equiv) in the manner al-
Horner–Wadsworth–Emmons (HWE) macrocyclization
was carried out using six equivalents of potassium carbon-
ate/[18]crown-6 (12 equiv) in toluene at 608C as described
for the synthesis of natural amphotericin B aglycone.[7]
Under these conditions, the red-colored unsaturated ketone
is produced in a yield of 48–55%. A legion of other condi-
[43]
tions were examined, for example, LiCl/DBU,[42] Ba(OH)2
(CF3)2),[44] but unfortunately these methods af-
or LiACTHNUTRGENN(UG OCHACHUTNGTRENNGUN
forded the product in lower yields (12–30%). Reversing the
order of steps, that is, coupling the two fragments through a
HWE reaction, employing and closing the macrocycle by a
macrolactonization reaction, also did not offer any advan-
tages in terms of yield. Reduction of the ketone with sodum
borohydride (10 equiv) to the unsaturated alcohol 42
(Scheme 8) proceeded in 73% yield with excellent stereose-
lectivity and was accompanied by a gratifying, readily ob-
servable color change from bright red to pale yellow.[7a]
A final obstacle remained: the forging of the bond be-
tween the aglycone and the mycosamine appendage. The
challenge of forming the glycosidic bond is partially a prod-
uct of the poor nucleophilicity of the C19 alcohol which is
both electronically deactivated and sterically hindered by
virtue of its position inside the concave grove of the agly-
Chem. Eur. J. 2009, 15, 7117 – 7128
ꢁ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
7123

E. M. Carreira et al.
since trichloroacetamide is highly insoluble in hexane and
precipitates from the reaction mixture. This protocol stood
out by its ability to successfully afford glycoside 44b, albeit
in very low yield of 4%. Additionally, the reaction was
marred by low conversion (75% of the starting alcohol 43
was recovered from the reaction) and formation of signifi-
cant amounts (12%) of the orthoester 45b as a side-product
(Table 3, entry 1). Rearrangement[47] of orthoester 45b to
glycoside 44b was attempted using a variety of acids ranging
from BF3 to PPTS. However, the inherent acid sensitivity of
the aglycone conspired with the stability of the orthoester
functionality to foil these efforts. Weak acids such as PPTS
(0.3–2 equivalents) in CDCl3 were unable to induce rear-
rangement while stronger acids such as methanesulfonic
(pKa ꢀ2.6) or triflic acid (pKa ꢀ14) led to decomposition of
the orthoester. The latter was usually accompanied by a
transient blue color, possibly indicative of the formation of
a polyene stabilized carbenium ion by the breaking of the
Scheme 8. a) 2.5 equiv 2,4,6-trichlorobenzoylchloride, 5.1 equiv pyridine,
3 equiv 7b; CH2Cl2, 08C, 48%; b) 6 equiv K2CO3, 12 equiv [18]crown-6,
toluene, 608C, 52%; c) 10 equiv NaBH4, MeOH, 08C, 73%.
cone. It is further compounded by the fact that mycosamine
has a mannoside configuration and is attached as the ther-
modynamically less favored b-anomer. Indeed, the introduc-
tion of b-mannoside residues remains one of the greatest
general challenges to carbohydrate chemistry.[45]
ꢀ
C19 O bond.
Careful examination of the problem including mechanistic
considerations pointed us in a productive direction. In the
Schmidt reaction, the first step is the extrusion of trichloroa-
cetamide by the action of a Lewis acid with concomitant for-
mation of an oxocarbenium ion. When a C2’ ester group ca-
pable of providing anchimeric assistance is present, an ace-
toxonium ion (e.g., 46) is formed. In the normal manifold
attack of the aglycone on C1’ of the pyran ring leads to the
formation of the glycoside (e.g., 44b). However, it is well
known that in the case of sterically hindered or electronical-
ly deactivated alcohols, such as amphotericin B aglycone 43,
attack may take place at the sterically less encumbered ace-
toxonium carbon leading to formation of orthoesters. We
postulated that by increasing the spatial demand of the ester
substituent R in 6 it would be possible to inhibit orthoester
formation in favor of glycoside.[48] This approach was not
without some potential problems, because of the need for
subsequent hydrolysis of the directing ester group in the
presence of the macrocyclic lactone. Accordingly, we turned
our attention to esters possessing significant steric bulk
We brought a number of the powerful and widely applica-
ble glycosidation methods that have appeared in the litera-
ture in the last two decades to bear on this problem with
little to show for our efforts.[46] While the lack of success
was distressing, it led us to re-examine the union of ampho-
teronolide 43 and trichloroacetimidate 6a (Scheme 9,
Table 3).[7d] This method employed 30 mol% PPTS (pKa 5.2)
as a mild-acid activator of the trichloroacetimidate and
hexane as the solvent. The use of hexane inhibits formation
of trichloroacetamide sugar adducts, a common problem,
Table 3. Introduction of the mycosamine moiety (Scheme 9).
Acceptor Donor Activator Aglycone [mm] 44 [%] 45 [%] 42 [%]
43
43
43
42
6a
6b
6b
6b
PPTS[a]
PPTS[a]
7[7d]
20
4
12
11
23
30
75
57
–
27
43
45
CMPT[b] 20
CMPT[c] 20
–
[a] 30 mol% [b] 50 mol% CMPT and 100 mol% 2-chloro-6-methylpyri-
dine. [c] 10 mol% CMPT and 20 mol% 2-chloro-6-methylpyridine.
CMPT=2-chloro-6-methylpyridinium triflate.
Scheme 9. Glycosidation of amphoteronolide 43 and 35-deoxy amphoteronolide 42 with trichloroacetimidates 6 (see Table 3).
7124
www.chemeurj.org
ꢁ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2009, 15, 7117 – 7128

Amphotericin B
FULL PAPER
while being electronically disposed for hydrolysis under
mild conditions (Table 3).
To our consternation reaction of 6b with pyridinium tri-
flate (30 mol%) and 43 only proceeded to about 30% con-
version. NMR studies revealed the source of inhibition to
be pyridine. Thus, one equivalent of pyridinium triflate was
able to fully activate 6b within 50 min leading to formation
of a mixture of the epimeric pyridinium species 49/50. The
Accordingly we screened a range of esters as ancillary
steering groups in the glycosidation reaction still using 30
mol% of PPTS as the activator and hexane as the solvent.
The determination of yields was complicated by the fact
that often the orthoester and glycoside products would elute
together. However, the glycoside to orthoester ratio could
be determined from the crude NMR since the characteristic
H1’ signal of the glycoside typically at around 4.5 ppm could
be clearly distinguished from the orthoester H1’ signal that
commonly appeared around 5.6 ppm. The ability to cleave
the steering group under mild conditions after glycosidation
was a chief parameter in selection of esters for examination.
Dominant formation of orthoester 45b resulted in the
case of R2 being ClCH2- or CH3CH2-. A benzoyl ester in 6
affords a 1:1 mixture of orthoester 45 (R=Bz) and glycoside
44 (R=Bz). In contrast reaction of the more bulky 6b (R=
longer time required for activation using pyridinium triflate
compared to PPTS was traced to the lower solubility of pyri-
dinium triflate in CDCl3. The triflate salts 49/50 proved
highly unstable, decomposing slowly over the course of 24 h.
Furthermore these species were unreactive towards agly-
cone 43.
Accordingly, we investigated the use of sterically hindered
pyridinium triflates as catalysts for the reaction. Extensive
experimentation led to the identification of 2-chloro-6-meth-
ylpyridinium triflate (pKa 2.8) as a highly efficient catalyst
for the reaction. In contrast, 2,6-lutidinum triflate (pKa 6.9)
was not acidic enough to catalyze the reaction. As expected
based on a previous experience with the activation of or-
thoesters, attempts to use triflic acid (10 mol%) led to de-
composition of the aglycone.
Exposure of 6b and 43 to 10 mol% of 2-chloro-6-methyl-
pyridinum triflate in the presence of 20 mol% of 2-chloro-6-
methylpyridine led to rapid and complete conversion of 43
into glycoside 44b (43%) and orthoester 45b (23%). Impor-
tantly, glycosidation of 42 under identical conditions afford-
ed the desired glycoside 33b in 45% yield (Table 3, entry 4).
The CMP ester could be hydrolyzed under mild conditions
(K2CO3, methanol/THF) to afford the C2’ alcohol 51 in
83% within 2 h (Scheme 10).
ClACHTUNGTRENNUNG(CH3)2C- with 30 mol% PPTS in the presence of 43 af-
forded the desired glycoside 44b as the major product
(entry 2). An additional finding of critical importance was
that the 2-chloro-2-methylpropanoic (CMP) ester could be
cleaved under mild basic condition (2 equiv K2CO3 in meth-
anol) in 69% yield.[29]
Glycoside donor 6b proved less reactive than 6a towards
aglycone 43. This necessitated performing the glycosidation
reaction at higher concentration to achieve conversion.[7d] In
order to address the issue of low conversion, the activation
of 6b by PPTS was studied by NMR. Within two minutes of
addition of 1 equivalent of PPTS to trichloroacetimidate 6b
in CDCl3 the formation of b-tosylate 47 was observed. In
the course of a few hours, 46 slowly converted into the more
stable a-anomer 48. While glucotosylates are able to glycosi-
date simple alcohols,[49] exposure of 47/48 to 43 did not lead
to formation of product. As tosylate appeared to inhibit
conversion, we surmised that substituting the tosylate for
the less nucleophilic triflate as the acid counter anion should
increase the reactivity of the activator/glycoside donor
couple.
The stereogenic center at C2’ of 51 was now inverted by
oxidation to the ketone and immediate reduction with
Scheme 10. a) 2 equiv K2CO3, MeOH/THF, 83% yield; b) 3 equiv (CF3CO)2O, 6 equiv DMSO, 6 equiv (Me2N)2CO, 6 equiv Et3N, CH2Cl2/Et2O, ꢀ788C to
RT; c) 2 equiv NaBH4, MeOH, 82% yield over 2 steps based on recovered 50; d) excess HF·4py, MeOH, 408C, 72%; e) 20 mol% CSA, MeCN/water,
37% (6 cycles); f) 1.2 equiv Bu3P, THF/MeOH/water, 67%.
Chem. Eur. J. 2009, 15, 7117 – 7128
ꢁ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
7125

E. M. Carreira et al.
NaBH4 to afford protected 35-deoxy amphotericin B 52
(Scheme 10).[7d] The three TBS protecting groups were re-
moved by the action of HF·4py complex in methanol. Re-
moval of the acetonide groups proved problematic due to
the acid sensitive nature of the unsaturated glycoside func-
tion. Best yields were obtained by performing the deacetali-
zation using 20 mol% CSA in acetonitrile/water. The initial-
ly formed mixture of mono-acetonides was isolated by flash
chromatography and resubmitted to the same conditions
several times. Unfortunately, the overall yield of deprotect-
ed compound remained low. Importantly sufficient material
was produced for completion of the synthesis. Thus, Stau-
dinger reduction of the C3’ azide afforded 35-deoxy ampho-
tericin B methyl ester (53). Reduction using propane dithiol/
triethylamine (10 equivalents) proceeded in low yields and
formation of by products. Attempts to hydrolyze the methyl
ester to give free 35-deoxy amphotericin B were unsuccess-
ful.[50] Ultimately, this was of no consequence as the biologi-
cal profile of amphotericin B methyl ester (AME) is compa-
rable to that of amphotericin B (1). Thus, the importance of
the 35-hydroxy function for biological activity could be eval-
uated by direct comparison of the properties of methyl ester
53 to those of AME.
The antifungal activity of 35-deoxy amphotericin B
methyl ester 53 and AME was evaluated against Saccharo-
myces cerevisiae (Table 4). Strikingly, 53 is more than 18-
fold less active than AME. Further evaluation was done
using the CAF2-1 strain of Candida albicans. Against this
strain, 53 is less than 26-fold active compared to the parent
AME. These results illustrate the importance of the 35-hy-
droxy group for activity. To further examine its role in the
formation of ion channels, we studied the ability of 35-
deoxy amphotericin B methyl ester (53) to cause K+ efflux
from large unilamellar vesicles (LUV), as monitored using
potassium ion selective electrodes.[17,51] We utilized LUV
with a membrane composed of 1-palmitoyl-2-oleoyl-sn-glyc-
ero-3-phosphocholine (POPC) with or without ergosterol as
a component (Figure 3 and Supporting Information).[7] At
1 mm concentration 35-deoxy amphotericin B methyl ester
(53) showed severely diminished ability to induce the leak-
age of K+ compared with amphotericin B methyl ester. At
the concentration of 0.1 mm or in pure POPC LUV no efflux
could be observed. Only at exceedingly high concentrations
(10 mm) could a weak efflux be detected (see Supporting In-
formation).
Table 4. Antifungal activity (MIC) of amphotericin B methyl ester
(AME) and 35-deoxy amphotericin B methyl ester.
Entry Compound S. cerevisiae BY4741
[mm]
C. albicans CAF2-1
[mm]
1
2
AME
53
0.25
4.60
0.10
2.60
[a] Measured according to the NCCLS protocol. See ref. [51] for experi-
mental details.
Collectively these results underscore the pivotal role of
the C35-hydroxyl group for the biological mechanism of
action as well as lend support for ion channel formation as a
necessary condition for the fungicidal effect of amphoteri-
cin B (1).[52,53] Importantly, it provides the first direct experi-
mental support for the formation of tail-to-tail dimeric ion
channels. Furthermore, the hydroxyl group is unlikely to
play a role in oxidative processes related to the polyene
function.[1b] Hence, the fact that the absence of this specific
hydroxy group is of importance, argues against the oxidative
damage mechanism.
Conclusion
In summary, we have demonstrated the molecular editing of
amphotericin B (1) through diverted total synthesis.[20] An
important feature of this stratagem was the development of
an efficient synthetic strategy that allowed access all the
moieties of amphotericin B (1) on a multi-gram scale. Using
the synthetic strategy described we have studied the impor-
tance of the C35-hydroxy group of amphotericin B (1). Our
results reveal the pivotal role of this group and are consis-
tent with the formation of dimeric ion channels and their
importance for the mechanism of action of amphotericin B
(1). Furthermore, we have developed
a glycosidation
method that give improved yields of glycoside over orthoest-
ers. Recent studies in our labs show the generality of this ap-
proach.[54]
Acknowledgements
This research was supported by ETH. Generous postdoctoral scholar-
ships were provided by the Carlsberg foundation (A.M.S.), Fonds
NATEQ (J.M.M.) and NSF (N.R.W.). We thank Professor Samuel Dani-
shefsky for a stimulating discussion.
Figure 3. KCl efflux from LUV induced by 53 (c) and amphotericin B
methyl ester (a) measured via potentiometry. The substrate was added
externally (as a DMSO solution) to afford a final concentration of 1 mm.
LUV with a diameter of 100 nm and containing 13% ergosterol and
87% POPC in their membranes were utilized. LUV=large unilamellar
vesicle, POPC=1-palmitoyl-2-oleoyl-sn-glycero-phosphocholine.
[1] a) Macrolide Antibiotics, Chemistry, Biology and Practice (Ed.: S.
Omura), Academic Press, New York, 1984; b) for a recent review
see: D. M. Cereghetti, E. M. Carreira, Synthesis 2006, 914–942.
[2] T. H. Sternberg, E. T. Wright, M. Oura, Antibiot. Annu. 1956, 566–
573.
7126
www.chemeurj.org
ꢁ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2009, 15, 7117 – 7128

Amphotericin B
FULL PAPER
[3] a) H. A. Gallis, R. H. Drew, W. W. Pickard, Rev. Infect. Dis. 1990,
12, 308–329; b) I. M. Hann, H. G. Prentice, Int. J. Antimicrob.
Agents 2001, 17, 161–169.
[4] A. Lemke, A. F. Kiderlen, O. Kayser, Appl. Microbiol. Biotechnol.
2005, 68, 151–162.
116, 5293–5297; Angew. Chem. Int. Ed. 2004, 43, 5181–5185; b) A.
Zumbuehl, P. Stano, M. Sohrmann, M. Peter, P. Walde, E. M. Car-
reira, Org. Biomol. Chem. 2007, 5, 1339–1342; c) V. Paquet, A.
Zumbuehl, E. M. Carreira, Bioconjugate Chem. 2006, 17, 1460–
1463; d) A. Zumbuehl, P. Stano, D. Heer, P. Walde, E. M. Carreira,
Org. Lett. 2004, 6, 3683–3686.
[5] P. Ganis, G. Avitabile, W. Mechlin, C. P. Schaffner, J. Am. Chem.
Soc. 1971, 93, 4560–4564.
[18] For a preliminary report, see: a) A. M. Szpilman, D. M. Cereghetti,
N. R. Wurtz, J. M. Manthorpe, E. M. Carreira, Angew. Chem. 2008,
120, 4407–4410; Angew. Chem. Int. Ed. 2008, 47, 4335–4338;
b) A. M. Szpilman, J. M. Manthorpe, E. M. Carreira, Angew. Chem.
2008, 120, 4411–4414; Angew. Chem. Int. Ed. 2008, 47, 4339–4341.
[19] We have previously disclosed full accounts of our gram scale synthe-
ses of the C33–C38 fragment: J. Tholander, E. M. Carreira, Helv.
Chim. Acta 2001, 84, 613–622; and the mycosamine subunit: J. M.
Manthorpe, A. M. Szpilman, E. M. Carreira, Synthesis 2005, 3380–
3388.
[6] Selected synthetic approaches towards amphotericin B: a) S. Hanes-
sian, S. P. Sahoo, M. Botta, Tetrahedron Lett. 1987, 28, 1143–1146;
b) T. Berkenbusch, R. Brueckner, Chem. Eur. J. 2004, 10, 1545–
1557; c) C. Bonini, L. Chiummiento, A. Martuscelli, L. Viggiani, Tet-
rahedron Lett. 2004, 45, 2177–2179; d) K. Karisalmi, K. Rissanen,
A. M. P. Koskinen, Org. Biomol. Chem. 2003, 1, 3193–3196; e) K.
Karisalmi, A. M. P. Koskinen, M. Nissinen, K. Rissanen, Tetrahedron
2003, 59, 1421–1427; f) G. Solladiꢄ, N. Wilb, C. Bauder, Eur. J. Org.
Chem. 1999, 3021–3026; g) G. Solladie, J. Hutt, Tetrahedron Lett.
1987, 28, 797–800; h) G. J. McGarvey, J. M. Williams, R. N. Hiner,
Y. Matsubara, T. Oh, J. Am. Chem. Soc. 1986, 108, 4943–4952;
i) J. M. Williams, G. J. McGarvey, Tetrahedron Lett. 1985, 26, 4891–
4894.
[20] Professor Danishefsky suggested that diverted total synthesis could
be used as way to perform “molecular editing” of undesired or un-
necessary structural features of natural products: R. M. Wilson, S. J.
Danishefsky, J. Org. Chem. 2006, 71, 8329–8351.
[7] See: a) K. C. Nicolaou, T. K. Chakraborty, Y. Ogawa, R. A. Daines,
N. S. Simpkins, G. T. Furst, J. Am. Chem. Soc. 1988, 110, 4660–4672;
b) K. C. Nicolaou, R. A. Daines, J. Uenishi, W. S. Li, D. P. Papahat-
jis, T. K. Chakraborty, J. Am. Chem. Soc. 1988, 110, 4672–4685;
c) K. C. Nicolaou, R. A. Daines, T. K. Chakraborty, Y. Ogawa, J.
Am. Chem. Soc. 1988, 110, 4685–4696; d) K. C. Nicolaou, R. A.
Daines, Y. Ogawa, T. K. Chakraborty, J. Am. Chem. Soc. 1988, 110,
4696–4705; e) for another seminal contribution see the synthesis of
the aglycone: R. M. Kennedy, A. Abiko, T. Takemasa, H. Okumoto,
S. Masamune, Tetrahedron Lett. 1988, 29, 451–454.
[8] For the synthesis of the candidin aglycon, see: I. Kadota, Y. Hu,
G. K. Packard, S. D. Rychnovsky, Proc. Natl. Acad. Sci. USA 2004,
101, 11992–11995; for the synthesis of the rimocidin aglycon, see:
G. K. Packard, Y. Hu, A. Vescovi, S. D. Rychnovsky, Angew. Chem.
2004, 116, 2882–2886; Angew. Chem. Int. Ed. 2004, 43, 2822–2826.
[9] For a leading review, see: S. C. Hartsel, C. Hatch, W. Ayenew, J.
Liposome Res. 1993, 3, 377–408.
[10] a) A. Finkelstein, R. Holz in Membranes (Ed.: G. Eisenman),
Marcel Dekker, New York, 1973, pp. 377–408; b) A. Marty, A. Fin-
kelstein, J. Gen. Physiol. 1975, 65, 515–526; c) R. W. Holz, Ann. N.
Y. Acad. Sci. 1974, 235, 469–479.
[11] a) B. De Kruijff, W. J. Gerritsen, A. Oerlemans, R. A. Demel,
L. L. M. Van Deenen, Biochim. Biophys. Acta Biomembr. 1974, 339,
30–43; b) B. De Kruijff, W. J. Gerritsen, A. Oerlemans, P. W. M. Van
Dijck, R. A. Demel, L. L. M. Van Deenen, Biochim. Biophys. Acta
Biomembr. 1974, 339, 44–56; c) B. De Kruijff, R. A. Demel, Bio-
chim. Biophys. Acta Biomembr. 1974, 339, 57–70; d) A. W. Norman,
Biochim. Biophys. Acta Biomembr. 1972, 290, 1–14.
[21] a) R. A. Singer, E. M. Carreira, J. Am. Chem. Soc. 1995, 117,
12360–12361; b) E. M. Carreira, R. A. Singer, Drug Discovery
Today 1996, 1, 145–150; c) J. Krꢀger, E. M. Carreira, J. Am. Chem.
Soc. 1998, 120, 837–838; d) J. Krꢀger, E. M. Carreira, Tetrahedron
Lett. 1998, 39, 7013–7016; e) Y. Kim, R. A. Singer, E. M. Carreira,
Angew. Chem. 1998, 110, 1261–1263; f) B. L. Pagenkopf, J. Krꢀger,
A. Stojanovic, E. M. Carreira, Angew. Chem. 1998, 110, 3312–3314;
Angew. Chem. Int. Ed. 1998, 37, 3124–3126; g) R. Singer, J. Brock,
E. M. Carreira, Helv. Chim. Acta 2003, 86, 1040–1044.
[22] K.-M. Chen, G. E. Hardtmann, K. Prasad, O. Repic, M. J. Shapiro,
Tetrahedron Lett. 1987, 28, 155–158.
[23] D. B. Dess, J. C. Martin, J. Am. Chem. Soc. 1991, 113, 7277–7287.
[24] P. L. Anelli, C. Biffi, F. Montanari, S. Quici, J. Org. Chem. 1987, 52,
2559–2562.
[25] S. Ohira, Synth. Commun. 1989, 19, 561–564.
[26] a) S. Saito, T. Hasegawa, M. Inaba, R. Nishida, T. Fujii, S. Nomizu,
T. Moriwake, Chem. Lett. 1984, 1389–1392; b) S. Saito, T. Ishikawa,
A. Kuroda, K. Koga, T. Moriwake, Tetrahedron 1992, 48, 4067–
4086.
[27] a) G. Wess, K. Kesseler, E. Baader, W. Bartmann, G. Beck, A. Berg-
mann, H. Jendralla, K. Bock, G. Holzstein, H. Kleine, M. Schnierer,
Tetrahedron Lett. 1990, 31, 2545–2548; b) J. L. Hubbs, C. H. Heath-
cock, J. Am. Chem. Soc. 2003, 125, 12836–12843.
[28] a) J. A. Dale, H. S. Mosher, J. Am. Chem. Soc. 1973, 95, 512–519;
b) G. R. Sullivan, J. A. Dale, H. S. Mosher, J. Org. Chem. 1973, 38,
2143–2147; c) I. Ohtani, T. Kusumi, Y. Kashman, H. Kakisawa, J.
Am. Chem. Soc. 1991, 113, 4092–4096.
[29] See the Supporting Information.
[12] a) J. Brajtburg, S. Elberg, D. R. Schwartz, A. Vertut-Croquin, D.
Schlessinger, G. S. Kobayashi, G. Medoff, Antimicrob. Agents Che-
mother. 1985, 27, 172–176; b) M. L. Sokol-Anderson, J. Brajtburg,
G. Medoff, J. Infect. Dis. 1986, 154, 76–83.
[13] M. Baginski, H. Resat, J. A. Mccammon, Mol. Pharmacol. 1997, 52,
560–570.
[14] a) For the preparation of an analogue of amphotericin B lacking the
carboxylic acid group, see: M. Carmody, B. Murphy, B. Byrne, P.
Power, D. Rai, B. Rawlings, P. Caffrey, J. Biol. Chem. 2005, 280,
34420–34426; b) A semisynthetic approach to this same molecule
has also recently been reported: D. S. Palacios, T. M. Anderson,
M. D. Burke, J. Am. Chem. Soc. 2007, 129, 13804–13805.
[15] a) B. N. Rogers, M. E. Selsted, S. D. Rychnovsky, Bioorg. Med.
Chem. 1997, 5, 3177–3182; b) For a fluorinated polyene analogue
whose biological profile has not been determined, see: H. Tsuchika-
wa, N. Matsushita, N. Matsumori, M. Murata, T. Oishi, Tetrahedron
Lett. 2006, 47, 6187–6191.
[30] a) D. E. Frantz, R. Fꢅssler, E. M. Carreira, J. Am. Chem. Soc. 2000,
122, 1806–1807; b) N. K. Anand, E. M. Carreira, J. Am. Chem. Soc.
2001, 123, 9687–9688; c) D. Boyall, F. Lopez, H. Sasaki, D. Frantz,
E. M. Carreira, Org. Lett. 2000, 2, 4233–4236; d) H. Sasaki, D.
Boyall, E. M. Carreira, Helv. Chim. Acta 2001, 84, 964–971; e) R.
Sans-Diez, B. Adger, E. M. Carreira, Tetrahedron 2002, 58, 8341–
8344; f) For an application in total synthesis, see: A. Fettes, E. M.
Carreira, Angew. Chem. 2002, 114, 4272–4275; Angew. Chem. Int.
Ed. 2002, 41, 4098–4101; g) S. Reber, T. F. Knoepfel, E. M. Carreira,
Tetrahedron 2003, 59, 6813–6817; h) F. Kleinbeck, E. M. Carreira,
Angew. Chem. 2009, 121, 586–589; Angew. Chem. Int. Ed. 2009, 48,
578–581.
[31] D. A. Evans, E. B. Sjogren, J. Bartroli, R. L. Dow, Tetrahedron Lett.
1986, 27, 4957–4960.
[32] a) B. O. Lindgren, T. Nilsson, Acta Chem. Scand. 1973, 27, 888–890;
b) B. S. Bal, W. E. Childers, H. W. Pinnick, Tetrahedron 1981, 37,
2091–2096.
[16] V. Paquet, E. M. Carreira, Org. Lett. 2006, 8, 1807–1809, and refer-
ences therein.
[33] D. A. Evans, T. C. Britton, J. A. Ellman, Tetrahedron Lett. 1987, 28,
6141–6144.
[17] a) A. Zumbuehl, P. Stano, M. Sohrmann, T. Vigassy, E. Pretsch, M.
Peter, P. Walde, D. Jeannerat E. M. Carreira, Angew. Chem. 2004,
[34] a) J. W. Bode, N. Frafel, D. Muri, E. M. Carreira, Angew. Chem.
2001, 113, 2128–2131; Angew. Chem. Int. Ed. 2001, 40, 2082–2085;
Chem. Eur. J. 2009, 15, 7117 – 7128
ꢁ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
7127

E. M. Carreira et al.
b) L. Fader, E. M. Carreira, Org. Lett. 2004, 6, 2485–2488; c) N.
Lohse-Fraefel, E. M. Carreira, Org. Lett. 2005, 7, 2011–2014; d) N.
Becker, E. M. Carreira, Org. Lett. 2007, 9, 3857–3860; For applica-
tions in total synthesis, see: e) D. Muri, N. Lohse, E. M. Carreira,
Angew. Chem. 2005, 117, 4104–4106; Angew. Chem. Int. Ed. 2005,
44, 4036–4038. F. Kleinbeck, E. M. Carreira, Angew. Chem. 2008,
120, 596–598; Angew. Chem. Int. Ed. 2008, 47, 586–589.
[35] This method was reported by the McGarvey group in their synthetic
studies towards amphotericin B (1) using a different homoallylic
acohol. See: a) G. J. McGarvey, J. A. Mathys, K. J. Wilson, J. Org.
Chem. 1996, 61, 5704–5705; b) G. J. McGarvey, J. A. Mathys, K. J.
Wilson, Tetrahedron Lett. 2000, 41, 6011–6015; NMR analysis of
hemiketal 32 and 33 indicated that the expected anti product was
formed.
[36] P. G. Baraldi, A. Barco, S. Benneti, S. Manfredini, D. Simoni, Syn-
thesis 1987, 276–278.
[37] A. Guarna, A. Guidi, A. Goti, A. Brandi, F. De Sarlo, Synthesis
1989, 175–178.
[38] a) G. Frꢆter, Helv. Chim. Acta 1979, 62, 2825–2828; b) D. Seebach,
D. Wasmuth, Helv. Chim. Acta 1980, 63, 197–200.
[39] A. G. Myers, B. H. Yang, H. Chen, L. McKinstry, D. J. Kopecky, J. L.
Gleason, J. Am. Chem. Soc. 1997, 119, 6496–6511.
Am. Chem. Soc. 1989, 111, 6881–6882; c) G. K. Packard, S. D. Rych-
novsky, Org. Lett. 2001, 3, 3393–3396.
[47] a) N. K. Kochetkov, A. Y. Khorlin, A. F. Bochkov, Tetrahedron 1967,
23, 693–707; b) For a recent review, see: F. Kong, Carbohydr. Res.
2007, 342, 345–373.
[48] a) A related approach was recently published, see: H. Yu, D. L. Wil-
liams H. E. Ensley, Tetrahedron Lett. 2005, 46, 3417–3421; See also:
b) G. Desmares, D. Lefebvre, G. Renevret, C. Le Drian, Helv. Chim.
Acta 2001, 84, 880–889.
[49] Glycoside tosylates have been reported to be capable of glycosidat-
ing simple alcohols: R. Eby, C. Schuerch, Carbohydr. Res. 1974, 34,
79–90.
[50] Treating 53 with LiOH led to formation of a complex mixture of
products. A complex mixture of byproducts is also formed during
azide reduction with basic triethylamine and 1,3-propanedithiol. The
base sensitivity of amphotericin B derivatives is highly dependant on
structure. Chemical evidence in support of this supposition is the
fact that while saponification of amphotericin B methyl ester
(AME) with LiOH proceeds cleanly, hydrolysis of the acetonide
protected amphotericin B methyl ester leads to epimerization, pre-
sumably at C34.
[51] V. Paquet, A. A. Volmer, E. M. Carreira, Chem. Eur. J. 2008, 14,
2465–2481.
[52] a) P. van Hoogevest, B. de Kruijff, Biochim. Biophys. Acta Bio-
membr. 1978, 511, 397–407; b) D. S. Orlov, T. Nguyen and R. I.
Lehrer, J. Microbiol. Methods 2002, 49, 325–328.
[40] J. Inanaga, K. Hirata, H. Saeki, T. Katsuki, M. Yamaguchi, Bull.
Chem. Soc. Jpn. 1979, 52, 1989–1993.
[41] I. Shiina, R. Ibuka, M. Kubota, Chem. Lett. 2002, 286–287.
[42] M. A. Blanchette, W. Choy, J. T. Davis, A. P. Essenfeld, S. Masa-
mune, W. Roush, T. Sakai, Tetrahedron Lett. 1984, 25, 2183–2186.
[43] I. Paterson, K.-S. Yeung, J. B. Smaill, Synlett 1993, 774–776.
[44] L. K. Blasdel, A. G. Myers, Org. Lett. 2005, 7, 4281–4283.
[45] a) A. V. Demchenko, Synlett 2003, 1225–1240; b) E. S. H. El Ashry,
N. Rashed, E. S. I. Ibrahim, Curr. Org. Synth. 2005, 2, 175–213.
[46] a) R. L. Halcomb, S. J. Danishefsky, J. Am. Chem. Soc. 1989, 111,
6661–6666; b) D. Kahne, S. Walker, Y. Cheng, D. Van Engen, J.
[53] The width of the phospholipid bilayer of these LUVs is 37 ꢂ and
thus any formation of single barrel ion channels should be preclud-
ed. See supporting information for additional measurements.
[54] A. M. Szpilman, E. M. Carreira, Org. Lett. 2009, 11, 1305–1307.
Received: January 27, 2009
Published online: June 19, 2009
7128
www.chemeurj.org
ꢁ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2009, 15, 7117 – 7128