Formation of high-axial-ratio-microstructures from natural and synthetic sphingolipids

Article abstract of DOI:10.1016/S0009-3084(97)00042-X

Amphiphiles that form high-axial-ratio-microstructures (HARMs) are being considered as novel materials for controlled release of drugs and other biologically functional molecules. HARMs consisting of tubules, ribbons, solid rods and helices are formed from sphingolipids by addition of water to a solution of amphiphile in DMF. Single molecular species of galactocerebroside (GalCer) containing long unsaturated fatty acid chains or natural GalCer containing mixed-length, non-hydroxy fatty acids (NFA-GalCer) or α-hydroxy fatty acids (HFA-GalCer) form cylindrical structures. In contrast, single molecular species of GalCer containing long saturated fatty acids form ribbons and helices. GalCer HARMs are typically under 100 nm in diameter and have lengths of several microns. The importance of the amide of GalCer for HARM formation was evaluated using psychosine, which forms solid fibers, whereas sphingosine and an analog of GalCer in which the amide is reduced to a secondary amine form amorphous aggregates. Single molecular species of ceramide containing long unsaturated fatty acid chains form cylindrical structures, whereas those with long saturated fatty acids form ribbons and helices. Short chain saturated ceramide also forms cylindrical structures. GalCer analogs with N-acetyl-glycine in place of the galactose form fibers whereas those with N-acetyl-proline yield amorphous material. The N-acetyl-proline-containing amphiphile can de doped into pure GalCer or NFA-GalCer without perturbing tubule formation.

Full text of DOI:10.1016/S0009-3084(97)00042-X

Chemistry and Physics of Lipids
88 (1997) 21–36
Formation of high-axial-ratio-microstructures from natural and
synthetic sphingolipids
Alex S. Goldstein a, Anatoly N. Lukyanov b, Paul A. Carlson b, Paul Yager b,*,
Michael H. Gelb a
a Departments of Chemistry and Biochemistry, Uni6ersity of Washington, Box 351700, Seattle, WA 98195-1700, USA
b Uni6ersity of Washington, Molecular Bioengineering Program, Center for Bioengineering, Box 352255, Seattle,
WA 98195-2255, USA
Received 19 March 1997; received in revised form 9 June 1997; accepted 9 June 1997
Abstract
Amphiphiles that form high-axial-ratio-microstructures (HARMs) are being considered as novel materials for
controlled release of drugs and other biologically functional molecules. HARMs consisting of tubules, ribbons, solid
rods and helices are formed from sphingolipids by addition of water to a solution of amphiphile in DMF. Single
molecular species of galactocerebroside (GalCer) containing long unsaturated fatty acid chains or natural GalCer
containing mixed-length, non-hydroxy fatty acids (NFA-GalCer) or h-hydroxy fatty acids (HFA-GalCer) form
cylindrical structures. In contrast, single molecular species of GalCer containing long saturated fatty acids form
ribbons and helices. GalCer HARMs are typically under 100 nm in diameter and have lengths of several microns. The
importance of the amide of GalCer for HARM formation was evaluated using psychosine, which forms solid fibers,
whereas sphingosine and an analog of GalCer in which the amide is reduced to a secondary amine form amorphous
aggregates. Single molecular species of ceramide containing long unsaturated fatty acid chains form cylindrical
structures, whereas those with long saturated fatty acids form ribbons and helices. Short chain saturated ceramide
also forms cylindrical structures. GalCer analogs with N-acetyl-glycine in place of the galactose form fibers whereas
those with N-acetyl-proline yield amorphous material. The N-acetyl-proline-containing amphiphile can de doped into
pure GalCer or NFA-GalCer without perturbing tubule formation. © 1997 Elsevier Science Ireland Ltd.
Keywords: Cerebroside; Ceramide; Nanostructure; Tubule; Helix; Drug delivery
1. Introduction
Recently there have been reports demonstrating
* Corresponding author.
that two-chain amphiphiles can self-organize into
0009-3084/97/$17.00 © 1997 Elsevier Science Ireland Ltd. All rights reserved.
PII S0009-3084(97)00042-X

22
A.S. Goldstein et al. / Chemistry and Physics of Lipids 88 (1997) 21–36
Fig. 1. Schematic images of tubules, cochleate cylinders, ribbons and helices.
stable, crystalline, nonliposomal microstructures
when suspended in aqueous media (Yamada et
al., 1984; Nakashima et al., 1984; Fuhrhop et al.,
1988; Shimizu and Hato, 1993; Yager and
Schoen, 1984; Georger et al., 1987). Such high-ax-
provide better shielding of the drug from prema-
ture metabolic degradation.
Sphingolipids have been reported to form
HARMs (Kulkarni et al., 1995; Archibald and
Mann, 1993; Archibald and Yager, 1992; Cura-
tolo and Neuringer, 1986). Cerebrosides are the
simplest mammalian glycosphingolipids. These
lipids consist of a galactose or glucose headgroup
attached to a nonpolar ceramide. Due to order-
disorder transition temperatures (Tm) well above
body temperature, cerebrosides are thought to
impart order to membranes (Curatolo and Neu-
ringer, 1986). In lipid storage diseases such as
Gaucher’s and Krabbe’s disease, cerebroside accu-
mulation disrupts normal membrane function and
forms intracellular lipid deposits (Naito et al.,
1988). Such deposits give rise to microstructures
within the afflicted cells (Yunis and Lee, 1970).
Furthermore, galactocerebroside (GalCer) is a
major component of myelin and the intestinal
brush border and thus contributes to these mem-
brane’s physical properties. Ceramide (Cer), the
product of hydrolytic removal of the sugar from
cerebroside, is thought to serve in cell signal
transduction (Bielawska et al., 1996). Further hy-
drolysis yields, in addition to fatty acids, sphin-
gosine, which also participates in cell signaling
(Bielawska et al., 1996).
ial-ratio-microstructures
(HARMs)
include
tubules, twisted ribbons, helices, and cochleate
cylinders (Fig. 1). Theories on microstructure for-
mation have been advanced for synthetic lipid
systems such as phosphatidylcholine with bu-
tadiyne-containing hydrocarbon chains, and such
theories have been applied to other HARM sys-
tems (Selinger et al., 1996; Nandi and Bagchi,
1996 and references therein). In short, such theo-
ries propose that HARM formation is the result
of intrinsic bending of rectangular bilayer lipid
sheets due to chiral packing of molecules in a
membrane. Interest in natural and synthetic
HARMs has stemmed primarily from their poten-
tial use as templates for mineralization and metal-
lation, for their appearance in biological
structures, for their implication in lipid storage
diseases and as drug-delivery systems (Schnur et
al., 1990; Behroozi et al., 1990; Chappell and
Yager, 1992; Schnur et al., 1994; Mann et al.,
1993; Rudolph et al., 1988; Archibald and Mann,
1994; Schnur et al., 1987). Such lipid molecular
assemblies should have several advantages over
current delivery systems: they require neither a
macroscopic matrix nor a pump to achieve con-
tinuous drug release (by a zero-order kinetic dis-
solution process), they can limit drug exposure to
the desired site of action and such assemblies may
Earlier investigations into the ability of sphin-
golipids to form microstructures focused on the
development of new formation techniques and, to
a limited extent, pure GalCer species. In this
paper, we demonstrate that it is possible to form

A.S. Goldstein et al. / Chemistry and Physics of Lipids 88 (1997) 21–36
23
HARMs from various pure GalCer and Cer spe-
2.2. Chemical synthesis
cies using a new formation technique. Further-
more, synthesized GalCer analogs were converted
to microstructures either as single molecular spe-
cies or as a component of binary mixtures. These
studies are a necessary prelude to understanding
the basic structural requirements for sphingolipid
molecular assemblies.
The chemical synthesis of amphiphiles not com-
mercially available is shown in Fig. 2.
2.2.1. N-hydroxysuccinimide ester of ner6onic
acid
The N-hydroxysuccinimide ester of nervonic
acid was prepared according to Lapidot et al.
(1967): Nervonic acid (0.558 g, 1.52 mmol) and
N-hydroxysuccinimide (0.175 g, 1.52 mmol) in 60
ml anhydrous EtOAc were stirred overnight with
dicyclohexylcarbodiimide (0.314 g, 1.52 mmol).
The white precipitate was removed, and the super-
natant evaporated in vacuo. The residue was re-
crystallized from EtOH to provide the
N-hydroxysuccinimide ester of nervonic acid as
fine white needles (0.539 g, 76%): m.p. 58–60°C;
Rf (CHCl3) 0.24; 1H NMR (500 MHz) 5.35 (t, 2H,
C-15, C-16, J=5.0 Hz), 2.81 (d, 4H, succinimide,
J=4.5 Hz), 2.60 (t, 2H, C-2, J=7.6 Hz), 2.01 (m,
4H, C-14, C-17), 1.74 (t, 2H, C-3, J=5.5 Hz),
0.88 (t, 3H, C-24, J=7.0 Hz).
2. Experimental procedures
2.1. Materials
All materials were of reagent grade purity and
used as received.
D-erythro-sphingosine was pur-
chased from Avanti Polar Lipids. All other non-
synthesized lipids were purchased from Sigma (St.
1
Louis, MO). H NMR spectra were obtained in
CDCl3 using a Bruker 300 MHz or 500 MHz
NMR spectrometer with tetramethylsilane as an
internal standard. Silica gel (EM Science Silica
Gel 60, 230-400 Mesh) was used for all flash
chromatography. Thin layer chromatography was
performed using Silica Gel 60 F254 (EM Science).
2.2.2. N-ner6onoyl ceramide (24:1-cer)
N-nervonoyl ceramide was prepared according
to Ong and Brady (1972): The N-hydroxysuccin-
imide ester of nervonic acid (0.092 g, 198.4 vmol)
2.1.1. Microscopy
Phase contrast optical micrographs were taken
using a Zeiss ICM 405 microscope (Carl Zeiss,
Thornwood, NY) with 40× (NA 0.75) phase
contrast lenses. Transmission electron micro-
graphs (TEM) were obtained using a Philips EM
410 electron microscope operating at an accelera-
tion potential of 80 kV. Samples were applied to
Formvar-coated 150 mesh copper TEM sample
grids with and without negative stain (2%
aqueous ammonium molybdate pH 5.0).
2.1.2. Calorimetery
Differential scanning calorimetery was per-
formed using a Seiko DSC-100 high sensitivity
calorimeter. Known concentrations of aqueous
lipid suspensions (50 vl) were heated from 1–
95°C at 1°C/min in 70 vl silver calorimetery pans.
In order to insure proper hydration of the lipids,
each sample was pre-heated under the same con-
ditions before the actual run. After calorimetery,
the pan’s contents were analyzed for decomposi-
tion by thin layer chromatography.
Fig. 2. Chemical synthesis of amphiphiles not commercially
purchased.

24
A.S. Goldstein et al. / Chemistry and Physics of Lipids 88 (1997) 21–36
vided the title compound as a white solid (0.090 g,
66%): Rf (3:1 hexane:EtOAc) 0.66; H NMR (300
and sphingosine (0.062 g, 207.0 vmol) were dis-
solved in 10 ml anhydrous THF and stirred
overnight under Ar. Flash chromatography
(1:0:0-90:10:1 CHCl3:MeOH:NH4OH) provided
N-nervonoyl ceramide as a white solid (0.118 g,
1
MHz) 7.70–7.23 (m, 25H), 5.36–5.25 (m, 5H,
NH, C-4, C-5, C-15%, C-16%), 4.39 (t, 1H, C-3,
J=5.4 Hz), 4.18 (m, 1H, C-2), 3.94 (dd, 1H, C-1,
J=5.1, 10.4 Hz), 3.70 (dd, 1H, C-1, J=5.1, 10.4
Hz), 2.00 (m, 4H, C-14%, C-17%), 1.86 (m, 2H,
C-2%), 1.72 (m, 2H, C-6), 1.44 (m, 2H, C-3%), 1.04
(s, 9H, t-Bu), 0.88 (t, 6H, C-18, C-24%, J=7.3
Hz).
1
91%): Rf (9:1 CHCl3:MeOH) 0.47; H NMR (500
MHz) 6.22 (d, 1H, NH, J=7.1 Hz), 5.73 (m, 1H,
C-5), 5.53 (dd, 1H, C-4, J=6.3, 15.4 Hz), 5.34 (t,
2H, C-15%, C-16%, J=4.6 Hz), 4.30 (t, 1H, C-3,
J=3.8 Hz), 3.96 (dd, 1H, C-1, J=3.1, 11.0 Hz),
3.91 (m, 1H, C-2), 3.71 (dd, 1H, C-1, J=3.1, 11.0
Hz), 2.22 (t, 2H, C-2%, J=7.4 Hz), 2.00 (m, 6H,
C-6, C-14%, C-17%), 1.60 (t, 2H, C-3%, J=7.8 Hz),
0.88 (t, 6H, C-18, C-24%, J=6.3 Hz).
2.2.5. N-ner6onoyl-3-O-[t-butyldiphenylsilyl]
ceramide
Compound 2.2.4. was detritylated according to
Koike et al. (1986): N-nervonoyl-1-0-triphenyl-
methyl-3-O-[t-butyldiphenylsilyl] ceramide (0.093
g, 82.4 vmol) was stirred for 4 h with p-toluene-
sulfonic acid monohydrate (0.010 g, 49.4 vmol) in
20 ml 1:1 MeOH:CH2Cl2. Et2O (40 ml) was
added, and the solution was washed with 10 ml
5% NaHCO3 (aq) and 10 ml H2O. Flash chro-
matography (6:1-0:1 hexane:EtOAc) provided
product as a white solid (0.034 g, 47%): Rf (3:1
2.2.3. N-ner6onoyl-1-0-triphenylmethyl ceramide
N-nervonoyl ceramide (0.018 g, 27.8 mmol),
triphenylmethyl chloride (0.015 g, 55.5 mmol) and
N,N-dimethyl-4-aminopyridine (0.007 g, 55.5
mmol) in 20 ml anhydrous toluene were refluxed
for 16 h under Ar. The solvent was removed by
rotary evaporation and the residue purified by
flash chromatography (9:1-1:1 hexane:EtOAc) to
provide N-nervonoyl-1-0-triphenylmethyl ce-
ramide as a white solid (0.018 g, 72%): Rf (3:1
1
hexane:EtOAc) 0.15; H NMR (500 MHz) 7.67–
7.30 (m, 10H), 5.93 (d, 1H, NH, J=7.1 Hz),
5.42–5.33 (m, 4H, C-4, C-5, C-15%, C-16%), 4.34 (t,
1H, C-3, J=4.5 Hz), 3.97-3.82 (m, 2H, C-1, C-2),
3.60 (m, 1H, C-1), 3.14 (m, 1H, OH), 1.98 (m, 6H,
C-2%, C-14%, C-17%), 1.86 (m, 2H, C-6), 1.55 (m,
2H, C-3%), 1.07 (s, 9H, t-Bu), 0.88 (t, 6H, C-18,
C-24%, J=7.0 Hz).
1
hexane:EtOAc) 0.21; H NMR (300 MHz) 7.42-
7.22 (m, 15H), 6.06 (d, 1H, NH, J=7.9 Hz), 5.63
(m, 1H, C-5), 5.35 (t, 2H, C-15%, C-16%, J=5.2
Hz), 5.25 (dd, 1H, C-4, J=6.2, 15.5 Hz), 4.18 (m,
1H, C-2), 3.69 (dd, 1H, C-3, J=3.9, 7.8 Hz), 3.32
(m, 2H, C-1), 2.20 (t, 2H, C-2%, J=8.1 Hz), 2.00
(m, 4H, C-14%, C-17%), 1.91 (m, 2H, C-6), 1.64 (m,
2H, C-3%), 0.88 (t, 6H, C-18, C-24%, J=6.5 Hz).
2.2.6. N-ner6onoyl-1-O-(N-acetyl-glycine)-3-
O-[t-butyldiphenylsilyl] ceramide
2.2.4. N-ner6onoyl-1-0-triphenylmethyl-3-O-[t-
butyldiphenylsilyl] ceramide
Amino acylation was adapted from the proce-
dure of Neises and Steglich (1985): N-nervonoyl-
3-O-[t-butyldiphenylsilyl] ceramide (0.021 g, 23.7
vmol), N-acetyl-glycine (0.006 g, 47.4 vmol), and
N,N-dimethyl-4-aminopyridine (0.06 g, 47.4
mmol) in 21 ml 2:5 CH3CN:CH2Cl2 (anhydrous)
were stirred for 2 h under Ar. Dicyclohexylcar-
bodiimide (0.010 g, 47.4 mmol) was added and
the reaction stirred for 24 h under Ar. The sol-
vents were removed in vacuo. Flash chromatogra-
phy (5:1-0:1 hexane:EtOAc) of the residue
provided product as a white solid (0.016 g, 70%):
The title compound was prepared according to
Numata et al. (1988): N-nervonoyl-1-O-triphenyl-
methyl ceramide (0.108 g, 0.12 mmol), imidazole
(0.066 g, 0.97 mmol), and t-butylchlorodiphenyl-
silane (0.79 ml, 3.03 mmol) were stirred 19.5 h in
25 ml anhydrous DMF under Ar. Twenty five ml
of H2O was added and the mixture extracted with
Et2O (3×15 ml). The ether extracts were washed
with 10 ml H2O and 10 ml saturated NaCl (aq).
Flash chromatography (15:1-2:1 hexane:EtOAc
and 1 ml triethylamine/100 ml of solvent) pro-

A.S. Goldstein et al. / Chemistry and Physics of Lipids 88 (1997) 21–36
25
1
NH), 6.01 (bs, 1H, NH), 5.76 (dt, 1H, C-5, J=
6.7, 15.5), 5.48 (dd, 1H, C-4, J=6.2, 15.5 Hz),
5.33 (t, 2H, C-15%, C-16%, J=5.0 Hz), 4.33 (d, 2H,
gly), 4.15 (m, 2H, C-2, C-3), 4.00 (m, 2H, C-1),
2.17 (t, 2H, C-2%, J=4.4 Hz), 2.03 (s, 3H, NAc),
0.86 (t, 6H, C-18, C-24%, J=6.6 Hz).
Rf (1:1 hexane:EtOAc) 0.23; H NMR (300 MHz)
7.67–7.57 (dd, 4H), 7.46–7.33 (m, 6H), 6.09 (bs,
1H, NH), 5.51–5.29 (m, 4H, C-4, C-5, C-15%,
C-16%), 4.40 (dd, 1H, C-3, J=2.9, 10.8 Hz), 4.26
(bs, 2H, C-1), 4.12 (m, 1H, C-2), 3.93 (t, 2H,
glycine, J=11.3 Hz), 2.00 (s, 3H, NAc), 1.05 (s,
9H, t-Bu), 0.88 (t, 6H, C-18, C-24%, J=6.4 Hz).
2.2.9. N-ner6onoyl-1-O-(N-acetyl-
ceramide (NAcPro-24:1-Cer)
L-proline)
2.2.7. N-ner6onoyl-1-O-(N-acetyl-l-proline)-
3-O-[t-butyldiphenylsilyl] ceramide
N-nervonoyl-3-O-[t-butyldiphenylsilyl]
mide (0.034 g, 38.4 vmol), N-acetyl-
Desilylation was performed according to
Hanessian and Lavallee (1975): N-nervonoyl-1-O-
(N-acetyl-L-proline)-3-O-[t-butyldiphenylsilyl] ce-
cera-
-proline
L
ramide (0.021 g, 20.5 vmol) in 12 ml anhydrous
THF and 0.01 ml 1.0 M n-butylammonium
fluoride (in THF) were stirred for 2 h under Ar.
The solvent was removed by rotary evaporation
and the residue purified by flash chromatography
(3:1-0:1 hexane:EtOAc) to provide product as a
(0.010 g, 63.6 vmol), and N,N-dimethyl-4-
aminopyridine (0.011 g, 90.0 vmol) in 15 ml 1:2
CH3CN:CH2Cl2 (anhydrous) were stirred for 30
min under Ar. Dicyclohexylcarbodiimide (0.012 g,
57.5 vmol) was added and the reaction stirred for
24 h under Ar. The white precipitate was removed
by vacuum filtration and the solvents evaporated
in vacuo. Flash chromatography (6:1-0:1 hex-
ane:EtOAc) of the residue provided product as a
white solid (0.029 g, 74%): Rf (1:1 hexane:EtOAc)
1
white solid (0.011 g, 69%): Rf (EtOAc) 0.31; H
NMR (500 MHz) 6.66 (d, 1H, NH, J=7.7 Hz),
5.70 (dt, 1H, C-5, J=6.7, 15.5), 5.47 (dd, 1H,
C-4, J=6.2, 15.5 Hz), 5.32 (t, 2H, C-15%, C-16%,
J=4.6 Hz), 4.47-4.26 (m, 4H, h, C-2, C-3), 4.06
(bs, 2H, C-1), 3.64-3.50 (dm, 2H, l), 3.30 (bs, 1H,
OH), 2.18 (m, 2H, i), 2.07 (s, 3H, NAc), 1.99 (m,
10H, C-6, C-2%, C-14%, C-17%, k), 1.59 (m, 2H,
C-3%), 0.86 (t, 6H, C-18, C-24%, J=7.0 Hz).
1
0.29; H NMR (300 MHz) 7.68–7.59 (dd, 4H),
7.43–7.26 (m, 6H), 6.14 (d, 1H, NH, J=8.8 Hz),
5.41–5.29 (m, 3H, C-4, C-15%, C-16%), 5.14 (dt,
1H, C-5, J=4.0, 8.8 Hz), 4.69 (d, 1H, h, J=7.7
Hz), 4.39 (dd, 1H, C-3, J=3.6, 8.1 Hz), 4.27 (d,
2H, C-1, J=12.4 Hz), 4.02 (t, 1H, C-2, J=7.3
Hz), 3.44 (t, 2H, l, J=6.4 Hz), 2.16 (m, 2H, i),
2.02-1.91 (m, 13H, C-6, C-2%, C-14%, C-17%, k,
NAc), 1.49 (m, 2H, C-3%), 1.03 (s, 9H, t-Bu), 0.88
(t, 6H, C-18, C-24%, J=6.6 Hz).
2.2.10. N-tetracos-15(Z)-enyl-sphingosine
(24:1-Amine)
Nervonoyl ceramide (0.015 g, 23.1 vmol) was
dissolved in 12 ml anhydrous Et2O. Lithium alu-
minum hydride (0.021 g, 485.8 vmol) was added,
and the mixture was refluxed for 1.5 h and then
stirred at room temperature for 48 h. Excess
hydride reagent was quenched by addition of 3 ml
EtOAc followed by 1 drop of saturated Na2SO4
(aq). The resultant white precipitate was filtered
and the solution purified by flash chromatography
(1:0-20:1 CHCl3:MeOH) to provide 24:1-Amine
2.2.8. N-ner6onoyl-1-O-(N-acetyl-glycine)
ceramide (NAcGly-24:1-Cer)
Desilylation was performed according to
Hanessian and Lavallee (1975): N-nervonoyl-1-O-
(N-acetyl-glycine)-3-O-[t-butyldiphenylsilyl]
ce-
as
a white solid (0.051 g, 33%): Rf (9:1
ramide (0.009 g, 9.1 vmol) in 10 ml anhydrous
THF and 0.01 ml 1.0 M n-butylammonium
fluoride in THF were stirred for 1 h under Ar.
The solvent was removed by rotary evaporation
and the residue purified by flash chromatography
(2:1-0:1 hexane:EtOAc) to provide compound
product as a white solid (0.002 g, 29%): Rf
1
CHCl3:MeOH) 0.19; H NMR (500 MHz) 5.87
(dt, 1H, C-5, J=6.7, 15.5), 5.45 (dd, 1H, C-4,
J=5.7, 15.5 Hz), 5.35 (t, 2H, C-15%, C-16%, J=4.9
Hz), 4.69 (bs, C-3), 4.05 (dd, 1H, C-1, J=4.6,
12.9 Hz), 3.90 (dd, 1H, C-1, J=4.6, 12.9 Hz),
3.46 (m, 1H, NH), 2.95 (m, 1H, C-2), 2.08-1.99
(m, 6H, C-6, C-14%, C-17%), 0.88 (t, 6H, C-18,
C-24%, J=6.7 Hz).
1
(EtOAc) 0.25; H NMR (500 MHz) 6.11 (bs, 1H,

26
A.S. Goldstein et al. / Chemistry and Physics of Lipids 88 (1997) 21–36
2.2.11. N-acetyl ceramide (2:0-Cer)
2.3.3. Freeze-thaw
Sphingosine (0.008 g, 26.7 vmol) in 10 ml 1:1
CH2Cl2:pyridine and 1 ml acetic anhydride were
stirred for 4.5 h and then evaporated in vacuo.
The residue was dissolved in 10 ml MeOH, and
K2CO3 (0.025 g, 180.8 vmol) was added. After
stirring overnight, the solvent was removed in
vacuo and the residue purified by flash chro-
matography (1:0-20:1 CHCl3:MeOH) to provide
product as a white solid (0.004 g, 44%): Rf (9:1
This procedure follows the method of Kulkarni
et al. (1995): Amphiphile (0.1 mg) was suspended
in 1 ml of aqueous buffer (10 mM KH2PO4, 100
mM NaCl, 1.5 mM NaN3, pH=6.6). The suspen-
sion was thrice incubated for 3 min at 90°C,
vortexed for 20 s at room temperature and then
sonicated (Laboratory Supplies and Company,
Hicksville, NY, output 80 KC) for 20 s at room
temperature. Next, the suspension was placed in
i-PrOH/dry ice for 2 min, thawed rapidly (:20 s)
and then vortexed 20 s. The freeze-thaw proce-
dure was repeated three times except that after the
last freeze the sample was allowed to warm to
room temperature over 1.5 h.
1
CHCl3:MeOH) 0.35; H NMR (300 MHz) 6.32
(d, 1H, NH), 5.78 (dt, 1H, C-5, J=6.7, 15.5), 5.50
(dd, 1H, C-4, J=3.6, 15.5 Hz), 4.35 (bt, 1H, C-3,
J=5.8 Hz), 3.97 (dd, 1H, C-1, J=3.5, 11.2 Hz),
3.93 (m, 1H, C-2), 3.71 (dd, 1H, C-1, J=3.5, 11.2
Hz), 2.30 (t, 3H, C-6, J=9.0 Hz), 2.05 (m, 7H,
NAc, C-14%, C-17%), 0.89 (t, 3H, C-18 J=3.4 Hz).
2.3.4. Thermal cycling
This procedure follows the method of
Archibald and Yager (1992): amphiphile was
placed in ethylene glycol:water (either 19:1) to a
final concentration of 1 mg/ml. The suspension
was thrice incubated for 10 min at 99°C and
sonicated at 50°C (12×30 s pulses with 30 s
pauses). After the final sonication, the suspension
was allowed to cool from 99°C to room tempera-
ture over 2.5 h.
2.3. Microstructure formation studies
Several methods were employed to determine
the tendency of the lipids to form HARMs.
2.3.1. DMF/H2O
Amphiphile (0.1 mg) was dissolved in anhy-
drous DMF so that the concentration was 1.0
mM. Water was added in :10 vl increments
until the solution became cloudy. The resulting
suspensions were incubated at 20°C for 2–24 h.
For larger amounts of amphiphile, water was
added with vortex mixing (:3 s) between addi-
tions.
2.4. Mixed lipid studies
2.4.1. NAcPro-24:1-Cer: 24:1-GalCer
Particles from the mixtures of NAcPro-24:1-Cer
and 24:1-GalCer (1:1, 1:3, 1:5, 1:7 mol:mol) were
prepared as in Section 2.3.1 and incubated for 24
h at 20°C. In order to determine the ratio of the
lipids incorporated in the HARMs, samples were
pelleted (2500 g for 30 min) and a known percent-
age of the supernatants removed. Supernatants
and pellets were dried in vacuo and the residues
weighed (mass corrections were made for the por-
tion of supernatant that remained with the pellet).
2.3.1.1. DMF/saline. To NFA-GalCer dissolved in
DMF (0.8 mg/ml) was added a saturated saline
solution of LiBr, LiCl, NaCl or MgCl2 (50% by
volume) while hand swirling. The resultant white
gel was incubated at 20°C for 16 h.
2.3.2. Pyridine e6aporation
2.4.1.1. Qualitati6e assessment of incorporation.
Supernatant and pellet material (from Section 2.4.1)
was heavily spotted on silica TLC plates (4 cm
length) and developed in EtOAc. The plates were
stained with phosphomolybdic acid reagent (Ald-
rich, Milwaukee, WI) and then heated (NAcPro-
24:1-Cer Rf=0.31 and 24:1-GalCer Rf=0.0).
Amphiphile (0.1 mg) was dissolved in pyridine
so that the concentration was 1.0 mM. Water was
added in :10 vl increments until the solution
became cloudy. The samples were incubated at
20°C so that solvent evaporated and a precipitate
formed (typically 24–48 h).

A.S. Goldstein et al. / Chemistry and Physics of Lipids 88 (1997) 21–36
27
in the formation of cochleate cylinders of dimen-
sions 5–30×100–300 nm, whereas NFA-GalCer
formed tubules and helical ribbons using
pyridine evaporation or thermal cycling
(Archibald and Yager, 1992). Brown and
coworkers, using a freeze-thaw method, found
that NFA-GalCer formed tubules and multil-
amellar liposomes (Kulkarni et al., 1995). All of
these observations were confirmed in the present
study.
Fig. 3. The subfractions of Galactocerebroside.
In the present study, we developed a DMF/
H2O precipitation method for the exploration of
HARM formation. Application of this technique
to NFA-GalCer yields cylindrical structures (Fig.
4a). Submission of HFA-GalCer to aggregate
formation using DMF/H2O provides cylindrical
structures with some helical ribbon content (Fig.
4b). This is in stark contrast to the cochleate
cylinders produced by thermal cycling of the
same amphiphile (Fig. 4c). Depending on the
2.4.1.2. A quantitati6e assessment of incorpora-
tion. The dried and weighed 1:3 doped superna-
tant and pellet (Section 2.4.1) were dissolved in
CD3OD and the amphiphile ratio was deter-
mined by comparison of the integrands of the
NAcPro-24:1-Cer hH (l 4.5) and vinyl NMR
signals.
formation
methodology
the
HFA-GalCer
2.4.2. NAcPro-24:1-Cer : 24:1-Cer
HARMs have different dimensions (Table 1).
This implies that the formation methodology/
solvent can affect HARM morphology.
The procedure was identical to the NAcPro-
24:1-Cer: nervonoyl-GalCer mixed lipid study
(Section 2.4.1). Compositional NMR analysis of
the amphiphile ratio was effected by comparison
of the integrands of NAcPro-24:1-Cer hH (l
4.5) and C-4 vinyl (l 5.4) in CD3OD.
The differences in tubule diameter and length
may be influenced by solvent, electrolytes, lipid
concentration and variations in composition of
NFA-GalCer and HFA-GalCer. The freeze-thaw
method uses 100 mM NaCl whereas the DMF/
H2O method is carried out in the absence of salt.
An early tubule formation theory argues that a
tilted phase of chiral molecules must be electro-
statically polarized which, in turn, can cause a
membrane strip to collapse upon itself and form a
cylinder (de Gennes, 1987). In this theory salts
should increase the tubule radius because electro-
static interactions would be shielded. Along this
line, Archibald noted that the dimensions of
HFA-GalCer cochleate cylinders formed by ther-
mal cycling were larger when formed in the pres-
ence of salt (Archibald and Yager, 1992).
Furthermore, in the better studied HARMs com-
posed of phosphatidylcholine with butadiyne con-
taining hydrocarbon chains, the tubule outer
diameter increased as the salt concentration was
increased in the formation medium, but this is
possibly the result of more bilayers in the tubule
2.4.3. NAcPro-24:1-Cer : NFA-GalCer
Performed as per NAcPro-24:1-Cer: ner-
vonoyl-GalCer mixing studies (Section 2.4.1).
3. Results and discussion
3.1. HFA-GalCer and NFA-GalCer
GalCer can be divided into two major sub-
fractions, that containing non-hydroxy fatty
acids (NFA-GalCer) and that containing more
polar h-hydroxy fatty acids (HFA-GalCer) (Fig.
3). The ability of GalCer to self-organize has
been previously investigated using pyridine evap-
oration, thermal and freeze-thaw cycling tech-
niques to form microstructures. Archibald and
Yager found that submission of HFA-GalCer to
pyridine evaporation or thermal cycling resulted

28
A.S. Goldstein et al. / Chemistry and Physics of Lipids 88 (1997) 21–36
Table 1
Effect of formation methodology/solvent on HARM morphology
Amphiphile
Tm (°C)a Methodb
Structure
Size (nm)c
Reference
NFA-GalCer
69.5
70.5
Pyridine
Thermal
Freeze
Freeze
DMF(35)
Tube/ribbon
Tube/ribbon
Tube/liposome
Cochleate
85×]10 000
85×]10 000
40–100×500–5000
100–150×]1000
25–50
Archibald and Yager (1992)
Archibald and Yager (1992)
Kulkarni et al. (1995)
Kulkarni and Brown (1996)
Tube
HFA-GalCer
24:1-GalCer
Pyridine
Thermal
DMF(35)
Cochleate
Cochleate
Tube/ribbon
5–30×10 000–30 000
5–30×10 000–30 000
40–100×2000–7000
Archibald and Yager (1992)
Archibald and Yager (1992)
66.1
59.3
Freeze
DMF(12)
Tube
Tube
25–30×250–400
20×600
Kulkarni et al. (1995)
22:1-GalCer
20:1-GalCer
18:1-GalCer
Freeze
Freeze
Freeze
Tube
Ribbon
Liposome/sheet/
multilamellar tubes
Tube
25–35
40–50
Kulkarni and Brown (1996)
Kulkarni and Brown (1996)
Kulkarni and Brown (1996)
45.4
84.1
DMF(35)
Freeze
30–60
24:0-GalCer
18:0-GalCer
16:0-GalCer
Ribbon
100–250×2000–10 000 Kulkarni et al. (1995)
81.1
DMF(35)
Ribbon
8–16×1250
H2O
DMF(35)
Tube
Ribbon
100
24
Curatolo and Neuringer (1986)
Psychosine
Thermal
DMF(105) Fiber
Amorphous
Archibald and Yager (1992)
Archibald and Mann (1993)
Archibald and Yager (1992)
40.1
42.3
50×13 000
Sphingosine
MeOH(12) Cochleate
DMF(150) Amorphous
50
Ceramide
24:1-Cer
Thermal
Amorphous
Thermal
Pyridine
DMF(3)
Amorphous
Amorphous
Tube
71.8
48.3
55–140×40 000
18:1-Cer
18:0-Cer
16:0-Cer
6:0-Cer
2:0-Cer
24:1-Amine
DMF(70)
DMF(35)
DMF(35)
DMF(70)
DMF(30)
DMF(20)
Tube
19–24
8–11
11–25
50–75
Ribbon
Ribbon
Tube
Amorphous
Amorphous
51.2
49.3
36.5
NAcPro-24:1-Cer 42.0
Pyridine
Thermal
Freeze
Amorphous
Amorphous
Amorphous
Amorphous
DMF(16)
NAcGly-24:1-Cer
DMF(30)
Tube
110–200
a Tm measurements refer to the major transition.
b Pyridine refers to pyridine evaporation method. Thermal and freeze refer to thermal and freeze-thaw cycling methodologies,
respectively. DMF refers to DMF/H2O precipitation. MeOH refers to MeOH/H2O precipitation. H2O refers to suspension of
lyophilized powder in water. The number in parenthesis refers to the volume percentage of water added.
c If no second measurement is provided then the length was indeterminate.

A.S. Goldstein et al. / Chemistry and Physics of Lipids 88 (1997) 21–36
29
Fig. 4. Negatively stained TEM images of: (a) NFA-GalCer HARMs formed by DMF/H2O precipitation; (b) HFA-GalCer HARMs
formed by DMF/H2O precipitation shows some helical ribbon content in addition to cylindrical structures; (c) HFA-GalCer formed
by thermal cycling.
(Chappell and Yager, 1991). In the present
study, it was found that precipitation of NFA-
GalCer from DMF by addition of saturated sa-
line solutions (LiBr, LiCl, NaCl, MgCl2) rather
than pure water still forms cylindrical structures;
however, the tubules are irregularly shaped but,
on average, the dimensions remain unchanged
(not shown).
The lipid concentration should not perturb
tubule formation as long as the lipid concentra-
tion is above the critical micelle concentration
(CMC). In all cases, cerebroside concentration
was above 12.3 vM. Although not measured,
cerebroside CMC values should be vanishingly
low (less than nanomolar) due to high, long-
chain fatty acid content, thus transfer of lipid to
crystalline HARMs from other fluid lipid struc-
tures, at room temperature, should be pro-
hibitively slow. On the other hand, lipid
concentration may affect tubule wall thickness
(Spector et al., 1996). As shown in studies in-
volving solvent precipitation of phosphatidyl-
choline containing butadiyne hydrocarbon
chains, more concentrated lipid solutions formed
wider tubules due to multiple bilayer wrappings.
Contrary to this, 24:1-GalCer tubules prepared
by DMF/H2O precipitation are narrower than
Brown’s freeze-thaw tubules (described later) de-
spite being formed from a more concentrated
solution (Kulkarni et al., 1995). Lastly, natural
variations in HFA- and NFA-GalCer fatty acid
composition and unsaturation may affect
HARM formation and dimensions.
3.2. Pure galactocerebrosides
To better understand the structural require-
ments of cerebrosides to form HARMs, single
molecular specifies of GalCer-based amphiphiles
were studied. Using DMF/H2O precipitation,
pure 24:1-GalCer forms cylindrical nanostruc-
tures of dimensions 20×600 nm as evidenced
by negative stained TEM (Fig. 5a, Table 1). Us-
ing freeze-thaw cycling, Brown found that 24:1-
GalCer formed tubules, but these are shorter
and wider than those formed by the DMF/H2O
method (Table 1) (Kulkarni et al., 1995).

30
A.S. Goldstein et al. / Chemistry and Physics of Lipids 88 (1997) 21–36
Replacement of the mono-unsaturated, 24-car-
Neuringer, 1986). In general, the widths of DMF/
H2O precipitated ribbons are narrower than
Brown’s freeze-thawed 24:0-GalCer ribbons. The
lengths were also many microns but intertwining
of the fibers made this difficult to accurately
determine.
bon fatty acid chain with the shorter mono-unsat-
urated oleoyl chain (18:1-GalCer) results in an
amphiphile that also forms cylindrical structures
using DMF/H2O precipitation (Fig. 5b, Table 1).
In contrast, Brown found that 18:1-GalCer
formed multilamellar cylindrical bilayers, sheets,
and liposomes when prepared by freeze-thaw cy-
cling (Kulkarni and Brown, 1996). Brown also
studied 22:1-GalCer which formed cylindrical mi-
crostructures whereas 20:1-GalCer formed helical
ribbons (Kulkarni and Brown, 1996).
The cerebroside lignoceroyl-GalCer (24:0-Gal-
Cer), which possesses the long saturated 24:0
chain formed a mixture of ribbons and liposomes
using freeze-thaw cycling (Kulkarni et al., 1995).
In the present study we found that stearoyl-Gal-
Cer (18:0-GalCer) and palmitoyl-GalCer (16:0-
GalCer) also form twisted ribbons using the
DMF/H2O method (Fig. 5c–d, Table 1). In con-
trast, Curatolo reported that lyophilized 16:0-Gal-
Cer formed 100 nm diameter multilamellar
tubules when suspended in water (Curatolo and
The unsaturated fatty acyl containing GalCer,
24:1-GalCer and 18:1-GalCer, as well as NFA-
GalCer and HFA-GalCer HARMs are suspected
to be hollow based on negative stained TEM. A
negatively stained TEM of a lipid HARM should
appear uniformly dark if the object is solid
whereas two parallel white lines would be preva-
lent against a dark background if the object is
hollow. Brown’s freeze-fracture electron mi-
croscopy on freeze-thaw prepared 24:1-GalCer
which is of similar size to that of DMF/H2O
precipitated 24:1-GalCer was unable to unam-
biguously determine if the tubules were hollow or
filled. The observation that some HFA-GalCer
tubules terminate in a helical ribbon (upper left of
Fig. 4b) supports the idea that these structures
may be hollow.
Collectively, these results suggest that a mono-
unsaturated fatty acyl chain GalCer is required
for cylindrical tubule formation. Furthermore, the
unsaturation requirement is not alleviated by
shortening the acyl chain. Perhaps the cis double
bond kink in 24:1-GalCer and 18:1-GalCer fatty
acids relaxes the rigidity of the system allowing
tubule formation as opposed to ribbons (reduces
curvature frustration).
3.3. Pure ceramides
Despite earlier reports that ceramide containing
a mixture of fatty acid chains was unable to form
HARMs by thermal cycling, (Archibald and
Yager, 1992) we are able to show that pure single
molecular species of ceramide form HARMs us-
ing DMF/H2O precipitation. Nervonoyl-Cer
(24:1-Cer) forms cylindrical HARMs when viewed
by TEM (Fig. 6a, Table 1). The density appears
uniform across the diameter of the fibers suggest-
ing that solid or cochleate cylinders are being
formed. This is in contrast to the cylindrical
nanostructures formed from 24:1-GalCer pre-
pared under identical conditions (compare Fig. 5a
Fig. 5. Negatively stained, DMF/H2O precipitated TEM im-
ages of: (a) 24:1-GalCer: (b) 18:1-GalCer: (c) 18:0-GalCer: (d)
16:0-GalCer. The unsaturated sphingolipids form cylindrical
structures whereas the saturated lipids form ribbons.

A.S. Goldstein et al. / Chemistry and Physics of Lipids 88 (1997) 21–36
31
Fig. 6. TEM images of DMF/H2O precipitated: (a) unstained 24:1-Cer; (b) negatively stained 18:1-Cer; (c) negatively stained
18:0-Cer; (d) negatively stained 16:0-Cer; (e) unstained 6:0-Cer.
and Fig. 6a). Similar dense HARMs are formed
from pure oleoyl-Cer (18:1-Cer, Fig. 6b, Table 1).
On the other hand, long saturated fatty acid
chain-containing ceramides, stearoyl-Cer (18:0-
Cer) and palmitoyl-Cer (16:0-Cer) form twisted
ribbons (Fig. 6(c and d), Table 1). The short chain
hexanoyl-Cer (6:0-Cer) forms cylindrical struc-
tures (Fig. 6e, Table 1), and acetyl-Cer (2:0-Cer)
did not form any discernible HARMs (not
shown).
These results imply that the hydrophilic galac-
tose headgroup is not required for HARM forma-
tion, although the presence of the carbohydrate
influences HARM morphology. Furthermore, the
ceramide study supports the notion that the pres-
ence of saturated fatty acid chains favors twisted
ribbons whereas the presence of mono-unsatu-
rated chains favors cylindrical HARMs (except
for 6:0-Cer). In the case of 6:0-Cer, the fatty acid
side chain is much shorter than the sphingosine
hydrocarbon chain. This high degree of chain
mismatch may cause the headgroup to have a
larger tilt relative to the hydrophobic alkyl region
which may cause a significant packing perturba-
tion.
3.4. Amide importance
The importance of an amide versus a secondary
amine in the ceramide structure for HARM for-
mation was studied by preparing the secondary
amine formed by hydride reduction of the amide
in 24:1-Cer (24:1-Amine). DMF/H2O precipita-
tion of 24:1-amine produced amorphous struc-
tures with no discernible HARMs (not shown).
Amphiphiles that lack the fatty acid chain were
also studied. Archibald and Mann reported that
sphingosine formed cochleate cylinders by precipi-
tation from water by addition of methanol
(Archibald and Mann, 1993). In the present

32
A.S. Goldstein et al. / Chemistry and Physics of Lipids 88 (1997) 21–36
study, DMF/H2O precipitation of sphingosine
fails to form any HARMs. On the other hand,
psychosine (GalCer lacking the fatty acid chain)
forms long thin HARMs when submitted to
DMF/H2O treatment (Fig. 7, Table 1). In con-
trast, thermal cycling of aqueous dispersions of
psychosine did not lead to HARMs (Archibald
and Yager, 1992). Collectively, these results sug-
gest that two-chain sphingolipids require the pres-
ence of an amide carbonyl to form HARMs;
however, a single-chain sphingolipid does not re-
quire the carbonyl for HARM formation.
3.5. Amino acid headgroups
One of the long term goals of these studies is to
fabricate HARMs composed of lipidated drugs,
such as lipidated peptides, for novel drug delivery
strategies. In this context, we examined cere-
broside-based amphiphiles that contain amino
acids in place of the polar galactose headgroup.
Modeling studies using proline in place of galac-
tose suggested that microstructure formation
would not be perturbed. Proline, like galactose, is
cyclic, has hydrogen bond donors/acceptors and is
of comparable size. Surprisingly, NAcPro-24:1-
Cer did not form any discrete HARMs using
Fig. 8. TEM image of negatively stained, DMF/H2O precipi-
tated NAcGly-24:1-Cer.
DMF/H2O precipitation, thermal or freeze-thaw
cycling, or pyridine evaporation (not shown).
NAcGly-24:1-Cer, containing
a smaller head
group, forms long thin fibers (Fig. 8). Clearly the
structure of the head group greatly influences
HARM morphology.
3.6. Mixed lipid systems
Despite the inability of NAcPro-24:1-Cer to
independently form discrete structures, we tried
doping NAcPro-24:1-Cer into HARMs formed
from 24:1-Cer, 24:1-GalCer, or NFA-GalCer (ma-
trix). An intimate mixture of NAcPro-24:1-Cer
with the HARM-forming amphiphile was first
made using DMF. Water was then added to cause
precipitation (Section 2). TEM images of the
structures are similar to those obtained from
DMF/H2O precipitation of the pure matrix am-
phiphiles. The HARMs were isolated and submit-
ted to compositional analysis by NMR (Section
2). NAcPro-24:1-Cer could be doped into either
24:1-GalCer or NFA-GalCer HARMs up to 25
mole% without affecting tubule formation. On
average, the tubules were very long (micrometers)
Fig. 7. TEM image of unstained, DMF/H2O precipitated
psychosine.

A.S. Goldstein et al. / Chemistry and Physics of Lipids 88 (1997) 21–36
33
Fig. 9. TEM images of negatively stained, DMF/H2O precipitated: (a) 1:3 NAcPro-24:1-Cer : 24:1-GalCer mixed lipid system: (b)
1:3 NAcPro-24:1-Cer : NFA-GalCer mixed lipid system.
with diameters of 13-20 nm (Fig. 9, Table 1).
Increasing the NAcPro-24:1-Cer component be-
yond 25% resulted in the formation of vesicles
in addition to tubules (not shown).
component alone. As evident, the doped systems
exhibit two transitions near the Tm of pure
lipids implying non-ideal mixing. The presence
of small transitions before and after the main
NFA-GalCer transition (52 and 75°C) may be
due to the presence of previously reported
metastable states (Maggio et al., 1985a,b; Haas
and Shipley, 1995; Curatolo, 1982). In pure
NFA-GalCer, these metastable states are related
to the hydration state of the sphingolipid. Fur-
thermore, the magnitude and onset temperature
of these transitions are dependent on heating
and cooling rates.
NMR analysis of the tubules formed from the
1:3 NAcPro-24:1-Cer
:
24:1-GalCer or 1:3
NAcPro-24:1-Cer NFA-GalCer mixed lipid
:
systems retain the original solution composition.
In the former case, differential scanning
calorimetery was employed to determine if the
lipids were ideally mixed (Fig. 10). Intimately
mixed binary systems should show a single
exothermic peak between the Tm of either pure

34
A.S. Goldstein et al. / Chemistry and Physics of Lipids 88 (1997) 21–36
Since differential scanning calorimetry indi-
cated non-ideal mixing, a second set of experi-
ments with NAcPro-24:1-Cer and NFA-GalCer
was carried out to help determine the spatial
relationship of the two lipids. Three mixing possi-
bilities are: (1) both lipids coexist in the HARM
as an intimate mixture; (2) both lipids coexist in
the HARM but one lipid coats the other lipid’s
microstructure as it precipitates from solution; or
(3) both lipids independently precipitate from so-
lution. NFA-GalCer and NAcPro-24:1-Cer mi-
crostructures were independently formed by
DMF/H2O precipitation (Fig. 11). The cere-
broside, when visualized by optical microscopy,
appear as thread-like structures with occasional
‘crystalline’ patches whereas NAcPro-24 : 1-Cer
forms only spherical aggregates. The indepen-
dently precipitated species were then combined so
that their concentration was identical to that of
the 1:3 NAcPro-24:1-Cer : NFA-GalCer mixed
lipid system. Optical micrographs of the pre-
formed then mixed lipids show a mixture of
threads with spherical aggregates, whereas inti-
mately mixed, then precipitated lipids, show only
threads. Thus possibility (3) is eliminated.
Fig. 11. Optical micrographs of DMF/H2O precipitated: (a)
NAcPro-24:1-Cer; (b) NFA-GalCer; (c) NAcPro-24:1-Cer:
NFA-GalCer aggregates independently formed then mixed; (d)
NAcPro-24:1-Cer : NFA-GalCer mixed and then precipitated.
Solubility arguments may be able to distinguish
between coating and coprecipitation. A more hy-
drophobic lipid should precipitate first upon wa-
ter addition to an organic solution of mixed
lipids. Although not measured, NAcPro-24:1-Cer
should be more hydrophobic than NFA-GalCer;
therefore, the amino acylated lipid should fall out
of solution and form circular aggregates which
would then be coated by NFA-GalCer. However,
this is clearly not the case since only thread-like
HARMs were seen in the mixed lipid precipita-
tion.
In contrast, HARMs formed from NAcPro-
24:1-Cer mixed into 24:1-Cer exclude the amino
acid-containing amphiphile as revealed by NMR
analysis and thin layer chromatography. Further-
more, optical microscopy showed that mixtures
containing NAcPro-24:1-Cer in excess of 17
mole% also contained liposomes in addition to
HARMs (not shown). Although it is known that
sphingolipids can partition into fluid membranes
to some degree, the precipitated HARMs are crys-
talline (Merrill et al., 1991; Maggio et al.,
1985a,b). Furthermore, exclusion from the nanos-
tructures during precipitation may be caused by
significant differences in the amphiphile’s aqueous
solubility.
Fig. 10. Differential scanning calorimetery heating curves of
NAcPro-24:1-Cer : NFA-GalCer mixed lipid systems showing
that the lipids are not ideally mixed.

A.S. Goldstein et al. / Chemistry and Physics of Lipids 88 (1997) 21–36
35
Bielawska, A., Greenberg, M., Perry, D., Jayadev, S., Shay-
man, J., McKay, C., Hannun, Y., 1996. (1S,2R)- -erythro-
4. Conclusion
D
2-(N-myristoylamino)-1-phenyl-1-propanol as an inhibitor
of ceramidase. J. Biol. Chem. 271, 12646–12654.
Chappell, J., Yager, P., 1991. Electrolyte effects on bilayer
tubule formation by a diacetylenic phospholipid. Biophys.
J. 60, 952–956.
Chappell, J.S., Yager, P., 1992. Formation of mineral mi-
crostructures with a high aspect ratio from phospholipid
bilayer tubules. J. Mat. Sci. Lett. 11, 633–636.
Curatolo, W., 1982. Thermal behavior of fractionated and
unfractionated bovine brain cerebrosides. Biochemistry 21,
1761–1764.
Curatolo, W., Neuringer, L., 1986. The effects of cerebrosides
on model membrane shape. J. Biol. Chem. 261, 17177–
17182.
de Gennes, P.-G., 1987. Electrostatic buckling of chiral lipid
bilayers. C.R. Acad. Sci. Paris 304, 259.
Fuhrhop, J., Schnieder, P., Boekema, E., Helfrich, W., 1988.
Lipid bilayer fibers from diastereomeric and enantiomeric
N-octylaldonamides. J. Am. Chem. Soc. 107, 510–512.
Georger, J., Price, R., Singh, A., Schnur, J.M., Schoen, P.E.,
Yager, P., 1987. Helical and tubular microstructures
formed by polymerizable phosphatidylcholines. J. Am.
Chem. Soc. 109, 6169–6175.
The demonstrated fact that ceramides with
amino acid headgroups can independently or as a
component of a binary mixture form HARMs is
encouraging from the drug delivery point of view.
It may be possible to attach peptides of therapeu-
tic interest to the ceramide and still form
HARMs. Due to tight crystal packing, such ag-
gregates may be protected from proteolysis and
provide a means for constant drug release via
dissolution only from the ends of the HARMs. In
addition, such microstructures are unlikely to be
rapidly cleared from the body due to their
lipophilicity, and thus may be immobilized at the
desired site of action. This potentially long in vivo
lifetime may make such a delivery system an ideal
way to induce a strong immune response, and this
may be useful for vaccine development. In mixed
lipid systems, the inexpensive mixed NFA-GalCer
may be used in place of the more costly single
molecular galactocerebroside species. A final ad-
vantage of using such a natural tubule-forming
lipid is that the human body already has enzymes
that can metabolize the drug-stripped lipid core.
Haas, N., Shipley, G., 1995. Structure and properties of
N-palmitoleoylgalactosylsphingosine
(cerebroside).
Biochim. Biophys. Acta 1240, 133–141.
Hanessian, S., Lavallee, P., 1975. The Preparation and Syn-
thetic Utility of tert-Butyldiphenylsilyl Ethers. Can. J.
Chem. 53, 2975–2977.
Koike, K., Numata, M., Sugimoto, M., Nakahara, Y., Ogawa,
T., 1986. A Highly Stereoselective Synthesis of 2(S), 3(R),
4E-and 2(S), 3(R), 4Z-N-Tetracosanoylsphingenine from
Acknowledgements
D
-Glucose. Carbohydr. Res. 158, 113–123.
The authors wish to express their thanks to Ms.
K. Lee for helpful discussions. This work was
supported by a grant from the Whitaker Founda-
tion.
Kulkarni, V.S., Anderson, W.H., Brown, R.E., 1995. Bilayer
nanotubes and helical ribbons formed by hydrated galacto-
sylceramides: Acyl chain and headgroup effects. Biophys.
J. 69, 1976–1986.
Kulkarni, V.S., Brown, R.E., 1996. Modulation of tubular
microstructural self-assembly in galactosylceramides: Influ-
ence of N-linked fatty acyl chains. Proc. Microsc. Mi-
croanal., 936–937.
References
Lapidot, Y., Rappoport, S., Wolman, Y., 1967. Use of esters
of N-hydroxysuccinimide in the synthesis of N-acylamino
acids. J. Lipid Res. 8, 142–145.
Archibald, D.D., Mann, S., 1993. Structural studies of lipid
fibers formed by sphingosine. Biochim. Biophys. Acta
1166, 154–162.
Maggio, B., Ariga, T., Sturtevant, J., Yu, K., 1985a. Ther-
motropic behavior of glycosphingolipids in aqueous disper-
sions. Biochemistry 24, 1084–1092.
Archibald, D.D., Mann, S., 1994. Self-assembled microstruc-
tures from 1,2-ethanediol suspensions of pure and binary
mixtures of neutral and acidic biological galactosylce-
ramides. Chem. Phys. Lipids 69, 51–64.
Archibald, D.D., Yager, P., 1992. Microstructural polymor-
phism in bovine brain galactocerebroside and its two major
subfractions. Biochemistry 31, 9045–9055.
Behroozi, F., Orman, M., Reese, R., Stockton, W., Calvert, J.,
Rachford, F., Schoen, P., 1990. Interaction of metallized
tubules with electromagnetic radiation. J. Appl. Phys. 68
(7), 3688–3693.
Maggio, B., Ariga, T., Sturtevant, J., Yu, K., 1985b. Ther-
motropic behavior of binary mixtures of dipalmitoylphos-
phatidyl choline and glycosphingolipids in aqueous
dispersions. Biochim. Biophys. Acta 818, 1–12.
Mann, S., Archibald, D.D., Didymus, J.M., Douglas, T.,
Heywood, B.R., Meldrum, F.C., Reeves, N.J., 1993. Crys-
tallization at inorganic-organic interfaces: Biominerals and
biomimetic synthesis. Science 261, 1286–1292.

36
A.S. Goldstein et al. / Chemistry and Physics of Lipids 88 (1997) 21–36
Merrill, A., Wang, E., Reddy, R., Menaldino, D., Geisler, V.,
Liotta, D., 1991. Federation of American Societies for
Experimental Biology, 75th Annual Meeting, Atlanta,
Georgia (April 21–25, 1991) Part II, Abs. 4032.
Naito, M., Takahashi, K., Hojo, H., 1988. An ultrastructural
and experimental study on the development of tubular
structures in the lysosomes of gaucher cells. Lab. Invest.
58, 590–598.
Nakashima, N., Asakuma, S., Kim, J., Kunitake, T., 1984.
Helical superstructures are formed from chiral ammonium
bilayers. Chem. Lett., 1709–1712.
Nandi, N., Bagchi, B., 1996. Molecular origin of the intrinsic
bending force for helical morphology observed in chiral
amphiphilic assemblies: Concentration and size depen-
dence. J. Am. Chem. Soc. 118, 11208–11216.
Neises, B., Steglich, W., 1985. Esterification of carboxylic
acids with Dicyclohexylcarbodiimide/4-Dimethylaminopy-
ridine: tert-Butyl Ethyl Fumarate. In: Saucy, G. (ed.),
Organic Syntheses, vol. 63. Wiley, New York, pp. 183-187.
Numata, M., Sugimoto, M., Shibayama, S., Ogawa, T., 1988.
A total synthesis of Hematoside, h-NeuGc-(2“3)-i-Gal-
(1“4)-i-Glc-(1“1)-Cer. Carbohydr. Res. 174, 73–85.
Ong, D., Brady, R., 1972. Synthesis of ceramides using N-hy-
droxysuccinimide esters. J. Lipid Res. 13, 819–822.
Rudolph, A.S., Calvert, J.M., Schoen, P.E., Schnur, J.M.,
1988. In: Gaber, B.P., Schnur, J.M., Chapman, D. (eds.),
Biotechnological Applications of Lipid Microstructures,
Plenum, New York, pp. 305–320.
Schnur, J.M., Price, R., Rudolph, A.S., 1994. Biologically
engineered microstructures: controlled release applications.
J. Control. Release 28, 3–13.
Schnur, J.M., Price, R., Schoen, P., Yager, P., Calvert, M.,
Georger, J., Singh, A., 1987. Lipid-based tubule mi-
crostructures. Thin Solid Films 152, 181–206.
Schnur, J.M., Schoen, P.E., Yager, P., Calvert, J.M., Georger,
J.H., Price, R., 1990. US Patent 4911981.
Selinger, J., MacKintosh, F., Schnur, J., 1996. Theory of
cylindrical tubules and helical ribbons of chiral lipid mem-
branes. Phys. Rev. E 53, 3804–3818.
Shimizu, T., Hato, M., 1993. Self-assembling properties of
synthetic peptidic lipids. Biochim. Biophys. Acta 1147,
50–58.
Spector, M., Easwaran, K., Jyothi, G., Selinger, J., Singh, A.,
Schnur, J., 1996. Chiral molecular self-assembly of phos-
pholipid tubules: A circular dichroism study. Proc. Natl.
Acad. Sci. USA 93, 12943–12946.
Yager, P., Schoen, P.E., 1984. Formation of tubules by a
polymerizable surfactant. Mol. Cryst. Liq. Cryst. 106,
371–381.
Yamada, K., Ihara, H., Ide, T., Fukumoto, T., Hirayama, C.,
1984. Formation of helical super structure from single-
walled bilayers by amphiphiles with oligo-
head group. Chem. Lett. 10, 1713–1716.
L-glutamic acid-
Yunis, E.J., Lee, R.E., 1970. Tubules of globoid leukodystro-
phy: a right handed helix. Science 169, 64–66.
.
.