Nuclear magnetic resonance of lipid A - The influence of solvents on spin relaxation and spectral quality

Article abstract of DOI:10.1016/S0009-3084(03)00055-0

Nuclear magnetic resonance (NMR) spectroscopy of lipid A is limited by rapid transversal relaxation and subsequent line broadening caused by the tendency of these glycolipids to form aggregates in all solvents. To examine the influence of solvents on NMR spectra, hexa-acyl lipid A from Escherichia coli F515 was investigated. Line widths at half height, longitudinal relaxation times, and transversal relaxation times were measured in different solvents, lipid A concentrations, and temperatures. Chloroform-d, dioxane-d₈, and pyridine-d₅ each mixed with 25% methanol-d₄ as well as sole DMSO-d₆ and 0.1M triethylamine-d₁₅ (TEA-d ₁₅) in D₂O caused good spectral resolutions and allowed structure analysis. ROESY and HMBC spectra gave an insight into the influence of transversal relaxation times on spectral quality in two-dimensional spectra. Solvent depending differences of interglycosidic NOEs indicated dissimilarities of the conformations in the interglycosidic linkage and allowed conclusions about the lipid A solution state.

Full text of DOI:10.1016/S0009-3084(03)00055-0

Chemistry and Physics of Lipids 125 (2003) 27–39
Nuclear magnetic resonance of lipid A—the influence
of solvents on spin relaxation and spectral quality
Lothar Brecker∗
Research Center Borstel, Center for Medicine and Biosciences, D-23845 Borstel, Germany
Received 28 October 2002; received in revised form 16 April 2003; accepted 23 April 2003
Abstract
Nuclear magnetic resonance (NMR) spectroscopy of lipid A is limited by rapid transversal relaxation and subsequent line
broadening caused by the tendency of these glycolipids to form aggregates in all solvents. To examine the influence of sol-
vents on NMR spectra, hexa-acyl lipid A from Escherichia coli F515 was investigated. Line widths at half height, longitudinal
relaxation times, and transversal relaxation times were measured in different solvents, lipid A concentrations, and temper-
atures. Chloroform-d, dioxane-d8, and pyridine-d5 each mixed with 25% methanol-d4 as well as sole DMSO-d6 and 0.1 M
triethylamine-d15 (TEA-d15) in D2O caused good spectral resolutions and allowed structure analysis. ROESY and HMBC spec-
tra gave an insight into the influence of transversal relaxation times on spectral quality in two-dimensional spectra. Solvent
depending differences of interglycosidic NOEs indicated dissimilarities of the conformations in the interglycosidic linkage and
allowed conclusions about the lipid A solution state.
© 2003 Elsevier Science Ireland Ltd. All rights reserved.
Keywords: NMR; Lipid A; Deuterated solvents; Spin relaxation; Spectral resolution
1. Introduction
Whitfield, 2002). Slight variations in the lipid A acy-
lation pattern, however, cause extensive differences
of the effect. Some isolated lipid A show endotoxic
activities when tested in human mononuclear cells,
whereas others have LPS antagonistic activities or
rather no effects (Zähringer et al., 1994; Alexander
and Rietschel, 2001). These different stimulations are
supposed to be caused by varying lipid A intrinsic
conformations (Seydel et al., 2000). The actual focus
of interest is the mechanism of lipid A binding to
TLR4, which likely involves other proteins like CD14
and MD2 (Raetz and Whitfield, 2002; Henneke and
Golenbock, 2002). To elucidate this molecular binding
mechanism, a systematic knowledge about configura-
tion, conformation, and aggregation state of different
Lipopolysaccharides (LPS, endotoxins), located in
the outer membrane of Gram-negative bacteria, are
potent stimulators of the mammalian immune system.
Lipid A, which constitutes its lipid component, is the
endotoxically active moiety of LPS. It has recently
been shown that lipid A activation of the Toll-like
receptor 4 (TLR4) induces mammalian endotoxin
signaling (Alexander and Rietschel, 2001; Raetz and
∗
Present address: University of Vienna, Organic Chemistry,
Währinger Straße 38, A-1090 Wien, Austria.
Tel.: +43-1-4277-52131; fax: +43-1-4277-9521.
E-mail address: lothar.brecker@univie.ac.at (L. Brecker).
0009-3084/$ – see front matter © 2003 Elsevier Science Ireland Ltd. All rights reserved.
doi:10.1016/S0009-3084(03)00055-0

28
L. Brecker / Chemistry and Physics of Lipids 125 (2003) 27–39
lipid A from several bacterial strains is necessary. To
investigate such structural features nuclear magnetic
resonance (NMR) spectroscopy is a powerful analyt-
ical method. The common NMR-methods for such
lipid A structure analysis were described (Agrawal
et al., 1994) and most recent advances in this field
have been reviewed (Zähringer et al., 1999). However,
strong tendency of the amphiphilic lipid A to form ag-
gregates restricts the NMR spectroscopic resolutions.
Three approaches were used to circumvent this
problem. Aqueous solutions containing sodium do-
decyl sulfate were utilized to minimize the micellar
size, which, however, only caused minor improve-
ments of the spectral resolution (Strain et al., 1983a,b;
Takayama et al., 1983a,b; Oikawa et al., 2000).
Therefore in a second approach lipid A was deriva-
tized to yield more soluble products, which allowed
an indirect determination of the lipid A configura-
tion (Takayama et al., 1983a,b; Oikawa et al., 2000).
In the third approach intact lipid A was dissolved
in mixtures of organic solvents like chloroform-d,
methanol-d4, DMSO-d6, or pyridine-d5 leading to
a reasonable line resolution, which made structure
determination possible (Baltzer and Mattsby-Baltzer,
1986; Wang and Hollingsworth, 1995; Zähringer
et al., 1999, 2001; Ribeiro et al., 1999). However,
no systematic correlation between the used solvent
mixtures and the spectral resolution was determined.
Therefore in this paper several solvents are system-
atically investigated with respect to numerous NMR
parameters and conditions for an optimal line resolu-
tion of lipid A spectra were determined. Furthermore,
variation of lipid A configuration in some solvents are
investigated and an explanation for unreliable signal
intensity in HMBC as reported by Ribeiro et al. (1999)
is elaborated.
petroleum ether method and the last precipitate was
isolated (Galanos et al., 1969). Crude lipid A was ob-
tained from 250 mg LPS batches by hydrolysis (0.1 M
acetate, pH 4.4, 100 ◦C, 1 h). The hydrolysate was
cooled on ice, adjusted to pH 8.5 with aqueous triethy-
lamine (TEA), dialyzed, and freeze-dried. Hexa-acyl
lipid A was obtained from fractions of 20.0 mg crude
lipid A by preparative layer chromatography on silica
gel plates (2.0 mm, 20 cm × 20 cm, silica gel 60F254
,
Merck) with chloroform/methanol/water 20:15:3
(v/v/v). The separated hexa-acyl lipid A fraction was
resuspended in 2.0 ml distilled water, adjusted to pH
8.5 by addition of 0.36 M TEA, and dialyzed six
times for 12 h against distilled water. The retentate
was lyophilized to give purified hexa-acyl lipid A.
2.2. Sample preparation for NMR spectroscopy
Purified hexa-acyl lipid A (60.0 mg) was sus-
pended in 30 ml distilled water at 0 ◦C, dissolved by
addition of 0.36 M aqueous TEA, and precipitated
adding 0.1 M HCl until pH dropped to ∼2. After cen-
trifugation (2500 × g, 15 min, 4 ◦C) the residue was
dissolved in 15 ml chloroform/methanol 4:1 (v/v) and
washed five times with distilled water. The solvent
was removed in vacuo and the residue was dried over
P4O10. The fully protonated lipid A (2.0 mg each)
was dissolved in 0.5 ml of the deuterated solvent and
transferred into 5 mm high precision NMR sample
tubes (Promochem). Triethylamine-d15 (TEA-d15) or
α,α,α-tris-(hydoxymethyl)-methylamine-d8 (Tris-d8)
was added to the D2O solution and the pH was ad-
justed adding a few microliter 1.0 M NaOD or 1.0 M
DCl in D2O, respectively. During addition the pH
value of the solution was controlled and did not
change to inadequate high or low pH values. The to-
tal volume of the sample was only slightly increased.
All samples were not degassed, but treated in an ultra
sonic bath (Bandolin Sanorex RX100) for 15–30 min
directly before measurement.
2. Materials and methods
2.1. Growth of bacteria, LPS extraction, lipid A
preparation, and purification
2.3. NMR spectroscopy
The E. coli F515 strain was grown aerobically in
a casein peptone medium using standard conditions
and the bacteria were killed, washed, and dried as
described earlier (Schmidt et al., 1970; Schlecht,
1975). LPS was extracted by the phenol–chloroform–
All spectra were recorded on an AVANCE DRX
600 spectrometer (Bruker) using the XWINNMR 2.6
software. Irradiation and measurement frequencies
1
were 600.1 MHz for H and 125.1 MHz for 13C and
the temperature was adjusted between 280 0.05 K

L. Brecker / Chemistry and Physics of Lipids 125 (2003) 27–39
29
and 330 0.05 K. Proton spectra were recorded with
1H-pulse angle of 90◦ and an acquisition of 16k
cols and its primary structure had been unequivocally
determined by analysis and synthesis (Imoto et al.,
1985; Wang and Hollingsworth, 1995). Therefore it
was used as model compound in this study. Its struc-
ture is shown in Scheme 1 where several atoms are
indicated to assign their signals in the NMR spectra.
For line form and relaxation analysis the protons H-1,
H-1ꢀ and H-d2, and all methyl group protons in the
fatty acids were investigated, as they represent the hy-
drophilic and hydrophobic regions in the molecule,
respectively.
a
data points. After zero filling to 32k data points
the free induction decay was Fourier transformed
to spectra with a spectral range of 6000 Hz. For the
determination of spectral resolution the line width
at half height was determined for each single spin
isochromat in the investigated multiplets. The aver-
age spin isochromat line width of half height was
calculated for each multiplet. Longitudinal relaxation
time constants (T1) were determined by the inver-
sion recovery method (Vold et al., 1968) and the
transversal relaxation times (T2) were measured with
the spin-echo sequence (CPMG; Meiboom and Gill,
1958). All two-dimensional homonuclear spectra were
recorded with 2k data points in the F2-dimension
and 256–512 experiments in F1-dimension (Bax and
Davis, 1985; Neuhaus and Williamson, 1989; Hurd,
1990). For TOCSY and ROESY mixing times of
100 and 225 ms were used, respectively (Bax and
Davis, 1985; Neuhaus and Williamson, 1989). Af-
ter zero filling to 512 data-points in F2-dimension
and appropriate sinusoidal multiplication in both
dimensions a Fourier transformation led to spectra
with a range of 4800 Hz in both dimensions. The
heteronuclear two-dimensional 1H/13C-HMQC and
1H/13C-HMBC spectra were measured with 2k data
points in the F2-dimension and 128–384 experiments
in F1-dimension (Willker et al., 1993). The heteronu-
clear spectra were processed and transformed like the
homonuclear spectra and had a final spectral range of
30,000 Hz in the F1-dimension (13C) and 4800 Hz in
the F2-dimension (1H).
3.1. Solitary organic solvents
From the variety of deuterated organic solvents
only chloroform-d and DMSO-d6 had been used as
single solvents for NMR measurements of lipid A
(Wang and Hollingsworth, 1995; Zähringer et al.,
2001). Whereas chloroform-d caused very poor line
resolution, the spectra in DMSO-d6 showed rea-
sonable line width at half height and good signal
separation.
In this study it was found that DMSO-d6 at 330 K
was the best single solvent for lipid A NMR spectra,
causing line widths at half height of ∼3.5 Hz of sin-
gle isochromats and transversal relaxation (T2) times
of ∼0.2 s (Fig. 1, Table 1a). Traces of water in the
solution, however, caused a broad signal at ∼3.0 ppm
and the DMSO signal at 2.49 ppm overlapped the
H-d2 signal (Fig. 1). Spectra of lipid A in DMF-d7 at
330 K showed slightly poorer resolution and shorter
transversal relaxation times. The water now res-
onated at 3.5 ppm and two solvent signals overlapped
some signals of the lipid A spectrum. Lipid A spec-
tra recorded in dioxane-d8 and pyridine-d5 at 300 K
showed similar line resolution as in DMF-d7 at 330 K.
In dioxane-d8 the spectra were disturbed by water and
dioxane signals in the region of carbohydrate proton
signals, whereas in pyridine-d5 the pyridine signals
did not overlap any lipid A signal. In pyridine-d5
traces of OD−-ions caused slight saponification of
esters and amides, which was detected by TLC after
48 h. The detected line widths at half height as well as
the transversal and longitudinal relaxation times of the
investigated lipid A protons in all these single solvents
are summarized in Table 1a. In agreement with ear-
lier reported results (Wang and Hollingsworth, 1995)
spectra taken in chloroform-d led to much poorer
The following internal standards for 1H and
13C were used: chloroform-d (chloroform, 1H:
7.24 ppm, 13C: 77.0 ppm); DMSO-d6 (DMSO, 1H:
2.49 ppm, 13C: 39.5 ppm); DMF-d7 (DMF, 1H:
1
2.74 ppm, 13C: 30.1 ppm); dioxane-d8 (dioxane, H:
3.53 ppm, 13C: 66.6 ppm); pyridine-d5 (pyridine,
1H: 7.19 ppm [H-2], 13C: 123.5 ppm [C-2]); D2O
(3-(trimethylsilyl)-propanoic acid-d5, 1H: 0.0 ppm,
13C: 0.0 ppm).
3. Results
Hexa-acyl lipid A from E. coli F515 was easy to
obtain in 50 mg scale using standard isolation proto-

30
L. Brecker / Chemistry and Physics of Lipids 125 (2003) 27–39
Scheme 1.
line resolution. Hexa-acyl lipid A is furthermore
only slightly soluble in acetone-d6, acetonitrile-d3,
2,2,2-trifluoroethanol-d3, and methanol-d4 which
were therefore not suitable for NMR measurements
of lipid A.
used several times and caused reasonable line widths
at half height (Baltzer and Mattsby-Baltzer, 1986;
Zähringer et al., 1999; Ribeiro et al., 1999).
Hence chloroform-d/methanol-d4 was tested with
lipid A concentration of 4.0 mg ml−1 and caused rea-
sonable line resolution in solvent ratios from 9:1 to
1:1 (v/v) at 300 K. Within these solvent ratios lipid
A signals showed ∼3.5 Hz line width at half height
and the transversal relaxation times were ∼0.2 s
(Fig. 1, Table 1b). In comparison the longitudinal
relaxation times (T1) were about one magnitude
3.2. Mixtures of chloroform-d with protic solvents
Although hexa-acyl lipid A spectra taken in
chloroform-d did not show sufficient line resolution,
mixtures of chloroform-d with protic solvents were

L. Brecker / Chemistry and Physics of Lipids 125 (2003) 27–39
31
Fig. 1. Parts of the hexa-acyl lipid A 1H-NMR spectra in five selected solvents: (a) DMSO-d6, 330 K; (b) chloroform-d/methanol-d4 4:1
(v/v), 280 K; (c) dioxane-d8/methanol-d4 4:1 (v/v), 300 K; (d) pyridine-d5/methanol-d4 4:1 (v/v), 300 K; (e) D2O (0.1 M TEA-d15), 310 K.
Shown are the protons signals H-1, H-1ꢀ, H-3, and H-3ꢀ from the backbone and H-a2 to H-d2, H-cꢀ2, H-dꢀ2; H-c3, and H-d3 from the
fatty acids, which represent lipophilic and hydrophilic parts in lipid A, respectively.

32
L. Brecker / Chemistry and Physics of Lipids 125 (2003) 27–39
Fig. 1. (Continued ).
longer. The T1 times of the investigated protons
were in the range of about 2.1–5.0 s (Table 1b). At a
chloroform-d/methanol-d4 ratio of 4:1 (v/v) increas-
ing lipid A concentrations from 0.5 to 16.0 mg ml−1
were tested. The concentration rise caused a decrease
of the T2 time and consequently an increase of the
line width at half height, whereas the longitudinal
relaxation times remained constant. A temperature de-
crease from 300 to 280 K did not influence the spectral
resolution and all lipid A and solvent signals showed
nearly constant chemical shifts. The signal of residual
HDO in the spectrum, however, shifted like HDO in
a methanol/water NMR thermometer and could there-
fore be “adjusted” to a spectral region where it did not
overlap other signals. Traces of DCl in the solution,
however, led to slight saponification of esters and
amides, which was detected by TLC after 48 h. The
measured T1 times, T2 times, and line width at half
height of lipid A protons in chloroform-d/methanol-d4
mixtures under varying conditions are summarized in
Table 1b.
3.3. Mixtures of aprotic solvents with methanol-d4
As the addition of 25% (vol.) methanol-d4 to a lipid
A solution in chloroform-d considerably improved the
spectral quality, a similar addition to lipid A solutions
in other aprotic deuterated solvents was inevitable.
In dioxane-d8 and pyridine-d5 this addition caused
a distinct improvement of the spectral resolution,
comparable to the one in chloroform-d/methanol-d4
(Fig. 1, Table 1d). Remarkable, the signal-to-noise
ratio in all of these mixtures was worse than in
DMSO-d6. The determined line width at half heights
as well as the longitudinal and transversal relaxation
times of the investigated lipid A protons are summa-
rized in Table 1d. An addition of methanol-d4 to lipid
A solutions in DMSO-d6 or DMF-d7 did not affect
the spectral quality and the solubility of lipid A in
acetonitrile-d3 and acetone-d6 was not enhanced by
addition of methanol-d4.
3.4. Lipid A in aqueous solutions of amines
A mixture of chloroform-d, methanol-d4, and D2O
was optimized earlier to a ratio of 2:3:1 (v/v/v) for
NMR investigations of lipid A (Ribeiro et al., 1999).
Therefore the influence of a varying D2O amount
in this solvent system was tested in this study. The
variation was kept in the ratio from 2:3:0 (v/v/v) to
2:3:1 (v/v/v) to remain a monophasic mixture (Bligh
and Dyer, 1959). While the HDO signal shifted
from 3.90 ppm in 2:3:0 (v/v/v) to 4.26 ppm in 2:3:1
(v/v/v), the transversal relaxation times of all in-
vestigated nuclei decreased and in accordance the
line widths at half height consequently increased
(Table 1c).
Recording lipid A spectra in D2O seemed to be
obvious, as water is the natural solvent, but the am-
phiphilic glycolipids formed micelles in water, which
made the quality of the spectra rather poor. Never-
theless, the spectral resolution could be improved by
addition of deuterated Tris-d8 or TEA-d15. In both
cases a 0.1 M concentration of the amine at 310 K and
30 min ultra sonification led to the best spectral reso-
lution in D2O (Fig. 1, Table 1e). The water signal did
not overlap any lipid A signal, but small signals of the
amines appeared. The 50 times molar surplus of the
amine caused pD values of 10.9 in Tris-d8 and 11.3

Table 1
Measured line width at half height (median of all single isochromates in a multiplet), longitudinal relaxation time (T1), and transversal relaxation time (T2) of lipid A in
several solvents
Solvent(s)
Ratio
(v/v)
Concentration
(mg)
Temperature
(K)
Line width (Hz)
T1 (s)
T2 (s)
H-1
H-1ꢀ
H-d2
Me
H-1
H-1ꢀ
H-d2
Me
H-1
H-1ꢀ
H-d2
Me
Part a
DMSO-d6
DMF-d7
CDCl3
Dioxane-d8
Pyridine-d5
4.0
4.0
4.0
4.0
4.0
330
330
300
300
300
3.0
3.6
3.2
4.2
n.d.a
n.d.b
6.0
4.8
2.1
7.2
n.d.a
3.0
3.6
2.8
6.4
n.d.a
3.4
3.6
4.3
3.6
n.d.a
4.0
5.0
2.3
1.0
2.8
2.1
2.1
0.10
>0.05
>0.05
0.07
0.20
0.20
0.25
>0.05
0.15
0.60
0.60
0.50
0.50
0.60
n.d.a
n.d.a
n.d.a
n.d.a
n.d.a
n.d.a
3.2
5.1
>0.05
>0.05
0.15
n.d.a
2.3
6.0
n.d.b
0.07
0.20
0.20
Part b
CDCl3/CD3OD
4:1
3:2
2:3
4:1
4:1
4:1
4:1
4:1
4:1
4.0
4.0
4.0
0.5
2.0
8.0
16.0
4.0
4.0
300
300
300
300
300
300
300
280
290
n.d.a
n.d.a
n.d.a
2.7
3.3
3.8
n.d.b
3.6
3.5
4.1
6.4
3.4
3.2
2.9
2.9
3.3
2.8
2.6
3.7
5.4
3.0
3.0
2.4
3.0
3.6
2.1
2.0
2.2
2.3
2.1
2.1
3.8
4.2
4.7
3.7
3.7
3.4
3.4
2.6
2.9
4.2
4.5
n.d.b
3.8
3.3
3.3
3.3
3.1
3.7
5.0
5.0
5.0
5.0
5.0
5.0
5.0
4.4
4.8
2.1
2.1
2.1
2.1
2.1
2.2
2.3
2.9
2.3
0.07
0.07
0.07
n.d.c
0.10
0.07
0.06
0.10
0.07
0.20
0.20
n.d.b
n.d.c
0.20
0.17
0.16
0.20
0.20
0.22
0.22
0.20
n.d.c
0.20
0.18
0.17
0.17
0.22
0.75
0.65
0.65
0.75
0.75
0.75
0.75
0.50
0.70
n.d.a
n.d.a
n.d.a
2.7
n.d.a
Part c
CDCl3/CD3OD/D2O
2:3:1
4.0
300
n.d.a
7.0
3.8
4.5
10.0
n.d.b
5.3
2.1
>0.05
>0.05
0.08
0.40
Part d
Dioxane-d8/CD3OD
Pyridine-d5/CD3OD
4:1
4:1
4.0
4.0
300
300
2.5
n.d.a
2.5
3.1
3.4
2.4
2.5
n.d.b
2.5
2.9
2.9
3.6
4.3
5.2
2.1
2.5
0.15
0.10
0.23
0.23
0.20
0.30
0.70
0.60
Part e
D2O (0.1 M Tris)
D2O (0.1 M TEA)
4.0
4.0
310
310
n.d.a
n.d.a
n.d.a
7.0
n.d.a
n.d.a
n.d.a
n.d.a
4.3
2.6
4.0
1.9
n.d.a
4.3
3.3
2.6
>0.05
0.07
>0.05
0.15
0.10
0.15
0.30
0.50
The spectral parameters are listed for the protons H-1, H-1ꢀ, H-d2 and the methyl groups.
a
Not detectable because of small coupling constants.
Not detectable because of signal overlap with other signals.
Not detectable because of low signal-to-noise ratio.
b
c

34
L. Brecker / Chemistry and Physics of Lipids 125 (2003) 27–39
in TEA-d15. Addition of DCl to adjust the pD to ∼7
caused a deterioration of the spectral resolution. De-
spite the high pD value no NMR detectable decompo-
sition of lipid A was observed during 24 h. However, a
thin layer chromatogram, which was taken after 48 h,
indicated that the hexa-acyl lipid A was slightly de-
composed.
of ∼150 ms. Fig. 2 shows the ROEs between the pro-
ton H-1ꢀ and protons H-5ꢀ, H-6a, and H-6b at differ-
ent mixing times with an optimal signal intensity at
∼200 ms.
A more serious problem was the “lack” of sig-
nals in HMBC spectra of lipid A that was re-
ported earlier (Ribeiro et al., 1999). In the present
study the HMBC spectra of lipid A (16 mg ml−1
,
chloroform-d/methanol-d4 4:1 (v/v)) showed de-
3.5. Influence of transversal relaxation time on
signal intensity
2
3
tectable J- and J-couplings between carbon atoms
and the according protons that possessed transversal
relaxation times above ∼0.1 s. These protons were
mainly located in the fatty acids of the hexa-acyl lipid
A. In opposite protons with transversal relaxation
times below ∼0.1 s showed low or undetectable cross
correlation to the according carbon atoms. All pro-
tons in the backbone possessed such low T2 times and
therefore did not show any reasonable signals in the
The short transversal relaxation times caused prob-
lems recording two-dimensional lipid A spectra that
require long transversal magnetization transfer. This
was demonstrated on the signal intensity in ROESY
spectra where the initial linear rate approximation of
the growing ROE at short mixing times was blotted
out by cross relaxation at relative short mixing times
Fig. 2. Parts from ROESY spectra of lipid A in chloroform-d/methanol-d4 4:1 (v/v) at different mixing times that indicate the correlation
between mixing time and signal intensity. Shown are the ROEs between H-1ꢀ and the protons H-5ꢀ, H-6a, and H-6b.

L. Brecker / Chemistry and Physics of Lipids 125 (2003) 27–39
35
3.6. Influence of solvents on secondary structure
The conformation and aggregation state of hexa-
acyl lipid A varied distinctly in solvents that caused
good spectral resolution. These structural alterations
are exemplary indicated by the interglycosidic NOEs
of H-1ꢀ to the protons in glucosamine I (Fig. 4). While
NOEs from H-1ꢀ to H-6a and H-6b possessed different
intensities in all tested solvents, in DMSO-d6 one of
these two NOEs was undetectable. Furthermore, in
pyridine-d5/methanol-d4 4:1 (v/v) an additional NOE
from H-1ꢀ to H-5 was detected, indicating a higher
flexibility of lipid A in this solvent.
4. Discussion
All protons in hexa-acyl lipid A possessed transver-
sal relaxation (T2) times that were about one mag-
nitude shorter than their longitudinal relaxation (T1)
times. As the T2 times of all lipid A protons were be-
littled in a similar extend, the triggering effect equably
influenced the whole lipid A molecule. Scalar relax-
ation of the first kind cannot effect the T2 times in
such an intensive and consistent extend, as only a few
protons in lipid A showed a chemical exchange in
protic solvents. Also scalar relaxation of the second
kind cannot be the cardinal effect, because the inves-
tigated protons did not show significant couplings to
the 14N nuclei in the glucoseamines, which are the
only nuclei in the molecule that possess very rapid T1
relaxation. Furthermore, the portion of proton scalar
relaxation caused by scalar coupled nuclei with spin
I = 1/2 is negligible in case of the longitudinal relax-
ation. However, the transversal relaxation generated
by these coupled nuclei constitutes an enumerable part
of the total transversal relaxation. Based on the ap-
proximate relaxation times of the coupled nuclei an
average of about 5–15% scalar relaxation can be esti-
mated for the observed protons. Therefore the major
part of the transversal relaxation in lipid A was caused
by dipole–dipole interactions.
Fig. 3. 13C-Traces of the HMBC spectrum in chloroform-d/
methanol-d4 4:1 (v/v) indicating the 2J- and 3J-couplings of se-
lected protons to according carbon atoms. (a) Protons of the ter-
minal methyl groups (T2 = 0.75 s) to carbon atoms ω − 1 and
ω − 2 with an excellent S/N ratio. (b) H-d2ꢀ (T2 = 0.60 s) to car-
bon atoms C-d1ꢀ, C-d3ꢀ and C-d4ꢀ with a good S/N ratio. (c) H-d2
(T2 = 0.22 s) to carbon atoms C-d1, C-d3 and C-d4 with poor
S/N ratio. (d) H-3ꢀ (T2 = 0.01 s) to carbon atoms C-d1, C-2ꢀ, C-4ꢀ
and C-5ꢀ with a very poor S/N ratio.
The transversal relaxation time of large, slowly tum-
bling molecules and aggregates is shorter than their
longitudinal relaxation time, whereas nuclei of small,
rapidly tumbling molecules possess nearly identical
transversal and longitudinal relaxation times in low
viscosity solvents. Therefore the reciprocal correlation
HMBC spectra. Fig. 3 presents HMBC traces of 13C
atoms, which show correlations to some investigated
protons. A coherence between the signal intensities
and the transverse relaxation time of the affiliated
protons is obvious.

36
L. Brecker / Chemistry and Physics of Lipids 125 (2003) 27–39
Fig. 4. 1H-Traces of H-1ꢀ from the ROESY spectra in five selected solvents: (a) DMSO-d6; (b) chloroform-d/methanol-d4 4:1 (v/v); (c)
dioxane-d8/methanol-d4 4:1 (v/v); (d) pyridine-d5/methanol-d4 4:1 (v/v); (e) D2O (0.1 M TEA-d15). Mixing time was 225 ms in all spectra.
Indicated are the intra- and interglycosidic NOEs of H-1ꢀ to other protons in the lipid A backbone.
between the transversal relaxation and the line width
at half height cause inevitably poor line resolution in
spectra of large molecules and aggregates. Hexa-acyl
lipid A in its fully protonated form, however, has
a molecular mass of 1798.39 Da, not leading to
any distinct differences between the two relaxation
times when the molecule is completely solvated.
Therefore the 10-fold difference between them in-
dicated an aggregation of the lipid A in all tested
solvents.

L. Brecker / Chemistry and Physics of Lipids 125 (2003) 27–39
37
The total transversal relaxation times (T2∗) calcu-
lated from the line width at half height and the mea-
sured transversal relaxation times (T2), which were
caused by genuine relaxation processes, allow a cal-
culation of the transversal relaxation time portions
(v/v) was the most sensible unnatural solvent, because
all signals were well resolved. No solvent signals
overlapped the lipid A spectrum when using an appro-
priate temperature and recycling of the analyzed ma-
terial was easy by removal of the solvents in a gentle
vacuum.
caused by field inhomogeneities (T2(ꢀB)). For that cal-
culation the equation (T2∗)−1 = (T2)−1 + (T2(ꢀB)
)
−1
As water is the natural solvent, the most interest-
ing solvent for lipid A analysis was 0.1 M TEA-d15
was used (Claridge, 1999). The resulting T2(ꢀB) por-
tions of the lipid A protons varied from 0.1 to 0.4 s
depending on the different solvent systems and inves-
tigated lipid A protons. These T2(ꢀB) portions were
significantly smaller than those of solvent protons,
which had all been longer than 2 s. This finding in-
dicates local field inhomogeneities around the lipid
A molecules, explainable when lipid A forms aggre-
gates.
in D2O. The substitution of D+-ions by DN(C2D5)3
+
led to good line resolution and allowed structure de-
termination in aqueous solvents, although a high pD
caused slow decomposition of lipid A. These findings
are in agreement with pH effects on solubility of LPS
in H2O reported by Mukerjee et al. (1999). Such inves-
tigations in the natural solvent are of actual interest,
as the lipid A molecular binding mechanism to TLR4
has not yet been elucidated. After the importance of
soluble receptor proteins CD14 and MD2 in the bind-
ing process has recently been shown (Henneke and
Golenbock, 2002), the significance of the lipid A ag-
gregation state and the involvement of other molecular
species in this process have to be investigated. There-
fore NMR of glycolipids in D2O is a helpful task to
discover the secondary lipid A structure and the dy-
namic process in binding it to TLR4 on mammalian
macrophages.
These results indicate that lipid A is not completely
solved but present as small micelles or aggregates in
all tested solvents. Hence an improvement of spectral
resolution requires an optimization of lipid A disag-
gregation.
4.1. Criteria for solvent selection
Five of the tested solvents disaggregated hexa-acyl
lipid A quite well and caused reasonable NMR spec-
tra. Single DMSO-d6 had to be heated to 330 K to get
well resolved spectra and the lipid A was therefore
thermally affected. Furthermore, HDO and solvent
signals partially overlapped the spectrum and the high
boiling point of the solvent hindered recovery of the
analyzed material. However, in DMSO-d6 the best
signal-to-noise ration of the lipid A proton signals
was detected. Furthermore, a slow chemical exchange
rate of the amide protons in DMSO-d6 caused well
resolved N–H proton signals suitable for detailed
analysis of the lipid A acylation pattern (Janusch
et al., 2002). In mixtures of dioxane-d8/methanol-d4
4:1 (v/v) the water signal and two solvent signals
overlapped several lipid A signals and made structure
determination difficult. Nevertheless the spectral res-
olution of the anomeric proton signals was the best
in this solvent. Pyridine-d5/methanol-d4 4:1 (v/v)
caused no overlapping signals and furthermore spread
the lipid A spectrum as it is common for aromatic
solvents. However, traces of water led to a high con-
centration of OD−-ions that catalyzed slow decom-
position of lipid A. Chloroform-d/methanol-d4 4:1
4.2. Signal intensities in two-dimensional spectra
The short transversal relaxation time of hexa-acyl
lipid A aggregates caused rapid transversal relaxation
during evolution and mixing times in two-dimensional
spectra and consequently diminished the signal inten-
sities. COSY and HMQC spectra of lipid A were only
slightly effected because of their short evolution peri-
ods. However, TOCSY and ROESY spectra were more
intensively concerned, because they required mixing
times of ∼100 and ∼250 ms, respectively. As shown
in Fig. 2, an exact regulation of the mixing time is
necessary to get optimal signal intensities.
Recording sensible HMBC spectra was nearly im-
possible. Like all 1H/13C heteronuclear correlation
spectra they possessed a much lower signal intensity
compared to the one in 1H/1H correlations, which was
caused by the low absolute sensitivity of 13C nuclei.
The fast transversal relaxation of several lipid A pro-
tons during the HMBC evolution period of 60–100 ms
furthermore diminished the magnetization to almost

38
L. Brecker / Chemistry and Physics of Lipids 125 (2003) 27–39
undetectable amounts before the acquisition started.
As all protons of the lipid A backbone showed such
rapid transversal relaxation, they did not cause any
reasonable signals in HMBC spectra and led to the
lack of several cross peaks, which is shown in Fig. 3
and was reported earlier (Ribeiro et al., 1999). There-
fore interglycosidic linkages in lipid A were superiorly
determined by ROESY spectra (Ribeiro et al., 1999).
Protons from the fatty acids in lipid A had a much
slower transversal relaxation and showed signals of
2J- and 3J-couplings that can be helpful to determine
fatty acid substitution pattern.
Acknowledgements
The author thanks U. Zähringer for substantial
contribution as well as S. Müller-Loennies and D.W.
Ribbons for proofreading. H.P. Cordes is gratefully
acknowledged for excellent technical assistance
recording the NMR spectra. The author is thankful to
N. Grohmann and K. Jakob for lipid A preparation.
References
Agrawal, P.K., Bush, C.A., Qureshi, N., Takayama, K., 1994.
Structural analysis of lipid A and Re-lipopolysaccharide by
NMR spectroscopic methods. Adv. Biophys. Chem. 4, 179–
236.
Alexander, C., Rietschel, E.T., 2001. Bacterial lipopolysaccharides
and innate immunity. J. Endotoxin Res. 7, 167–202.
Baltzer, L.H., Mattsby-Baltzer, I., 1986. Heterogeneity of lipid
A: structural determination by 13C and 31P-NMR of lipid
4.3. The secondary structure
For a comprehensive investigation of lipid A con-
formation molecular modeling is obligatory. How-
ever, NMR spectra provide several valuable and easy
derivable information about the secondary structure
and the aggregation state. For example, differences
between the longitudinal and transversal relaxation
times indicated that aggregates were formed in all
solvents. Furthermore, the transversal relaxation
time variation of some protons in different solvents
(Table 1) indicated dissimilar tumbling of different
molecular parts. For example, H-1 had a longer T2
time in DMSO-d6 than in pyridine-d5/methanol-d4
4:1 (v/v), while it is the other way round with the
H-1ꢀ and H-d2. These differences indicated dissimilar
lipid A aggregation forms, as molecular parts on the
aggregate surface possessed a higher flexibility and
longer T2 time than those in the center of the aggre-
gate.
A
fractions from lipopolysaccharide of Escherichia coli.
Biochemistry 25, 3570–3575.
Bax, A., Davis, D.G., 1985. MLEV-17 based two-dimensional
homonuclear magnetisation transfer spectroscopy. J. Magn.
Reson. 65, 355–360.
Bligh, E.G., Dyer, W.J., 1959. A rapid method of total lipid
extraction and purification. Can. J. Biochem. 37, 911–917.
Claridge, T.W.D., 1999. High-Resolution NMR Techniques in
Organic Chemistry. Pergamon Press, Amsterdam.
Galanos, C., Lüderitz, O., Westphal, O., 1969. A new method for
the extraction of R lipopolysaccharides. Eur. J. Biochem. 9,
245–249.
Henneke, P., Golenbock, D.T., 2002. Innate immune recognition
of lipopolysaccharide by endothelial cells. Crit. Care Med. 30,
S207–S213.
Hurd, R.E., 1990. Gradient-enhanced spectroscopy. J. Magn.
Reson. 87, 422–428.
Particular NOEs gave insight into some details of
lipid A conformation and aggregation state. In all sol-
vents, for example, the intensity of the H-1ꢀ/H-6a-NOE
differed from the intensity of the H-1ꢀ/H-6b-NOE,
indicating a hindered flexibility of the interglyco-
sidic bonds. However, in pyridine-d5/methanol-d4 4:1
(v/v) there was an additional interglycosidic NOE
from H-1ꢀ to H-5. Conformational prerequisite for
the closeness between these nuclei was a higher flex-
ibility of the interglycosidic linkage. This required a
separate solvatization of the fatty acid chains at glu-
cosamine I and at glucosamine II, indicating a lipid
A solution state in pyridine-d5/methanol-d4 4:1 (v/v)
that differed fundamentally from them in the other
solvents.
Imoto, M., Yoshimura, H., Sakaguchi, N., Kusumoto, S., Shiba, T.,
1985. Total synthesis of Escherichia coli lipid A. Tetrahedron
Lett. 26, 1545–1548.
Janusch, H., Brecker, L., Lindner, B., Alexander, C., Gronow, S.,
Heine, H., Ulmer, A.J., Rietschel, E.T., Zähringer, U., 2002.
Structural and biological characterization of highly purified
hepta-acyl lipid A present in the lipopolysaccharide of the
Salmonella enterica sv. Minnesota Re mutant strain R595. J.
Endotoxin Res. 8, 343–356.
Meiboom, S., Gill, D., 1958. Modified spin-echo method for
measuring nuclear relaxation times. Rev. Sci. Instrum. 29, 688–
691.
Mukerjee, P., Kastrowsky, M., Obst, S., Takayama, K., 1999.
Lipopolysaccharide preparations in aqueous media: implication
for solution versus suspension. In: Brade, H., Morrison, D.C.,
Opal, S., Vogel, S. (Eds.), Endotoxin in Health and Disease.
Marcel Dekker, New York, pp. 221–228.

L. Brecker / Chemistry and Physics of Lipids 125 (2003) 27–39
39
Neuhaus, D., Williamson, M., 1989. The Nuclear Overhauser
Effect in Structural and Conformational Analysis. VCH,
Weinheim.
Oikawa, M., Fukase, K., Suda, Y., Kusumoto, S., 2000. Bioactivity
and solution conformation of lipid A analogues. J. Endotoxin
Res. 6, 155.
Takayama, K., Qureshi, N., Mascagni, P., Nashed, M.A., Anderson,
L., Raetz, C.R.H., 1983a. Fatty acyl derivatives of glucosamine
1-phosphate in Escherichia coli and their relation to lipid A.
J. Biol. Chem. 258, 7379–7385.
Takayama, K., Qureshi, N., Mascagni, P., 1983b. Complete
structure of lipid A obtained from the lipopolysaccharide of
the heptoseless mutant of Salmonella typhimurium. J. Biol.
Chem. 258, 12801–12803.
Raetz, C.R.H., Whitfield, C., 2002. Lipopolysaccharide endo-
toxines. Annu. Rev. Biochem. 71, 635–700.
Ribeiro, A.A., Zhou, Z.M., Raetz, C.R.H., 1999. Multi-dimensional
NMR structural analyses of purified lipid X and lipid A
(endotoxin). Magn. Reson. Chem. 37, 620–630.
Schlecht, S., 1975. Growth and lipopolysaccharide content of
Salmonellae grown in submersed cultures according to the
batch-method. 2. Communication: influence of growth on
lipopolysaccharide synthesis. Zbl. Bakt. Hyg., I Abt. Orig. A
232, 61–72.
Schmidt, G., Jann, B., Jann, K., 1970. Immunochemistry of R
lipopolysaccharides of Escherichia coli. Eur. J. Biochem. 16,
382–392.
Seydel, U., Oikawa, M., Fukase, K., Kusumoto, S., Brandenburg,
K., 2000. Intrinsic conformation of lipid A is responsible for
agonistic and antagonistic activity. Eur. J. Biochem. 267, 3032–
3039.
Vold, R.L., Waugh, J.S., Klein, M.P., Phelps, D.E., 1968.
Measurement of spin relaxation in complex systems. J. Chem.
Phys. 48, 3831–3832.
Wang, Y., Hollingsworth, R.I., 1995. A solvent system for the
high-resolution proton nuclear magnetic resonance spectro-
scopy of membrane lipids. Anal. Biochem. 225, 242–251.
Willker, W., Leibfritz, D., Kerssebaum, R., Bermel, W.,
1993. Gradient selection in inverse heteronuclear correlation
spectroscopy. Magn. Reson. Chem. 31, 287–292.
Zähringer, U., Lindner, B., Rietschel, E.T., 1994. Molecular
structure of lipid A, the endotoxic center of bacterial
lipopolysaccharides. Adv. Carbohydrate Chem. Biochem. 50,
211–276.
Zähringer, U., Lindner, B., Rietschel, E.T., 1999. Chemical
structure of lipid. A. Recent methodical advances towards the
complete structural analysis of a biologically active molecule.
In: Brade, H., Morrison, D.C., Opal, S., Vogel, S. (Eds.),
Endotoxin in Health and Disease. Marcel Dekker, New York,
pp. 93–114.
Strain, S.M., Fesik, S.W., Armitage, I.M., 1983a. Characterization
of lipopolysaccharide from a heptoseless mutant of Escherichia
coli by carbon 13 nuclear magnetic resonance. J. Biol. Chem.
258, 2906–2910.
Strain, S.M., Fesik, S.W., Armitage, I.M., 1983b. Structure
and metal-binding properties of lipopolysaccharides from
heptoseless mutants of Escherichia coli studied by 13C and
31P nuclear magnetic resonance. J. Biol. Chem. 258, 13466–
13477.
Zähringer, U., Salvetzki, R., Wagner, F., Lindner, B., Ulmer,
A.J., 2001. Structural and biological characterisation of
a
novel tetra-acyl lipid A from Escherichia coli F515
lipopolysaccharide acting as endotoxin antagonist in human
monocytes. J. Endotoxin Res. 7, 133–146.