Protein structure determination by high-resolution solid-state NMR spectroscopy: application to microcrystalline ubiquitin.

Abstract

High-resolution solid-state NMR spectroscopy has become a promising method for the determination of three-dimensional protein structures for systems which are difficult to crystallize or exhibit low solubility. Here we describe the structure determination of microcrystalline ubiquitin using 2D (13)C-(13)C correlation spectroscopy under magic angle spinning conditions. High-resolution (13)C spectra have been acquired from hydrated microcrystals of site-directed (13)C-enriched ubiquitin. Inter-residue carbon-carbon distance constraints defining the global protein structure have been evaluated from 'dipolar-assisted rotational resonance' experiments recorded at various mixing times. Additional constraints on the backbone torsion angles have been derived from chemical shift analysis. Using both distance and dihedral angle constraints, the structure of microcrystalline ubiquitin has been refined to a root-mean-square deviation of about 1 A. The structure determination strategies for solid samples described herein are likely to be generally applicable to many proteins that cannot be studied by X-ray crystallography or solution NMR spectroscopy.

Full text

Page 1

Protein Structure Determination by High-Resolution
Solid-State NMR Spectroscopy: Application to
Microcrystalline Ubiquitin
Stephan G. Zech,*,†,§A. Joshua Wand,‡and Ann E. McDermott*,†
Contribution from the Department of Chemistry, Columbia UniVersity,
3000 Broadway Mail Code 3113, New York, New York 10027, and Department of Biochemistry
and Biophysics, UniVersity of PennsylVania, The Johnson Research Foundation,
Philadelphia, PennsylVania 19104
Received January 17, 2005; E-mail: Stephan.Zech@web.de; aem5@columbia.edu.
Abstract: High-resolution solid-state NMR spectroscopy has become a promising method for the
determination of three-dimensional protein structures for systems which are difficult to crystallize or exhibit
low solubility. Here we describe the structure determination of microcrystalline ubiquitin using 2D13C-13C
correlation spectroscopy under magic angle spinning conditions. High-resolution13C spectra have been
acquired from hydrated microcrystals of site-directed13C-enriched ubiquitin. Interresidue carbon-carbon
distance constraints defining the global protein structure have been evaluated from ‘dipolar-assisted rotational
resonance’ experiments recorded at various mixing times. Additional constraints on the backbone torsion
angles have been derived from chemical shift analysis. Using both distance and dihedral angle constraints,
the structure of microcrystalline ubiquitin has been refined to a root-mean-square deviation of about 1 Å.
The structure determination strategies for solid samples described herein are likely to be generally applicable
to many proteins that cannot be studied by X-ray crystallography or solution NMR spectroscopy.
Introduction
Over the past few years, there have been remarkable advances
in solid-state NMR (ssNMR) experiments for characterization
of protein structure and function. In a wide variety of systems
the protein’s insolubility made X-ray crystallography or solution
NMR unsuitable, while questions on structure and dynamics
can be addressed with ssNMR.1-5These efforts typically involve
structurally homogeneous samples and utilize recently developed
pulse sequences for sequential correlation of resonances, detec-
tion of tertiary contacts, and characterization of torsion angles.
Excellent NMR line widths can be achieved for microcrys-
talline or precipitated hydrated globular systems using magic
angle spinning (MAS) methods. More recent advances in high-
field instrumentation and pulse sequences for chemical shift
correlation experiments led to more efficient methods of
assigning solid-state proteins with extensive isotopic enrichment
and a rapid succession of studies on small globular proteins
has become evident, including BPTI,6the R-spectrin SH3
domain,7,8,9the catabolite repression phosphocarrier protein
(Crh),10human ubiquitin,11,12thioredoxin,13the immunoglobulin
binding domain ?1 of streptococcal protein G (GB1),14and
peptides such as neurotensin,15mastoparan-X,16and a fibrillar
peptide fragment of transthyretin.17
For tertiary structure determination, or fold definition, long-
range constraints have typically served a critical role. It has been
recently demonstrated that simple spin diffusion experiments,
if combined with strategic labeling schemes, are sufficient to
determine a moderate resolution structure of a protein. This was
first shown for the R-spectrin SH3 domain using site-directed
selective carbon labeling schemes and a ‘proton-driven spin
diffusion’ (PDSD) pulse sequence.8
†Columbia University.
‡University of Pennsylvania.
§Present address: EPIX Pharmaceuticals, Inc., 67 Rogers St., Cambridge,
MA 02142.
(1) Thompson, L. K. Curr. Opin. Struct. Biol. 2002, 12, 661-669.
(2) Williamson, P. F.; Ernst, M.; Meier, B. H. MAS solid-state NMR of
isotopically enriched biological samples. In BioNMR in Drug Research;
Zerbe, O., Ed.; Wiley-VCH: Weinheim, 2003; pp 243-282.
(3) Opella, S. J.; Marassi, F. M. Chem. ReV. 2004, 104, 3587-3606.
(4) Tycko, R. Curr. Opin. Struct. Biol. 2004, 14, 96-103.
(5) McDermott, A. E. Curr. Opin. Struct. Biol. 2004, 14, 554-561.
(6) McDermott, A.; Polenova, T.; Bockmann, A.; Zilm, K. W.; Paulson, E.
K.; Martin, R. W.; Montellione, G. T. J. Biomol. NMR 2000, 16, 209-
219.
(7) Pauli, J.; Baldus, M.; van Rossum, B.; de Groot, H.; Oschkinat, H.
Chembiochem. 2001, 2, 272-281.
(8) Castellani, F.; van Rossum, B.; Diehl, A.; Schubert, M.; Rehbein, K.;
Oschkinat, H. Nature 2002, 420, 98-102.
(9) Castellani, F.; van Rossum, B.; Diehl, A.; Rehbein, K.; Oschkinat, H.
Biochemistry 2003, 42, 11476-11483.
(10) Bo ¨ckmann, A.; Lange, A.; Galinier, A.; Luca, S.; Giraud, N.; Juy, M.;
Heise, H.; Montserret, R.; Penin, F.; Baldus, M. J. Biomol. NMR 2003, 27,
323-339.
(11) Igumenova, T.; Wand, A. J.; McDermott, A. J. Am. Chem. Soc. 2004, 126,
5323-5331.
(12) Igumenova, T. I.; McDermott, A. E.; Zilm, K. W.; Martin, R. W.; Paulson,
E. K.; Wand, A. J. J. Am. Chem. Soc. 2004, 126, 6720-6727.
(13) Marulanda, D.; Tasayco, M. L.; McDermott, A.; Cataldi, M.; Arriaran, V.;
Polenova, T. J. Am. Chem. Soc. 2004, 125, 16608-16620.
(14) Franks, W. T.; Zhou, D. H.; Money, B. G.; Graesser, D. T.; Sahota, G.;
Rienstra, C. M. J. Am. Chem. Soc. 2005, in press.
(15) Luca, S.; White, J. F.; Sohal, A. K.; Filippov, D. V.; van Boom, J. H.;
Grisshammer, R.; Baldus, M. Proc. Natl. Acad. Sci., U.S.A. 2003, 100,
10706-10711.
(16) Fujiwara, T.; Todokoro, Y.; Yanagishita, H.; Tawarayama, M.; Kohno, T.;
Wakamatsu, K.; Akutsu, H. J. Biomol. NMR 2004, 28, 311-325.
(17) Jaroniec, C. P.; MacPhee, C.; Bajaj, V. S.; McMahon, M. T.; Dobson, C.
M.; Griffin, R. G. Proc. Natl. Acad. Sci., U.S.A. 2004, 101, 711-716.
Published on Web 05/28/2005
8618 9 J. AM. CHEM. SOC. 2005, 127, 8618-8626
10.1021/ja0503128 CCC: $30.25 © 2005 American Chemical Society

Page 2

Advances in fast MAS-based magnetization transfer schemes
allow for directed, highly efficient homonuclear (13C-13C)
recoupling during the mixing times of chemical shift correlation
spectra.2Several homonuclear methods have been developed
recently which achieve broadband (i.e. nonspectrally selective)
efficient recoupling. The ‘1H-13C dipolar-assisted rotational
resonance’ (DARR) experiment18or the related ‘radiofrequency
field-assisted diffusion’ sequence19have been shown to exhibit
an excellent efficiency in recoupling of distant carbon atoms.
In this study we use the DARR carbon-carbon correlation
experiments on selectively13C-enriched microcrystals of human
ubiquitin to obtain a large number of interresidue constraints
defining secondary structure and global fold of the protein. In
line with previous studies,11,12,20,21hydrated ubiquitin microc-
rystals give rise to highly resolved ssNMR spectra, facilitating
site-specific assignment of resonances. The line width can be
further reduced if selective13C labeling schemes are involved.22,8
Furthermore, significant information is available from analysis
of chemical shifts alone. CR and C? isotropic shifts provide
information on the secondary structure context. This information
can be crucial in defining the fold of a protein or validating the
sample conditions. Here we apply the database provided by the
program TALOS23to obtain backbone dihedral angle restraints.
Together with the distance restraints gathered from13C-13C
correlation spectroscopy, the 3D structure of ubiquitin is refined
to a great level of molecular detail.
Materials and Methods
Sample Preparation. Uniformly [13C,15N] labeled human ubiquitin
(referred to as [U-13C]-Ubq) has been prepared as described previously.24
Selectively13C-enriched and uniformly15N-enriched ubiquitin largely
devoid of adjacent13C nuclei (referred to as [2-13C]-Ubq) has been
isolated from E. coli grown on [2-13C]-glycerol (Cambridge Isotopes)
as carbon source and15NH4Cl as nitrogen source.25In contrast to the
original paper, the E. coli strain used here did not lack succinate
dehydrogenase and malate dehydrogenase. Furthermore, NaH13CO3has
not been added to the growth medium in our study. Isolation and
purification of the protein has been done as described previously.24
About 8 mg of lyophilized ubiquitin has been dissolved to 25 mg/
mL in a sodium citrate buffer (20 mM, pH 4.1) and crystallized as
described previously11by slow addition of 2-methyl-2,4-pentanediol
(∼60% final volume). A white precipitate appeared within minutes and
showed needle-shaped microcrystals under the microscope. The emul-
sion has been stored overnight at 4 °C and centrifuged at 104rpm for
20 min. Most of the supernatant has been removed, and the hydrated
solid (ca. 25-30 mg) has been filled into a 4 mm Bruker MAS rotor.
The sample has been confined to the center of the rotor using Teflon
spacers. For the diluted sample, 6 mg of [2-13C]-Ubq and 12 mg of
isotopically unenriched ubiquitin (Aldrich) have been cocrystallized,
yielding about 55 mg of hydrated microcrystals. All rotors have been
stored at -80 °C until use with no changes in the spectra observable
for various experiments recorded over a period of several months.
NMR Experiments. All experiments have been performed on a
Bruker Avance DRX-750 spectrometer, operating at 750.22 MHz proton
and 188.65 MHz carbon frequency. A double resonance (H-F/X)
widebore MAS probe with 4.0 mm rotor diameter has been used. The
sample has been cooled by a flow of dry air (ca. 2000 L/h) with an
inlet temperature between 233 and 245 K as indicated in the figure
captions. The spinning frequency has been adjusted to 10.0 kHz. 1D
CP-MAS experiments have been acquired with a ramped RF field
(80-100%) on the proton channel, applying an RF field strength of
76 kHz on the proton and 45 kHz on the carbon channel during the
cross-polarization time of 2.0 ms. The proton RF field has been adjusted
to 100 kHz during the initial 90° pulse and 95 kHz during the acquisition
time of about 25 ms using the XiX sequence for proton decoupling.26
For the 2D carbon-carbon correlation spectra, the DARR pulse
sequence18as depicted in Figure 1 has been used. RF power levels
have been identical to 1D CP-MAS experiments with a RF strength of
50 kHz for the 90° pulses on the carbon channel. During the mixing
time (10-500 ms), the RF field on the proton channel matched the
spinning speed of 10 kHz. The spectral width has been set to 377 ppm
in ω2 and 150 ppm in ω1 with the transmitter frequency at 75 ppm.
Typically 350 points in t1with 80-240 averages per point have been
acquired. The TPPI method has been used for phase sensitive detection
in t1by incrementing φ2.27The total acquisition time has been about
10 min for 1D spectra and between 20 and 72 h for the 2D data sets.
All data have been processed with NMRpipe.28A phase-shifted
sinebell function has been used for apodization before zero filling to
2048 points in t2 and 1024 in t1 prior to FT in each dimension. The
magnetic field has been referenced to DSS, using the13C methylene
peak in solid adamantane as an external standard.29Analysis and
assignment of the 2D data sets has been carried out using Sparky version
3.1.30
Structure Calculations. The structure of ubiquitin has been
calculated using CNS version 1.131by applying a molecular dynamics
simulated annealing protocol with torsion angles as internal degrees
of freedom. All parameters were identical to the default protocol (see
‘NMR Tutorials’ section of CNS) which involves a high-temperature
annealing at 50 000 K for 1000 steps with 15 ps per step and a
subsequent cooling stage with decreasing temperature from 50 000 K
to 0 K in 1000 steps. Experimental carbon-carbon distance constraints
and backbone dihedral angle restraints obtained by TALOS23have been
(18) Takegoshi, K.; Nakamura, S.; Terao, T. Chem. Phys. Lett. 2001, 344, 631-
637.
(19) Morcombe, C. R.; Gaponenko, V.; Byrd, R. A.; Zilm, K. W. J. Am. Chem.
Soc. 2004, 126, 7196-7197.
(20) Paulson, E. K.; Morcombe, C. R.; Gaponenko, V.; Dancheck, B.; Byrd, R.
A.; Zilm, K. W. J. Am. Chem. Soc. 2003, 125, 14222-14223.
(21) Paulson, E. K.; Morcombe, C. R.; Gaponenko, V.; Dancheck, B.; Byrd, R.
A.; Zilm, K. W. J. Am. Chem. Soc. 2003, 125, 15831-15836.
(22) Hong, M. J. Magn. Reson. 1999, 139, 389-401.
(23) Cornilescu, G.; Delaglio, F.; Bax, A. J. Biomol. NMR 1999, 13, 289-302.
(24) Wand, A. J.; Urbauer, J. L.; McEvoy, R. P.; Bieber, R. J. Biochemistry
1996, 35, 6116-6125.
(25) LeMaster, D. M.; Kushlan, D. M. J. Am. Chem. Soc. 1996, 118, 9255-
9264.
(26) Detken, A.; Hardy, E. H.; Ernst, M.; Meier, B. H. Chem. Phys. Lett. 2002,
356, 298-304.
(27) Cavanagh, J.; Fairbrother, W. J.; Palmer, A. G., III; Skelton, N. J. Protein
NMR Spectroscopy - Principles and Practise; Academic Press: San Diego,
1996.
(28) Delaglio, F.; Grzesiek, S.; Vuister, G.; Zhu, G.; Pfeifer, J.; Bax, A. J. Biomol.
NMR 1995, 6, 277-293.
(29) Morcombe, C.; Zilm, K. W. J. Magn. Reson. 2003, 162, 479-486.
(30) Goddard, T. D.; Kneller, D. G. Sparky 3.106; University of California,
San Francisco, 2002.
(31) Brunger, A.; Adams, P.; Clore, G.; Delano, W.; Gros, P.; Grosse-Kunstleve,
R.; Liang, J.-S.; Kuszewski, J.; Nilges, N.; Pannu, N.; Read, R.; Rice, L.;
Simonson, T.; Warren, G. Acta Crystallogr. D 1998, 54, 905-921.
Figure 1. Pulse sequence for carbon-carbon correlation experiments using
the DARR scheme for magnetization transfer during the mixing time. The
parameters for RF field strengths and pulse durations are described in the
text. The following phase cycle has been used: φ1) (yy j)8, φ2) (xxx jx j)4,
φ3) (y4y j4)2, φ4) x8x j8, φrec.) xx jx jx (x jxxx j)2xx jx jx.
Structure Determination of Microcrystalline Ubiquitin
A R T I C L E S
J. AM. CHEM. SOC. 9 VOL. 127, NO. 24, 2005 8619

Page 3

used as input for the structure refinement. All structures have been
energy minimized and the 10 lowest energy structures have been
selected and analyzed.
Results and Discussion
Improved Resolution in Sparsely Labeled Samples. Figure
2 compares the 1D CP-MAS spectra of [U-13C]-Ubq (A) and
[2-13C]-Ubq (B). The spectrum for [U-13C]-Ubq is almost
identical to the one obtained previously for a similar preparation
of microcrystals recorded under different experimental condi-
tions.12The13C line width of this sample can be estimated to
be in the 0.5-0.6 ppm range. The spectrum of the [2-13C]-Ubq
sample (Figure 2B) shows reduced signal intensities in the
methyl group region (5-25 ppm) and in the carbonyl region
(∼174 ppm) due to the low amount of13C enrichment predicted
for this labeling scheme. More importantly, the spectral resolu-
tion of the [2-13C]-Ubq sample is considerably improved mainly
due to removal of one-bond J and D couplings.8,25This is easily
seen by comparison of the CR region of both spectra (Figure
2C and D). The labeling scheme predicts a high degree of13C
enrichment in the CR atoms of most amino acids (except Glu,
Arg, Pro, Leu), whereas the C? and carbonyl atoms are expected
to have low13C content. Due to the resulting absence of J and
D couplings, the line width of the [2-13C]-Ubq sample decreases
to about 0.2-0.3 ppm (Figure 2D).
Identification of Mobile Residues at Low Temperature.
Homonuclear13C chemical shift correlation spectra of micro-
crystals of the [U-13C]-Ubq sample have been measured at 233
and 245 K with mixing times of 10, 15, and 20 ms. Figure 3A
shows the spectrum of the [U-13C]-Ubq sample at 233 K with
a mixing time of 20 ms. It turned out that the spectra taken
between 233 and 245 K exhibit many peaks that have not been
observed in the previous studies which had been conducted at
higher temperatures (∼278-288 K).12The13C line width in
the 2D spectra, however, appears to be nearly independent of
temperature within the range investigated. This indicates that
the temperature dependent observation of some residues is
influenced by their mobility rather than due to disorder or sample
inhomogeneity.
Identification of the spin system of the additional resonances
has been achieved by a side chain walk as described previ-
ously.6,7,12Site-specific assignments have been initially guided
by solution NMR shifts and subsequently corroborated utilizing
sequential two- and three-bond couplings observed at 20 ms
mixing time (Figure 3A). For most residues, the site specific
assignment has been further confirmed by sequential contacts
in the DARR spectra of the [2-13C]-Ubq sample described
below. Thereby, several residues which have been missing in
the previous experiments12could be identified. Those comprise
Leu8(CR, C?, Cδ1, CO), Thr9(CR, C?, Cγ2), Gly10(CR, CO),
and Lys11(CR, C?, Cγ, CO) located in a mobile loop region as
well as Arg72(CR, C?, Cδ) and Leu73(CR, C?, Cγ) near the
flexible C-terminus of the protein. For many residues for which
only partial assignment has been obtained previously,12more
atoms comprising the spin system could be identified (e.g. for
Glu24, Glu34, Arg42, Gln49, Asp52).
A likely reason for the temperature dependence is that the
initial proton-carbon magnetization transfer is not efficient for
mobile residues. For example, the rate of relaxation or dissipa-
tion of proton magnetization, which can be measured through
the line intensity in the13C spectrum after CP, is controlled by
thermal motion and can therefore be used to probe the mobility
of individual regions within a complex structure. This demon-
strates that the experimental condition, and in particular the
temperature, have a profound influence on the number of
observable amino acids especially for mobile residues. While
lower temperatures are more demanding on the long time
spectrometer stability and potentially lead to line broadening,32
experiments over a range of temperatures might be required to
identify amino acids in mobile regions of the protein.
Observation of Long-Range Contacts for Selectively13C-
Labeled Ubiquitin. Structure determination by NMR spectros-
copy relies to a large degree on the observation of long-range
contacts which provide a large number of interresidue restraints.
While with solution state NMR methods, mainly proton-proton
restraints are obtained, for the first 3D protein structure
determination based on MAS solid-state NMR techniques
predominantly carbon-carbon contacts and a number of15N-
(32) Martin, R. W.; Zilm, K. W. J. Magn. Reson. 2003, 165, 162-174.
Figure 2. 1D13C-CP-MAS spectra acquired at 245 K from ubiquitin
microcrystals with uniform13C enrichment (A) and with selective13C
enrichment (B) obtained with [2-13C]-glycerol as carbon source. Experi-
mental conditions are described in the text. Panels C and D show the
expanded aliphatic region of the spectra in A and B, respectively. Spectrum
C exhibits considerable narrower lines for most CR atoms due to the absence
of one-bond13C-13C couplings in the selective labeling scheme.
A R T I C L E S
Zech et al.
8620 J. AM. CHEM. SOC.9VOL. 127, NO. 24, 2005

Page 4

15N contacts have been used for structure refinement.8To
recouple distant carbon atoms in ubiquitin, the DARR experi-
ment has been applied to the [2-13C]-Ubq sample. Since in this
sample, adjacent13C atoms are present only for few residues,
dominant one-bond dipolar couplings are sufficiently suppressed
and dipolar truncation effects are reduced.8The ability of the
DARR sequence to efficiently recouple distant carbon atoms
(up to ∼6 Å) has been recently demonstrated on sparsely labeled
samples of the anti-apoptotic protein Bcl-xL (∼20 kDa)33and
for extensively
ubiquitin.19
Various DARR spectra of [2-13C]-Ubq have been obtained
at 245 K with mixing times ranging between 10 ms and 500
ms. Figure 3B-D show the spectra obtained at 50 ms (B), 200
ms (C), and 500 ms (D). The removal of one-bond couplings
due to site-directed selective labeling of the [2-13C]-Ubq sample
is obvious by comparison of Figure 3A and B. Since at short
mixing times (∼10 ms) mainly one-bond carbon-carbon
correlations are observed, the DARR spectrum of [2-13C]-Ubq
appears greatly simplified compared to the [U-13C]-Ubq sample.
In line with the proposed labeling scheme,8,25the peaks in th
aliphatic region can be assigned to intraresidue correlation for
Val, Leu and Ile which exhibit adjacent13C-enriched atoms.
Comparison of the carbonyl peak integrals shown Figure 2A
and B reveal an amount of13C enrichment in the carbonyl
2H and uniformly
13C-enriched samples of
groups of the [2-13C]-Ubq sample that is roughly in agreement
with the proposed labeling scheme.8,25However, closer inspec-
tion of Figures 3B-D shows that carbon atoms without
proposed13C enrichment25are observed at mixing times g50
ms. This suggests that the13C enrichment present in our sample
is not identical to the one proposed previously.25A likely
explanation for this might be the use of a different strain of E.
coli and the lack of13C-enriched sodium carbonate which leads
to increased scrambling of13C labels. As shown below, the
larger13C enrichment for carbonyl groups in some amino acid
types (Asn, Thr, Lys) allows us to obtain significant structural
information also from the carbonyl atoms in this sample.
However, the absence of dominant CR-CO correlations at
short mixing times indicates a very low degree of simultaneous
labeling of CR and CO. In fact, the peaks observed in the CR-
CO region in the spectra at mixing times of 10 ms (not shown)
and 50 ms (Figure 3B) are primarily assigned to sequential
contacts (∼2.5 Å). These can be used to confirm the site-specific
assignments which have been obtained previously by 3D15N-
13C-13C correlation experiments.11Apart from the CO groups,
fractional13C enrichment is also found for the methyl groups
of Val, Leu, and Ile(Cγ2).
At larger mixing times of 200 and 500 ms (Figure 3C-D),
an increasing number of correlations are observed due to
recoupling of distant carbons. More than 700 peaks can be
resolved in the 500 ms spectrum. A summary of connectivities
observed as function of the mixing time is given in the
(33) Zech, S. G.; Olejniczak, E.; Hajduk, P.; Mack, J.; McDermott, A. E. J.
Am. Chem. Soc. 2004, 126, 13948-13956.
Figure 3. Aliphatic and carbonyl regions of the13C-13C DARR spectra of uniformly labeled microcrystalline ubiquitin (A) and selectively labeled [2-13C]-
Ubq (B-D). Spectrum A has been obtained at 233 K with a mixing time of 20 ms and shows mainly one-bond intraresidue correlations. Spectra B-D have
been measured at 245 K with mixing times of 50, 200, and 500 ms, respectively. With increasing mixing time, a larger number of long-range correlations
is observed.
Structure Determination of Microcrystalline Ubiquitin
A R T I C L E S
J. AM. CHEM. SOC. 9 VOL. 127, NO. 24, 2005 8621

Page 5

Supporting Information as Supplemental Figure 2. In principle,
from the analysis of the cross-peak buildup curves (i.e. peak
volume as function of the mixing time) it is possible to obtain
(semi-)quantitative distance information. However, many prob-
lems such as fractional labeling and peak overlap have to be
overcome to derive reliable information from buildup curves.
An alternative strategy for assignment of the long-range peaks
and distance classification is given below with the resulting
structure refinement based on the interresidue constraints.
Comparison of DARR and PDSD Experiments. In the
structure determination of the R-spectrin SH3 domain, the simple
PDSD experiment has been very successful in the observation
of long-range distance correlation up to about 7 Å.8This
experiment has lenient power requirements for recoupling carbon
atoms, i.e., no RF irradiation is applied during the mixing time.
In contrast, other sequences such as RFDR34require high-power
proton decoupling during the mixing time and cause sample
heating and potentially degradation if small dipolar couplings
need to be observed.
In13C-13C spin diffusion experiments, the two major factors
governing the magnetization transfer are the dipolar coupling
between the carbon atoms and the spectral overlap between the
spins involved. One way of increasing the spectral overlap
between13C spins is to recouple the1H-13C dipolar interaction
which then assists the magnetization exchange between carbon
atoms. This can be facilitated by applying a RF field on the
proton channel which is matched with multiples of the spinning
speed (ωror 2ωr).18Since an efficient magnetization transfer is
crucial for the detection of small dipolar couplings between
distant spins, it is important to compare the performance of the
DARR sequence with the previously used PDSD experiment.
Figure 4 shows spectral regions for the DARR experiment
(blue) overlayed with the PDSD experiment (magenta) both
measured at 500 ms mixing time. The two spectra are shown
with the same contour levels and have been measured back-
to-back under the same experimental conditions. They differ
only with respect to the proton irradiation during the mixing
time which has been matched with the spinning speed in case
for the DARR experiment (ω1
been omitted in the PDSD experiment (ω1
both experiments give very similar results with respect to the
peak positions and intensities of the diagonal peaks. However,
for many cross-peaks, the intensity is stronger in the DARR
experiment than for the PDSD experiment. Peaks with consider-
able difference in volume between the two experiments have
been assigned in Figure 4.
The distances in the X-ray structure (PBD entry 1ubq35)
between the assigned carbon atoms in Figure 4 range between
4.5 Å (Ser20-CR to Pro19-C?) to 6.7 Å (Val70-C? to Leu71-
Cγ). Distances in that range are very important for the
identification of tertiary contacts and secondary structure
elements. In fact, at 500 ms mixing the tertiary contact between
Ala28-CR and Pro38C? (5.5 Å) is much more pronounced under
DARR conditions and is already present at 200 ms mixing time.
Many other tertiary contacts have been identified only in the
DARR experiment, but are not present under PDSD conditions.
We therefore conclude that under the experimental conditions
H) ωr) 2π × 10 kHz) but has
H) 0). As expected,
chosen here, the DARR experiment has a better performance
in recoupling distant carbon atoms.
It is important to note that simply increasing the mixing time
in a PDSD experiment beyond 500 ms is unlikely to yield the
same correlations than observed with the DARR sequence at
500 ms, because both experiments behave similarly with respect
to the relaxation parameters for13C spins. Therefore, a PDSD
experiment with longer mixing time would yield significantly
lower overall signal intensity with many important correlation
remaining below the S/N level. Furthermore, extensive mixing
times would diffuse the spin magnetization randomly over larger
distances, but not necessarily reveal more structurally important
long-range correlations.
The improved recoupling efficiency of the DARR experiment
is expected to become even more pronounced at higher spinning
speeds where the1H-1H dipolar interactions are averaged out
more efficiently. However, since only low power1H irradiation
is required during the mixing time, sample heating is not a severe
problem for the DARR experiment, making it also a practical
choice for temperature sensitive protein samples, even if long
mixing times are required.
Assignment Strategy for Long-Range Correlations. With
the assignments obtained on the [U-13C]-Ubq sample, it is
straightforward to assign the spectrum of the [2-13C]-Ubq sample
at short mixing times (∼10 ms). Since [2-13C]-Ubq contains
(34) Bennett, A. E.; Ok, J. H.; Griffin, R. G.; Vega, S. J. Chem. Phys. 1992,
96, 8624-8627.
(35) Vijay-Kumar, S.; Bugg, C. E.; Cook, W. J. J. Mol. Biol. 1987, 194, 531-
544.
Figure 4. Comparison of magnetization transfer efficiency for the DARR
experiment (blue) with the PDSD experiment (magenta). Both spectra are
measured on [2-13C]-Ubq microcrystals at 245 K under identical conditions,
except for the1H-RF field applied during the 500 ms mixing time which
was 10 kHz for the DARR and 0 for the PDSD experiment. Both spectra
have been processed identically and are shown with the same contour levels.
Peaks with considerable difference in volume are shown with assignment.
A R T I C L E S
Zech et al.
8622 J. AM. CHEM. SOC.9VOL. 127, NO. 24, 2005

Page 6

only a small number of residues with adjacent13C-enriched
atoms, only intraresidue peaks are observed at this mixing time,
including e.g. Leu C?-Cγ, Val C?-Cγ, and Ile CR-C?. In
addition, a number of two-bond couplings is already observed
at 10 ms mixing time. The 50 ms spectrum shown in Figure
3B allowed us to identify most of the amino acids on the basis
to their chemical shifts due to intraresidue (one to three bond)
couplings between highly
spectrum exhibits also sequential contacts (see supplemental
Figure 2 in Supporting Information), particularly in the CR-
CO region which are used to confirm site specific assignment
for the residues observed only at low temperature (see above).
A statistical analysis of chemical shifts obtained at mixing
times of 10 and 50 ms revealed that for the majority of13C
atoms the chemical shifts are very well defined with rms
deviations in the order of 0.03-0.08 ppm. This has been
confirmed for various samples prepared from [2-13C]-Ubq and
hence demonstrates the reproducibility of the preparation
protocol. For some atoms, however, a deviation of the13C
chemical shift observed for the [2-13C]-Ubq sample compared
to the uniformly labeled sample has been observed and is
interpreted as residual dipolar shift due to the presence of one
bond13C-13C couplings in [U-13C]-Ubq.
The intraresidue and sequential assignments obtained at 10
and 50 ms have been copied to the 200 and 500 ms spectra.
For the assignment of the remaining peaks in the long mixing
time spectra, we used the following guidelines: (i) The
agreement of the chemical shifts must be better than (0.1 ppm,
which is range slightly larger than the average rms deviation
of the13C atoms and also closely resembles the line width for
the [2-13C]-Ubq sample. (ii) Only atoms which are labeled to
high degree in [2-13C]-Ubq are considered. (iii) Since sequential
contacts are usually shorter than medium or long-range contacts,
preference is given to the assignment of sequential peaks over
a medium or long-range contacts if both fulfill points i and ii.
This strategy allowed us to assign about 122 interresidue
peaks in the 500 ms spectrum unambiguously. Medium and long
range distances are depicted in supplemental Figure 1 in
Supporting Information. It turns out, that more than half of the
remaining peaks can be assigned unambiguously in one dimen-
sion. In many cases, the ambiguity in the second dimension
can be addressed by an analysis of the secondary structure
elements which are obtained separately e.g. from secondary
chemical shifts analysis. Furthermore, the assignment can be
assisted by ‘network anchoring techniques’ which evaluate the
self-consistency of long-range assignment independently of
knowledge of the three-dimensional structure.36,37Contacts
between secondary structure elements which are close in the
3D structure can thus be identified by generating a connectivity
matrix as shown in supplemental Figure 2 in Supporting
Information.
Moreover, paralleling many solution NMR structure refine-
ment procedures, ambiguities can be resolved by taking distance
information from a preliminary three-dimensional structure
model into account.36-38The peak list obtained so far contained
13C-enriched atoms. The 50 ms
already a significant number of unambiguous medium (35) and
long-range (28) contacts defining the global fold of the protein.
These are depicted in supplemental Figure 1 in Supporting
Information. We decided to use these restraints in a structure
calculation with the program CNS (see below for details). This
preliminary structure already showed the correct fold of the
protein within secondary structure elements, although many
regions were flexible due to lack of restraints. Nonetheless, most
of the one-dimensional ambiguities and many two-dimensional
could be resolved with the help of the preliminary structure
obtained in the first refinement step. A very similar strategy
for resolving ambiguities is used by the programs ARIA38or
CYANA36in an more elegant, automated fashion. This iterative
process has been repeated twice with the structure showing a
great improvement in the backbone rmsd values. For the last
cycle, also atoms with lower13C enrichment such as the CO
groups have been taken into account. Thereby, it was possible
to assign the wast majority of peaks in the 200, 300 and 500
ms spectra. A total of 336 unambiguous interresidue contacts
have been identified and were used for the subsequent structure
refinement.
A critical milestone for this procedure is the identification
of a sufficient set of unambiguous long-range correlations
defining the global protein fold. If the initial assignment contains
false constraints, the result of successive cycles could be biased
by the erroneous global fold of structures generated in the first
calculation. In this respect, the novel PASD algorithm39recently
implemented in the structure determination package Xplor-
NIH,40provides a highly error-tolerant approach for automated
peak-picking, constraint identification, and structure calculation
cycles. Once modified for the specific needs of ssNMR
applications (e.g. including13C-13C or15N-15N constraints,
calibration of distance ranges from dipolar recoupling experi-
ments, and proper handling of fractional labeling and intermo-
lecular contacts), this algorithm could dramatically speed up
the protein structure refinement from ssNMR data.
Identification of Intermolecular Contacts. In a highly
ordered system such as microcrystalline material, it should be
possible to observe contacts between ubiquitin molecules at the
crystal contact sites. If not identified properly, these could
potentially lead to a misinterpretation of the intramolecular
restraints and subsequently lead to problems in the structure
refinement. As demonstrated previously, intermolecular contacts
can be suppressed by diluting the isotopically enriched sample
with natural abundance material, mostly at the expense of
reduced signal intensity.6,8As a compromise between signal-
to-noise ratio and suppression of crystal contacts, we prepared
a sample containing 1/3 labeled and 2/3 unlabeled material.
Intermolecular contacts can thus be identified by comparing
signal intensities between the fully labeled and the ‘diluted’
sample.
As a representative example, the CR-CR region of the
spectrum of the diluted sample (green) is shown in Figure 5A
overlayed on the spectrum of the labeled sample (red), both
measured in a DARR experiment at 500 ms mixing time.
Relative peak intensities for all sequential and medium range
distances assigned in this spectral region are very similar, except
(36) Gu ¨ntert, P. Prog. Nucl. Magn. Reson. Spectrosc. 2003, 43, 105-125.
(37) Gu ¨ntert, P. Structure calculation using automated techniques. In BioNMR
in Drug Research; Zerbe, O., Ed.; Wiley-VCH: Weinheim, 2003; pp 39-
66.
(38) Linge, J.; Habeck, M.; Rieping, W.; Nilges, M. Bioinformatics 2003, 19,
315-316.
(39) Kuszewski, J.; Schwieters, C. D.; Garrett, D. S.; Byrd, R. A.; Tjandra, N.;
Clore, G. M. J. Am. Chem. Soc. 2004, 126, 6258-6273.
(40) Schwieters, C.; Kuszewski, J.; Tjandra, N.; Clore, G. J. Magn. Reson. 2003,
160, 66-74.
Structure Determination of Microcrystalline Ubiquitin
A R T I C L E S
J. AM. CHEM. SOC. 9 VOL. 127, NO. 24, 2005 8623

Page 7

for the pair of cross-peaks at 56.1/51.7 ppm (indicated by
arrows). This pair shows a considerable differences in peak
intensities on both sides of the diagonal with the signal being
hardly above the noise level for the diluted sample. It turns out
that those peaks can be unambiguously assigned in one
dimension to the CR atom of Ala46. In the second dimension,
only three atoms (all CR) fall into a (0.1 ppm range with their
distances to the CR atom of Ala46 given in parentheses: Asp32
(24.7 Å), Glu41 (16.6 Å), and Lys11 (19.3 Å).
Figure 5B shows the location of these residues in the X-ray
structure (1ubq). Indeed, Ala46 is located within a loop region
with its CR facing to the outside of the protein, hence making
intermolecular contacts feasible. The residues Asp32 and Lys11
are located on opposite sides of the protein and can act as
potential coupling partner by means of intermolecular contacts.
On the other hand, a contact with Glu41 seems to be unlikely
since (i) the CR of Glu is not13C enriched to a high degree in
the [2-13C]-Ubq sample and (ii) Glu41 faces to the inside of
the protein making the CR atom difficult to access from the
outside. Along those lines, several other crystal contacts have
been identified. Indeed, a large number of weak peaks which
could not be assigned in the spectra of the fully labeled sample
are not present in the diluted sample and are therefore likely to
represent intermolecular contacts. Consequently, those have been
excluded from the restraint list.
It is worth noting that problems with intermolecular contacts
may not arise if amorphous precipitates are used, i.e., if no long
range order is present. In this case intermolecular contacts are
not expected to be predominant. Nonetheless, hydrated protein
precipitates can show line widths similar to those observed for
microcrystals,32,33indicating a very defined environment (local
order) for both microcrystals and amorphous precipitates. On
the other hand, the observation of well-defined crystal contacts
proves that even under tough experimental conditions such as
high g-force due to fast MAS, or periodically heating of the
sample due to high power proton decoupling and subsequent
cooling, the microcrystals stay intact over a long period of time
with a large degree of long-range order present.
Structure Refinement of Microcrystalline Ubiquitin. Using
the iterative procedure described above, it was possible to
identify 336 interresidue contacts for the [2-13C]-Ubq sample.
Out of those, 149 were sequential, 74 were medium range (1 <
|i - j| e 4) and 113 were long range contacts. For the
calculation of the protein structure, these restraints need to be
assigned to specific distance ranges preferentially on an
experimental basis. In principle, the relative cross-peak volumes
are expected to correlate with the internuclear distances and with
the mixing time. Unfortunately, a comparison of the peak
volumes in the spectra at long mixing times with the distances
in the X-ray structure (1ubq) showed only poor correlation. Even
(short) sequential or intraresidues peaks were found to have a
wide range of peak volumes in the DARR spectra. We ascribe
this phenomenon to fractional labeling or insufficient suppres-
sion of dipolar couplings as discussed previously.8
We therefore decided to assign distance ranges on the basis
of the well-defined secondary structure elements8,9and the first
appearance of the peaks at particular mixing times. Sequential
contacts are restraint to the appropriate distances given by the
peptide geometry which include e.g. sequential CO-CR: 2.1-
2.9 Å, sequential CR-CR: 2.5-5.0 Å, sequential CR-C?:
2.5-5.0 Å, sequential C?-C?: 2.5-6.0 Å. Distances in
secondary structure elements are assigned according to their
average values as analyzed previously8,9and include e.g. the
CRi-C Ri+3distances in an R-helix which are restrained to 2.5-
5.0 Å. All other medium and long range peaks in the 500 ms
spectrum are initially assigned to a range between 2.5 and 7.5
Å. If one of those peaks is identified already at 300 ms or 200
ms mixing time, the distance range is reduced to 2.5-6.5 Å
and 2.5-5.5 Å, respectively. These upper limits for the distance
bins have been motivated by histograms (see supplemental
Figure 3 in Supporting Information) depicting the number of
identified contacts within a certain X-ray distance range as
function of the mixing time. Nonetheless, even without informa-
tion from the X-ray structure, a refinement would have been
possible with uncalibrated distances ranges, although with likely
larger rmsd’s for the resulting NMR structure. A more quantita-
Figure 5. (A) Comparison of peaks intensities at 500 ms mixing time for
a sample containing 100% [2-13C]-Ubq (red) with a diluted sample
containing 33% [2-13C]-Ubq and 67% natural abundance ubiquitin (green).
Peaks with considerable difference in intensity are likely to represent
intermolecular contacts. The peaks marked with an arrow are assigned to
CR of Ala46 in one dimension. (B) Location of the CR atoms (spheres) of
Ala46, Asp32, Lys11. and Glu41 in the X-ray structure of ubiquitin (1ubq).
With respect to Ala46, the CR atoms of Asp32 and Lys11 are located on
the opposite site of the protein, making a crystal contact with Ala46 possible.
A R T I C L E S
Zech et al.
8624 J. AM. CHEM. SOC.9VOL. 127, NO. 24, 2005

Page 8

tive range for the distance bins could then be obtained by
detailed, thus time-consuming, analysis of the restraint and van
der Waals violations.
Interestingly, many of the contacts close to the upper distance
limits involve amino acids with two adjacent enriched13C-atoms
(i.e. Val, Leu, Ile). Is is likely that these contacts represent a
two-step magnetization transfer, i.e., a ‘fast’ one-bond transfer
followed by a long-range transfer over a shorter distance than
the indicated one. This raises an additional concern for the
quantitative analysis of cross-peak buildup curves, since very
detailed experiments would be required to distinguish between
one-step and multistep transfer. It should be noted that the upper
limit of observable distances depends not only on the mixing
time, but particularly on the S/N of the spectra (i.e. sample
amount, labeling scheme, magnetic field, spectrometer sensitiv-
ity, etc.) and might therefore vary from one experiment or
sample to another.
The resulting list of interresidue distance restraints has been
subjected to a structure calculation using the default protocols
of CNS and a linear conformer as the initial structure.31More
than 600 structures were calculated, which showed no more than
one violation in any of the following categories: bond lengths
(>0.1 Å), bond angles (>5°), improper dihedral angles (>5°)
and van der Waals contacts (<1.6 Å). The 10 lowest energy
structures have been selected for further analysis. The rmsd for
the backbone heavy atoms of this structure ensemble is shown
in Figure 6C as dashed line. It turned out that the structure
bundle obtained sorely from distance constraints exhibits the
correct global fold of the protein, but was still flexible even in
secondary structure elements with respective backbone rmsd’s
between 1 and 3 Å.
To further improve the quality of the structure, we included
restraints for the backbone dihedral angles, θ and φ, obtained
from secondary chemical shift analysis performed by TALOS.23
The TALOS database used here contains 78 proteins with more
than 9000 residue triplets. To avoid a bias, we deleted the
ubiquitin entry from the database. Chemical shifts obtained from
the ssNMR assignment for backbone15N, CR, C?, and CO were
used as input.11,12TALOS generated restraints classified as
‘good’ for 61 out of 76 residues. Accordingly, 122 restraints
for θ and φ including respective error margins given by TALOS
have been used for the subsequent structure refinement. For this
final step, both distance and dihedral angle restraints have been
used as input with the average structure obtained from distance
restraints only used as starting structure. The resulting bundle
of 10 structures lowest in energy is shown in Figure 6A and B
as represented by the backbone heavy atoms. For comparison,
the X-ray structure (1ubq) is depicted as thick line.
The structure based on distance and dihedral angle constraints
shows very well-defined secondary structure elements such as
the R-helices with rms deviations in the backbone heavy atoms
between 0.7 and 1.1 Å (Figure 6C). It furthermore closely
resembles the X-ray structure (1ubq). Larger flexibility is
observed only for regions with small number of restraints such
as the loop between Leu8 and Lys11, and the residues near the
C-terminus. This correlates well with the temperature-dependent
observation of residues discussed above. Residues in mobile
regions often lack intensity due to inefficient cross-polarization,
making their observation by ssNMR methods more difficult.
Deviations between X-ray and ssNMR structure are therefore
likely due to lack of experimental restraints rather than caused
by different crystal packing.
Conclusion
The 3D structure of microcrystalline ubiquitin has been
obtained from solid-state MAS NMR data providing a large
Figure 6. (A) Side and (B) top view of the structure ensemble for the 10
lowest-energy structures of ubiquitin calculated using distance constraints
obtained from the DARR experiments and backbone dihedral angle
constraints provided by TALOS. The thick line represents the backbone of
the X-ray structure 1ubq for comparison. Secondary structure elements such
as R-helices and ?-sheets are shown in red and blue, respectively. The
backbone atoms of well-defined regions of secondary structure have been
fitted to minimum rmsd using MOLMOL 2K.1.44(C) Backbone heavy atom
rmsd values obtained for the 10 lowest energy structures as function of
residue number. The dashed line represents the structure obtained from
distance constraints only. The solid line shows the rmsd’s obtained for the
structure bundle shown in A and B based on both distance and dihedral
angle constraints. Regions with well-defined secondary structures show
rmsd’s between 0.7 and 1.1 Å while mobile regions such as the C-terminus
are poorly defined.
Structure Determination of Microcrystalline Ubiquitin
A R T I C L E S
J. AM. CHEM. SOC. 9 VOL. 127, NO. 24, 2005 8625

Page 9

number of interresidue carbon-carbon restraints. The quality
of the structure could be improved by an analysis of secondary
chemical shifts providing backbone dihedral angle restrains.
Essentially, only two separate sample preparations with different
13C labeling schemes and a small number of experiments have
been used for both resonance assignment11,12and structure
determination of ubiquitin.
This study represents the second example of a protein
structure derived ab initio from high-resolution ssNMR data
obtained on hydrated microcrystals. Our approach shows large
similarities to the structure determination of the SH3 domain
with respect to labeling schemes and experiments used.8In the
initial SH3 structure refinement, about 300 distance restraints
were obtained mainly from 2D carbon correlation experiments
on selectively13C-enriched SH3 microcrystals, grown with either
[2-13C]-glycerol or [1,3-13C]-glycerol as a carbon source. Both
samples give rise to complementary labeling schemes, which
simplifies the identification of long-range contacts.8
In contrast, in our study ambiguities in the assignment of
carbon-carbon restraints have been resolved using an iterative
approach between assignment and structure refinement. This
represents an alternative route which can be easily automated
using recently developed software. Eventually, 336 interresidue
distance restraints complemented by 122 dihedral angle restraints
obtained from chemical shifts analysis, gave rise to a 3D
structure with rmsd’s of ∼1 Å or less for secondary structure
elements. Hence, the quality of the ubiquitin structure presented
here is similar to that of the SH3 domain with backbone rmsd’s
of ∼1.6 Å for secondary structure elements.8This demonstrates
the general applicability of this method to determine 3D
structures of proteins with moderate size.
Clearly, our study represents only a first step toward
refinement of protein structures on the basis of ssNMR data.
As demonstrated recently, increasing the spectral resolution by
means of 3D15N-13C-13C correlation experiments will increase
the number of distance restraints dramatically.9Furthermore,
by using novel experiments and advanced labeling schemes, it
is feasible to complement the13C-13C restraints by15N-15N
and1H-1H and backbone dihedral angle restraints.8,20-22,41
Higher dimensional spectra and novel labeling schemes will
increase the number of unambiguous restraints necessary for
the iterative assignment/structure calculation approach discussed
above and will therefore extend the applicability of this method
to more complex systems.
Soon, the combination of those techniques will enable a
refinement of 3D structures for solid protein formulations to a
high quality. This has been demonstrated already for the
structure of peptides which could be resolved at truly atomic
detail.42,17Furthermore, it is likely that the complexity of the
systems accessible to ssNMR studies at high-resolution will
increase beyond the 12-15 kDa range soon. Since the tech-
niques discussed above are independent of molecular tumbling
times, ssNMR can be ideally applied to large molecular weight
systems of medium complexity, such as smaller membrane
bound proteins in their natural environment, or for the charac-
terization of protein-protein and protein-peptide interactions
in high molecular weight matrixes.
The experiments described here can be performed on solid
sample formulations, such as microcrystalline material or
hydrated precipitates and are therefore not limited to highly
soluble globular proteins. Although still in its infancy, high-
resolution MAS solid-state NMR spectroscopy provides an
exciting opportunity to solve structures of proteins not amenable
to solution NMR studies or X-ray crystallography, such as
insoluble proteins, protein aggregates, and membrane proteins
for which considerable progress has been reported recently.43
Acknowledgment. We thank T. Igumenova (U. Pennsylvania)
for providing the assignments of microcrystalline ubiquitin prior
to publication. S.G.Z. acknowledges the Alexander-von-Hum-
boldt Foundation for financial support. This work was supported
by grants from NSF, MCB 0316248 (to A.E.M.) and NIH, DK
39806 (to A.J.W.). A.E.M. is member of the New York
Structural Biology Center. Support for the NYSBC has been
provided by NIH/NIGMS through grant P41 GM66354.
Supporting Information Available: Supplemental Figures
1-3. This material is available free of charge via the Internet
at http://pubs.acs.org.
JA0503128
(41) Reif, B.; van Rossum, B. J.; Castellani, F.; Rehbein, K.; Diehl, A.;
Oschkinat, H. J. Am. Chem. Soc. 2003, 125, 1488-1489.
(42) Rienstra, C. M.; Tucker-Kellogg, L.; Jaroniec, C. P.; Hohwy, M.; Reif, B.;
McMahon, M. T.; Tidor, B.; T. L.-P.; Griffin, R. G. Proc. Natl. Acad.
Sci., U.S.A. 2002, 99, 10260-10265.
(43) Krabben, L.; van Rossum, B.-J.; Castellani, F.; Bocharov, E.; Schulga, A.
A.; Arseniev, A. S.; Weise, C.; Hucho, F.; Oschkinat, H. FEBS Lett. 2004,
564, 319-324.
(44) Konradi, R.; Billeter, M.; Wu ¨thrich, K. J. Mol. Graphics 1996, 14, 51-
55.
A R T I C L E S
Zech et al.
8626 J. AM. CHEM. SOC.9VOL. 127, NO. 24, 2005