eth-22701-02 research collection
TRANSCRIPT
ETH Library
Anisotropic dynamics in molecularsystemsstudied by NMR relaxation
Doctoral Thesis
Author(s):Lienin, Stephan Frank
Publication date:1998
Permanent link:https://doi.org/10.3929/ethz-a-001990510
Rights / license:In Copyright - Non-Commercial Use Permitted
This page was generated automatically upon download from the ETH Zurich Research Collection.For more information, please consult the Terms of use.
Diss. ETH «*• 3Diss. ETH No. 12871
Anisotropic Dynamics in Molecular Systems
Studied by NMR Relaxation
Dissertation
for the degree of
Doctor of Natural Sciences
submitted to the
Eidgenossische Technische Hochschule
Zurich
presented by
Stephan Frank Lienin
Dipl. Chem. ETH
born March 23, 1970
citizen of Germany
accepted on the recommendation of
Prof. Dr. Richard R. Ernst, examiner
Prof. Dr. Rafael Bruschweiler, co-examiner
Zurich, 1998
Fur meine Eltern
Contents 1
Contents
Contents 1
Symbols and Abbreviations 5
Abstract 11
Zusammenfassung 13
1 Introduction 15
2 Spin Relaxation and Molecular Motion 21
2 1 Fundamentals of spin relaxation 21
211 Equation of motion 21
2 1 2 Semiclassical relaxation theory 22
2 1 3 Relaxation-active interactions 24
2 14 Operator representation of the relaxation superoperator 27
2 2 Relaxation-active molecular motion 28
2 2 1 Correlation functions for molecular motion 28
222 Overall rotational diffusion 30
2 2 2 1 Isotropic rotational diffusion in a homogeneous
environment 30
2 2 2 2 Rotational diffusion in a heterogeneous environment 34
2 2 3 Intramolecular motion 35
2 2 3 1 Intramolecular backbone motion in proteins 35
2 2 3 2 Internal correlation functions and model-free approach 38
2 2 3 3 Extraction of amsotropy of intramolecular motion 41
2 2 3 4 Analytical treatment of the 3D GAF model 45
2 3 Calculation of relaxation-rate constants 47
2 4 Molecular dynamics simulation 53
24 1 The force field 53
2 4 2 Calculation of correlation functions 54
2 4 3 Extraction of 3D GAF fluctuation amplitudes 54
2 5 Motion-induced fluctuations of the CSA 57
2 5 1 Analysis of CSA trajectory 58
25 2 CSA averaging due to 3D GAF motion 60
3 Anisotropic Intramolecular Backbone Dynamics of Ubiquitin 63
3 1 Introduction 63
3 2 Ubiquitin 64
3 3 Experimental 67
3 3 1 Sample preparation 67
2 Contents
3 3 2 NMR experiments 68
3 3 2 1 General setup and techniques 68
3 3 2 2 2D N relaxation experiments 68
3 3 2 3 2D C relaxation experiments 72
3 3 3 Extraction of relaxation data 73
3 4 Molecular dynamics simulation 75
3 4 1 Generation of molecular dynamics trajectory 75
3 4 2 Processing of trajectory 75
3 4 2 1 Determination of equilibrium peptide plane 75
3 4 2 2 Extraction of fluctuation amplitudes from the trajectory 76
3 4 3 Analysis of MD trajectory 77
3 4 3 1 Selection of peptide planes with 3D GAF motion 77
3 4 3 2 Results for the extracted fluctuation amplitudes 80
3 4 3 3 Dihedral angles and plananty of the peptide plane 81
3 4 3 4 Orientation of the peptide plane frame 82
3 5 Analysis of experimental relaxation data 83
3 5 1 Raw data analysis 83
3 5 2 Spin relaxation mechanisms in the peptide plane 87
35 3 Which parameters can be extracted from the experimental data9 91
354 Fit results for 3D GAF model 95
3 5 5 Uncertainty estimates of fit parameters 97
3 5 6 Comparison between the 3D GAF analysis and a 15N
model-free analysis 102
3 6 Discussion 104
36 1 Amsotropy of peptide-plane dynamics 104
3 6 2 Correlation with secondary structure 107
3 7 Conclusions 110
4 CSA Fluctuations Studied by MD simulation and DFT Calculations 111
4 1 Introduction 111
4 2 MD simulation and DFT calculations 112
4 3 Average 15N and 13C CSA tensors 116
43 1 Processing of the CSA trajectory 117
432 Analysis of 15N CSA tensors 118
4 3 3 Analysis of 13C CSA tensors 121
4 3 4 Comparison of different averaging approaches 123
4 4 Fluctuations of the CSA tensors 126
4 4 1 Fluctuation of CSA amsotropy and asymmetry 126
4 4 2 Onentational fluctuations of the CSA tensors 128
4 5 15N and 13C CSA relaxation of the protein backbone 131
4 5 1 Implications of CSA fluctuations on CSA relaxation 131
4 5 2 CSA relaxation based on average CSA tensors 137
Contents 3
4 5 3 Role of the antisymmetric part of CSA tensor 140
4 6 Conclusions 142
5 Rotational Motion of a Solute Molecule in a Highly Viscous Liquid 145
5 1 Introduction 145
5 2 Experimental 148
52 1 Charactenzation of the system and sample preparation 148
5 2 2 NMR relaxation measurements 150
5 3 Results and Discussion 152
53 1 Test of a single-correlation-time model 152
5 3 2 Distribution of correlation times 154
5 3 3 Analysis of the temperature-dependent relaxation data 159
5 4 Conclusions 162
6 Appendix 163
References 179
Acknowledgements 191
Curriculum Vitae 193
4 Contents
Symbols and Abbreviations 5
Symbols and Abbreviations
Symbols
''„ <v<7?V
A Eigenoperator of the static Hamiltoman H0
AM) Scalar vanable of stochastic process it
A Angstrom, 10 10m
a^ Coefficients for the expansion of the tensor operator T^ mof the
interaction (i in terms of eigenoperators
B, Bm, Bn Spin operators
B0 External magnetic field vector
B0 Strength of the magnetic field along z direction
,(?) General cross-correlation function between the irreducible tensor
operators of the interactions it and v
C„v(0 , Cnv(t) Liquid state cross-correlation functions for rank 2 interactions |i and v
in the laboratory and molecular frame, respectively
ct(t) Time-dependent coefficient for expansion of state vector |\|/(f))
cmnO Time-dependent coefficient for expansion of probability /(<P, t)
c, cv Strengths of the interactions it, v, respectively
D Diffusion constant
D Axially symmetric dipolar coupling tensor
Dmn Wigner rotation matrix element of rank /
dmn Reduced Wigner rotation matrix element of rank /
Et Eigenvalue of the static Hamiltoman H
ew en, e Pnncipal axis system of the 3D GAF model
e Unit vector along the direction of the interaction pnncipal axis \i
/(O, f) Probability of finding a director at orientation O at time t
H Static Hamiltoman of a quantum mechanical system
H(t) Time-dependent Hamiltoman of the spin system including the
interaction with the environment
HCSA Hamiltoman of the antisymmetric part of the CSA interaction of the
spin system
6 Symbols and Abbreviations
Ts\m
HCSA Hamiltoman of the symmetnc part of the CSA interaction of the spin
system
HD Dipolar Hamiltoman of the spin system
Hz Zeeman Hamiltoman of the spin system
H0 Static Hamiltoman of the spin system
H, (t) Time-dependent Hamiltoman representing a stochastic perturbation
h Planck constant, h = 6 6261 10~34 Is
/ Spin operator
^11 a a({0) General cross-spectral density function between irreducible tensor
compounds of the interactions \i and v
k Boltzmann constant, k = 1 3806 10"23JK l
L Rank of the irreducible tensor operator of interaction |0.
P/(Q) Legendre polynomial of rank /
P(O0|<I>(f)) Conditional probability to find a director at the orientation c£> at time t
if it was at position <t>0 at time t = 0
p(%<_) Distribution function for overall tumbling correlation times Tc
Rh Hydrodynamic radius
R Correlation coefficient
rls Vector between the / and the S spin
rIS Distance between the 7 and the S spin
/ff Effiwt^o 15W_1
J\i\
Effective N-'H distance including averaging over NH stretching and
bending vibrations
£ Spin operator
2S Generalized rank 2 order parameter of the interactions |i and v
Absolute temperature in Kelvin
T^ Irreducible tensor operator of interaction \i
T, Longitudinal spin-lattice relaxation time
T| p
Rotating-frame transverse relaxation time
T2 Transverse spin-spin relaxation time
Tr{A) Trace of matrix A
1 Time
tJ Spin operator
P* Spin operator
Symbols and Abbreviations 7
Yt Normalized spherical harmonic of rank /
a, p, y Fluctuation angles about the principal axes ew e», e of the 3D GAF
model
PCD Parameter of Cole-Davidson spectral density
$kkw Parameter of the Kohlrausch-Wilhams-Watts stretched exponential
function
f Relaxation superoperator
Tmn Matnx element of the relaxation superoperator f" representing a
transfer rate constant between the base operators Bm and Bnis
^lons Matnx containing the rate constants describing longitudinal relaxation
processes of a spin system IS
is
Ttrans Matrix containing the rate constants descnbing transverse relaxation
processes of a spin system IS
y7, ys Gyromagnetic ratios of spins / and 5, respectively
Ac Amsotropy of the CSA tensor g
2
^anti Quadratic sum over antisymmetnc matrix elements of the CSA tensor
5(f) Delta function
ti Macroscopic viscosity
T| Asymmetry of the CSA tensor g
9„v(0 Angle between the interaction pnncipal axes |i and v
Xc Scaling factor for the 13C CSA principal values
XN Scaling factor for the 15N CSA pnncipal values
(i Global multiplier for all interaction strengths in the peptide plane
|i0 Permeability of vacuum, |i0 = 4ti 10
% Rotational fnction coefficient
p (t) Density operator of the spin system
p0 Equihbnum density operator of the spin system
g Chemical shielding amsotropy tensor
a1,0|| Pnncipal values of an axially symmetnc CSA tensor The
corresponding pnncipal axes are orthogonal and parallel to the
symmetry axis of the tensor
gsym Symmetnc part of the chemical shielding amsotropy tensor g
0'
Antisymmetnc part of the chemical shielding amsotropy tensor g
Symbols and Abbreviations
ax Difference between <5XX and azz
G„ Difference between o,„ and a„y yy zz
axx, a , azz Pnncipal values of the chemical shielding anisotropy tensor g
o, axz, o~VT Matnx elements of the antisymmetric part of the chemical shielding
anisotropy tensor g
oa, Oq, ay Fluctuation amplitudes for fluctuations about the axes ea, eB, e of
the 3D GAF model
aj,
Variance of the normalized spherical harmonic Y2 r
t Time
xc Overall tumbling correlation time
x^Jr Effective correlation time for cross-correlation function of interactions
(i and v
Tmf Internal correlation time
v^: Correlation time for internal cross-correlation function of interactions
(X and v
xm Mixing time in relaxation experiment
<t> Set of Euler angles relating the molecular and the laboratory frame
cp( Backbone dihedral angle of the fragment C\ _,- N, - C" - C\
2
X Least squares fitting error
\\lL Backbone dihedral angle of the fragment Nl-Cl - C\ - N, + ,
\\\i(t)) State function for a closed quantum mechanical system
1^) Eigenvectors of the static Hamiltoman H
Q (?) Time-dependent set of polar angles of the direction of the interaction
pnncipal axis (X in the defined reference system
ii°
(0 Time-dependent set of polar angles of the direction of the interaction
principal axis |x in the laboratory frame
O!"" (t) Time-dependent set of polar angles of the direction of the interaction
pnncipal axis |U in the molecular frame
co. Backbone dihedral angle of the fragment C, _, -C'(_ , -r^-C,
co Angular frequency in rads-l
co Eigenfrequency of Hamiltoman H0 in rads
COy Larmor frequency of spin X in rads
Symbols and Abbreviations 9
Abbreviations
2D Two-Dimensional
3D Three-Dimensional
CHARMM Chemistry at HARvard MacroMolecular dynamics
CSA Chemical Shielding Amsotropy
D Dipole-dipole interaction
GAF Gaussian Axial Fluctuation
HSQC Heteronuclear Single-Quantum Correlation
INEPT Insensitive Nuclei Enhanced by Polanzation Transfer
LRM Long-Range Motion
lab Laboratory frame
MD Molecular Dynamics
mol Molecular frame
NMR Nuclear Magnetic Resonance
NOE Nuclear Overhauser Enhancement factor
PP Peptide plane
ppf Peptide plane frame
ppm Parts per million
rf Radio frequency
SRM Short-Range Motion
10 Symbols and Abbreviations
Abstract 11
Abstract
In this thesis, experimental NMR relaxation methods and theoretical
techniques are combined to extract information about anisotropic dynamics
of molecular systems in the liquid state.
In the first part, the anisotropy of rapid fluctuations of the peptide planes15 13
in ubiquitin is explored by combined N and C nuclear spin relaxation
measurements and molecular dynamics computer simulation. T{, T2, and
NOE data are collected at B0 -field strenghts corresponding to 400 and
600 MHz proton resonance. A 1.5 ns simulation of ubiquitin in an explicit
water environment is performed using CHARMM 24. The simulation
suggests that, for 76% of the peptide planes, the relaxation-active motion of15 13
the backbone N and C spins is dominated by anisotropic Gaussian
axial fluctuations of the peptide planes about three orthogonal axes. The
dominant fluctuation axes are nearly parallel to the C{ _ Y- Cj axes. The
remaining peptide planes belong to more flexible regions of the backbone
and cannot be described by this type of motion alone. Based on the results
of the computer simulation, an analytical 3D GAF motional model is
applied to the experimental relaxation data. The fluctuation amplitudes of
the peptide planes show a significant anisotropy of the internal motion. This
15 13analysis demonstrates that a combined interpretation of N and C
relaxation data by a model derived from a computer simulation may provide
detailed insight into the fast time-scale backbone dynamics that goes
beyond the results of a standard model-free analysis.
In the second part, density functional theory calculations of chemical
shielding anisotropics (CSA) and a molecular dynamics simulation are
combined to a CSA trajectory which contains the fluctuations of the CSA15 13
tensor induced by intramolecular motion. For the N and C CSA
tensors, located in three different peptide planes in the a helix, a ($ strand,
and a loop region in ubiquitin, it is found that the fluctuations characterized
as standard deviations of the chemical shielding anisotropy distributions do
not exceed 10%. The anisotropics of the calculated average CSA tensors
differ only slightly between the different secondary structure elements. An
analysis of the CSA trajectories indicates that the fluctuations of the CSA
12 Abstract
tensors do not have to be explicitly taken into account for a description of
CSA relaxation and thus the use of a locally averaged CSA tensor is
sufficient for the back-calculation of relaxation rate constants. The effect of
motional averaging of the CSA tensors, which should be considered when
CSA tensors obtained by solid state NMR experiments at room temperature
are used for the interpretation of CSA relaxation in the liquid state, is
estimated for different values of the motional parameters. This analysis is a
first step towards an improved understanding of CSA interaction strengths
of backbone spins in proteins which is important to reduce uncertainties of
extracted motional parameters in NMR relaxation studies.
In the third part, it is investigated whether the viscosity-dependent
retarding effect of a polymeric solvent on the rotation of solute molecules
can be used to shift the NMR observation window for the timescale of
intramolecular motion. It is found that the 13C NMR relaxation
measurements of the model system 1,3-dibromoadamantane in highly
viscous polymeric chlorotrifluoroethene can be explained neither by
isotropic nor by realistic anisotropic tumbling in a single environment. The
experimental data are rationalized in terms of fast exchange between at least
two environments with correlation times differing by up to two orders of
magnitude. This demonstrates that a uniform retardation of molecular
tumbling by a polymeric solvent is not always feasible.
Zusammenfassung 13
Zusammenfassung
In der vorliegenden Arbeit werden NMR-Relaxationsexperimente und
theoretische Methoden miteinander kombiniert, um die anisotrope Dynamik
in molekularen Systemen in fliissiger Phase zu charakterisieren.
Im ersten Teil wird die Anisotropic schneller Fluktuationen der
Peptidebenen im Protein Ubiquitin untersucht, indem man 15N- und 13C-
Kernspinrelaxationsmessungen und eine Molekuldynamik-Computer-
simulation miteinander verkniipft. Hierbei werden Tl, T2 und NOE-
Relaxationsdaten bei Z?0 -Feldstarken entsprechend 400 und 600 MHz fur
die Protonenresonanz gemessen. Desweiteren wird eine 1.5 ns lange
Simulation von Ubiquitin einschlieBlich expliziter Wasserumgebung mittels
CHARMM24 durchgefuhrt. Die Simulation zeigt, daB bei 76% aller15 13
Peptidebenen die relaxationsaktive Bewegung der N - und C -Kernspins
im Backbone durch anisotrope Fluktuationen der Peptidebenen um drei
orthogonale Achsen beschrieben werden kann, wobei die Verteilung der
Fluktuationswinkel um jede der Achsen einer GauB-Veiteilungsfunktion
geniigt (3D GAF-Modell). Dabei liegen die Achsen, um welche die groBten
Fluktuationen stattfinden, nahezu parallel zu den C1 _ t- C; -Achsen. Die
restlichen Peptidebenen gehoren zu flexibleren Regionen des Backbone und
konnen nicht mittels dieser Art von Bewegung beschrieben werden.
Aufgrund der Resultate der Computersimulation wird eine analytische Form
des 3D GAF-Modells verwendet, um die experimentellen Relaxationsdaten
zu interpretieren. Die extrahierten Fluktuationsamplituden der Peptidebenen
spiegeln eine signifikante Anisotropic der internen Bewegung wider. Die15 13
Analyse zeigt, daB die kombinierte Interpretation von N- und C-
Relaxationsdaten mittels eines Modells, welches anhand einer
Computersimulation entwickelt wurde, detaillierte Einsicht in die schnelle
Backbonedynamik ermoglicht und so iiber die Resultate einer modellfreien
Analyse hinausgeht.
Im zweiten Teil werden Berechnungen der chemischen Verschiebungs-
anisotropie (CSA) mittels Dichtefunktional-Theorie und eine Molekiil-
dynamik-Simulation kombiniert. Es resultiert eine CSA-Trajektorie, welche
die durch die intramolekulare Bewegung induzierten Fluktuationen des
14 Zusammenfassung
CSA-Tensors enthalt. Im Falle der N - und C -CSA-Tensoren von drei
Peptidebenen in der a-Helix, in einem P-Faltblattstrukturelement und in
einer Schleifenregion von Ubiquitin iiberschreiten die Fluktuationen,
welche durch die Standardabweichungen der Verteilungen der chemischen
Verschiebungsanisotropie gekennzeichnet sind, nicht die 10%-Marke. Die
Anisotropien der berechneten, durchschnittlichen CSA-Tensoren
unterscheiden sich kaum in den unterschiedlichen Sekundarstruktur-
elementen. Eine Analyse der CSA-Trajektorien zeigt, daB die Fluktuationen
der CSA-Tensoren nicht explizit fiir die Beschreibung der CSA-Relaxation
beriicksichtigt werden miissen. Stattdessen geniigt es, lokal gemittelte CSA-
Tensoren fur die Berechnung der Relaxationsratenkonstanten zu verwenden.
Weitergehend wird die Bewegungsmittelung von CSA-Tensoren fiir
verschiedene Werte der Bewegungsparameter abgeschatzt. Dieser Effekt
sollte beriicksichtigt werden, wenn die bei Raumtemperatur mittels
Festkorper-NMR ermittelten CSA-Tensoren fiir die Interpretation von CSA-
Relaxation in fliissiger Phase verwendet werden. Die vorgelegte Analyse
stellt einen ersten Schritt in Richtung eines verbesserten Verstandnisses der
CSA-Interaktionsstarke fiir Kernspins im Backbone von Proteinen dar. Dies
ist wichtig, urn die Unbestimmtheit der extrahierten Bewegungsparameter
in NMR-Relaxationstudien zu verringern.
Im dritten Teil wird untersucht, ob der viskositatsabhangige
Verlangsamungseffekt eines polymeren Lbsungsmittels auf die Rotation
eines gelosten Molekiils dazu verwendet werden kann, das NMR-
Beobachtungsfenster fiir die Zeitskala intramolekularer Bewegung zu
verschieben. Es zeigt sich, daB die C-NMR-RelaxationsmeBdaten des
Modellsystems 1,3-Dibromadamantan in hochviskosem Polychlortrifluor-
ethylen weder mit isotroper noch realistisch anisotroper rotatorischer
Diffusion in einer Umgebung erklart werden konnen. Die experimentellen
Daten werden durch einen schnellen Austausch zwischen mindestens zwei
Umgebungen rationalisiert, wobei die Korrelationszeiten sich um bis zu
zwei GroBenordnungen unterscheiden. Dies zeigt, daB die einheitliche
Verlangsamung der rotatorischen Diffusion durch polymere Losungsmittel
nicht immer moglich ist.
Introduction 15
1 Introduction
For a complete description of a molecule and its interaction partners it is
crucial to characterize both structure and dynamics. Molecules are normally
described and visualized by a static arrangement of atoms in space, which
however represents only an average structure or a single snapshot of a
molecule on the move. Intramolecular motion and, in liquid systems, also the
overall tumbling motion are inherently present and have to be taken into
account for the interactions between molecules. This is exemplified best for
proteins which play a fundamental role in biological systems [1,2]. Proteins
show intramolecular dynamics within a wide time window ranging from sub-
picoseconds to seconds. Their functional role relies on intramolecular
flexibility as for example in molecular recognition processes to achieve an
optimum match [3] or in biochemical transformations where the enzymes
adopt different conformational states [1].
Some highly successful experimental methods have been established for
the elucidation of a molecular structure. In particular, X-ray crystallography
[4,5], neutron diffraction [6], and nuclear magnetic resonance (NMR) [7,8]
provide accurate "images" of molecules. Intramolecular motion is only
indirectly contained in the elucidated structures which, in fact, result from
averaging over the ensemble of molecules and different conformations
present during the experiment. For studying the molecular motion
experimentally, NMR is at present the most versatile and powerful tool [7,9-
11]. Fast motional processes in the picosecond to microsecond range can be
sampled by relaxation experiments. Here, a certain state of polarization or
coherence [7] is generated first and its decay towards an equilibrium state is
monitored subsequently.
On a microscopic level, nuclear spin relaxation is the consequence of a
weak coupling of the observed spins to their surrounding which is often
termed the "lattice" and includes essentially all other degrees of freedom
within the molecular system. Stochastic motional processes of the lattice as
the overall tumbling and intramolecular motion of a molecule lead to
16 Introduction
reorientation of internuclear vectors and molecular axes [12] and therefore
modulate magnetic interactions as, for example, the dipole-dipole interaction
or the chemical shielding anisotropy (CSA) interaction [7] The resulting
time-dependent magnetic fields induce transitions between populations or
lead to dephasing of coherences of the observed spin system The measured
relaxation rate constants therefore contain information about the frequencies
by which the local magnetic fields are modulated and can be analyzed in
terms of an appropnate motional model for the underlying stochastic
processes
NMR studies of the intramolecular motion in proteins concentrate
normally on a local description of the reorientation of internuclear vectors in
either the backbone domain or in the amino acid sidechams Improved
labeling techniques [13] enable the investigation of an increasing amount of13 15
proteins by C and N nuclei as "spies" for the motional modes
Information about backbone motion can therefore be extracted from N,
C and C relaxation data by descnbing the reorientation of mainly the1 N 15 1 a 13 a
H - Nor H - C bond vectors Hereby, the so-called "model-free"
approach of Lipan and Szabo [14,15] is often used where intramolecular
motion is parametrized by a measure for the time scale of the process and an
"order parameter" which is an abstract measure for the spatial restnction of
the relaxation-active interaction at the site of the monitored nucleus This
simple approach does not provide a very detailed physical insight into the
reorientation processes but circumvents some intrinsic problems connected
with the application of a detailed motional model First, the correctness of
such a motional model cannot be proved by relaxation measurements alone
Second, it is often not possible to discriminate between different motional
models, such as rotational diffusion or jump-like processes, which often fit
equally well to the relaxation data
In Chapter 3 of this thesis, it is tried to describe the backbone motion of the
protein ubiquitin in a way that goes beyond the description of the
reorientation of local vectors in terms of the standard model-free analysis
The proposed analysis [16] takes into account that the peptide bond, which
Introduction 17
is the central building block of the protein backbone, defines, to a good
approximation, a planar and rigid fragment The extraction of the motional
parameters, which characterize the peptide plane motion, from experimental
relaxation data is the mam goal of this chapter In the rigid fragment, the
dipolar and CSA interaction tensors, which dominate the relaxation of the15 13N and C peptide bond spins, have fixed relative orientations Since these
interaction principal axes are generally not parallel, they probe
reonentational processes about different directions in space A combined
15 13
interpretation of N and C relaxation data should therefore allow for the
exploration of locally anisotropic motion of the whole peptide plane
fragment Although it is well known that intramolecular motion of
biomolecules is generally anisotropic, which can be seen from anisotropic
crystallographic B-factors [17], the magnitude of amsotropy for the motion
of the peptide plane has never been extracted from experimental data before
The peptide plane reonentation is discussed in detail using a motional model
which is defined by the orientation of three orthogonal principal axes about
which the peptide plane fluctuates and the corresponding distribution
functions for the fluctuation angles about each principal axis Such a
motional model cannot be extracted from the expenmental relaxation data
itself but must be provided by molecular dynamics (MD) computer
simulations [18,19] which introduce independent knowledge in form of an
empirical molecular force field parametrized based on quantum chemical
calculations and mainly optical spectroscopic data Bremi and Bruschweiler
[20] have recently proposed a 3D Gaussian Axial Fluctuation (GAF) model
for the anisotropic reonentation of the peptide plane based on the analysis of
a MD trajectory of a decapeptide The pnncipal axes of reonentation consist
of the Cj _ [- C, axis and two orthogonal axes lying in the peptide plane and
orthogonal to the plane The fluctuations are charactenzed by Gaussian
distnbutions In Chapter 3, it is tested by an extended MD simulation
whether the 3D GAF model applies for a protein Furthermore, it is analyzed
under which conditions the 3D GAF model is appropnate for fitting the
relaxation data
The extraction of the amsotropy of the peptide plane motion relies on a
18 Introduction
detailed knowledge of the magnitude and orientation of the relaxation-active13 15
dipolar and the C and N CSA interaction tensors. X-ray
crystallography, neutron scattering, and optical spectroscopy can provide
experimental data on the peptide plane geometry [21] defining the dipolar
interaction strengths. CSA parameters can be extracted from solid-state
NMR experiments of small peptides [22]. The set of interaction strengths
used in the analysis of Chapter 3 is based on these experimentally determined
interaction parameters which leads to several problems. First, it is not clear
whether the interaction strengths, which have often been measured for model
compounds, can be transferred to proteins and how much these parameters
vary from residue to residue. Second, the use of interaction strengths with
fixed magnitude and orientation may not be appropriate since intramolecular
motion induces changes in the molecular geometry and in the electronic
environment which results in fluctuations of the dipolar and CSA interaction
strengths altering thereby spin relaxation. These problems are addressed in
Chapter 4, where a novel approach is presented for the interpretation of CSA
relaxation in proteins. It is based on the combination of MD simulation for
the description of the dynamics in the protein ubiquitin and a quantum
chemical method, the density functional theory (DFT) [23], for the13 15
calculation of CSA parameters. CSA tensors of C and N backbone spins
are calculated for each snapshot of a MD trajectory leading to a "CSA
trajectory" which directly contains the fluctuations of the CSA interaction
tensor due to intramolecular motion. In Chapter 4, these fluctuations are
characterized and their contribution to CSA relaxation of protein backbone
spins is assessed. It is checked whether CSA relaxation can be described with
CSA parameters averaged over the time-dependent fluctuations. The
averaged CSA parameters are compared to experimental results by solid state
NMR.
The conceptual framework underlying for Chapters 3 and 4 is summarized
in Fig. 1.1. It describes how MD simulation techniques, analytical motional
models, and quantum chemical methods have to be combined for a detailed
interpretation of NMR relaxation data in terms of intramolecular motion.
Introduction 19
NMR Relaxation Experiment
X-ray diffraction
solid-state NMR
Interaction
StrengthsQuantum Chemical
Methods
i '
,.
Order Parameters
Time Scales
f Intra- \molecular
y MotionJ
Analytical Molecular
M(3d els Dynamics
Fig 1 1 Investigation of intramolecular motion by interpretation of experimental NMR
relaxation data using a combination of quantum chemical methods molecular
dynamics computer simulation and analytical motional models
In liquid systems, the intramolecular motion is often masked by the overall
tumbling motion of the molecule The decomposition of both relaxation-
active processes could be improved if the timescale of the overall tumblingmotion can be changed selectively without affecting the internal motional
processes In Chapter 5 it is investigated [24] by NMR relaxation
experiments whether a highly viscous, polymeric solvent is an appropriatetool for the isotropic retardation of the overall tumbling motion of solute
molecules
20 Introduction
2.1 Fundamentals of spin relaxation 21
2 Spin Relaxation and Molecular Motion
In this chapter, the theoretical framework for the following chapters is
presented. First, the fundamentals of spin relaxation in terms of semiclassical
relaxation theory are shortly summarized followed by a description of the
treatment of the overall tumbling and intramolecular motion in liquid
systems as relaxation-active stochastic processes. Special emphasis is laid on
the methodology to extract information about the anisotropy of
intramolecular motion from NMR relaxation data which is exemplified for
the anisotropic peptide plane fluctuations in the protein backbone. In
addition, MD simulation techniques and quantum chemical DFT methods
are shortly introduced.
2.1 Fundamentals of spin relaxation
2.1.1 Equation of motion
The time evolution of a quantum mechanical system, e.g. a single nuclear
spin system, under the influence of a static Hamiltonian H is given by the
time-dependent Schrodinger equation [25,26]:
ijfNit)) = ff|i|/(0>, (2-1)
where H is expressed in units of h/27t and the time-dependent state function
\y(t)) contains all information about the system. It can be decomposed into
a linear combination of time-dependent coefficients c((f) and eigenvectors
III/,) of the static Hamiltonian H which obey the time-independent
Schrodinger equation:
M
IV(0> = X ^(f)IV,). with H\y) = E,ty,), (2.2)
i= l
where El is the energy of the eigenstate |t|/(). Since NMR is an ensemble
spectroscopy, it is convenient to describe an ensemble of spin systems with
its time-dependent density operator p(f) [7]:
22 Spin Relaxation and Molecular Motion
P(0 = lY(0><V(OI, (2-3)
where the bar indicates averaging over the ensemble. The density operator is
2one element of the M -dimensional Liouville space and can be represented
as M x M matrix with respect to a basis of the Hilbert space of dimension M
[27]. Then, the diagonal elements of the density matrix correspond to the
populations of the different energy levels Et of the Hamiltonian H whereas
the off-diagonal elements represent coherences resulting from phase
correlations of pairs of different eigenstates averaged over the ensemble [7].
The time evolution of the density operator under the influence of the
Hamiltonian H is given by the Liouville-von Neumann equation [7]
jtp(t) = -i[H, p(0] , (2.4)
which can be derived by combination of the time-dependent Schrodinger
equation and the definition of the density operator for an ensemble assuming
that there are no interactions among the subsystems in the ensemble.
2.1.2 Semiclassical relaxation theory
A closed quantum mechanical system shows a unitary and reversible time
evolution. The concept of relaxation comes in when a subsystem is regarded
which is coupled to the environment treated as a perturbation in the case of
weak coupling. A microscopic semiclassical theory of spin relaxation was
formulated by Bloch, Wangsness, and Redfield (BWR) [28-31] and has
proved to be useful for the description of experimental relaxation data in
many cases [7,9]. In the semiclassical approach the spin system is treated
quantum-mechanically, and the surroundings are treated classically. This
means that the density of energy levels of the lattice is assumed to be quasi-
continuous with populations that are described by a Boltzmann distribution.
Furthermore, the lattice is assumed to be in thermal equilibrium at all times.
The total spin Hamiltonian of the spin system is described by the sum of
the static Hamiltonian H0 and a stochastic Hamiltonian HAt) which
2 1 Fundamentals of spm relaxation 23
represents the coupling of the spin system to the motional degrees of freedom
in the lattice (see Chapter 1)
H{t) = HQ + H{{t), (2 5)
where Hq = Hz + Hj consists of a Zeeman term for the coupling of the
nuclear spins to the external static magnetic field and a term for the time-
independent scalar spin-spin coupling [7] Chemical exchange processes,
rendering these terms time-dependent, and a coupling of the spin system to a
radio frequency field shall not be considered at this place The stochastic part
#j(f) represents a stochastic process with a time average of zero and is
regarded as a weak perturbation with respect to H0
The derivation of the master equation of spin relaxation in an ensemble of
spin systems by second-order perturbation theory is described in detail in the
textbook of Abragam [9] Starting by inserting Eq (2 5) into the Liouville-
von Neumann equation, the final form of the master equation is the main
result of the BWR relaxation theory and takes the form
jtP(t) = - i[H0, p(t)] - f{p(0 - p0} , (2 6)
with the relaxation superoperator i defined by the operation
-iHn1
dx (2 7)fB = J Hx{t),[e°
Hl{t-x)e°
,B]
where the bar indicates averaging over the ensemble of spin systems It is
important to note that the validity of the master equation relies on several
assumptions First, it is only valid in the so-called weak collision regime
implying that the correlation times of the involved stochastic processes (see
Section 2 2 1) are significantly shorter than the time-scale on which the
density operator p(t) evolves [9] It is also required that the magnitude of the
perturbation Hamiltoman \HA is much smaller than the inverse of the
corresponding correlation times [9] Thus, Eq (2 6) is not appropnate for
24 Spin Relaxation and Molecular Motion
strong perturbations that are slowly modulated. In the derivation of Eq. (2.6)
the density operator p(f) has to be replaced by the difference p(f) - p0 [9]
to ensure a relaxation towards the thermal equilibrium represented by the
density operator p0. Therefore, the given master equation is an
inhomogeneous differential equation only valid in the high temperature
approximation [32].
2.1.3 Relaxation-active interactions
Spin relaxation results from the stochastically modulated reorientation of
an interaction tensor inducing a time-dependent local magnetic field at the
locus of the observed nucleus. There are a variety of relaxation-active
interactions like the electric quadrupole interaction, random field interaction,
spin rotation interaction, and interactions leading to scalar relaxation of the
first and second kind [7,9,33]. In this work, however, the discussion is
restricted to the chemical shielding anisotropy (CSA) interaction and the
dipolar interaction for spin systems consisting of spin 1 /2 nuclei.
The stochastic Hamiltonian H^ is given as the sum over the relaxation-
active interactions present, each of them expressed in terms of irreducible
tensor operators for the spin part and irreducible representations of the 3D
rotation group for the spatial part [34]:
V
"i(0 = 1^(0 I (-»\,-,«y0)7? , (2.8)
where it is assumed that each interaction \i is described by one rank / only.
|X runs over the interactions and c is a characteristic constant measuring the
strength of the interaction n. The angular momentum number / is also the
rank of the tensor and q is the magnetic quantum number. The spatial part is
represented by the normalized spherical harmonics Y, (Q..At)), where
D, (t) is a time-dependent set of polar angles defining the orientation of the
interaction u. in a chosen coordinate frame. T, are the irreducible tensor
operators for the spin part of the interaction |i.
2.1 Fundamentals of spin relaxation 25
The dipole-dipole interaction. The Hamiltonian of the magnetic dipole-
dipole interaction between the spins ) and $ connected by the vector rIS is
given in its bilinear form by HD = -)£>$, where D is the rank 2 dipolar
coupling tensor which is axially symmetric with its symmetry axis parallel to
rJS. The Hamiltonian for the laboratory frame can be decomposed according
to Eq. (2.8):
HD = -!~Vs4jf £ (-1)V«(<W)^(U), (2.9)
rIS q = -1
where the time-dependent orientation of r]S in the laboratory frame is given
by Q.D (t). The normalized second rank spherical harmonics Y2 „
and the
irreducible spin tensor operators T2 „of rank 2 are given by [34]:
r2,0(e><P) = Jii^3^086)2-1) •>r2,o(tf.^) = \{3Uzvz-^y>
r2i±1(9,<p) = T^sinecose/^; r2i±1(^, t) = t\(u±vz+ UzV±);
y2,±2(q,<p) = J§(sine)2e±2,ip;r2,±2(^^) = ^±y±) • <2-10)
The chemical shielding anisotropy interaction. The Hamiltonian for an
interaction between a spin S and the static magnetic field BQ is given by
HCSA = ysSgB0. For nuclei situated at sites with low local symmetry (e.g.
in biomolecules), the chemical shielding tensor a reflects the anisotropic
and antisymmetric nature of its electronic environment. As a general second
rank tensor, g can be decomposed into a sum of irreducible tensors of rank
0,1, and 2 [35]:
g = g(0) + g(1) + g(2), (2.11)
where g is the isotropic chemical shift which is not relaxation-active and
can be neglected here [36]. g and g corresponds to the antisymmetric
component g = (g-g )/2 and the symmetric (but not necessarily
axially symmetric) component qsym = (g + g )/2, respectively. In the
26 Spin Relaxation and Molecular Motion
principal axis frame of g they take the form:
0 a*y axZ
cxy
0 °>*
°« -°vZ 0
,and (2.12)
sym
°xx 0 0
0 oyy 01
=
3°*
2 0 0
0-101
+ 3°.
-10 0
0 2 0
0 o °«J 0 0 -1 0 0-1
(2.13)
where ox = oxx-ozz and a = a -az,, and it has been used that any
non-axially symmetric tensor can be split into two axially symmetric,
orthogonal parts [37]. The time-dependent Hamiltonian for the rank 2
interaction of the symmetric part can be decomposed according to Eq. (2.8)2
Hsym
CSA (t) = -JsL
q = -2
+ ay X (-\)qYx_q(Q!ya\t))T2q{lBQ)q = -2
(2.14)
where Q (t), £2X (t) denote the time-dependent orientations of the CSA
principal axes x and y in the laboratory frame, respectively. For the
antisymmetric part, which cannot be diagonalized, it is convenient to
transform a first from the principal axis frame of the symmetric part to
the laboratory frame using a Wigner rotation matrix with a set of time-
dependent Euler angles D. = (9, <|), %). The Hamiltonian is then given by
[35]
l/l \
*0') = YsX ("1)1 I Si.XVW) k*&^>' <2'15>
q = -1 Vm = -1 /
where Dm_ (Q.) denote Wigner rotation matrix elements of rank 1 [34]. The
rank 1 parameter olm and the irreducible rank 1 spin tensor operators Txare given by [34,35]:
2 1 Fundamentals of spin relaxation 27
cl0 = -,j2cxy,Tl0($,B0) = 0,
°i ±1= °« ± "^ • Ti ±A Bo) = -\s±B0 (2 16)
2.1.4 Operator representation of the relaxation superoperator
Taking into account the symmetry properties Yl = (-\)qYl * and
Tl = {-\)qTl + [34], an explicit form of the relaxation superoperator
can be calculated by inserting Eq (2 8) into Eq (2 7)
= xi(-ir
x J c^)cv(? - t)y1 _?(iy o)y/y; _9.*("v(? -T))^' (217)
0
where it is assumed that each interaction \i is descnbed by one rank / only
This is possible for the CSA interaction since the contributing rank 1 and 2
interactions of the antisymmetric and the symmetric part, respectively, can be
treated independently due to missing cross-correlation between both parts of
different rank (see Section 2 3)
The inner commutator in Eq (2 17) can be transformed by expanding the
tensor operators for the interaction (J. in terms of eigenoperators A of the
static Hamiltoman H0 [9]
Ti,q = ^UA> (2 18)
p
where the eigenoperators are defined by the operation [H0, A ] = (oA
Replacing both tensor operators in Eq (2 17) yields the final form for the
relaxation superoperator in the laboratory frame
V*'-in0i v t 0X
,B
28 Spin Relaxation and Molecular Motion
fHii.'-'^'z.-UA.*/U, Vq, q' p, p
-t
x/rj/v>-,,^K)K,tA;,,5]], (2.19)
where the power spectral density functions J, , , have beenV' v' 9' "
introduced:
J'X-^'{(0)= lcZ',-^{x)e"mdx' (2-20)
—oo
which are the Fourier transform of the ensemble averaged correlation
functions of the spherical harmonics
<Vv,_,,^(t) = c^t)cv(t -
t)fv ^(fyo)^,VW'- *))
(2.21)
For ii = v, C, , „ _„.defines an autocorrelation function, for (i ^ v, a
cross-correlation function.
In Eq. (2.19), all combinations between the quantum numbers p and p' were
included. Terms with p £ p' are non-secular in the sense of perturbation
theory and do not affect the long-time behavior of p (r) since they show a fast
oscillation in the rotating frame which averages to zero much more rapidly
than relaxation occurs [9]. In the "secular approximation" terms with p * p'
are neglected.
2.2 Relaxation-active molecular motion
2.2.1 Correlation functions for molecular motion
The overall tumbling and intramolecular motion of a dissolved molecule
with N atoms can be described by an ensemble of dynamic trajectories in a
conformational space of dimension 3N. From an experimental point of view,
however, it is not possible to characterize the reorientation of a single
molecule in a time-resolved manner, since the measurement results from an
2.2 Relaxation-active molecular motion 29
ensemble average and the variations with time contained in each of these
trajectories show a complicated and stochastic behavior [38] on a time scale
which is too rapid to be sampled experimentally. One has to rely on the fact
that the stochastic processes can be described by some averaged motional
constants. This information can be extracted by averaging over the ensemble
which is done by calculating a set of correlation functions. In principle, one
can calculate a set of 3iV autocorrelation functions
^uu(r'T) = AJt)AM + x), 3N(3N-l)/2 pair correlation functions
C v(f, x) = A (t)Av(t + x), and a large number of higher-order correlation
functions [38], where the bar indicates the average over the ensemble of
molecules in the system and {Aj(f),..., A^N(t)} represent the 3N
stochastic scalar variables describing the molecule's motion in its
configurational space. However, such a complete set of correlation functions
cannot be extracted from experimental data but only few auto- and pair-
correlation functions can be characterized. In studies of intramolecular
mobility, one concentrates on motional modes which can be described
locally for a subunit of the molecule in terms of a motional model with only
few dynamical parameters. Other simplifications comprise the assumption
that the overall tumbling motion of the molecule is independent from the
intramolecular motion. It is assumed that the molecule has a certain rigid
shape when investigating the overall rotational diffusion.
Rapid reorientational processes, which modulate relaxation-active
interactions on time scales accessible for laboratory frame relaxation
experiments, enter the relaxation superoperator via Eq. (2.19) in form of
spectral density functions, which are Fourier transforms of the ensemble
averages of the angular correlation functions in Eq. (2.21). Since the
stationarity condition applies for a system in the equilibrium, there is no
origin in time and the correlation function does only depend on the time
difference X. The correlation functions are then given as ensemble averages
over the initial times A (0)Av(x) = A (t)Av(t + x). In addition, it is
usually assumed that the reorientation processes are ergodic, which means
that a property of a reorientational process is equally given by the time
average over one molecule's trajectory and the average over the whole
30 Spin Relaxation and Molecular Motion
ensemble of molecules at the same time and vice versa:
(A (0)Av(t)) = A (0)Av(t). Therefore, the ensemble average in the
correlation function of Eq. (2.21) can be replaced by an average over the
time-behavior of one single molecule
<vv,W> = <M0>MT>%ViV0))yw*<£W)>- (2-22)
This is an important result needed for the calculation of a correlation function
from the MD trajectory of a single molecule or from a model describing a
motional process in a single molecule.
2.2.2 Overall rotational diffusion
2.2.2.1 Isotropic rotational diffusion in a homogeneous environment
The Debye small-step rotational diffusion model [39,40] is an appropriate
description for an overall reorientation process of a molecule which is far
greater in size than the solvent molecules representing a quasi-continuous
and homogeneous medium. This is fulfilled, to a good approximation, for
proteins dissolved in water. It is often found that the model holds well even
for small solute molecules in case of low viscous solvents [41]. The model
views the reorientational motion of a molecule as being impeded by a
viscosity-related frictional force. If the rotational friction coefficient
operating at the surface of the molecule regarded as a sphere with radius Rhis that represented by the macroscopic solvent viscosity r|, then the rotational
friction coefficient has the Stokes value [42,43] \ = inRhi\ and the
rotational diffusion rate constant for isotropic rotational diffusion is given by
the Stokes-Einstein-Debye (SED) relationship [39]
D = —£i—, (2.23)
%nRhT\
with D in rads.The rotational diffusion of a molecule in a homogeneous
solvent medium has been described by several authors [9,39,40,44-48]. The
derivation of the correlation function for this overall tumbling process shall
2.2 Relaxation-active molecular motion 31
be given in the following. The orientation of the molecular frame, exclusively
affected by the overall rotational diffusion, with respect to the laboratory
frame shall be given by the angular term $ which represents a set of three
Euler angles. For /(O, t) which is the probability of finding the molecular
frame at the orientation 4> at time / the following diffusion equation holds
with /($, 0) = 8(4> - 4>0) as initial condition. V2 is the angular part of the
Laplacian operator in spherical coordinates. Following Favro [44], the
solution of Eq. (2.24) is given by /($, t) = P(<I>0|<t>(/)), which is the
conditional probability of finding the molecule at orientation <&(£) at time t,
if it was at orientation 4>0 at time t = 0. In the following an explicit solution
for P(O0|$(r)) shall be derived. /(<!>, 0 can be expanded in terms of
Wigner rotation matrix elements Dmn
/(*.'>= I i(Oi(*)' (2-25)/, m, n
By substituting this expression into Eq. (2.24) and taking into account
V2Dlmn(^) = -1(1+ \)Dlmn{Q>) [34] as well as the orthogonality
relationships of the Wigner matrices [34]
2
K»*(*)Dm(0) = 27778"5^V (2-26)
it follows
pmn(t) = -Dl(l+ 1)4(0 with 4(0 = 4(0)*-fl/(/+1)'. (2.27)
Introducing of the initial condition by expanding the 8 function in Wigner
rotation matrix elements [34]
32 Spin Relaxation and Molecular Motion
5($-<D0) = X ^DL*(*o>0L(*). (2.28)
/, m, n oil
/ 21+ I I *
yields cmn(0) = r-Dmn (<J>0), and finally871
P(*0|O(0)= I ^^L*(*o)^L(^"D/(/+1)r- (2-29)
/, m, n oTC
This result has now to be incorporated in the correlation function of
Eq. (2.22). Assuming statistical independence of the overall tumbling and
the intramolecular motion, Eq. (2.22) can be transformed taking into account
the transformation properties of the spherical harmonics:
civd) = <^(0)cv(T)y,i0(flJlflfc(0))y/i0*(or('t))>= I<^(*o)^o*(*(^))>
r, r'
x (^(0)^(1)7, r(nJ""(0))y,F/(jC'(t)» , (2.30)
where the condition I = I = lv has been used since cross-correlations
between interactions of different rank / are not considered in this work. In
addition, C (x) is no longer dependent on the index q, which was set to
zero, since the correlation function is invariant under an arbitrary rotation in
the laboratory frame [49]. Q denotes polar angles (9 q> ) of the
principal axis of the interaction tensor |i in the molecular frame. The overall
tumbling and the intramolecular motion are averaged separately, and the
overall tumbling average can now be expressed in terms of the conditional
probability />($0|$(t)) [9]:
(Dlr0(%)Dl,0\®{T)))
= |Ji'(*0)D'0(*0)/'(*0|«&(x))£)J,0*(*(T))d*0<«>I (2.31)
2where P(®0) = 1/(8n ) is the probability of orientation <&Q at time
t = 0. Substitution of Eq. (2.29) into (2.31) and taking into account Eq.
(2.26) yields
2.2 Relaxation-active molecular motion 33
(Dlr0^0)Dlr,0*mx))) = ^-/(U l)°\f. (2-32)
Substitution of Eq. (2.32) into Eq. (2.30) leads to the correlation function
; _l -z(z+i)dt
SvW -
2l+\e
x £ {c^)cv{T)Ylr{n\0))Yl*(Qol{T))). (2.33)
For a rigid molecule affected by isotropic overall tumbling only,
c, cv, Q , Qv are time-independent. The interaction strengths c
, cv
can then be bracketed out from the correlation function and appear in the
constant prefactor when calculating relaxation rate constants (see Section
2.3). With the addition theorem of spherical harmonics [34]
^tlp^cose^) = X hWW^> (2-34)
q = -l
where P^cosQ^) = J4n/(2l+ l)Yj o(9nV) are Legendre polynomials
and 9 is the angle between the principal axes of the interactions \i and v,
one obtains for the correlation functions of rank /
-x/x(0
C^v(x) = e PticosQ^), (2.35)
with the definitions of the rotational tumbling correlation times
xlcl) = l/(2D) for rank 1 and T^xf1 = 1/(6D) = x^/3 forrank2. A
factor 4tc has been introduced in Eq. (2.35) to normalize the correlation
function for x = 0 to start at 1 for all auto-correlated processes (ft = v).
The corresponding spectral density function can be calculated as a Fourier
transform according to Eq. (2.20):
34 Spin Relaxation and Molecular Motion
./uv(a>) = P;(cos9^v;2i(')
(/)21 + (COT*0)
(2.36)
This spectral density function will be needed for the calculation of the
relaxation due to the antisymmetric part of the CSA (rank 1 interaction) in
Section 4.5.3 and of the dipolar relaxation (rank 2 interaction) in Chapter 5.
2.2.2.2 Rotational diffusion in a heterogeneous environment
It is well known that the experimental data of, in particular, highly viscous
systems, often cannot be explained by a mono-exponential correlation
function (see Eq. (2.35)) for isotropic rotational diffusion in a homogeneous
solvent medium. The non-exponential character of the correlation function
can be explained in two fundamentally different ways [50-53]. On the one
hand, local density fluctuations in the liquid influence the rotational
correlation function since they cause fluctuations in the environment of a
rotating molecule [54]. Rapid local density fluctuations, whose correlation
function decays on a short time-scale compared to the rotational correlation
time, would, again, result in an exponential correlation function. However,
local density fluctuations with a correlation time slower than the rotational
correlation time (but still faster than the relaxation times, see Chapter 5),
which occur in highly viscous liquids, might give rise to a distribution of
correlation times which results in the following correlation function for an
auto-correlated rank 2 interaction:
C(x) = jp(xc)e~Z/Xcdxc, (2.37)
o
where p(ic) defines the distribution function of the correlation times. On the
other hand, the assumption of independently reorienting molecules might not
be realistic in a highly viscous liquid. If the reorientational processes of
neighbored molecules are correlated with each other, one has to consider
cooperative motion of entire subsystems. A description of such motional
processes is extremely difficult and only very simplifying theoretical
2.2 Relaxation-active molecular motion 35
approaches exist [55]. This intrinsic non-exponential behavior is often
characterized by the empirically found Kohlrausch-Williams-Watts (KKW)
or stretched-exponential correlation function:
C(x) = exp(-(x/tc)Pw), (2.38)
with 0 < $KKw < 1. To decide on the question whether the description of
independently reorienting solute molecules in a heterogeneous environment
or of complex reorientation processes with a collective character (or both) is
realistic, it would be required to directly sample the correlation function of
small sub-ensembles in the system. This is possible for extremely viscous
liquids due to the slow time-scale of the reorientation processes and the local
density fluctuations [56]. However, for molecular systems with reorientation
processes faster than milliseconds which are studied by NMR relaxation
spectroscopy and other methods like dielectric relaxation, it is not possible to
decide on this question. Therefore, most of these studies rely on Eq. (2.37)
simply due to the fact that many (mainly empirical) models have been
developed to describe the distribution function of the correlation times p(lc)
[53]. Some of these models have been applied in Chapter 5.
2.2.3 Intramolecular motion
2.2.3.1 Intramolecular backbone motion in proteins
The entire backbone motion in proteins results from a superposition of a
large number of motional modes on different time-scales. For the fastest
time-scales with correlation times <50ps the contributions of different
modes can be classified based on normal mode analysis [57-59]. High
frequency modes ( go > 500cm ), such as bond stretching and bending, have
small amplitudes and a very local character. These very rapid fluctuations can
be included in an effective dipolar interaction strength resulting from
motional averaging of the equilibrium interaction strength (see Section
2.2.3.2). Collective low frequency modes (co<250c/n ,see ref. [59])
exhibit larger amplitudes. According to Briischweiler [59], two kinds of the
collective modes can be distinguished for the reorientation of the interaction
36 Spin Relaxation and Molecular Motion
principal axis under consideration (e.g. the NH vector in the peptide plane i):
Long-range motion (LRM) causes angular modulations of the interaction
principal axis by affecting a whole protein segment, which contains the
principal axis, but leaves the adjacent local backbone dihedral angles <p;,
V,_i (for definition see Fig. 2.1a and ref. [1]) unchanged. Short-range
motion (SRM) leads to fluctuations of the adjacent local dihedrals. In case of
compact proteins with lots of tertiary interactions it is this short-range motion
in dihedral angle space {(pr a>(, ij/;} which represents the dominant
relaxation-active motion for backbone spins in the sub-50 ps correlation-
time window [59].
Since the fluctuations about the dihedrals co are relatively small in
amplitude (see Section 3.4.3.3), peptide planes are well-defined fragments. It
is desirable to find a local description for the reorientation of a peptide plane
which results from the cumulative effect of projections of a large number of
the collective modes on the individual peptide-plane fragments. A local
description of peptide plane / in terms of fluctuations of the dihedral angles
vj/; jand <p(, however, is not appropriate since the correlation coefficients
p show large deviations from -1 (this is shown for the protein ubiquitin
in Fig. A.2a in the appendix) which demonstrates the dependence of the
dihedral angular motion also on the motion of adjacent peptide planes
[60,61]. Bremi and Briischweiler [20,62] have concluded from the analysis
of a MD trajectory of the decapeptide antamanide that it is preferable to
describe the local small-amplitude motion of a peptide plane by a
reorientational principal axis system fixed at the equilibrium peptide plane
and characterize the motion in terms of Gaussian distribution functions for
the fluctuations about and time constants with respect to these axes. This
results in the 3D Gaussian Axial Fluctuation (3D GAF) model [20,62] whose
principal axis frame is depicted in Fig. 2.1b. The three orthogonal principal
axes of reorientation are defined by the Cl_i-Cl axis, named e,the in-
plane principal axis ea, and the principal axis eg which is orthogonal to the
equilibrium peptide plane.
2 2 Relaxation-active molecular motion 37
Fig 2 1 (a) Definition of dihedral angles along the protein backbone (b) Definition of
the reference coordinate system ea, eo, e fixed to the peptide plane It defines the
principal axis system for the peptide plane reorientation according to the 3D GAF
model. The principal axes directions of the relevant spin interactions, expressed in this
frame, are given in Table 3 2
ct B ct o.The reonentation of the backbone C - C and C -H vectors cannot
be descnbed by the 3D GAF model since they are affected by the motion of
both adjacent peptide planes. A detailed motional model describing their
reorientation is missing so far. Intramolecular motion of the backbone on
time-scales with correlation times between 50 ps and 0.1 jxs occur mainly in
38 Spin Relaxation and Molecular Motion
unfolded proteins, or in loop and end regions of globular proteins In those
more flexible regions additional jump processes between different
conformations add on the underlying 3D GAF motion of die peptide planes
In pnnciple, these peptide planes can be descnbed by combined 3D GAF and
jump models Motional processes involving the backbone on even slower
time-scales in the fis to 10s range as e g hinge-bending motion, hydrogen
bond dynamics and exchange, helix-coil transitions, and cis-trans isomensm
of the peptide plane shall not be discussed at this place
2.2.3.2 Internal correlation functions and model-free approach
First, it is assumed that the intramolecular motion can be separated from
the isotropic overall tumbling motion
C^x) = Cltum\x) O), (2 39)
where according to Eq (2 33) the overall tumbling correlation function is
given by C"m
(x) = e The internal correlation function takes
the form
/
C^n'(x) =
27TT I <S<°>cvC)r7 r(a;o,(0))yl r*(<o/(t))>r = -l
= £-<^(0)cv(T)P;(cose^v(T))>, (2 40)
where the addition theorem for spherical harmonics in Eq (2 34) has been
used and 9„v(0 is the time-dependent angle between the pnncipal axes In
the following, only rank 2 interactions will be considered and the interaction
strengths c,c will be regarded as time-independent and therefore will be
bracketed out from the correlation function This leads to
2
c|Tv(T) =
y 2 <r2 r("7'(0))F2 ,*(nT'(T))> = (/V^Vx))>'r = -2
(2 41)
where u. and v refer to the pnncipal axes ofrank 2 interactions, e g a dipolar
2.2 Relaxation-active molecular motion 39
director or a CSA tensor principal axis, and P2(x) = (3x2- l)/2 is the
second order Legendre polynomial. A factor 4k has been introduced in
Eq. (2.41) to normalize the correlation function for x = 0 to start at 1 for all
auto-correlated processes (|X = v), since />2(cos(0)) = 1, and at
P2(cos9 ) with 8 s 0„V(O) for all cross-correlated processes. In case of
cross-correlation between two principal axes of the same CSA tensor, one
hasP2(cos(7t/2)) = -1/2.
The internal correlation function C (x) can be calculated explicitly for a
motional model as described in Section 2.2.3.4 for the 3D GAF model of the
peptide plane reorientation. Alternatively, an abstract, "model-free"
approach which corresponds to a simple parametrization of the spectral
density function can be used (see also Chapter 1). If the intramolecular
motion is Markovian (e.g., diffusive or jump-like), then C (x) can be
expressed as a series of exponentials Cuv(x) = Va;exp(-x/xj), where
x0-»°°, x1>x2>...>0, and 0<a(<l for all ;'= 1,..., Z with Z
dependent on the type of motion. It was shown by Lipari and Szabo [14,15]
that in case of isotropic overall tumbling motion, and internal motions with
ny Llv 2
correlation times %int m the extreme narrowing regime ((x;nfco) « 1) much
faster than the overall tumbling motion, the internal correlation function is
exactly given by
Ox) = Slv + {^(cose^-sjj/1 T"". (2.42)
2
5„v is the so-called generalized order parameter and corresponds to the
plateau-value of the correlation function. In case of auto-correlation (i = v,2
S„„ is a model-independent measure for the spatial restriction of the motion
which modulates the relaxation-active interaction \l. Absence of internal2
motion leads to S = 1, large-amplitude internal motion to small order
parameters. The plateau value can be calculated as
i = jl <y2,r(Q"0,)){y2jr*(<0/)>. (2.43)
40 Spin Relaxation and Molecular Motion
For an auto-correlated process \i = v, Briischweiler and Wright [63] have
derived an alternative formula in terms of variances of the spherical
harmonics
Sl=l-f I42/ (2-44)
r = -2
This formula is useful for the calculation of order parameters in case of
Gaussian axial fluctuations (see Section 2.2.3.3). A similar formula for the
calculation of the cross-correlated order parameter including the time-
dependence of the interaction strengths c, cv is given in Section 2.5.1.
The auto-correlated and cross-correlated spectral density functions can be
calculated from Eq. (2.42) taking into account Eq. (2.20) and the independent
overall tumbling with correlation time xc [64]:
r2 l%c. id ,^,c ^ e2 i
ZV/
1 +W 1 + (CO
where the effective correlation time xe^ is given by
(\ff) = lc + (^n/) •In case of fast intramolecular motion, e.g. 3D
GAF motion, the correlation times xint are in the sub-50-ps range, the second
term of Eq. (2.45) becomes very small and thus the Xm( values have only
little influence on the relaxation parameters.
The spectral density function of the model-free approach can be used for
the calculation of the relaxation rate constants (see Section 2.3). Then, the
order parameter and the internal correlation time can be extracted from
experimental relaxation data and represent an abstract measure for the
amount of intramolecular motion present. On the other hand, the extracted
order parameters can also deserve as connecting link between the
experimental raw data and the motional model, since often the order
parameters are determined in a first step, subsequently interpreted using a
specific model analytically describing a motional mode (see Section 2.2.3.4).
However, this approach does only work if motional processes which are not
2 2 Relaxation-active molecular motion 41
included in the motional model but, in pnnciple, do affect the overall order
parameter can be treated separately In Chapter 3, for instance, the order
parameters for the reorientation of the backbone ^N^H bond are descnbed
solely in terms of peptide plane reorientation according to the 3D GAF
model Such a description ignores fluctuations of the NH vector with respect
to the peptide plane frame which, in this case, consist of the stretching and
the in-plane and out-of-plane bending motion of the NH vector However,
these fluctuations are not correlated with the peptide plane reorientation and
act on a much faster timescale Thus, they can be included in an effective
dipolar interaction strength which is then defined by the effective distance
rfH [57,65]
eff -3 ~1/3
rNH = iSbend^rNH^stretch} >(2 46)
where rNH is the equilibrium NH distance Eq (2 46) results from an
independent treatment of the stretching and bending modes [65] Stretching-3 -1/3
vibrations lead to an averaged distance (rNH) t,= (rNH) The
averaging effect depends on the underlying potential A strictly harmonic
potential leads to a decreasing of the effective bond length This effect is only
partly compensated by anharmonic contributions to the potential [65]
According to the analysis of Henry and Szabo [65], the averaging effect of
the stretching vibration is almost completely independent of the environment
and the structure of the molecule which is a consequence of the fact that zero-_3
point N-H stretches are primary responsible for averaging (rNH) The
bending modes, which lead to a reorientation of the NH vector, add on the
faster stretching mode and can be taken into account by the order parameter
Sbencj leading to an increased effective NH distance The overall scaling
effect is rather small [65] due to the partial compensation of the averaging
effects of stretching and bending modes
2.2.3.3 Extraction of anisotropy of intramolecular motion
The anisotropy of intramolecular motion descnbed by reonentation about
three orthogonal pnncipal axes can be assessed expenmentally and by
42 Spin Relaxation and Molecular Motion
computer simulations It can be extracted from NMR relaxation data if the
relaxation-active spin interactions have different directions in space In the
following, this shall be illustrated by an extended version of the "correlated
order parameter" approach of Bremi [62] For Gaussian axial fluctuation of
the interaction pnncipal axis e about a single axis e,the autocorrelation
2 2order parameter 5^ = S
ucan be denved from Eq (2 44) In case of
fluctuations much smaller than % it is approximately given by [63]
s\ = \ 3<Vln 9ucx' (2 47)
where 9 is the angle between the axes e and ea, and ca is the variance
of the fluctuation distribution
Fig 2 2 Fluctuation ellipsoid describing anisotropic intramolecular motion The
orientation of the interaction principal axis eu in the principal axis frame ew ea, e for
reorientation is given by the polar angles (9 <p )
In case of fluctuations about the three orthogonal pnncipal axes ea, ea, e,
the auto-correlated order parameter S is given by
sl=l~3(°asm V + °psin enp + Vin 9HY)2 22 2 22 22
= l-3(aa(l-sin 6 cos (p ) + Oq(1 - sin G^sin (p^) + oysin 9^),
(2 48)
2.2 Relaxation-active molecular motion 43
where (0 , q> ) are the polar coordinates of the principal axis e in the
principal axis frame ea, ea, e which is shown in Fig. 2.2.
2 2 2If the order parameters S
, Sv, S^ for the three interaction principal axes
eu' ev eX> wnich are au" affected by three-dimensional reorientation about
the principal axes ea, en, e are known, it is possible to extract the
fluctuation amplitudes and therefore the anisotropy information by applying
the following transformation:
[o2a,al,G2/ = l-A~\{l-sl),(l-S2v),{l-S2x)]T, (2.49)
where the matrix A contains the orientation information of the three
interaction principal axes:
2 2 2 2 21 - sin 0,.cos cp.. 1 - sin 0„sin (p., sin 0..
2 2 2 2 21 - sin 0vcos cpv 1 - sin 0vsin cpv sin 0V
2 2 2 2 21 - sin 0^cos q>^ 1 - sin 9^ sin (p^ sin 0^
(2.50)
The three fluctuation amplitudes can be extracted in case of a regular matrix
A, which means that e, ev, ex have to be linear independent. The presented
approach might be useful to generally decide whether a set of spin
interactions is well-conditioned enough to extract the fluctuation amplitudes
and to estimate their errors given the errors contained in the order parameters.
However, it has to be considered that Eq. (2.48) represents a good
approximation only in case of fluctuation amplitudes smaller than 10° which
prevents a general usage of this equation.
In principle, the fluctuation amplitudes can be obtained from relaxation
data of one single spin. However, such an approach does not apply in many
practical cases since the contributing spin interactions are often not "well-
conditioned". For example, the relaxation of a 15N nucleus of the protein
backbone is governed by the (axially symmetric) magnetic dipolar ^N-'H
44 Spin Relaxation and Molecular Motion
coupling and by the N chemical shielding anisotropy (CSA) interaction.
However, the CSA is approximately axially symmetric with the symmetry
axis nearly parallel to the ^N-1!! vector. Thus, fast intramolecular motion
scales the dipolar ^N-1!! and the 15N CSA interactions by nearly the same
2order parameter S
,the matrix A becomes nearly singular and the fluctuation
amplitudes cannot be accurately extracted from experimental data.
Fig 2 3 Orientations of the dipolar 15N-'H interaction and chemical shift anisotropy
tensors of 13C and 15N spins in the peptide plane.
The most promising approach for monitoring three-dimensional
anisotropic motion is the combination of relaxation data of different observer
nuclei fixed at a rigid molecular subunit in order to use their complementary
information to extract the anisotropy of the fragment's motion. This was first
shown by Bremi and Briischweiler [20] for the anisotropic peptide plane
reorientation using synthetic 15N and 13C relaxation data calculated from a
MD trajectory. The proposed analysis takes advantage of the fact that the
peptide-bond geometry remains, to a good approximation, planar at all times.
Thus, the relaxation-active CSA and dipolar interaction tensors have fixed
relative orientations shown in Fig. 2.3 and probe reorientational processes
about different directions in space.
2.2 Relaxation-active molecular motion 45
2.2.3.4 Analytical treatment of the 3D GAF model
The internal correlation function for interactions modulated by the 3D
GAF peptide plane reorientation can be calculated by transforming
Eq. (2.41) [12,66]:
/, m, rri = -2
(2.51)
where the transformation properties of spherical harmonics [34] were taken
into account
Y2J(n;°'{T)) = £ D^(Q(x))Y2tm(ep/). (2.52)
m = -2
The polar coordinate set e]f ~ (6 , <p ) defines the time-independent
orientation of the interaction principal axis Li in the instantaneous peptide
plane frame (pp) ea(t), eM), eAt) rigidly attached to the peptide plane
according to Fig. 2.1b. Thus, the motional behavior of this interaction
principal axis in the molecular frame is described by the transformation
Dml(Q.(%)) where Q. represents a set of Euler angles which describes the
time-dependent transformation between the instantaneous peptide plane
frame and the molecular frame. Without loss of generality, the molecular
frame is set to be the peptide plane frame ea, en, e of the equilibrium
peptide plane. Then, the transformation from the instantaneous peptide plane
frame en(t), eM), eJt) to the molecular frame can be carried out by three
subsequent rotations with angles -a, -(3, -y about ea, e», e, respectively.
However, when using Wigner matrices it is convenient to decompose the
transformation into a series of six rotations: (i) with angle -y about e, (ii)
with rc/2 about e», (iii) with n/2 about ey, (iv) with a about e, (v) with
n/2 about eg, and (vi) with -0 about e . Taking into account this series of
rotations one obtains from Eq. (2.52)
46 Spin Relaxation and Molecular Motion
.mol. (2)( 7t r,W2)f 11 K
y2, /(Q7(x)) = X *>${". \ -Pf^. £•^m(<") • (2-53)
k, m = -2
The Wigner rotation matrix elements can be decomposed [34]:
D?M^X)--e^d?Me-'lX, (2.54)
(2)where dkl (9) are the reduced Wigner matrix elements [34]. Substituting
Eqs. (2.53) and (2.54) into Eq. (2.51) yields the internal correlation function
2
C<*> =
y Zl2(kk)
, -i*'a(T) + i*a(0)v / i/p(x)-i/p(0Xe <e ){e )
I, k, k', m, rri = -2
, im,y(z)-imy(0\,(2)fn^,(2)fK\,{2)fn\,(2)fTt\v , p/\,7* , p/\
12
(2.55)In the next step the dihedral correlation functions of the type
(e'm ) have to be calculated for Gaussian axial fluctuation. Based
on the treatment by Chandrasekhar [67], Szabo [68] derived the dihedral
correlation function:
, im'y{x)-imy(0\ I 1 2
(e ')gaf = exp<j--a7
-
2 2 -Dx/a
m + rri -Imm'e (2.56)
where D is the diffusion constant for a diffusion process in a harmonic
potential and a is the standard deviation of the Gaussian distribution of
fluctuation angles y. Finally, the order parameter for the 3D GAF motion of
the peptide plane is given as the plateau value S = lim C'v(x) [20]:'Hv 'UA^
£.- % X'u.v exp
/ 2,,
2 .,2. 2,2 ,2,\i
oa(k +k' ) 22 cAm +m' y—~ <y -^—y—
/, k, k', m, rri
x <-oJ -'&\!>[?,(f>2f*>2i(?h Xp>rl j4">. asn\i
2.3 Calculation of relaxation-rate constants 47
where ca, cB, oy are the standard deviations (expressed in rad) of the
fluctuations about the principal axes ea, en, e„.
Equations (2.45) and (2.57) provide a convenient way to determine the
influence of 3D GAF motion on auto- and cross-correlated relaxation
parameters or, conversely, they allow the determination of the 3D GAF
parameters from experimental data as described in Chapter 3.
2.3 Calculation of relaxation-rate constants
For the calculation of relaxation rate constants of the different relaxation
pathways within a network of spins the master equation given in Eq. (2.6) is
transformed to the rotating frame [49]
j/(t) = -fr{pr(0-pj}, (2.58)
and then converted to the matrix form [49,69]
Jt =~lrnmK> (2-59)
m = 1
where it has been taken into account that T = F in case of restriction to
secular contributions only. Eq. (2.59) represents a linear system of first order
differential equations. The base coefficients bm result from decomposing the
deviation of the density operator from equilibrium in the rotating frameT T
|p (?) - p0) in terms of an orthogonal basis {Bm} in Liouville space:
<BJ(pT(0-pJ)>
(A\B) is the scalar product in Liouville space defined by the trace
(A\B) = Tr{A^B}. One has to select a basis {Bm} which is suitable for the
spin system under consideration. For X spins 1/2 the Liouville space of the
base operators has a dimension 2.In this thesis, a spin system IS of two
spins 1/2 has to be described and the 16 operators forming the shift-operator
48 Spin Relaxation and Molecular Motion
basis are convenient [49]:
E, 21z, 2SZ, 4IzSz, I+S_, I_S+, I+, S+, I, S_, 2I+SZ, 2IvS+,
2I_Sz,2IzS_,I+S+,LS_. (2.61)
The matrix elements Tnm of the superoperator T are the transfer rate
constants between the base operators Bn and Bm, and are given by
r-=(vJ' (2-62)
Diagonal elements Tnn are the rate constants for an auto-relaxation process
of Bn, off-diagonal elements Tnm represents the rate constants for an cross-
relaxation process between operators Bn and Bm. For a single relaxation
mechanism, the relaxation matrix can be simplified and a block-diagonal
matrix is obtained [7], since many cross-relaxation rate constants are zero. In
particular, cross-relaxation between operators with different coherence
orders is precluded as a consequence of the restriction to secular
contributions (see Section 2.1.4); for example, cross-relaxation does not
occur between zero- and single-quantum coherence. Furthermore, only
coherences that are degenerate or nearly degenerate show cross-relaxation
between off-diagonal elements of the density operator in the laboratory
frame. Thus, "transverse" cross-relaxation does not occur in a system of two
unlike spins IS [69]. In case of more than one relaxation-active interaction,
cross-correlation between the interactions can lead to relaxation pathways,
which are forbidden for a single relaxation mechanism [69]. In biomolecules,
interference effects between dipolar and CSA interactions or between dipolar
interactions of different pairs of spins have to be taken into account.
Longitudinal relaxation of a heteronuclear IS spin system. The
regarded system consists of a spin S with a non-axially symmetric CSA
tensor (e.g. N or C backbone spins) and a spin / without chemical shift
anisotropy (to good approximation represented by a proton). The dipolar
interaction between the spins / and S depends on the fixed distance ris. It
has been shown [69] that the longitudinal relaxation pathways resulting from
2.3 Calculation of relaxation-rate constants 49
dipolar and CSA interactions can be described by the subset of operators
{lz, S , 2/zSz} which corresponds to one block in the block-diagonalrelaxation matrix:
js
long
0 rsr2/lSzr2Vi,2/zSz
(2.63)
where relaxation described by the rate constant Fs 2j sresults only in the
case of cross-correlation between the dipolar and CSA interactions. The
auto-relaxation rate constant Fs scan be extracted from Tl experiments of
the spin S and is given as the sum of the dipolar (D) and CSA relaxation
contributions [37]:
sesz1,5
1
\,S^D
1
l,SJCSA(2.64)
with the dipolar relaxation rate constant
\,SJD
=lf^2 h_20U7t7 Un
2 2 -X2Y/Ys<>7s> {3/ (co5) + 7 (cd,-©^
+ 6/„ (a>7 + a>s)}, (2.65)
where y/( ys are the gyromagnetic ratios of the spins I and S, and (a,, co5
are the corresponding Larmor frequencies, h is Planck's constant and |X0 is
the magnetic field constant. /,,,.(«>) (with \i = {IS)) is the auto-correlated
spectral density function of the internuclear vector r]S. The longitudinal
CSA relaxation contributions of the symmetric part of the CSA can be
described as the sum of two auto-correlated CSA relaxation terms and one
cross-correlated CSA relaxation term according to Eq. (2.14)
V-)^ 1, S'CSA
(4i°x2jxx((as) + Oy^yy^s) + 2°xayJxy^s)} • (2-66)15
50 Spin Relaxation and Molecular Motion
The parameters ov and o„ are defined by cv = cv -a,T andx y ~ x xx ZZ
oy = a -
ozz, and oxx, o , azz are the pnncipal values of the CSA
tensor Jxx{<&), Jyy((d) are the auto-correlated spectral densities of the CSA
pnncipal axes x and y, and Jxy((0) is the corresponding cross-correlated
spectral density
The cross-relaxation rate constant r; sresults solely from the dipolar
interaction and is given by
F'^ = 55® it) Y?Ys<^3>2{6Vffl/ + (n5)-V0/-^^
(2 67)
T[ sis contained in the heteronuclear steady-state NOE (Nuclear
Overhauser Effect) of spin 5 which is obtained after presaturation of spin /
NOE = 1 + — —2—\ (2 68)Ys T5
Mi
where Tx sis the total relaxation time of the spin S including all
contnbutions The other relaxation rate constants r; j , Ts 2j $ >m^
T2/ i 21 scan ^e monitored selectively by different expenments and are
given elsewhere [70] These rate constants have not been measured dunng
this work but their existence had to be considered in the design of the
relaxation expenments In particular, the cross-relaxation descnbed by the
rate constant Ts 2/ shad to be suppressed in T^ expenments (see
Chapter 3) to yield a mono-exponential decay when monitonng the auto-
relaxation process of Sz
Transverse relaxation of a IS spin system. The transverse relaxation of
the IS spin system can be descnbed by the operator subset {/+, 5, 25+/,}
[36] with the following block diagonal of the relaxation matnx
2.3 Calculation of relaxation-rate constants 51
JS
_rv+ ° 0
o rs+;S+ TS+,2S+Iz0 r5+, 2S+IZ r2S+lz,2S+Il
(2.69)
The auto-relaxation rate constant Fs scan be extracted from T2
experiments of the spin S and can be decomposed into its dipolar and CSA
relaxation contributions:
s., s.
1
P2,S
1 1
T2,S*D ^T2,S'CSA
1 A lfVa\2fh\2 2 2, -3.2
+ /^((D; - cos) + 6y^(co7) + 67^(0); + ©s)} ,
t2,sjd 4ou«; wY/Y5<r's) {4V0)+37^(<°5)
(2.70)
(2.71)
r-)1^{^[4^(0) + 37^)] +
90
+ a/[4Jyy(0) + 3Jyy«os)] + 20,0^47^(0) + 3/^(a>5)]} . (2.72)
The cross-relaxation rate constant Ts SI= Fs 2s i
results from cross-
correlation between the dipolar and CSA interactions and takes the form [64]
*s,si = -^%S)f2^W<'-«>{^(4^/,*(0) + 3ys/>je(a)s))
+ cy(4JSIy(0) + 3JSIy((Os))} (2.73)
JS1 xand JSI are the spectral densities related to the cross-correlation
functions between the SI dipolar director and the principal axes x and y of
the CSA tensor of spin S, respectively. Cross-relaxation along this pathway
has to be suppressed in T2 experiments (see Chapter 3).
Recently, there was a revival of interest in dipole-CSA cross-correlated
52 Spin Relaxation and Molecular Motion
cross-relaxation experiments in biomolecules [64,71,72]. One of their most
interesting applications might be the identification of conformational
exchange processes in the (xs to ms range [73]. The cross-relaxation rate
constant rs SIleads to differential transverse relaxation of spin S of the two
multiplet components belonging to the a and p states of the J-coupled /
spin. Modulations of the chemical shift of spin 5 equally affect both
multiplet components and therefore do not lead to differential relaxation,
provided that the 1JSI coupling is constant which often can be safely assumed
[74]. On the other hand, it is well-known that conformational exchange, that
modulates the isotropic chemical shifts, also contributes to T2 [75,76].
Thus, conformational exchange is identified by comparison between the
T2 srate constants and the cross-relaxation rate constant Ts S].
In
Chapter 3, the cross-relaxation rate constant TN NHinduced by cross-
correlated relaxation between the ^N-1!! dipolar interaction and the 15N
chemical shift anisotropy tensor is used to identify peptide planes which are
involved in conformational exchange processes.
Relaxation due to antisymmetric part of CSA. In Section 2.1.3 the
antisymmetric part of the CSA, which is a rank 1 interaction, has been
separated from the symmetric part. Due to the orthogonality relationships of
the Wigner rotation matrix elements (see Eq. (2.26)) there is no cross-
correlation between the two interactions of different rank and the
antisymmetric part gives rise to a completely independent relaxation
mechanism. T{ relaxation of the spin S due to overall tumbling modulation
of the antisymmetric part of the CSA can be calculated by taking into account
Eq. (2.15) [35]:
1 \anti 1 2 2 (I)
r-j = Wsi^LA^s)' (2-74)1\,SJCSA J
2 2 2 2 f 1)where Aoantl = G + Gxz + Gvz and / (co) is the rank 1 spectral density
function for a rigid molecule given in Eq. (2.36).
2.4 Molecular dynamics simulation 53
2.4 Molecular dynamics simulation
2.4.1 The force field
Molecular dynamics (MD) computer simulations describe the classical
motion of atoms or molecular fragments in a force field which has been
developed taking into account experimental data, quantum chemical
calculations and also empirical considerations [77-80]. A MD trajectory is
calculated by starting at a given initial conformation and solving Newton's
equations of motion for all the atoms in the system:
mrV,(0 = " ^V(rv r2,...,rs), (2.75)dt 3r,
where S is the number of atoms in the system, m[ and r, the mass and
position of particle i. The force field of the simulation package CHARMM
[77,78] which has been used in Chapters 3 and 4 is given by the following
potential energy function V :
V({rl})= l\Kb(b-bQ)2+ £ \kq(Q-%)2Bonds Angles
2 -
Improper Dihedrals
Dihedrals
pairs(i,j)
Cn{i,j) C6(i,jj12 6
+ V ^,(2.76)— ,47ten£r„
pairsxi, j) U r U
where b, Kb, and b0 are the bond length, the bond stretching force constant,
and the equilibrium distance parameter, respectively; 0, Kq ,and G0 are the
bond angle, the angle bending force constant, and the equilibrium bond
angle, respectively; \, K^, and £0 are the improper dihedral, its force
constant, and the corresponding equilibrium value, respectively; (p, K , n,
and 8 are the dihedral angle, its force constant, multiplicity, and phase,
respectively; r, ql, e0 and er are the non-bonded distance between atoms
i and j, the charge of atom i, the electric field constant, and a dielectric
54 Spin Relaxation and Molecular Motion
parameter, respectively; Ci2(i, j) and C6(/, j) parametrize the Lennard-
Jones potential for the non-bonded interaction between atoms i and ;'.
MD simulations are of increasing importance for studying biomolecular
systems [18,19]. They represent the best theoretical technique to draw a
qualitative picture of biomolecular motions and to allow the derivation of
realistic motional models for intramolecular motion [37]. However, a
quantitative agreement between simulation and experiment is often not
observed. The differences partly result from the fact that liquid state NMR
relaxation measurements always reflect ensemble averages over a huge
number of molecules, whereas relaxation parameter extracted from a MD
trajectory result from a time average over a trajectory in the nanosecond
range. Results from MD simulation will be discussed in detail in Chapters 3
and 4.
2.4.2 Calculation of correlation functions
The internal correlation function c'"(t)= {P2(cos(6 (?)))) with the
angle 9 (f) between two axially symmetric interactions u. and v can be
calculated from the MD trajectory consisting of N snapshots with the time
increment At:
K 'k=\
where e^j. defines the orientation of the symmetry axis of the axially
symmetric interaction tensor ]i at snapshot k in the molecular frame. If the
correlation time of the observed motional process is sufficiently smaller than2
the simulation time, a plateau value of the internal correlation function, S ,
can be calculated by a single loop over the trajectory according to Eq. (2.87)
if o, avv are set to 1.
2.4.3 Extraction of 3D GAF fluctuation amplitudes
The principal axis directions and fluctuation amplitudes of the fluctuation
ellipsoid representing 3D GAF motion of a peptide plane can be determined
2.4 Molecular dynamics simulation 55
from the trajectory by an analytical method which was introduced by Bremi
and Briischweiler [20] and shall be derived in the following. A principal axis
frame consisting of three orthonormal vectors e{, e2, e3 is rigidly attached
to the equilibrium peptide plane. First, we assume the special case with axes
e[ = (1,0, 0), (t2 = (0, 1,0), (t3 = (0,0, 1) which are collinear to the
principal axes of the reference frame ea,en,e, respectively, defined for the
3D GAF model in Fig. 2.1b. Next, the peptide plane is reoriented by a
rotation with the matrix Rppf = Re (Y)^e(3(P)^ea(a) representing three
successive rotations, with angle a about ^, then with angle p about en, and
finally with angle y about e '
V/(0C'P'Y)
cos P cosysin a sin P cosy cos a sin P cosy
-cosasiny + sinasiny
„ .sin a sin 8 siny cosasinPsiny
cosPsiny+ cosacosy -sinacosy
-sinP sinacosP cosacosp
. (2.78)
The three orthogonal unit vectors are then given by et t= R Ja, P, y)e(
with i = 1,2, 3. Linear averaging over Gaussian distributions of the
reorientation fluctuations a, P, y yields the three vectors with respect to the
average peptide plane frame:
^ / -cl/2-al/2 \
(elt) = (<cospcosy)3OGAF,0,0) = [ep y
,0,0j,^ / -a2a/2-a2/2 \
<e2(> = (0, <cosacosy)3DGAf.,0) = \0,e ,0J,
(e~^t) = (0,0,<cosacosp)3DGAF) = (o, 0, e'^2~ °p/2j, (2.79)
where the Gaussian averaging is performed according to ref. [63]:
<cosmq>>GflIIM =
J(2ito(p2)
2-1/2 -<P /(2a ) -m*a;/2^
e cosm(paq> = e
_^ _^ _^
(2'80)
Note, that the linear averaged vectors (et t), (e2 (), (e3 ) ^^ neither
orthogonal to each other nor normalized to 1. In addition, these vectors do
56 Spin Relaxation and Molecular Motion
not depend on the order of rotations R„ (a), /? (B), R (y) which is a
consequence of the Gaussian averaging Hence, the descnption with Rppt is
possible without loss of generality A diagonal product matnx
M,, = (e, ,) (e, ,) (i, J = 1,2,3) of these averaged vectors can beIJ 11 J t
calculated
M =
2 2
0T0 0
0 e
2
-v2
°y0
0 0 e
2 2
°P
(2 81)
and the fluctuation amplitudes are finally given by a linear combination ofthe
diagonal elements Aj, A2, A3 of the matrix M
2v A^Ata
Klogd[A,3
2 1A A3^ =
2ll0gA^ (2 82)
Rea,e2= Re$>In the general case with the orthogonal unit vectors e^
e3 = Re not colhnear to the principal axes ea, ea, e,where the rotation
by the matrix R describes the onentational offset from colhneanty and the
columns of R contain the coordinates of e^, e2, e3 with respect to the
equilibrium peptide plane frame, a product matnx M' is obtained which is no
longer diagonal However, it can be shown by a few transformations that theT
diagonal matrix M is given by the operation RM'R
This leads to the following procedure for extracting the fluctuation
amplitudes oa, o», a from the MD trajectory First, a frame of three
orthogonal unit vectors Cj, e2, £3 is fixed at the peptide plane of each
snapshot The onentation of this frame with respect to the peptide plane has
to be the same for all snapshots but can be arbitrarily chosen Diagonahzation
of the calculated product matrix M' yields the fluctuation amplitudes from
Tthe eigenvalues of the matrix M = RM'R via Eq (2 82) The orientations
for the principal axes of the peptide-plane reonentation are given by R e,,
R e1,R e-.
2 5 Motion-induced fluctuations of the CSA 57
2.5 Motion-induced fluctuations of the CSA
The quantitative knowledge of chemical shielding anisotropy (CSA)
tensors is important in the context of many biomolecular NMR applications
CSA tensors provide important information on onentations of molecular
fragments and on the electronic environment of the nuclei, which depends on
the molecular geometry Moreover, the CSA contribution to spin relaxation
in liquids gives umque insight into overall motion as well as anisotropic
internal dynamics A quantitative interpretation of relaxation data greatly
benefits from the accurate knowledge of CSA tensors
CSA parameters can be extracted from solid-state NMR experiments [22]
In case of biological molecules, however, those studies are restricted to small
peptides Recent progress in quantum chemical methods for the calculation
of CSA tensors allows applications to molecules of moderate size with
increasingly quantitative agreement with experimental data [81-85] Density
functional theory (DFT) [23,86] has proven to be particularly useful to
calculate CSA tensors of model systems that represent essential fragments of
larger biomolecules
The CSA parameters of nuclear spins are functions of the molecular
geometry and its environment which are modulated by intramolecular
motion including vibrations, angular fluctuations, and fluctuations of near
contacts The effect of intramolecular motion on the CSA tensor might be
described by a two-level approximation In a first step, the CSA tensor can
be regarded as fixed at the molecular fragment of interest and reorients
according to the fragment's motional modes On a second level, the
magnitude and the orientation of the tensor with respect to the molecular
fragment are modified due to the changes in the chemical and electronic
environment induced by the intramolecular motion Rapid CSA tensor
modulations are sensed differently depending on the NMR experiment
While solid-state NMR experiments yield a linearly time-averaged CSA
tensor, NMR relaxation experiments reflect the variance of CSA tensor
fluctuations via second order pertubation theory
[87]:4n)factoraintroducing(by(2.40)Eq.from
calculatedbecanfunctioncorrelationtheCmt(-x),=Cint(x)accountinto
Takingterms.cross-correlatedallandauto-allofsumtheiswhichfunction
correlationrank2ndabydescribedbethencanframemoleculartheinvalues
principalandorientationstensorCSAtheoffluctuationstime-dependentThe
(2.83)
2
-1
-1
^ozzdiag+
-1
^oyydiag+
-1
-1
2
\f
diag1
vvdiag
tensors:symmetricaxiallyfree,
trace-"orthogonal",threeofsumaasrepresentedbenowcantensorCSA
symmetric)axiallynecessarilynot(butsymmetricThek°zz,k>°yy,k'axx,
eigenvaluestracelessandkez,
k'
c>,
k>
x,ek,ek,exeigenvectorsofseries
atoleadingdiagonalizedandsymmetrizedisframe,moleculartheingiven
trajectory,theof{Gk}tensorCSAEachfollowing:theinexplainedbewill
procedureThisfunction.correlationacalculatingbycharacterizedbetohave
trajectoryCSAtheincontainedparametersCSAtheoffluctuationsThe
trajectoryCSAofAnalysis2.5.1
analyzed.be
willubiquitinproteintheinspinsbackbone15NandCoftrajectoriesCSA
4,ChapterInmolecule.theofmotionintramoleculartheondependdirectly
fluctuationswhose...,N)I,=(k{ck}tensorsCSAofseriesatoleadsThis
trajectory.MDtheincontained"snapshot"everyforcalculatedisnucleus
certainafortensorCSAAimportance:centralofistrajectoryCSAthe
approach,thisIntensor.CSAtheoneffectsdynamicalstudytopossibilities
largeoffercalculationsDFTandsimulationMDofcombinationThe
MotionMolecularandRelaxationSpin58
2 5 Motion-induced fluctuations of the CSA 59
Cin,(T) = (oxx(0)Oxx{t)P2(ex(0) ex(x)))
+ (a (0)ayy(x)P2(e (0) • e (x))>
+ <c77(0)c77(T)P2(e7(0) • e,(x))>
+ 2<o„(0)ovv(T)P2(^(0)-ev(T))>
+ 2<oJCC(0)ozz(x)P2(eJt(0)-ez(x))>
+ 2(ayy(0)azz(T)P2(ey(0) ez(x))). (2.84)
Due to the functional form of Eq. (2.84) it is not necessary to assume any
ordering of the eigenvalues. They can be taken in arbitrary order, e.g., in the
order they are provided by the diagonalization subroutine. In other words, for
each snapshot the axes x, y, z together with the eigenvalues can be arbitrarily
permuted. The individual terms of Eq. (2.84) are rather meaningless, since
only their total sum gives rise to an experimentally observable relaxation
effect. The initial value of the internal correlation function contains the
interaction strength and is not normalized to 1:
c,„/(0) = <<4> + (o2yy) + (a2zz) - (axxoyy) - (axxazz) - (ayyozz).
(2.85)
2The plateau value of the correlation function, the order parameter 5
,can be
calculated as a sum of order parameters which correspond to the auto- and
cross-correlated terms in Eq. (2.84):
S2 = S2XX + S2yy + 4 + 2S% + 2S2XZ + 2S2yz, (2.86)
2where the 5„
vcan be calculated according to ref. [62] including the principal
value fluctuation:
60 Spin Relaxation and Molecular Motion
V = 4<0unV^0vvZv>-4(<annV^0vv)+<V^0vvZv))
+ \<V)(°w> + 3«VW<°vv(Zv*v)>
+ <VZHVKvW + <Wu»<°vvW)
+ 4(<VXJ)(°vvX5> + (V^^vv^v)
-<o^xJ><ovvyJ> - <aw^)(ovvxJ)), (2.87)
where (I, V = x, y, z are labels for the vectors of the corresponding CSA
principal axes and X, Y, Z are the cartesian coordinates of these vectors in
the molecular frame.
In case of very fast CSA fluctuations with correlation times below 20 ps the
CSA relaxation rate constants can be calculated based on the calculated order
parameters. This will be described in Chapter 4.5.1.
2.5.2 CSA averaging due to 3D GAF motion
CSA tensors which are extracted from solid-state NMR experiments at
room temperature are affected by motional averaging due to the
intramolecular motion present in the solid state. The motional averaging
effect can be calculated analytically for the 15N and 13C backbone spins,
assuming that the CSA tensor is fixed at a peptide plane undergoing 3D GAF
motion and that the motional reorientation does not lead to a modification of
the tensor's principal values and principal axis orientations with respect to
the peptide plane frame. The calculation is somehow analogous to the
procedure of Briischweiler and Case [88] who calculated the effect of
harmonic motion on the Karplus equation for spin-spin coupling.
A CSA tensor Z with an arbitrary orientation with respect to the
equilibrium peptide plane frame (ppf) ea, ea, e,can be calculated as
? = Rppf(cC, P', •f)qR~pf(a\ P', y'), where g = diag(axx, a azz) and
2 5 Motion-induced fluctuations of the CSA 61
the rotation matnx R * (see Eq (2 78)) descnbes a rotation with fixed
angles a', P', y' about the pnncipal axes ea, en, e, respectively The
motionally averaged CSA tensor 1! is given by averaging over the ensemble
of CSA tensors reonented by 3D GAF motion
r = <J?^a,p,Y)*w/^P\Y)gJ?^(2 88)
3DGAFwhere ( )a r Y
defines die average over the Gaussian distnbution of
reonentation angles a, P, y For an CSA tensor ? with an onentation
corresponding to (a1, p', y') £ (0, 0, 0) the explicit solution for X' is quite
lengthy (not shown) Here, the solution for the trivial case
(a', P', y') = (0, 0, 0) with the pnncipal axes ex, e, ez of the CSA tensor
£ collmear to the pnncipal axes ea, en, e of the peptide plane frame, shall
be given With the matnx R <• defined m Eq (2 78) and the relations
2 GAF 1 1 -2a2<(cos(p) >(p = - + -e Tsec((p),and
2, GAF 1 1 -2a.2.((sincp)2^ = V~y ^es(<p), (2 89)
where ec((p), es(<p) with (p = a, p, y are functions of the fluctuation
amplitudes aa, oR, a descnbing the peptide plane reonentation, the
pnncipal values of the diagonalized averaged CSA tensor S' are finally given
by
a'xx = <yyy(es(a)es{$)ec(Y) + ec(a)es{y))
+ 0zz(ec(oc)es(p)ec(y) + es(a)es(y)) + oxxec($)ec(y),
+ azz(ec(a)es($)es(y) + es(a)ec(y)) + axxec($)es(y),
°\z = ^xx^(P)+ <V»ec(P) + <VC(«)*C(P) (2 90)
The onentation of the averaged CSA tensor S' is not modified compared to
62 Spin Relaxation and Molecular Motion
the reference tensor Z The trivial case with (a', (3', y') = (0, 0,0) is useful
since the 13C CSA principal axes are approximately colhnear to the peptide
plane frame pnncipal axes (see Fig 2 3) An analogous calculation leads to
the "removal" of the 3D GAF motion from the averaged CSA tensor Z' in
order to obtain the tensor Z A detailed discussion of motional averaging of
CSA tensors using also this "3D GAF averaging" approach is given in
Chapter 4
3 1 Introduction 63
3 Anisotropic Intramolecular Backbone
Dynamics of Ubiquitin
3.1 Introduction
In this chapter, it is demonstrated experimentally that anisotropic
intramolecular backbone motion can be characterized by a combined
interpretation of relaxation data of 13C and 15N nuclei belonging to the same
peptide plane The procedure follows the general protocol originally
developed for the characterization of side-chain dynamics and subsequently
applied to the characterization of backbone dynamics, exemplified on the
cyclic decapeptide antamanide [20,37] The protocol uses information
gained from molecular dynamics (MD) computer simulations and from
analytical treatments to find a suitable parametnzation of the motion causing
nuclear spin relaxation From the detailed analysis of a 1 5 ns MD trajectory
of ubiquitin solvated in a box of water, the basic motional processes affecting
backbone spin relaxation are determined and expressed in terms of an
analytical motional model It is found that 76% of the backbone peptide
planes are not involved in conformational exchange processes and show
predominantly small-amplitude motion Their relaxation behavior can be
described by the 3D Gaussian axial fluctuation (GAF) model which has been
introduced in Section 2 2 3 Each peptide plane is treated as a rigid entity
exhibiting rapid reonentational motion about three orthogonal principal axes
with a Gaussian fluctuation distribution The axis of maximum angular
fluctuation turns out to be nearly parallel to the C, _ j- C; direction The
model, expressed in analytical terms, is applied to experimental T! and T2
data of the 13C and 15N backbone spins, and to {^HJ-^N NOE data
collected at two magnetic field strengths (400 and 600 MHz proton
resonance frequency)
The approach in this chapter differs from procedures previously proposed
for the dynamical interpretation of backbone relaxation data of proteins
Most studies focused on the interpretation of N relaxation data in terms of
the standard Lipan-Szabo approach Recently, experiments were proposed
64 Anisotropic Intramolecular Backbone Dynamics of Ubiquitm
for monitoring C relaxation. Engelke and Ruterjans [89] used a model-
free description assuming identical order parameters for the three C CSA
principal axes which is equivalent to assuming isotropic internal motion at
1 ^the C sites. They found for ribonuclease Tl significant discrepancies
between the derived C and N order parameters belonging to the same
peptide bond, but no physical explanation was provided. Allard and Hard
[90] recognized the importance of dipolar contributions of the neighboring
protons to 13C relaxation and the need for separate order parameters for the
different interactions. They modeled the backbone motion of the
thermostable Sso7d protein by two order parameters for each 13C, one
describing motion at the backbone c site and one describing the
(effective) motion of the internuclear vectors to the neighboring protons.
Similar to the treatment by Engelke and Ruterjans, their approach assumes
isotropic intramolecular motion at the C site. Dayie and Wagner [91,92]
derived information on the spectral densities of the carbonyl CSA interaction
in villin 14T. No interpretation in terms of anisotropic intramolecular motion
was undertaken. Zuiderweg and coworkers [93] measured numerous auto-
and cross-correlated relaxation rate constants involving 15N and 13C
backbone atoms of flavodoxin and determined the associated motional order
parameters. Differences in these order parameters were ascribed to
anisotropic peptide-plane motion, which was modeled in terms of uniaxial
restricted diffusion. No attempt was made to interpret the set of relaxation
data by a unified motional model.
In Section 3.2, the biological role and the structure of the studied protein
ubiquitin is characterized. In Section 3.3, details of the 15N and 13C NMR
relaxation experiments are given, whereas the generation, processing and
analysis of the MD trajectory is described in Section 3.4. In Section 3.5, the
extraction of the motional 3D GAF parameters from the NMR relaxation
data is outlined in detail, followed by a discussion in Section 3.6.
3.2 Ubiquitin
Ubiquitin is a highly conserved, 76-residue protein (molar mass of
8565 gmol ) which is present universally in eukaryotic cells. It is involved
3 2 Ubiquitin 65
in a variety of cellular processes, most importantly protein degradation
Proteins that are selected for degradation are covalently linked to multiple
molecules of ubiquitin The linkage is an isopeptide bond between the C-
terminal glycine in ubiquitin and side-chain amino groups of lysine on the
target protein The conjugates are then recognized by special proteases
within the cell Most m-vivo- degradation of abnormal or short-lived proteins
is mediated by the ubiquitin-dependent pathway More details concerning the
biochemical function of ubiquitin can be found in refs [94,95]
The primary structure of ubiquitin is given in Fig 3 la Ubiquitin lacks
cystem and tryptophan residues and includes only one histidme, one tyrosine,
and one methionine at its N terminal The isoelectric point is at 6 7 due to
11 acidic and 11 basic residues (plus one histidme) The crystal structure of
human ubiquitin was determined and refined to 1 8 A by Vijay-Kumar et al
[96] The structure in solution determined by NMR [98-100] is virtually
identical to the crystal structure As illustrated in Fig 3 lb, the structure is
rich in secondary structure elements, including a mixed five-strand (3 sheet,
which contains two strands (residues 1 to 7 and 64 to 72) in parallel and three
strands (residues 10 to 17, 40 to 45, 48 to 50) in antiparallel direction, an a
helix (residues 23-34), a short 310 helix (residues 56-59), and seven reverse
turns Nearly 90% of the backbone is involved m a hydrogen-bonded
secondary structure The curved p sheet and the flanking a helix enclose a
single core of densely packed hydrophobic side chains, which is likely to
contribute to the high stability of ubiquitin towards denaturation by heat,
extreme values of pH, and denaturing agents [101,102]
66 Anisotropic Intramolecular Backbone Dynamics of Ubiquitin
b)
1 5 10
Met-Gln-lle-Phe-Val-Lys-Thr-Leu-Thr-Gly-Lys-Thr-lle-15 20 25
Thr-Leu-Glu-Val-Glu-Pro-Ser-Asp-Thr-lle-Glu-Asn-Val-30 35
Lys-Aia-Lys-lle-Gln-Asp-Lys-Glu-Gly-lle-Pro-Pro-Asp-40 45 50
Gln-Gln-Arg-Leu-lle-Phe-Ala-Gly-Lys-Gln-Leu-Glu-Asp-55 60 65
Gly-Arg-Thr-Leu-Ser-Asp-Tyr-Asn-lle-Gln-Lys-Glu-Ser-70 75
Thr-Leu-His-Leu-Val-Leu-Arg-Leu-Arg-Gly-Gly
G76
Fig. 3.1 (a) Primary sequence and (b) schematic representation of the native state of
human ubiquitin [96] drawn by using Molscript [97],
3.3 Experimental 67
Ubiquitin has been subject of numerous NMR studies. H, C, and N
resonance assignments are available at pH 4.7 and 30°C [103], and at pH 5.7
and 30°C [104]. The backbone dynamics of ubiquitin at pH 5.7 and 30°C
was investigated by Schneider et al. [105] in terms of a standard analysis of
15N relaxation data. They found a correlation time of 4.1 ns for isotropic
overall tumbling motion and no apparent correlation between secondary2
structure and order parameters SNH for the NH vectors. The rotational
tumbling of ubiquitin at pH 4.7 and 30° C was analyzed in more detail by
Tjandra et al. [106]. An axially symmetric rotational diffusion tensor with a
diffusion anisotropy of ~ 17 % was extracted from N relaxation data
together with the order parameters for the NH vectors. Wand et al. [104] used
fractionally 13C-labeled ubiquitin to extract order parameters for the
oc ex
dynamics of backbone C - H vectors and methyl groups. They found, for
those residues within an element of secondary structure, the order parameters2 2
Sa a usually larger than the order parameters SNH of the same residue. A
C M
detailed investigation of the backbone dynamics of the partially folded A
state of ubiquitin was presented by Brutscher et al. [73]. The analysis of
cross-correlation relaxation data of the backbone of native ubiquitin was
subject of several NMR studies [64,71,107]. Ubiquitin was also used as
model compound in some molecular dynamics simulation studies. Braatz et
al. [108] compared a time-averaged simulated structure with the
experimental crystal structure. Abseher et al. [109] characterized partially
unfolded states of ubiquitin with MD simulation.
3.3 Experimental
3.3.1 Sample preparation
Fully 13C, 15N-labeled ubiquitin was kindly provided by A. J. Wand
(Buffalo, NY) and purchased from VLI (Southeastern, PA). For optimal
consistency, all C and N relaxation data used for the analysis were
measured on a single ubiquitin sample containing 2 mM ubiquitin in 90%
H20 and 10% D20 at pH 4.7 with a 45 mM sodium acetate buffer. The
sample was deoxygenated and sealed in a standard 5 mm NMR sample tube.
68 Anisotropic Intramolecular Backbone Dynamics of Ubiquitin
3.3.2 NMR experiments
3.3.2.1 General setup and techniques
NMR experiments were performed on Bruker AMX-600 and DMX-400
spectrometers, both with triple-resonance (1H, 15N, 13C) equipment
including shielded z gradients. The sample temperature was set to 300 K.
Special care was taken to achieve a temperature variation of less than 0.5 K
between different relaxation experiments and different spectrometers by
using the temperature dependence of differential cross-peak shifts in the 2D
HSQC correlation spectra as an internal probe for calibration of the actual
sample temperature. This is illustrated in Fig. 3.2 for a HSQC spectrum
resulting from a 15N Tt experiment. For all NMR experiments,
WATERGATE [110] and water flip back pulses [111] were inserted in the
pulse sequences to obtain a relaxed spin state of the water magnetization
before acquisition. Pulsed field z gradients were applied for coherence-
transfer-pathway selection [112]. Quadrature detection in the indirect
dimensions of the two-dimensional experiments was obtained by the TPPI-
States method [113]. The spectral widths and the carrier frequencies (in
parentheses) were set for *H to 12 ppm (4.76 ppm) and for 15N to 22 ppm
(118.76 ppm), the carrier frequencies for 13Ca to 57.2 ppm and for 13C to
173.9 ppm. The frequency for the selective 13C and 13Ca pulses was
switched by time-proportional phase incrementation during rf pulses.
Selective rectangular 90° pulses were used with pulse lengths of 74 u.s (400
MHz) and 57 |0.s (600 MHz). The tip angle was changed by adjusting the
power level.
3.3.2.2 2D 1SN relaxation experiments
Tj and NOE experiments were performed at 400 and 600 MHz, T2
experiments in the rotating frame at 600 MHz proton resonance frequency.
The used pulse sequences are shown in Fig. 3.3. Since a fully I3C, 15N -
labeled sample of ubiquitin was used, the Ca and C nuclei had to be
broadband decoupled during the evolution period tx. lH 180° pulses and
selective Ca and C 180° pulses were applied during the T{ and T2 mixing
3.3 Experimental 69
time delays to suppress cross-relaxation and effects caused by cross-
correlation between 15N CSA and ^N^H dipolar or 15N-13C dipolar (DD)
interactions. During the T2 relaxation period, chemical shift and JNH-
coupling evolution were suppressed by applying spin-lock rf fields with
17^2^1/271= 1.9 and 2.6 kHz. In these experiments the magnetization is
locked along an effective field vector with an orientation depending on the
frequency offset Av of the 15N nucleus. The measured relaxation rates jeas
were corrected for these offset effects using the relation
Tcorr = r^fli(riSin20)/(ri_r^ascos2e) with the angle 6 defined by
tan 9 = |yArB1|/2jtAv.
Two sets of Tj experiments were carried out both at 400 MHz and at
600 MHz to estimate statistical errors. The Trrelaxation decay was sampled
at ten time points: 20, 60, 140, 240 (2x), 340,480,660, 800, and 1000 ms at
400 MHz and 20 (2x), 60, 140, 240, 360, 480 (2x), 660, and 800 ms at
600 MHz. The relaxation decay of the two sets of T2 measurements in the
rotating frame was sampled at eight different time points: 8, 24, 48, 72, 96,
120, 160, and 200 ms at 600 MHz. The matrix size of the acquired complex
2D data was 128 x 512 with acquisition times of 143 ms (tj) and 107 ms (t2)
at 400 MHz and 96 ms (tj) and 71 ms (t2) at 600 MHz using 16 scans per
complex tj increment. For the {1H}I5N-NOE measurements, two spectra
with H saturation and one without were recorded in an interleaved manner.
Different H saturation schemes of 5 s length were applied during the recycle
delay of 7 s: either a windowless WALTZ-16 sequence or a train of 120°
pulses applied every 10 ms. The long recovery delay of 7 s was chosen to
allow the initial 15N magnetization to reach its equilibrium value despite
saturation transfer by water exchange [114]. The three acquired matrices
were 75x512 (400 MHz) and 128x512 (600 MHz), and 32 scans per
complex tj increment were used for each of the three data sets.
70 Anisotropic Intramolecular Backbone Dynamics of Ubiquitm
7 7 [ppm]
Fig 3 2 'H 15N correlation spectrum resulting from a 2D N 7", relaxation
experiment (see Fig 3 3a) with mixing time xm = ~^[ls Four pairs of two overlapping
peaks are sui rounded by circles The corresponding peptide planes were excluded from
further analysis (see Section 3 5 1) The region of three peaks surrounded by a square
belongs to the iesidues Val 5 Leu 8 and He 44 The measurement of differential
chemical shifts between the peaks of Val 5/Ile 44 and the reference peak (Leu 8) allows
to contiol the sample temperature which varies by less than 0 5 K between different
experiments at different spectrometers
3.3 Experimental 71
a) y y
1h m, n p -r.n.io„
-y o.
15nI n lAHAl Tm lAlv2 e I ;aHa| waltz
13C
PFG
JL GARP GARP
b) y y
1h IaQai. n
13,
I<£iy
15N I n U^ha^ W^ I lAHAl WALTZ
JL GARP GARP
PFG
C)Saturation: 5 s
1h nnnnnnnnnnnn
15N
13C
PFG
<J>1
JL
|Aptl/2; e | | Af| A | WALTZ
GARP GARP
Fig. 3.3 Pulse sequences used for (a) 15N T,, (b) 15N T2, and (c) {'H}-'5N-NOEmeasurements. Delays are set to A=2.2 ms. e=A+t|/2. Phase cycling: <!>] = [x, (-x)];
C>rec = [x, (-x)]. The C carrier is set between the 13Ca and C resonances to
decouple both nuclei during the tj evolution of N.
72 Anisotropic Intramolecular Backbone Dynamics of Ubiquitin
3.3.2.3 2D C relaxation experiments
2D 13C Tj and T2 relaxation experiments were carried out at 400 and 600
MHz using HNCO-type experiments [90-92,115]. The pulse sequences are
depicted in Fig. 3.4. With the evolution period tj on N one has the
possibility to directly compare the actual sample temperature in N and C
relaxation experiments by extracting differential chemical shifts in the
corresponding 1H- N correlation spectra (see Fig. 3.2). During the constant
time evolution period tt on 15N the protons are decoupled with WALTZ-16
and 13Ca with one selective 180° pulse; during the mixing time, *H and 15N
180° pulse trains were applied in both Tj and T2 experiments and a train of
selective Gaussian 180° pulses of 400 (is pulse length every 100 ms in the Tl
experiment to suppress both cross-relaxation between C and surrounding
spins and cross-correlation effects between c CSA and dipolar
interactions (15N-13C\ 1H-13C, 13Ca-13C). During the T2-relaxation period
a spin-lock with field strengths of 1.7 kHz (400 MHz) and 2.7 kHz
(600 MHz) was applied on 13C, but no pulses were applied on 13Ca. At
600 MHz the spin-lock experiment was compared with a CPMG experiment
using a train of selective C 180° pulses every 500 [is. No systematic
deviation between the two experiments was found.
Statistical errors were estimated from two sets of Tj experiments at
400 MHz and three sets of T] experiments at 600 MHz. The T{ relaxation
decay was sampled at 15 and 11 time points: 4,200, 300,400,500,600,700,
800,1000,1200,1500,1800,2200,2600, and 3000 at 400 MHz and 20,200,
400,600, 800,1000,1200,1400,1800,2200, and 2600 ms at 600 MHz. The
T2 relaxation decay was sampled at 11 and 12 different time points: 25, 50,
75, 100, 125, 150, 200, 250, 300 (2x), and 400 ms at 400 MHz and 25, 50,
75, 100, 125, 150, 200, 250, 300 (2x), 350, and 400 ms at 600 MHz. The
matrix sizes of the acquired complex 2D data were 25 x 512 (400 MHz) and
38x512 with acquisition times of 28 ms(ti)and 107 ms (t2) at 400 MHz and
28 ms (tt) and 71 ms (t2) at 600 MHz using 64 (400 MHz) and 32 (600 MHz)
scans per complex ti increment.
3 3 Experimental 73
a)y v
x 0n
1H |a[1aL|2A[ WALTZ I PI Pi l~l PI I WALTZ I2A, Jb'-Xn-X ^rec
£>0-,|
y <t>4
i5n| n i .a.n n n n n u^^i AnA^ALTzio2 y ^3
13C. |GARP|
I :v2fl13ca [GARP|
pfgOI I n 11 «
b)y y
'H I41aL|2A| WALTZ | innn I WALTZ |2A|
isn| n T a\\4 n n n pW-X
-V2|l--"'
,A||A|WALTZ|
«c- n l^i^&H n |GARP|
13C« ,V2fl |GARP|
pFGnn « n n n
Fig 3 4 Pulse sequences used for (a) nC T, and (b) l3C T2 relaxation measurements
Delays are set to A=2 2 ms, A[=8 5 ms, A2=14 ms, and A^=15 ms Constant time
evolution TN=A2 Phase cycling <1>|= [(-y), y], <t>2= !.4x,4( x)], <J>3= [16(-y), I6y],04= [8x, 8(-x)], Orec= [x, -x, x, -x, 2(-x, x, -x, x), x, -x, x, -x, -x, x, -x, x, 2(x, -x, x, -x),
-x, x, -x, x]
3.3.3 Extraction of relaxation data
The data were processed with the FELIX program version 95 0 (Biosym
Technologies) Prior to 2D Fourier transformation, the time-domain data
were zero-filled in the t2 dimension to 2048 complex points and multipliedwith a cosine-bell window For the
' 5N relaxation experiments, zero-filling
74 Anisotropic Intramolecular Backbone Dynamics of Ubiquitin
was applied in the ^ dimension to 1024 real points followed by
multiplication with a cosine-bell window. For the C relaxation
experiments, a mirror-image linear prediction procedure [116] was applied
along the constant-time ti dimension, which was followed by zero-filling to
1024 real points and multiplication with a Kaiser window. The resonance
assignments were taken from Wang et al. [103]. Peak intensities were
extracted from 2D spectra using a local grid search routine for each cross
peak. Tj and T2 values were determined by fitting the measured peak heights
to the mono-exponential function I(tm) = /0exp(-xm/7,1 2) with two fit
parameters, Tt 2and the peak intensity /0 at mixing time %m = 0.
Representative decay curves for C T[ and T2 measurements are shown in
Fig. 3.5. The { H}-15N steady-state NOE values were determined from the
ratios of the measured cross-peak intensities in the presence (Isal) and
absence (Iunsat) of proton saturation: NOEmeas = Isa/1unsat.
Fig. 3 5 Typical magnetization decay curves for peptide planes Lys 6 and Asp 32 in a
P strand and in the central a helix of ubiquitin, respectively Peak intensities were
measured as peak heights (a) The 13C Tj fit is based on 11 experimental data points
measured with the experiment of Fig 3 4a at 600 MHz proton resonance frequency
with mixing times of 20, 200, 400, 600, 800, 1000, 1200, 1400, 1800, 2200, and
2600 ms (b) The 13C T2 fit is based on 12 experimental data points measured with the
experiment of Fig 3 4b at 600 MHz proton resonance frequency with mixing times of
25, 50, 75, 100, 125, 150, 200, 250, 300 (2x), 350, and 400 ms
3.4 Molecular dynamics simulation 75
3.4 Molecular dynamics simulation
3.4.1 Generation of molecular dynamics trajectory
The coordinates of the X-ray structure of ubiquitin [96] in the Brookhaven
data bank (file lubq) were used for the starting conformation of the MD
simulation. All protons were added in their standard geometric positions
using the CHARMM program [77,78]. The resulting structure was then
energy-minimized in vacuo and immersed in a cubic box of a side length of
46.65 A containing a total of 2909 explicit water molecules. The simulation
was performed with the CHARMM force field version 24b2 under periodic
boundary conditions with an integration time step in the Verlet algorithm of
1 fs. The SHAKE algorithm [117] was applied to all bond lenghts involving
a hydrogen atom. A cutoff of 8 A was used for non-bonded interactions.
Truncation was done with a shifting function for electrostatic interactions
with a dielectric constant e = 1 and a switching function for van-der-Waals
interactions. The temperature was set to 300 K, and after an equilibration of
500 ps, snapshots were stored in intervals of 1 ps, leading to a total of 1500
conformations for the 1.5 ns simulation time. The snapshots were then
postprocessed by a mass-weighted least-square difference rotation and
translation of the protein backbone atoms with respect to a reference
conformation at 750 ps simulation time. In this way overall rotational and
translational diffusion of the molecule, that occurs during the simulation, is
eliminated yielding atomic coordinates for each snapshot in the same
molecular reference frame.
3.4.2 Processing of trajectory
3.4.2.1 Determination of equilibrium peptide plane
The following cluster analysis was performed to determine the equilibrium
orientation of a peptide plane:
1. For peptide plane i, the corresponding peptide plane frame with the
orthonormal axes ea k,e»
k, ey k,defined in Fig. 2.1b, is calculated for each
snapshot k = 1, ...,N by applying the following procedure:
76 Anisotropic Intramolecular Backbone Dynamics of Ubiquitin
i- (x (x- The axis e
kis given by the normalized Cx _ j Cl vector.
a „ax
The two vectors a =,
'
'—r and t> =
N.C',.! , C'.^C,., C^N,
N.C',.,, c ca CaNdefine
the peptide plane. The influence of the dihedral angle co( is averaged out
due to this definition. The axis enkwhich is orthogonal to the plane is
then given by normalization of the vector resulting from axb.
- Finally, the in-plane principal axis is given by ea k= en
kX ey k.
2. A matrix K of dimension N x N is build where the matrix element Kkl is
a measure for the difference in orientation of the two peptide plane frames of
snapshots k and I: Ku = (e^kx^) +(«j^x^) + (^xe~^{) .
3. Summation over each row of the matrix K yields N row sums, the smallest
of which corresponds to the snapshot with the equilibrium peptide plane.
4. In case of a 3D GAF motion with its Gaussian distribution of peptide plane
fluctuation amplitudes there are many peptide planes which lie very close to
the one with a minimal row sum. This was used to select an equilibrium
peptide plane which is nearly planar (o)( = 180°).
3.4.2.2 Extraction of fluctuation amplitudes from the trajectory
The fluctuation amplitudes of the 3D GAF model can be determined from
the trajectory by two different methods which should be regarded as
complementary. The first one (the analytical "matrix method") has been
derived in Section 2.4.3 and allows a fast extraction of the fluctuation
amplitudes and the orientations of the principal axes for the peptide plane
reorientation. However, it does not allow to check if the distributions of the
fluctuations about the three principal axes are Gaussian or not. This can be
achieved with a much more time-consuming "alignment method" [62]. Here,
the fluctuation amplitudes are calculated for each of the 1500 snapshots by
aligning the instantaneous peptide-plane axes ea k,«» fcand e
k by a
3 4 Molecular dynamics simulation 77
transformation with their corresponding equilibrium directions which were
calculated according to Section 3 4 2 1 The transformation is accomplished
by three successive rotations (l) rotation by the angle a about the axis ea,
(u) rotation by |3 about the axis en, and (in) rotation by y about the axis e
Due to the smallness of each of these rotations, the order of their application
is not crucial The analysis of the resulting distributions of the fluctuations
about all three axes is a main criterion for deciding whether a peptide plane
shows 3D GAF motion or not In case of Gaussian distributions the
fluctuation amplitudes are given as the corresponding standard deviations
However, this method is not appropriate for determmg the orientations of the
reonentation pnncipal axes which reveals the complementary character of
the "matnx" and the "alignment" method For all peptide planes with
dominant 3D GAF motion, the fluctuation amplitudes obtained by the two
methods are in good agreement Conversely, agreement between the two
procedures is a useful indicator for the dominance of the 3D GAF motion of
the peptide plane under consideration
3.4.3 Analysis of MD trajectory
3.4.3.1 Selection of peptide planes with 3D GAF motion
Throughout the whole analysis the peptide planes are labeled and
numbered by the amino acid residue that contnbutes the mtrogen atom The
plane of (a non-prolme) peptide bond i contains the atoms N:, H; ,
"expand O,.!
For each of the 72 non-prohne peptide planes in ubiquitin, the
reonentational probability distnbutions were obtained from the 1500
snapshots of the trajectory by using the "alignment method" desenbed in
Section 3 4 2 2 It is found that the motion of 57 non-prohne peptide planes
involves predominantly 3D Gaussian axial fluctuations In Fig 3 6a results
are shown for the representative plane of He 30 Nearly all peptide planes
located in the a helix or in one of the (3 strands belong to this group The
effective internal correlation times xint of the 3D GAF motion is for most of
the 57 peptide planes well below 20 ps
78 Anisotropic Intramolecular Backbone Dynamics of Ubiquitin
The remaining 15 peptide planes belonging to residues in the loop regions
including Thr 7 to Lys 11, Gly 35, He 36, Gly 47, Gly 53, Arg 54, and the
flexible C-terminus, Arg 72 to Gly 76, show additional jump processes on
slower time scales. This either leads to bimodal angular distributions with
two dominant conformations (see Fig. 3.6b for peptide plane Leu 8) or to
asymmetric distributions due to exchange between a larger number of sites.
They were excluded from further analysis which can be seen in Table 3.1.
Table 3 1 Exclusion of peptide planes whose motion cannot be described by a 3D GAF
model both in experiment and simulation
Reason for exclusion Excluded planesRemaining
planes
Proline residues Pro 19, Pro 37, Pro 38 72
Slow motion apparent in
MD trajectory
Thr 7 to Lys 11, Gly 35, He 36, Gly 47,
Gly 53, Arg 54, Arg 72 to Gly 76
57
NMR relaxation data
not available"
He 13, Asp 21, Glu 24, Ala 28, Gin 31,
Gly 53, Leu 67, Leu 69, Arg 72, Leu 73
50
Slow motion apparent in
relaxation datab
Leu 8, Thr 9, Gly 10, Lys 11, Asn 25,
Asp 52, Gin 62, Arg 74, Gly 75, Gly 76
47
Poor fits by 3D GAF
model
Gin 2, He 23 45
a Excluded peptide planes show either spectral overlap or very weak signals (see
Section 3 5 1)
Excluded peptide planes show either low NOE values or short T2 relaxation times
due to conformational or chemical exchange (see Section 3 5.1)
3.4 Molecular dynamics simulation 79
-40
-40
-105 -65 -25 15
\|/29 Ide9]
a30
40
[cleg]
P30
0 40
[deg]
-40 0 40
[deg]
-80 -40 0
-80 -40 0
125 175 225 275
V|/7 [deg]
-80 -40 0
oc8
40 80
[deg]
40 80
[deg]
Y8
40 80
[deg]
Fig. 3.6 Backbone dihedral angle distributions and peptide-plane reorientational
probability distributions about the axes ea, e=, e in ubiquitin derived from the 1.5 ns
molecular dynamics simulation in water at 300 K. (a) Probability distributions of
He 30, exhibiting a unimodal, nearly Gaussian behavior, (b) Probability distributions
of Leu 8, exhibiting multimodal and asymmetric behavior. Peptide plane motion of
Leu 8 cannot be described solely by a 3D GAF model.
80 Anisotropic Intramolecular Backbone Dynamics of Ubiquitin
3.4.3.2 Results for the extracted fluctuation amplitudes
The directions of the principal axes of the 3D GAF motion and the
fluctuation amplitudes were extracted according to Eq (2.82). Oa, 0g, and
o are plotted in Fig 3.7 for the 57 peptide planes. Numerical values are
given in Table A.l in the appendix. All 57 peptide planes show a significant
degree of intramolecular motional anisotropy with the largest fluctuation
about the axis e.Axial symmetry of the fluctuation ellipsoid with equal
fluctuation amplitudes about axes ea and Co (aa = Op = aa«) is fulfilled in
good approximation for nearly all 57 peptide planes as is visible in Fig. 3.7.
20
16
-o
12
en.
D
S
0
Mlllllllll iiiiiiiii«i
10 60 7020 30 40 50
Peptide-plane number
Fig 3 7 MD reonentational fluctuation amplitudes of the peptide planes of ubiquitin
about the principal axes ea, ea, e as functions of the peptide-plane number The
3D GAF fluctuation amplitudes aa (open circles), Op (stars), and o\, (filled circles)
were exti acted from the 1 5 ns MD trajectory by averaging over 1500 snapshots (see
Section 3 4.2 2). The secondaiy structure is indicated at the top
3.4 Molecular dynamics simulation 81
3.4.3.3 Dihedral angles and planarity of the peptide plane
The degree of non-planarity expressed by the dihedral angle co;, defined by
the atoms C,_ t
- C,_ j
- N: - C, of each peptide bond /, was determined
from the MD trajectory. The average value of co is for all peptide planes near
180° (177°±5°) and the average standard deviation is 6.8° ± 0.8°, which is
largely independent of the amino-acid type and the secondary structure.
Numerical values for average and standard deviation of co; of each peptide
plane are given in Table A.l in the appendix. The (dl fluctuations are in good
approximation statistically independent of the superimposed peptide plane
reorientation. The correlation coefficients for the correlation between the
dihedral angles co( and V,_i»<P,» which are plotted in Fig. A.2b in the
appendix for all residues in ubiquitin, are on average significantly smaller
than 0.5.
20 40 60 80 100"~
500 1000 1500 2000
time [fs] time [fs]
Fig. 3 8 Dihedral angle correlation functions of peptide plane 32 in ubiquitin. (a) Initial
decay of correlation functions for dihedrals a>i2, ty^ ¥31 extracted from a CHARMM
MD simulation of 4 ps length taking snapshots every 2fs. (b) Illustration of the damped
harmonic oscillation which is present in the dihedral angle correlation function
Ca (t) = ((a>32(0)-co32)(cfl32(r)-co32)) The oscillation has a wavelength of32
_|
approximately 200 cm.
Correlation functions of the dihedral angles co(, cp , i|/( _ lwere extracted
from a MD simulation of4 ps length taking snapshots every 2 fs. In Fig. 3.8a,
typical results are shown for peptide plane 32 in the a helix of ubiquitin. Due
0.5
_n r
82 Anisotropic Intramolecular Backbone Dynamics of Ubiquitin
to the shortness of the trajectory only the very fast motional modes about the
dihedral angles are sampled. Longer trajectories would falsify the
comparison since predominantly the correlation function of the dihedrals
<P,> V, _ iwould be affected by slower motion. The correlation times of the
dihedral angle ft)( are, on average, a factor 2 shorter than the ones for the
dihedral angles (p(, \y[ _ 1.The statistical independence and the shorter
correlation times for co; suggest, that an appropriate reference frame for the
3D GAF amplitudes would be the effective peptide plane with an averaged
co( angle. In Fig. 3.8b, it is illustrated that a harmonic oscillation about the
dihedral angle (Hl is still visible in the correlation function Cw (t) since me
coupling to the environment is weaker than in case of the other dihedrals. The
extracted wavelength ofapproximately 200 cm is in good agreement with
results from optical spectroscopy [118].
3.4.3.4 Orientation of the peptide plane frame
The orientations of the principal axes for the peptide-plane reorientation
were extracted for each peptide plane with the "matrix method" according to^ ^ x
Section 2.4.3. The calculated principal axes ea ,ea
pp,e differ
slightly from the principal axes ea, ea, e of the 3D GAF model defined in
Fig. 2.1b. The orientations of the principal axes e with respect to the
peptide plane frame ea, en, e are depicted in Fig. 3.9 for all 57 peptide
planes which show 3D GAF motion. On average, the e axis lies—^ (X ot
approximately in the plane and shows an angle to the e axis (C, _ j- C,
axis) of about 7°. Consequences of this small but systematic offset will be
discussed in Section 3.5.5.
3 5 Analysis of experimental relaxation data 83
Fig 3 9 Polar coordinate plot of
the orientations of the
calculated principal axes eywith respect to the principal axes
ew erj, ey, defined for the 3D
GAF model in Fig. 2.1b. The
orientations were extracted
according to Section 2 4.3 for all
57 peptide planes, which show a
3D GAF motion.
3.5 Analysis of experimental relaxation data
3.5.1 Raw data analysis
The experimental relaxation data of 62 peptide planes were analyzed. No
relaxation data were extracted for the remaining 13 peptide bonds belonging
(l) to Pro 19, Pro 37, and Pro 38 due to the absence of NH protons, (ii) to
Glu 24 and Gly 53 exhibiting peaks with low sensitivity caused by line
broadening, and (iii) to He 13, Asp 21, Ala 28, Gin 31, Leu 67, Leu 69,
Arg72, and Leu 73 due to cross-peak overlaps in the ^N-'H HSQC
spectrum (see Table 3.1). The relaxation data for the two B0-field strengths
are shown in Fig. 3.10 and Fig. 3.11 as functions of the peptide-plane
number. Numerical values for the experimental N Ti, T2, and NOE, and
13C Tt and T2 values at 400 and 600 MHz are given in Table A.2 in the
appendix. Comparison of repeated relaxation measurements yields the
following estimates for the statistical uncertainties: 1.5% for 15N Tj's at 400
and 600 MHz, 2% for 15N T2's at 600 MHz, 4% for 15N NOE's at 400 MHz,
2.5% for 15N NOE's at 600 MHz, 2.5% for 13C Tfs at 400 MHz, 2% for
13C Tys at 600 MHz, 5% for 13C T2's at 400 MHz, and 3% for 13C T2's
84 Anisotropic Intramolecular Backbone Dynamics of Ubiquitin
at 600 MHz
The peptide planes Leu 8, Thr 9, Gly 10, Lys 11, and Gin 62, which belong
to loop regions, and the peptide planes Arg 74, Gly 75, and Gly 76 forming
the C-terminus of ubiquitin show 15N NOE values below 0 52 at 400 MHz
and below 0 68 at 600 MHz, which are significantly smaller than those of
other peptide planes, which are in the range between 0 52 and 0 67 at
400 MHz and between 0 68 and 0 79 at 600 MHz This reflects the presence
of additional large-amplitude internal motions at slower time scales which
cannot be modeled solely by a 3D GAF motion of their peptide planes These
eight peptide planes were excluded from further analysis (see Table 3 1)
Peptide planes involved in conformational exchange processes were also
excluded (Table 3 1) These comprise peptide planes 25 and 52 which show
significantly reduced 13C T2 values (see Fig 3 10) The same effect is also
apparent when comparing the 15N T2 values at 600 MHz with ^N^H dipole-
15N CSA cross-correlation data obtained from CT-HSQC experiments [71]
(see Section 2 3), which is shown in Fig 3 11 The absence of a similar
anomaly for the peptide plane Asn 25 in 1 /TN NHindicates that the
anomaly in 5N T2 is caused by conformational exchange This leaves a
remainder of 47 peptide planes whose relaxation-active small-amplitude
motion can be modeled by a 3D GAF motion
3.5 Analysis of expenmental relaxation data 85
0.80
0.75
0.70
111
O 0.65
JF 0.60
5 10 15 20 25 30 35 40 45 50 55 60 65 70 75
Peptide plane number
{1H}-15N NOE
5 10 15 20 25 30 35 40 45 50 55 60 65 70 75
Peptide plane number
Fig. 3.10 First part.
86 Anisotropic Intramolecular Backbone Dynamics of Ubiquitin
10 15 20 25 30 35 40 45 50 55 60 65 70 75
Peptide plane number
10 15 20 25 30 35 40 45 50 55 60 65 70 75
Peptide plane number
Fig 3 10 Backbone (a) l5N T,, (b) {'H}-15N NOE, (c) 13C Th and (d) 13C T2
relaxation data of ubiquitin as functions of the backbone peptide-plane number
Relaxation data were measured at two B0-field strengths corresponding to the proton
resonance frequencies 400 MHz (open circles) and 600 MHz (filled circles) at 300 K
3 5 Analysis of experimental relaxation data 87
0 25
^ 020<N
I-
zin
015
0105 10 15 20 25 30 35 40 45 50 55 60 65 70 75
Peptide plane number
Fig 3 11 Significant conformational exchange contnbutions to 15N T2 measured at
600 MHz (filled circles) can be identified by comparison with the inverse ^N-'H
dipole-CSA cross correlation rate constants FN m at 600 MHz proton frequency (see
Section 2 3) that are also given in the figure (stars)
3.5.2 Spin relaxation mechanisms in the peptide plane
The dominant relaxation-active interactions and their principal axis
onentations with respect to the peptide-plane frame are depicted m Fig 2 3
Since the interactions probe different directions, their relaxation
contributions yield complementary information on the ngid-body motion of
the peptide plane The relative onentations of the interactions relevant for
15N and C relaxation and the relative magnitude of their contnbutions are
given in Table 3 2
With regard to 15N relaxation, the standard model-free analysis considers
only the dipolar ^N-1!! and the 15N CSA relaxation and descnbes the2
intramolecular motion by the order parameter SNH One finds, however, that
also the 15N-13C and 15N-13Ca dipolar relaxation lead to measurable
88 Anisotropic Intramolecular Backbone Dynamics of Ubiquitin
contributions (see upper part of Table 3 2) They can ngorously be taken into
account within the more comprehensive 3D GAF model
For the backbone 13C spins, CSA relaxation is dominant but dipolar
contributions have also to be considered [90] Even at 600 MHz proton
resonance the dipolar contributions are non-negligible, as is shown in the
lower part of Table 3 2 The largest dipolar contributions to 13C relaxation
originate from the directly bonded Ca and from close protein protons
While the motion of the in-plane dipolar 13C'-13Ca and 13C'-1HN
interactions can be descnbed by the 3D GAF model, the internuclear out-of-
plane 13C- 1Ha and 13C- 'Fr vectors are influenced also by other types of
motion The influence of all protons, except H of the same peptide plane,
was accounted for by an isotropic dipolar leakage term, which can be
expressed [90] as the dipolar relaxation contribution of a virtual proton at an
effective distance r - This effective distance was calculated for all 15N and
13C nuclei in ubiquitin from the energy-minimized X-ray structure The rg^values for 15N range from 1 72 to 1 87 A and for 13C from 1 69 to 1 90 A
The combination of all the relaxation-active interactions summarized in
Table 3 2 leads to the dependence of the overall relaxation rate constants Tlof the 15N and 13C backbone spins on the fluctuation amplitudes aag and
a depicted in Fig 3 12 Both 15N and 13C T{ show a larger dependence
on cag than on ay In case of N (see Fig 3 12a), the dependencies on can
and ay differ only slightly since the dominant dipolar ^N-'H interaction is
almost not affected by the motion about the ea axis On the other hand, in
case of C (see Fig 3 12b) the dependence on a is 3 to 4 times weaker
than the dependence on Gag This can be explained by the onentation of the
13C CSA pnncipal axes (see Fig 2 lb and Fig 2 3) which are nearly
colhnear to the principal axes ea, ea, e of the peptide plane frame Motion
about ea fully affects the modulation of the large CSA principal values oxx
and Gzz, and leads to the largest dependence of the C Tl on aa whereas
motion about the axes en and e both affect the smallest pnncipal value a
beside axx and Gzz, respectively
3 5 Analysis of experimental relaxation data 89
Table 3 2 Relevant parameters for dipolar and CSA interaction strengths and their
contributions to relaxation of the backbone 15N and 13C spins assuming isotropic overall
tumbling with a correlation time tc = 4 03 ns and intramolecular 3D GAF motion of the
peptide plane
Relaxation-
active
interaction
Geometrical
and CSA
parameters
(e,<p)'
[deg]
Contnbutions
tor;1at 400 MHz
Contnbutions
tor;1at 600 MHz
S2 '
"N-'H 102AC (101 3, 180) 81 l%g 70 4% s 081
15N.13C, 135 A (003)" (138 4, 0) 2 2% * 18%* 0 86
15xj 13r>a 1 47 A (0 03)d (14 4,0) 14%* 1 1%* 0 91
15NJHrest 1 80 A (003)' 3 3%« 2 9%« (1)'
15N CSA "
axx
°yy
°zz
58 3 ppm
-513 ppm
109 6 ppm
(33 3, 180)
(90, 90)
(123 3,180)
12 0% * 23 8% *
0 88
080*
0 83
13c-13c<x_, 152 A (003)d (159 6,180) 12 4% * 118%* 0 90
13C. 13c« imkd (12 3,180) 0 6%* 0 6%* 091
13C. 15N 1 35 A (0 03)d (138 4, 0) 5 5%* 3 5%* 0 86
13C,.1HN 2 06A*' (66 9,180) 8 0%* 4 6%* 0 82
13p> lxrrest 1 82 A (0 03)' 213%* 118%* (1)'
13C'CSA*
°xx
ay>
-74 4 ppm
-7 4 ppm
81 8 ppm
(2 3, 0)
(92 3, 0)
(90, 90)
52 2% * 67 7% *
091
0 80'
0 80
"The pnncipal values and pnncipal axis onentations of the N CSA tensor were
taken from ref [119] The angle between the CSA z axis and the ^N-'H bond
vector is 22° (see also Fig 2 3)bThe pnncipal values and pnncipal axis onentations of the C CSA were taken
from ref [120] The angle between the CSA y axis and the 13C=0 bond is 13°
(see also Fig 2 3)
90 Anisotropic Intramolecular Backbone Dynamics of Ubiquitin
cThe l5N-lK bond length was set to the standard value of 1 02 AdThe average bond lengths and their standard deviations given in brackets were
determined from the MD trajectory (see Section 3 5 3) as well as the six bond
angles defining the peptide plane 120 3° (NHNC), 115 7° (NCaHN), 124 0°
(NC'Ca), 120 9° (CON), 118 0° (C'NCa), 121 1° (C'CaO)eA virtual distance between the 15N and C nuclei and all protons outside of the
peptide plane was calculated for all peptide planes using the energy-minimized
X-ray structure [96] The average value with standard deviation is given
f 6, tp are the polar angles of the corresponding spin interaction in the peptide
plane frame defined in Pig 2 3
s The contnbutions are given as percentage of the 15N overall relaxation rate
constants 1/7, = 3 21 s1 (400 MHz) and 2 22 s1(600 MHz) assuming a 3D
GAF motion with ca^ = 7° and cy = 14°, Tc = 4 03ns, and xm, = 2 ps
The contnbutions are given as percentage of the C overall relaxation rate
constants 1/7, = 0 88 s'(400 MHz) and 0 73 s
'(600 MHz) assuming a 3D
GAF motion with oap = 7° and c = 14°, xe = 4 03 ns, and \mt = 2 ps
' The order parameter S^ has been computed for oap = 7° and o = 14°
according to Eq (2 57)
1 A static approximation (S =1) was assumed to calculate the contribution of the
protons outside of the peptide plane since these interactions cannot be modeled
by a 3D GAF motion
The order parameter for the cross-correlation function between the y and
2z principal axes is Su= -0 38
'The order parameter for the cross-correlation function between y and z principal
axes is S = -0 35
3.5 Analysis of experimental relaxation data 91
10 15
aap, ay
20 25
[deg]20 25
[deg]
10 15
aap, ay
Fig 3 12 Dependence of (a) 15N 7", and (b) 13C T\ on the fluctuation amplitudes
aaa (solid line, a is set to ay = 0°) and cy (dashed line, aaa is set to aan = 0°).
The parameters for the interaction strengths were taken from Table 3 2. The correlation
times for the overall tumbling and the internal motion were set to xc = 4.03 ns and
xmr= 20 ps, respectively
3.5.3 Which parameters can be extracted from the
experimental data?
In principle, it is conceivable to extract all motional parameters and all
parameters defining the interaction strengths of each peptide plane in the
protein by fitting a sufficiently large number of relaxation measurements.
However, the presently available measurements do not permit such a general
approach and the extraction of the motional parameters for each peptide
plane becomes the primary goal. Several problems arise concerning the
interaction strengths. The geometry of the peptide plane has to be extracted
from results by different techniques (as X-ray, neutron scattering, and
quantum chemical calculations). The CSA tensors have to be taken from
solid-state NMR measurements of some model compounds, and it is not clear
whether these results can be transferred to proteins. In addition, it has to be
assumed that the interaction strengths do not vary significantly between
different peptide planes. In doing so, a "picture" is created where differences
between residues only show up in the motional parameters. Significant
residue-wise variation of the interaction strengths is possible and would
deteriorate the fit results for the motional parameters (see Section 3.5.5). In
case of a peptide-plane study the most critical parameters are the 'H-^N
92 Anisotropic Intramolecular Backbone Dynamics of Ubiquitin
dipolar interaction strength and the N and C CSA tensors. It was tried to
fit scaling factors for all of these interaction strengths but this was only
possible in case of the 13C CSA tensor (see below).
Motional parameters. The experimental data of the considered 47 peptide
planes were evaluated to determine their 3D GAF fluctuation amplitudes.
Based on the MD analysis that showed nearly axially symmetric fluctuation
ellipsoids for almost all peptide planes, it was assumed throughout the
2
analysis that oa = o» = aag. The order parameters 5 of Eq. (2.57) are
then determined for each peptide plane by a pair of aag, a values.
Additional parameters entering the spectral densities (Eq. (2.45)) are the
overall tumbling correlation time xc and the correlation times xwt. For
simplicity, an isotropic internal correlation time is assumed, i.e. in a given
peptide plane xint = xint for all pairs \i, v = a, (3, y. This leads to a
description of the motion of each peptide plane by the four model parameters
aap, a , xc, xint which can be extracted from the experimental relaxation
data.
For each peptide plane i, an overall tumbling correlation time xc ,was
extracted from the 15N data alone by a standard model-free analysis, fitting
the parameters xc ,, xint ,,and 5r to the experimental 15N Tl, NOE (400
and 600 MHz), and T2 (600 MHz) data. Averaging over all 47 residues (see
above) yields xc = 4.03 ns, which is well comparable with a previously
determined value of 4.09 ns obtained under similar experimental conditions
[106].
Parameters defining dipolar interaction strengths. The average bond
lengths and bond angles were extracted for a number of representative
peptide planes from the MD trajectory by averaging over the 1500 snapshots.
The results are given in Table 3.2. The standard deviations for the bond
lengths and bond angles for the ensemble of representative peptide planes do
not exceed 3%. The N-H distance was set to the standard value of
rNH = 1^2 A which has often been used in previous analyses [105,106]. It
is significantly higher than the value obtained from the MD trajectory:
(rNH) = 0.997 A. Note that reNH is an effective distance that should reflect
3.5 Analysis of experimental relaxation data 93
the averaging by rapid local stretching and bending motion of the N - H
bond, rNH = {Sbend(r'^H)Uretch} , including also zero-pointeff
vibrational effects (see Section 2.2.3.2). The value for rNH might need to be
modified in the future when more precise information from solid-state NMR
and from quantum chemical calculations is available.
Parameters defining CSA interaction strengths. It is possible that the
CSA tensors vary with the amino-acid type, with the presence of hydrogen
bonds, and with the local backbone \|/( _ jand (p( dihedral angles. Since there
are no experimental CSA values of ubiquitin available, the CSA tensors were
taken from solid-state NMR studies of small peptide fragments: The 15N
CSA tensor of Boc-Gly-Gly-[15N]Gly-OBz was determined by Hiyama et al.
[119] and the 13C CSA tensor of [l-13C]Glycyl-Gly-HCl by Stark et al.
[120]. The selected 13C CSA tensor is similar to other experimental tensors
[121-124], while for 15N CSA tensors larger variations are observed [124-
128]. It should be noted that the CSA tensors determined by solid-state NMR
correspond to tensors that are partially averaged due to intramolecular
motion. Considering this fact, 15N and 13C CSA scaling factors XN and Xcare introduced, which isotropically upscale the experimental CSA tensors.
The exact knowledge of the N CSA tensor is less crucial than the one of the
13C tensor, since at field strengths corresponding to 400 and 600 MHz
proton resonance 15N relaxation is dominated by dipolar relaxation. The
experimental value for the I5N CSA tensor of Ac = ah - o± = 164 ppm
[119] (assuming an axially symmetric tensor) was upscaled by a fixed value
of XN = 1.07 to compensate for the motional averaging in the solid-state
NMR study. This corresponds to a rigid-molecule value of Aa = 176 ppm
which is used in the following for all peptide planes. Similar values have
been applied in other NMR studies [107,129].
If the c CSA tensor is regarded as known and fast internal motion,
xmt < 20 ps, is assumed, it is possible to extract the fluctuation amplitudes
Gaa, a from N T{ and C T{ relaxation data at one magnetic field
strength alone. This is illustrated in Fig. 3.13. For fast internal motion, a
certain relaxation rate constant 15N Tx corresponds to an overall order
94 Anisotropic Intramolecular Backbone Dynamics of Ubiquitin
parameter S2NH which is almost equal (see also Section 3.5.6) to the order
parameter SNH Nfifor the dominant dipolar interaction. A certain order
parameter SNH NHcan be described in terms of combinations of the 3D
GAF fluctuation amplitudes aan,a according to Eq. (2.57). This is
2illustrated in Fig. 3.13a for different order parameters SNH NH ranging from
0.6 to 1.0. The a
ellipsoid:
ap,a combinations define, to a good approximation, an
A sinp, a = B cosp. (3.1)
The parameters A and B define the form of the ellipsoid and the parameter
p represents an angle between 0 and n/2. In Fig. 3.13b, the 13C relaxation
rate constant Tx is calculated for combinations (aan,a ) defined by the
ellipsoid which corresponds to SNH NH= 0.80. In contrary to the
corresponding constant 15N Tj ,the calculated 13C Tx cover a large range
from 0.86s"1 (for p = 0°) to 0.69s"1 (for p = 90°). The very distinct
dependence of 15N and 13C T\ on the CTap,o combinations enables a
fitting procedure with a unique solution for the fluctuation amplitudes.
a) b) 0 90
25^3NH, NH0 60^
D) 20
-0 65
-0 70
JO, -0 75
«i15-0 80
ea U0 85
10-0 90
5
"0 95
10 15 20 25 300 65,
10 20 30 40 50 60 70 80 90
oY [deg] P [deg]2
Fig. 3.13 (a) Description of the order parameter SNH NHin terms of the 3D GAF
fluctuation amplitudes aa a and a according to Eq (2 57) Different values of the
order parameter, for which the solutions (Oa n,ay) are given, are indicated (b)
Dependence of relaxation rate constant nC T\ on the parameter p which defines the
2
position on the ellipsoid which corresponds to SNH NH= 0 8 in Fig 3.13a (bold
line). The ellipsoid of the (aaa,ay) solutions is defined according to Eq (3 1) with
A = 15 5° and B = 16 5° Xc is set to 1 083 (see Section 3 5 4)
3 5 Analysis of experimental relaxation data 95
Relaxation data at multiple B0 fields enable a separation of the CSA and
the dipolar contributions based on their distinct B0-field dependence which
can be used to extract additional information about the CSA tensor from the
relaxation data. First, it was tried to individually scale the different C CSA
tensors by scaling factors Xc (for each peptide plane i. This would allow
one to partially account for the structural dependence of the CSA tensors.
Although a sufficient number of relaxation measurements was available, the
fitting procedure turned ou. to be unstable due to a strong correlation of Xc t
with the motional parameters o"ag, a .On the other hand, a stable fitting
procedure (see Section 3.5.5) results when fitting a global scaling factor Xc,which uniformly scales the assumed C CSA tensor for all peptide planes,
together with the motional parameters.
3.5.4 Fit results for 3D GAF model
The experimental data of the 47 peptide planes considered were evaluated
to determine their 3D GAF fluctuation amplitudes. The fluctuation
amplitudes oaB, <3 and the internal correlation time xmt were fitted to the
nine experimental parameters 15N Tj, T2, and NOE and 13C Tl5 and T2 at
400 and 600 MHz for each of the 47 peptide planes. The peptide planes of
Gin 2 at the N terminal of ubiquitin and He 23 were excluded (see Table 3.1)
due to a large least-square fit error (four times larger than the average one).
The relaxation data of Gin 2 reflect increased flexibility of the N terminus
which cannot be described by 3D GAF motion. The scaling factor Xc for the
13C CSA tensor was determined as a global fit parameter for the remaining
45 peptide planes. A value Xc = 1.083 ± 0.004 was obtained that reflects
the motional averaging of the CSA tensor in the solid-state measurement
[120], since Xc> 1.0. The overall tumbling correlation time xc was kept
fixed at 4.03 ns. The fit results for the fluctuation amplitudes are shown in
Fig. 3.14 as functions of the peptide-plane number. Numerical values for all
fit parameters are given in Table A.3 in the appendix.
96 Anisotropic Intramolecular Backbone Dynamics of Ubiquitin
fflllllllfflilfflll
20 30 40 50
Peptide-plane number
Fig 3 14 Fit of the 3D GAF model using Eqs (2 45) and (2 57) to the experimental
data consisting of nine auto relaxation parameters (see Fig 3 10 and Fig 3 11) for the
45 peptide planes which can be described by a 3D GAF motion The optimized
parameter set consists of a global scaling factor Xc for the principal values of the C
CSA tensor and three parameters o"ag, o ,and imt for each peptide plane The
overall correlation time T was set to 4 03 ns and the principal values of the N CSA
tensor, given in Table 3 2, were upscaled with XN = 1 07 (see Section 3 5 3) The
optimum scaling factor is Xc = 1 083 The optimum values o"ag (open circles) and
ov (filled circles) are given in the figure for each peptide plane The error limits of the
fitted parameters were determined by a Monte Carlo procedure consisting of 60 fits
with random Gaussian errors added to the relaxation parameters according to the
experimental standard deviations The secondary structure elements are indicated at
the top
Nearly all of the 45 peptide planes exhibit 3D GAF motion with a
substantial degree of anisotropy with ay > aan Thus the dominant axial
fluctuation of the peptide planes takes place about the e axis connecting
-l-land Cj On average, the experimentally determined fluctuation
amplitudes, given in Fig 3 14, are about a factor 1 4 (oan) and a factor 1 6
3 5 Analysis of experimental relaxation data 97
(o ) larger than the ones observed in the MD trajectory (see Fig. 3.7). As can
be seen in Fig. 3.15, the internal correlation time xint converges for all
peptide planes to small values xint < 30 ps, consistent with the MD results,
rendering the relaxation data largely insensitive to xr
35
30 o
25
•
•
Xmt 20
tps]15 • •
•
••
•
•
10
5
n
••
•
• •
<
.•• •
•
• •
•
• •• •
•
•
• • •
10 7020 30 40 50 60
Peptide plane number
Fig 3.15 Results for the fitted internal correlation times xmt of the 45 peptide planes
which can be described by a 3D GAF motion The two representative error bars of
peptide planes 12 and 30 were determined by a Monte Carlo procedure consisting of
60 fits with random Gaussian errors added to the relaxation parameters with the
experimentally determined standard deviations
The fitted fluctuation amplitudes were correlated with the experimental
raw data. All N relaxation parameters are highly correlated with ay
(correlation coefficient R ranges from 0.5 to 0.65) but show no significant
correlation with aag (/?<0.1). On the other hand, the 13C relaxation
parameters are highly correlated with o~ag (R ranges from 0.5 to 0.8) and less
correlated with ay {R ranges from 0.2 to 0.5). This behavior reflects the
strong dependence of the C relaxation parameters on aan as illustrated in
Fig. 3.12b.
3.5.5 Uncertainty estimates of fit parameters
Experimental random errors. To estimate the influence of experimental
random errors on the fitted fluctuation amplitudes oao and ay, a Monte
Carlo error analysis was performed. In a series of 60 runs, random Gaussian
errors with the experimentally determined standard deviations (see Section
98 Anisotropic Intramolecular Backbone Dynamics of Ubiquitin
3 5 1) were added to the experimental relaxation data For each of the 60 data
sets, a simultaneous fit of the individual fluctuation amplitudes of the 45
peptide planes and of a global scaling factor Xc for the 13C CSA tensors was
performed The resulting errors of the fluctuation amplitudes are indicated as
error bars in Fig 3 14, the numerical values are given in Table A 3 in the
appendix
Dependence of 13C'-CSA tensors on secondary structure. To further
investigate the structural dependence of the 13C CSA tensor, the 45 peptide
planes are divided into the three categories "oc-hehx", "p-sheet", and "other"
The individual scaling factors were fitted for each category
jhdix= t 076±0008) x<geet= i 070±0010, X°cher= 1 096±0010 The
resulting fluctuation amplitudes of the 45 peptide planes are shown in
Fig A 3 in the appendix The similarity of the three scaling factors seems to
confirm that the 13C CSA tensors of ubiquitin vary only slightly This result
is supported by the density functional theory (DFT) calculations in Chapter 4
indicating the limited variation of the anisotropy of the C CSA located in
the a helix and in a (3 strand
Effect of interaction strengths on fit results. In Fig 3 16a, the
dependence of the fitted fluctuation amplitudes caa and o on the global
scaling factor Xc is illustrated for the peptide plane He 30 The figure shows
that Xc has a strong influence on the motional anisotropy of the peptide
plane (for definition see Eq (3 2)) Xc is uniquely defined as can be seen
from the error function of the fit The interplay between interaction strengths
and intramolecular motion is illustrated in Fig 3 16b for fit results of peptide
plane 30 A global multiplier |i which uniformly scales all dipolar and CSA
interaction strengths of the N and C spins is vaned between 0 96 and
1 09 (ft = 10 corresponds to the interaction strengths which were used in
Section 3 5 4) Increasing interaction strengths are compensated by an
increasing amount of intramolecular motion However, the very distinct
dependencies of the C relaxation rate constants on oag, a (see
Fig 3 12b) lead to a non-linear dependence of aan on |j. This prevents the
introduction of reduced variables oao = oao /u. In Fig 3 16c, the NH
3 5 Analysis of experimental relaxation data 99
bond length rNH is varied between 1.00 and 1.04 A (rj{/H = 1-02 A was used
in Section 3.5.4). An increasing NH bond length leads to a strictly decreasing
fluctuation amplitude a and a moderately increasing fluctuation amplitude
cao. As the 15N relaxation rate constants show a similar dependence on
°afl' °v (see Fig- 3.12a), the overall effect compensates the increasing rNH.
On the other hand, there is no overall effect on the C relaxation rate
constants due to the dominant dependence on oao (see Fig. 3.12b).
a)
2,
*"^»^fc.-
-* cy-
13*10 °u^__
~"
CO.
aO
:
5 x\-
610
590
'570 5C
550
1 04 1.06 1 08 1 10 1.12
0 96 1 00
H
1 04 1 08 1 00 1 01 1 02
rNH [A]1 03 1 04
Fig 3.16 Dependence of the fit parameters aag (solid line) and oy (dashed line) of
peptide plane He 30 on (a) The global scaling factor Xc for the 13C CSA tensor The2
corresponding overall fitting error function % (sum of the error contributions of all 45
peptide planes) for the global fit of Xc is given as dotted line, (b) The global multiplier
u which scales all dipolar and CSA interaction strengths of the 15N and 13C
backbone spins (c) The NH bond length The scaling factor \g was kept fixed at
1 083 in (b) and (c)
100 Anisotropic Intramolecular Backbone Dynamics of Ubiquitm
Offset of peptide plane frame from assumptions in the 3D GAF model.
In Section 3.4.3.4, the offset of the principal axis frame for peptide plane
reorientation from the assumptions in the 3D GAF model was extracted from
the MD trajectory. Test calculations which estimate the dependence of the
fitted fluctuation amplitudes on such an offset are summarized in Table 3.3.
First, a relaxation data set was calculated assuming fluctuation amplitudes of
aan = 7° and o = 14° for reorientation about the principal axis frame
.For an axially symmetric 3D GAF fluctuation ellipsoid,
'a, pp'e^PP'' y,pp
the offset of this principal axis frame from assumptions in the 3D GAF model
is uniquely defined by the polar coordinates of the assumed symmetry axisi —^ —x _j^
e with respect to the peptide plane frame ea, en, e (see Table 3.3).
Table 3 3 Estimation of the systematic error in the fitted fluctuation amplitudes aao, aywhich results from an offset of the fluctuation tensor orientation from assumptions in the
3D GAF model For details of the calculation procedure see text
orientation of e
(9, cp) [deg]a
fitted aap [deg]*
fitted ay [deg]*
(0,0) 70 14 0
(10, 0) 75 130
(10,45) 73 134
(10,90) 70 14 0
(10, 135) 70 14 3
(10, 180) 71 14 3
a Polar coordinates of assumed principal axis e with respect to the equilibrium
peptide plane frame ew ea, e
In the fitting procedure isotropic overall tumbling with a correlation time
xc = 4 03 ns, and 15N and 13C CSA scaling factors XN = 1.07 and
Xc = 1 083 are assumed
In the next step, fluctuation amplitudes were fitted to this relaxation data
set assuming the principal axis frame ea, en, e with onentations defined in
the 3D GAF model. Comparison of the fitted fluctuation amplitudes with
3.5 Analysis of experimental relaxation data 101
°aB = 7° anc^ Gy = ^° yi^^s the magnitude of the systematic error due
to the offset of the principal axis frame. The largest effect on o is observed
if e lies in the plane and is slightly oriented towards ea (corresponds to
0 > 0°, cp = 0°). The resulting systematic error might be relevant since this
case corresponds approximately to the average orientation of e extracted
from the MD trajectory (see Fig. 3.9). A slight modification of the orientation
of the principal axes ea, en, e in the 3D GAF model might be appropriate if
MD simulations of other proteins confirm the results which were found for
ubiquitin (see Fig. 3.9).
Anisotropic overall tumbling. In the present analysis it was assumed that
ubiquitin undergoes isotropic overall tumbling (see Section 3.5.4). The
tumbling anisotropy of ubiquitin has been determined by Tjandra et al. [106]
to be = 18%. An analysis of the influence of an 20% anisotropy of tumbling
on the fitted 3D GAF fluctuation amplitudes is summarized in Fig. 3.17.
First, a synthetic relaxation data set is calculated assuming an axially
symmetric rotational tumbling tensor with the correlation times
xc xx= xc = 4.29 ns and xc zz= 3.58 ns for the tumbling about the three
principal axes (please mind that (xc xx+ xc> + xc zz)/3 = 4.03 ns) and the
3D GAF fluctuation amplitudes aa(3 = 7° and ay = 14° for the
intramolecular motion. The relaxation data set is then fitted as described in
Section 3.5.4 assuming isotropic tumbling with the correlation time
xc= 4.03 ns. The fitted fluctuation amplitudes can be compared with the
values aan = 7° and c = 14° to estimate the systematic error connected
with the assumption of isotropic overall tumbling. In Fig. 3.17, it is shown
that an effect of less than 1.2° on the fitted fluctuation amplitudes aaa and
ay occurs depending on the orientation of the overall tumbling diffusion
tensor with respect to the equilibrium peptide plane where the orientation is
defined by the polar coordinates (9, (p) of the symmetry axis of the rotational
diffusion tensor in the equilibrium peptide plane frame. The maximal effect
corresponds to a symmetry axis of the rotational diffusion tensor that is
perpendicular to the average peptide plane (corresponds to
(6, cp) = (90°, 90°) for the solid line in Fig. 3.17).
102 Anisotropic Intramolecular Backbone Dynamics of Ubiquitin
aap, ay[deg]
Fig 3 17 Dependence of the fitted fluctuation amplitudes 0"ag, Gy on the orientation
of the symmetry axis of the anisotropic (but axially symmetric) rotational diffusion
tensor with respect to the equilibrium peptide plane frame The orientation is given by
the polar coordinate 0 which is varied between 0° and 180° while (p is kept fixed at
0° (dashed line) and 90° (solid line) For details of the calculation procedure see text
3.5.6 Comparison between the 3D GAF analysis and a 15N
model-free analysis
Based on the results of the 3D GAF analysis and by using Eq. (2.57), it is2
possible to calculate order parameters S for arbitrary pairs of spin
interactions within the peptide plane. In Table 3.2, the order parameters for
different interactions are given for the following values of the fluctuation
amplitudes: aag = 7,a = 14
.The calculated order parameters are
largest for the dipolar i5N-13Ca and 13C'-13Ca interactions which are lessot ex
modulated by the dominant fluctuation about the Ct _,- Ct axis than, for
example, the 15N-]H interaction.
The order parameters SNH NHof the dipolar N- H interaction were
calculated for all 45 evaluated peptide planes using Eq. (2.57) together with
the fitted 3D GAF fluctuation amplitudes from Fig. 3.14. In Fig. 3.18 these
values are compared with order parameters obtained directly from a standard
model-free analysis of the 15N Tt, T2, and NOE data at both field strengths.
3 5 Analysis of experimental relaxation data 103
These order parameters are termed here SNH. Fig. 3.18 shows a good overall
agreement between the order parameters SNH and SNH NHcalculated by the
two approaches. It illustrates the compatibility of the 3D GAF analysis with
the model-free description. The systematic offset between the two sets is due
to the assumption in the model-free analysis of an equal order parameter
SNH for the dipolar NH and for the 15N CSA interactions. This is, however,
only justified for isotropic internal motion or for an axially symmetric CSA
tensor with the unique axis collinear to the NH vector. These assumptions are
not realistic (see Table 3.2) and are not required for the 3D GAF analysis.
0.85
20 30 40 50 60
Peptide-plane number
Fig. 3 18 Compatibility of 3D GAF model with a standard 15N model-free analysis
The SNH NHorder parameters for the dipolar 15N-'H interaction derived from the 3D
GAF model data of Fig 3 14 and Eq (2.57) (filled circles) are compared with the 15N
SNH order parameters determined by a model-free analysis from 15N relaxation data
at 400 and 600 MHz, (open circles). xc was set to 4.03 ns and a symmetric 15N CSA
tensor with Aa=176ppm was assumed Additional relaxation contributions from
other protons (assuming a rigid molecular frame), from the 15N-13C dipolar
interaction (assuming an averaged order parameter SNC = 0 83), and from the 15N-
13Ca dipolar interaction (assuming an averaged order parameter SNC = 0 89) were
also included in the model-free analysis
104 Anisotropic Intramolecular Backbone Dynamics of Ubiquitin
A model-free analysis based exclusively on N relaxation data neither
yields axial fluctuation amplitudes nor information on the amsotropy of
intramolecular peptide-plane motion Such additional information is
provided by the complementary C relaxation data that form an integral
part of the 3D GAF analysis
3.6 Discussion
3.6.1 Anisotropy of peptide-plane dynamics
The 3D GAF fluctuation amplitudes extracted from the MD analysis and
from the relaxation data yield insight into the degree of anisotropy of the
peptide plane motions A convement measure for the degree of anisotropy is
the "anisotropy factor"
Ac=
ay-°ap(3 2)
c (rjY + 2aap)/3'
reflecting the difference between reonentational motion about the
C, _ j- Cj axis and motion about the orthogonal axes, weighted by the
average fluctuation amplitude The anisotropy factors for all peptide planes
obtained from the MD trajectory and from the expenmental results are shown
in Fig 3 19 Numencal values are given in Table A 3 in the appendix
It shall be mentioned at this place that there are qualitative and quantitative
differences between the results of the expenmental NMR relaxation study
and the MD simulation The quantitative differences in the fluctuation
amplitudes become apparent by companng Fig 3 7 and Fig 3 14 The
expenmental fluctuation amplitudes are on average larger than in the MD
simulation Reasons for this discrepancies may be suspected in the choice of
not accurately known geometnc parameters in the data evaluation On the
other hand it is also conceivable that the used CHARMM force field is
slightly too stiff, leading to low fluctuation amplitudes
The sets of peptide planes showing large-amplitude, non-3D GAF motion
apparent in the expenmental data and the MD simulation are not identical
3 6 Discussion 105
This particularly concerns the peptide planes of Asn 25 and Glu 62 showing
non-3D GAF motion in the experimental data, but 3D GAF motion in the
simulation The peptide planes of Thr 7, Gly 35, He 36, Gly 47, and Arg 54
exhibit non-3D GAF motion only in the simulation Due to the restncted
length of the trajectory, there is some uncertainty concerning the statistical
significance of jump processes that lead in Section 3 4 3 to the exclusion of
peptide planes from the 3D GAF analysis There is also little correlation
between the residue-specific 3D GAF fluctuation amplitudes in the
expenmental and the MD data In this context, one should remember that
even a rather long MD trajectory, covering more than 1 ns, represents in
essence an "extended snapshot" of the long-term motion of the protein and
does not comprehensively cover the ensemble conformations relevant for the
expenmental results For this reason, no agreement in all details can be
expected, even for a "perfect" force field
106 Anisotropic Intramolecular Backbone Dynamics of Ubiquitin
1.5
1.0
Aa— 0.5o
0.0
-0.5
b) 3.0
2.5
• «w«
••
•
• • •
MD results
10 20 30 40 50 60
Peptide-plane number
70
10 20 30 40 50
Peptide-plane number
60 70
Fig. 3.19 Anisotropy of peptide-plane motions of ubiquitin defined in Eq. (3.2). Panel
(a) gives the anisotropies calculated from the MD trajectory with the fluctuation
amplitudes given in Table A.l in the appendix using caa = (o"a + Oa)/2 .Panel (b)
gives the anisotropies determined by the experimental aag, a values of Table A.3 in
the appendix.
3.6 Discussion 107
3.6.2 Correlation with secondary structure
An attempt was made to correlate the fluctuation amplitudes with the
secondary structure elements of native ubiquitin in Fig. 3.1, which are
indicated also at the top of Fig. 3.7 and Fig. 3.14. Average fluctuation
amplitudes were calculated separately for a helical, P sheet, and other
regions from the MD trajectory: {<*a#)hehx = 4.1°, (<*y)hellx = 8-0°,<°«B>
u ,
= 4-7°. <<0u ,
= 110°> <<W ,u= 5-3°- and
» ap' sheet N T sheet x Up'other
(a) ,= 11.0
.In Fig. 3.20 it can be seen that the central a helix shows
a rather homogeneous behavior with fluctuation amplitudes that are
somewhat lower than for the rest of the protein. The loop regions and the p*sheet regions are more flexible but do not show significant differences when
compared to each other. As described in Section 3.4.3 and Table 3.1, only
peptide planes are taken into account that exhibit an exclusive 3D GAF
behavior.
The differences of the experimentally determined fluctuation amplitudes
between the three categories are even smaller. The average fluctuation
amplitudes are (c^)^ = 5-5°' <°y>aw« = 164°' (ct«P>sheet = 1J°'
<°T>'sheet = 16-5°' Mother = 58°'• «* Mother « 16.8°. Fig. 3.21
visualizes the (almost) absent correlation of the experimental parameters
Gaa and o with the secondary structure elements. Ellipses are shown in the
aaa,G plane that contain the pairs of values of the planes located in one of
the three categories. It is apparent that the ellipses strongly overlap and miss
a characteristic structure-related behavior. On the other hand, there is a
significant anticorrelation between aan and ay leading to similarly inclined
ellipses for all three structural elements.
As mentioned above, the absence of a significant correlation of the
fluctuation amplitudes with the secondary structure reflects to some extent
the fact that the peptide planes with large-amplitude motion have been
excluded from the analysis. On the other hand, it demonstrates that the most
rigid parts of a globular protein, like ubiquitin, show similar fluctuation
amplitudes irrespective of the secondary structure.
108 Anisotropic Intramolecular Backbone Dynamics of Ubiquitin
P sheets central a helix
5 10 15
°a,p [deg]20
other
(%')W
'G'o.^'other = 53°.
^Y other11 oL
5 10 15
<Vp tde9]20
5 10 15
<Vp [deg]
superposition of ranges
20
5 10 15
<Vp [deg]20
Fig 3 20 Loci of the pairs of MD fluctuation amplitudes aag = (aa + oB)/2, cy for
the different peptide planes and their assignment to the three categories "a helix", "P
sheet", and "others", indicated by ellipses
3.6 Discussion 109
P sheets central a helix
10 15 20 25
<Vp[deg]
25other
20 ft +
a? 15CDD
\ +
\ +
DM0
5 (°a$other ~ 5,8°
^Y other~ 16.8U
5 10 15 20 25
°"a,(3 tde9]
superposition of ranges
10 15 20 25 10 15 20
°a,p [deg]
Fig. 3.21 Loci of the pairs of experimental fluctuation amplitudes oag, aY for the
different peptide planes and their assignment to the three categories "a helix", "|3
sheet", and "others", indicated by ellipses.
110 Anisotropic Intramolecular Backbone Dynamics of Ubiquitin
3.7 Conclusions
So far mostly local order parameters have been used to characterize
intramolecular mobility in biopolymers. The author is convinced that future
investigations will concentrate increasingly on the description of the
anisotropic motion of entire molecular subunits, such as secondary structural
elements, aromatic rings, purine or pyrimidine bases, peptide planes, or
methylene groups. Considering these subunits as rigid objects, it is possible
to deduce the motional anisotropy of their motion from the auto- and cross-
relaxation properties of several observer nuclei rigidly attached to the same
fragment.
In this chapter, the focus was lied on fast-time-scale motional processes
exhibited by the rigid backbone peptide planes which are reflected in 13C
and N relaxation data. By comparison with an extended MD simulation,
those peptide planes were identified and characterized which are dominated
by anisotropic 3D Gaussian axial fluctuations. A significantly anisotropic
motional behavior was found that depends little on the secondary structure
elements. The extracted motional parameters, however, depend critically on
the magnitude and orientation of the dipolar and CSA interaction tensors,
responsible for relaxation. It is hoped that in the future solid-state NMR
studies of labeled proteins and refined quantum-chemical calculations [81-
85] will provide more accurate information on these tensors and will lead to
a better understanding of their dependence on the local environment. This
will lead to more accurate and more realistic descriptions of the dynamics of
peptide planes and other molecular subunits. Supported by the further
developments of NMR and MD methodologies, studies of this type will
provide information for a better understanding of the relation between
dynamics and biomolecular function.
4.1 Introduction 111
4 CSA Fluctuations Studied by MD
simulation and DFT Calculations
4.1 Introduction
In this chapter, a combination of quantum chemical CSA calculations and
molecular dynamics simulation is presented to gain more insight into the
magnitude and fluctuation of CSA tensors of the backbone 15N and 13C
spins in proteins and to discuss their relevance for nuclear spin relaxation.
Density functional theory (DFT) has proven to be useful to calculate CSA
tensors of essential fragments of larger biomolecules (see Section 2.5). For
the peptide-bond moiety, it is N-methylacetamide (NMA) which has been
used as a substitute. Despite its relative small size, NMA has allowed one to
gain valuable insight into the origins of the chemical shielding properties of
nuclei within the peptide plane [82,85]. The strategy for the CSA DFT
calculations presented in this chapter relies on results of a recent study [130]
where it was demonstrated that DFT methods yield CSA tensors for 15N
nuclei in the side chain of crystalline asparagine and in the peptide bond of
crystalline alanine-alanine dipeptide with an accuracy comparable to solid-
state NMR.
In Chapter 3, the interpretation was based on several ad hoc assumptions
for the 15N and 13C CSA tensors in order to calculate the significant CSA
relaxation contributions. The CSA tensors were taken from solid-state NMR
studies of small peptides and were assumed to be the same for all peptide
planes which show a 3D GAF motion. A scaling factor for the CSA tensor
was introduced to compensate for the intramolecular motion already
contained in the CSA tensors obtained by solid-state NMR. The uncertainties
connected with such a procedure and the need for additional fitting
parameters - the scaling factors \N and %c - resulted in increased
uncertainties for the extracted motional parameters, in particular the
fluctuation amplitudes for the peptide plane reorientation. With improved
knowledge about the CSA tensors of the backbone spins in the protein a
characterization of the protein's intramolecular motion would be more
112 CSA Fluctuations Studied by MD simulation and DFT Calculations
accurate. As a first step towards this goal, this analysis is concerned with the
investigation of the magnitudes and the fluctuation of the 15N and 13C CSA
tensors located in three different peptide planes of the protein ubiquitin.
To investigate assumptions that are commonly used in relaxation studies
including the one in Chapter 3, DFT calculations are combined with a MD
trajectory of ubiquitin. CSA tensors are calculated for selected "snapshots"
of the MD trajectory with a constant time increment. This leads to a series of
CSA tensors, the "CSA trajectory", whose fluctuations directly depend on
the molecular motion. The analysis of the extracted CSA parameters and
their fluctuations allow a critical test of several assumptions. This concerns
the transferability of CSA tensors obtained by solid-state NMR for small
peptides to liquid-state NMR relaxation experiments on proteins, the
uniformity of CSA tensors located in different secondary structure elements
of the protein, the influence of intramolecular motion on the CSA tensor, the
effect of CSA fluctuations on relaxation, and the contribution of the
antisymmetric part of CSA tensors to relaxation.
In Section 4.2, the DFT calculations and their combination with the MD
trajectory are described. In Section 4.3, averaged CSA tensors are calculated
which can be compared with results obtained by solid-state NMR, and in
Section 4.4, the fluctuations of the CSA tensors during the trajectory are
characterized. Finally, the implications of magnitude and fluctuation of the
calculated CSA's on NMR relaxation are discussed in Section 4.5.
4.2 MD simulation and DFT calculations
MD simulation. The MD simulation of ubiquitin solvated in a box of
water has been described in Section 3.4.1. For the DFT calculation an
extended MD trajectory of 2.5 ns has been used and 625 snapshot were
selected with a time increment of 4 ps.
General method of DFT calculations. For the chemical shielding
calculations, the Sum-Over-States Density Functional Perturbation Theory
(SOS-DFPT) [131,132] as implemented in the deMon-NMR program [133-
135] was used. In this method, Kohn-Sham orbitals are applied to the
4 2 MD simulation and DFT calculations 113
Ramsey formula [136] for chemical shielding, and the energy denominators
are approximated by differences in Kohn-Sham orbital energies, corrected
for changes upon excitation in the exchange correlation potential The
shieldings were calculated using the Loc 1 SOS-DFPT approximation [131],
with the molecular orbitals localized by the method of Boys [137] The
gradient-corrected Perdew-Wang-91 (PW91) exchange-correlation potential
[138] was used Numencal quadrature was earned out on FINE RANDOM
[133,135] angular gnds with 64 radial shells Calculations were performed
using the IGLO-II basis sets [139] Explicitly included fragments that
represent the environment were treated by the less expensive Gaussian
orbital basis set DZVP [140,141] In some calculations water solvent
molecules within a certain cutoff distance were explicitly included For more
details see ref [130]
Strategy for CSA DFT calculations in ubiquitin First, an appropriate
substitute, containing the nucleus (or nuclei) for which the CSA tensor is
calculated, is defined with a bonding topology that resembles the original
molecule as much as possible and that at the same time is small enough to be
treated by DFT The IGLO-II basis sets are used for the atoms of interest and
for all neighbonng atoms belonging to this fragment If the atom for which a
CSA is calculated is part of a delocahzed electron system, such as an amide
plane, all atoms in this system as well as their nearest neighbors are treated
in the same way For all other atoms a DZVP basis set is used In a second
step the close contacts are identified In biomolecules these are mainly
"strong" hydrogen bonds that involve the atom(s) of interest either directly
or via directly bonded neighbors A typical hydrogen bond of the type
C = O H - N can be considered as "strong" if the distance between the
heavy atoms O and N is below 3 2 A and if the angle N-H-0 is near 180°
The close contacts influence both the amsotropy and the onentation of the
CSA The contact partners are then substituted by small molecules In
particular, peptide planes are replaced by N-methyl-acetamides (NMA)
Close water molecules and other small fragments are treated by the DZVP
basis set In a third and final step the partial charge distnbution of the
environment is included Its effect is particularly important in the presence of
114 CSA Fluctuations Studied by MD simulation and DFT Calculations
surrounding fragments that carry net charges like deprotonated carboxyl- or
protonated amino-groups A 10-15 A cutoff distance for the surrounding,
which should be chosen such that the total charge inside the sphere is zero,
appears to be sufficient
The 15N and 13C CSA tensors of three different peptide planes in the
backbone of ubiquitm were examined with DFT calculations (a) peptide
plane 10 between Thr9 and Gly 10, located in a loop region, (b) peptide
plane 30 between Lys 29 and He 30, located in the central a helix, and (c)
peptide plane 43 between Arg 42 and Leu 43, located in a (3 strand The
geometry of the molecular fragments which were used to calculate the CSA
tensors in these three peptide planes are depicted in Fig 4 1 for one
representative snapshot of the MD trajectory which was modified as
descnbed in the following In all three cases, the central fragment consists of
the studied peptide plane i defined by the atoms -Cl _ j -C, _ jO-NH-C -
with an acetylated N terminus and an amide group for its C terminus The C^
centers are always replaced by a methyl group Peptide plane 10 (see
Fig 4 la), which is exposed to the solvent, shows, on average, for every
fourth snapshot one close contact to a water molecule The corresponding
water molecules were included in the DFT calculation In case of peptide
planes 30 and 43 (see Fig 4 lb and c, respectively) close contacts with
peptide planes 33 and 70, respectively, due to hydrogen bonding have to be
taken into account Peptide plane 30 shows no close contacts with water
molecules whereas the carbonyl of peptide plane 43 is involved in a hydrogen
bond with one water molecule which is trapped throughout the trajectory
CSA trajectory. DFT calculations were performed for all of the 625
modified snapshots and the resulting 15N and 13C CSA tensors were stored
as a "CSA trajectory", where each snapshot contains the full CSA tensors
expressed in the coordinate frame of the reoriented CHARMM coordinate
files (see Section 3 4 1)
4.2 MD simulation and DFT calculations 115
a)
<
®-& Gly10
water
water
b)
Asp 32
Fig. 4.1 First part.
116 CSA Fluctuations Studied by MD simulation and DFT Calculations
c)
'^9f Gin 41
Arg42^ Qj)
Val70
Fig 4 1 Molecular fragments which were used in the CSA DFT calculations for (a)
peptide plane 10 (loop region), (b) peptide plane 30 (a helix), and (c) peptide plane 43
(two antiparallel (3 strands) The shown geometries correspond to a single snapshot of
the MD trajectory The central fragment is modified by introducing methyl groups
replacing all C» centers, an amide group as C terminus and an acetylated N terminus
Hydrogen bonds are indicated as dashed lines
15 13,4.3 Average 13N and 1JC CSA tensors
In this section, different approaches for calculating average CSA tensors
from the CSA trajectory are outlined They all refer to very rapid fluctuations
of the CSA tensors with correlation times in the sub-100 ps range The
"global frame averaging" approach comprises a linear averaging over the
CSA tensors of the trajectory which contains the intramolecular motion of
the peptide plane If the amount of motion contained in the trajectory and
present in the solid state at room temperature is comparable, this approach
4.3 Average 15N and 13C CSA tensors 117
corresponds to a solid-state NMR measurement which also yields a CSA
tensor as a linear time-average over the motional degrees of freedom.
Obviously, such an effective CSA tensor should not be used to define the
CSA interaction strength entering the CSA relaxation rate constants. Two
other approaches are introduced to compensate this motional scaling. In the
"local frame averaging" approach the peptide planes with the fixed CSA
tensors are reoriented in order to remove the intramolecular motion. Only
those fluctuations of the CSA tensor remain which are induced by the
changing chemical and electronic environment during the trajectory.
Averaging over these fluctuations yields the "local frame average CSA
tensor". The "3D GAF averaging" which has already been introduced in
Section 2.5.2 is a procedure for calculating the motional scaling of a CSA
tensor which results exclusively from peptide plane reorientation by 3D GAF
motion. It is either possible to calculate a "3D GAF average" tensor starting
from the CSA tensor not affected by intramolecular motion or vice versa. In
this approach, it is assumed that the changing environment has no influence
on the CSA tensor. Whether this is a good assumption or not can be tested by
comparison with the "local frame averaging" approach.
4.3.1 Processing of the CSA trajectory
Global frame averaging. The CSA tensors of the CSA trajectory {g^}(k - l,...,N), given in the molecular frame, were averaged over all
snapshots (N = 625):
N
k= 1
The averaged CSA tensor was rendered traceless and the symmetric part was
separated from the anti-symmetric part: qav = g^m + g"" ', where
sym ,T
.-,
anti,
T ..„ sym ..
°av =(°av + °avy2 d °av = (°«y-Sflv)/2- <2av W3S then
diagonalised yielding the CSA principal values au (I = x, y, z) as
eigenvalues and the principal axes in the molecular frame as eigenvectors.
The antisymmetric part aav was transformed into the principal axis system
118 CSA Fluctuations Studied by MD simulation and DFT Calculations
of the symmetric tensor by the transformation V gav'V where the V matrix
columns contain the calculated eigenvectors.
Local frame averaging. First, the intramolecular motion of the peptide
plane is removed from the CSA trajectory by aligning the instantaneous
peptide plane of each snapshot with the equilibrium peptide plane. The CSA
tensor {ak} of each snapshot k is transformed accordingly by using the
rotation matrix extracted from the alignment procedure. The such modified
CSA trajectory is then processed in the same way as described for the "global
frame averaging" approach.
4.3.2 Analysis of 15N CSA tensors
Characteristic parameters of the averaged 15N CSA tensors of the peptide
planes 10, 30, and 43 are given in Table 4.1. The orientations of these CSA
tensors with respect to the equilibrium peptide planes can be found in Table
A.6 in the appendix. The equilibrium peptide plane belongs to one single
snapshot with an approximately planar structure (co=180°) and was
selected according to Section 3.4.2.1. However, the orientation of the NH
bond of this snapshot was modified by averaging over its out-of-plane and in-
plane bending motion contained in the trajectory.
The "global frame average" 15N CSA tensors of peptide planes 30 and 43
located in the central a helix and a P strand, respectively, are very similar
concerning their anisotropy and asymmetry whereas the calculated tensor
located in the loop region (peptide plane 10) shows a strongly reduced
anisotropy and an increased asymmetry. In this case, the motionally averaged
tensor reflects large-amplitude internal motion including jumps between
different conformations (see Section 4.4.1). The z principal axis (see Table
A.6 in the appendix) lies approximately in the plane in all three cases. The
angle 9^ 7between the z principal axis and the NH vector ranges from 16
to 22° (see Table 4.1). The y principal axis, normally assumed to be
orthogonal to the peptide plane, shows a significant deviation from this
orientation (see Table 4.1: angle 0j_ between 8 and 19°).
4 3 Average 15N and 13C CSA tensors 119
Table 4 1 Results for the pnncipal values (traceless), amsotropy Ao, asymmetry r\, and
the parameter Aa~nII of the 15N CSA tensors of the peptide planes (pp) 10, 30, and 43 in
ubiquitin using three different approaches of averaging, namely the "global frame
averaging", the "local frame averaging", and the "3D GAF averaging" approaches (see
text)
PP
no
averaging
approach [ppm] [ppm] [ppm]
Ao"
[ppm]
r\b °JVtf z
[deg] [deg] [ppm2]
10
global -59 4 -13 4 72 8 109 2 0 63 219 199 358
local -59 8 -38 0 97 8 146 7 0 22 20 4 10 6 457
30
global -64 3 -28 5 92 8 139 3 0 39 16 7 12 4 547
local -64 9 -29 6 94 5 1418 0 37 163 82 546
3D GAF/ -65 5 -29 9 95 4 143 1 0 37 16 7 124
43
global 65 3 -28 5 93 8 140 7 0 39 19 6 81 516
local -65 8 -29 8 95 6 143 3 0 38 180 72 527
3D GAF g -661 -30 6 96 7 145 1 0 37 19 6 81 —
flArj = Gzz-(oxx + Gyv)/2
r\ (°w-c«)/oz.°
9ww *ls tne angle between the z pnncipal axis and the NH bond
8j_ is the angle between the y principal axis and the normal to the peptide
plane
..
2 2 2 2
f Starting from the "global frame average" CSA tensor of this peptide plane, the
corresponding tensor not containing the 3D GAF motion is calculated under the
assumptions of the "3D GAF averaging approach" (see text) using the fluctuation
amplitudes aa = 3 90°, Op = 4 13°, g = 7 53° extracted for peptide plane
30 from the CSA trajectory
g Asf, but using the fluctuation amplitudes aa = 4 33°, Op = 2 87°, rjy = 8 95°
extracted for peptide plane 43 from the CSA trajectory
A comparison of the "global frame average" CSA tensors with results
obtained by solid-state NMR is difficult since exclusively small peptides with
mostly glycine residues as earner of the CSA tensor have been studied
120 CSA Fluctuations Studied by MD simulation and DFT Calculations
experimentally (see refs [119,124-128] and Table A 4 in the appendix) The
calculated anisotropics < 141ppm are smaller than the experimental ones,
which range from 144 to 165ppm The calculated CSA tensors show, on
average, a larger asymmetry than the experimental ones Interestingly, the
smallest experimentally determined anisotropy of 144ppm belongs to the
non-glycine peptide Ala-Ala [124] Preliminary DFT calculations (not
shown) indicate that glycine residues tend towards larger 15N CSA
anisotropics In the experimental studies, it is often assumed that the y
principal axis lies orthogonal to the plane, which is a rather crude
approximation according to DFT calculations Expenmental 0^ values
range between 12 and 24° which is comparable with the range found for the
calculated values The significance of differences between calculated and
expenmental CSA tensors for spin relaxation will be discussed in Section
452
There might be several reasons that lead to the discrepancy between
calculated and experimentally determined CSA tensors First, the
environment of the peptide planes in ubiquitin with its water environment
differs significantly from the one in small and crystallized peptides studied
experimentally Second, the structure of the peptide plane has a large
influence on the CSA tensor If the geometrical setup defined by the
CHARMM force field does not compare quantitatively with the
experimentally studied peptide plane structure significant differences
between experiment and DFT calculation would result In case of 15N CSA
tensors, it might be the NH bond length which is a most cntical parameter
Since the SHAKE algorithm [117] was applied to all bond lenghts involving
a hydrogen atom, the NH bond length is fixed at rNH = 0 997 A throughout
the MD trajectory The dependence of CSA parameters on the NH bond
length was estimated by test calculations for one single snapshot of peptide
plane 43 The results are summarized in Fig 4 2 A nearly perfect linear
dependence of the anisotropy and asymmetry of the 15N CSA tensor on the
NH bond length results This linear scaling for one snapshot can be directly
transferred to the ensemble parameters of the average CSA tensor in Table
4 1 In this case, an increased NH bond length would lead to a further
4.3 Average 15N and 13C CSA tensors 121
decreased anisotropy of the calculated CSA tensor. Finally, it is not known if
the intramolecular motion present in the solid state at room temperature is
comparable to the amount of motion contained in the CSA trajectory.
Different motional scaling of the CSA tensor would lead to a discrepancy
when comparing experimental results with CSA tensors obtained by
averaging over the trajectory.
145 0.56
130
0.54
r\
0.52
0.50
0 98 0.99 1.00 1.01 1.02 1.03 1.04 1.05
rNH [A]Fig 4.2 Dependence of the anisotropy Ao and the asymmetry n of the calculated I5N
CSA tensor on the NH bond length rNH. The NH bond length was systematically
vaned for a single snapshot of peptide plane 43. For an increasing bond length the
anisotropy decreases linearly by 151 ppm/A, while the asymmetry grows linearly by
0.54/A.
4.3.3 Analysis of 13C CSA tensors
Characteristic parameters of the averaged 13C CSA tensors of the peptide
planes 10, 30, and 43 are given in Table 4.2. The orientations of these CSA
tensors with respect to the equilibrium peptide planes can be found in Table
A.7 in the appendix.
The "global frame average" C CSA tensors of peptide planes 30 and 43
show a very similar anisotropy but a significant difference in their asymmetry2
and a large difference in Aa, describing the antisymmetric part of the
122 CSA Fluctuations Studied by MD simulation and DFT Calculations
CSA. The CSA anisotropy and asymmetry of peptide plane 10 reflects, again,
the large motional averaging in the loop region but the difference of the
anisotropy compared to peptide planes 30/43 is less pronounced than in the
case of the N CSA. The x principal axis is in all cases approximately
orthogonal to the peptide plane, the deviation from that direction is less than
7° (see angle 0± xin Table 4.2). The orientation of the y principal axis with
respect to the CO director shows different results for all three peptide planes:
The axis of plane 43 is nearly collinear to the CO bond whereas the axis of
planes 10 and 30 lie in the sectors defined by the atoms OC'Ca and NC'O,
respectively.
1 ^
Experimental results for the C CSA tensor of several peptides are given
in refs. [120-124] and are listed in Table A.5 in the appendix. The
anisotropics lie in the interval between 106 and 126 ppm, which includes the
calculated values of Table 4.2. Interestingly, it is again the CSA tensor
determined for the peptide Ala-Ala [124], whose anisotropy of
Aa = 112ppm fits best to the calculated values 111.7ppm and 113.7ppm
for peptide planes 30 and 43, respectively. The experimentally determined
orientations differ in nearly all cases from the calculated ones in so far as the
x and z principal axes have interchanged their orientations. This is not the
case for the peptide AcGlyTyrNH2 [121] and Ala-Ala [124] which shows an
asymmetry of x\ = 1 leading to underdetermined orientations of the x and z
principal axes in the plane orthogonal to the y principal axis. In the
experimental studies, the y principal axis lies exclusively in the sector
OC'Ca with angles 0CO ranging from 0 to 13°. The differences between
calculated and experimental CSA tensors and their significance for
relaxation will be discussed in Section 4.5.2.
4 3 Average 15N and 13C CSA tensors 123
Table 4 2 Results for the principal values (traceless), anisotropy Ao", asymmetry t|, and
the parameter Acanri (for definitions see Table 4 1) of the C CSA tensors of the
peptide planes 10, 30, and 43 in ubiquitin using three different approaches of averaging,
namely the "global frame averaging", the "local frame averaging", and the "3D GAF
averaging'
approaches (see text)
PP
no
averaging
approach [ppm] [ppm] [ppm]
Ao
[ppm]
T\A "
^CO y
[deg]
S.x"[deg]
AoL,[ppm2]
10
global -510 -20 0 710 106 5 0 44 101 68 41
local -67 6 113 78 9 118 3 071 39 13 57
30
global -72 0 -2 5 74 5 1117 0 93 -4 8 68 127
local 73 6 -2 1 75 7 1135 0 95 -2 5 25 129
3D GAFc -74 0 -17 75 7 1136 0 96 48 68
43
global -67 8 -8 0 75 8 1137 0 79 12 21 15
local -70 2 71 77 3 1160 0 82 21 24 15
3D GAFd -70 0 -7 0 77 0 115 5 0 82 12 21 —
"
9C0 is the angle between the y principal axis and the CO bond For
8C0 > 0°, the y principal axis lies m the sector defined by the atoms OC'Ca,
for 9C0 }< 0° in the sector NC'O
"i xls me atlS^s between the x principal axis and the normal to the peptide
planec
Starting from the "global frame average" CSA tensor of this peptide plane, the
corresponding tensor not containing the 3D GAF motion is calculated under the
assumptions of the "3D GAF averaging approach" (see text) using the fluctuation
amplitudes oa = 3 90°, op = 4 13°, oy = 7 53° extracted from the CSA
trajectory
dAsc but using the fluctuation amplitudes ca = 4 33°, Op = 2 87°,oY = 8 95°
extracted from the CSA trajectory
4.3.4 Comparison of different averaging approaches
The calculated "local frame average" CSA tensors for 15N and 13C are
given in Table 4 1 and Table 4 2, respectively The removal of the
intramolecular motion contained in the CSA trajectory leads in all cases to
124 CSA Fluctuations Studied by MD simulation and DFT Calculations
increased values for the amsotropy For the N tensor, the strongest effect is
found for the very flexible peptide plane 10 where the asymmetry is
significantly reduced and the amsotropy increases by a factor 1 34 This
indicates that the differences of the "global frame average" CSA tensors of
peptide plane 10 and 30/43 result mainly from the large motional averaging
in case of peptide plane 10 The peptide planes 30 and 43, which can be both
described by a 3D GAF motion (see Chapter 3), show a much smaller scaling1 ^
of the amsotropy by a factor of 1 02 For the c tensor of peptide plane 10,
the effect of motional averaging is smaller than in the N case, the
amsotropy increases by a factor 1 1 whereas a factor of 1 02 is found in case
of the peptide planes 30/43 with 3D GAF motion
The calculated "local frame average" CSA tensors of the peptide plane 30
and 43 can be compared with results for the "3D GAF averaging" approach
also contained in Table 4 1 and Table 4 2 Here, the "global frame average"
CSA tensor was used as the reference tensor from which the contained
3D GAF motion was "removed" using the fluctuation amplitudes for the 3D
GAF motion extracted from the CSA trajectory Differences between a CSA
tensor calculated this way and the "local frame average" CSA are only due
to CSA fluctuations caused by a changing chemical and electronic
environment during the trajectory, which appear exclusively in the "local
frame averaging" approach Different behavior can be seen for N and C
1 ^tensors In case of c, the averaging due to 3D GAF reorientation is
dominant and the tensors of the two approaches are nearly identical (see
Table 4 2) On the other hand, the 15N CSA tensors of the "local frame
averaging" approach show a significantly lower amsotropy than the tensors
calculated when assuming averaging due to reorientation only It can be
concluded that the N tensor shows a larger sensitivity on changes in the
environment than the 13C CSA tensor (see also Section 4 4)
4 3 Average 15N and 13C' CSA tensors 125
Table 4 3 Subtraction of 3D GAF motional scaling from 15N and 1JC CSA tensors
Based on the "global frame average" CSA tensor a CSA tensor not containing the
peptide plane reorientation is calculated from a given set of fluctuation amplitudes
according the assumptions of the "3D GAF averaging" approach (see text)
spin(o«p,o7)
[deg] [ppm] [ppm] [ppm]
Ac
[ppm]
Tl
15N
(0,0)" 64 3 -28 5 92 8 139 3 0 39
(5,5) 65 9 -29 2 95 1 142 7 0 39
(10,10) 70 6 -315 102 1 153 2 0 39
(7,14) -67 8 33 4 1012 1518 0 34
(10 20) -715 -39 5 1110 166 5 0 29
13C,
(0,0)b -67 8 -8 0 75 8 1137 0 79
(5,5) -69 4 -8 0 77 4 116 1 0 79
(10,10) -74 0 -9 0 83 0 124 5 0 78
(7,14) 73 9 -5 5 79 4 119 1 0 86
(10,20) -812 -2 2 83 4 125 1 0 95
" The given CSA tensor corresponds to the "global frame average" CSA tensor of
peptide plane 30 given in Table 4 1 The results given in the rows below are
calculated by using this tensor as a reference
The given CSA tensor corresponds to the "global frame average" CSA tensor of
peptide plane 43 given in Table 4 2 The results given in the rows below are
calculated by using this tensor as a reference
The "3D GAF averaging approach" can be useful to remove motional
scaling effects from an experimentally obtained CSA tensor in order to
calculate an amsotropically upscaled CSA tensor which is appropriate for the
interpretation of CSA relaxation in the liquid state Results for sets of
different fluctuation amplitudes are illustrated in Table 4 3 It can be seen that
anisotropic intramolecular motion has an anisotropic scaling effect on the
CSA tensor The asymmetry of the CSA tensor is increasingly modified with
increasing anisotropy of the peptide plane motion The anisotropy of the
126 CSA Fluctuations Studied by MD simulation and DFT Calculations
CSA tensor is significantly affected in case of fluctuation amplitudes
°aB Y> 10°
•In cases of additional jump motion the effects are even larger.
Here, also the orientation of the calculated CSA might differ significantly
from the reference tensor due to averaging between two (or more) distinct
conformations.
4.4 Fluctuations of the CSA tensors
In a next step, the fluctuations of the CSA parameters during the CSA
trajectory shall be described. The fluctuations of the orientation and of the
principal values of the tensor are described independently here. Both effects
come together and result in either motional averaging of the CSA tensor in
solid-state NMR measurements (see Section 4.3) or CSA relaxation (see
Section 4.5.1). For the analysis, the CSA tensor gk of each snapshot k is
symmetrized and diagonalised. The antisymmetric part is transformed for
each snapshot as described for the averaged CSA tensor in Section 4.3.1.
4.4.1 Fluctuation of CSA anisotropy and asymmetry
For the I5N and 13C CSA tensors of all peptide planes the average values
and their standard deviations for the anisotropy and asymmetry of the2
symmetric part and for the parameter Ac fof the antisymmetric part are
given in Table 4.4. An example for a distribution of Ac and r\ are depicted
in Fig. 4.3 for the N tensor of peptide plane 30. In all cases, the mean values
for the anisotropy are slightly increased when compared to the results for the
"local frame average" CSA (see Table 4.1 and Table 4.2), since the "local
frame average" approach includes averaging over orientational fluctuations
with respect to the molecular fragment. The 15N CSA parameters show
generally larger fluctuations than the 13C CSA parameters. The standard
deviations for the anisotropy values range from 7 to 15% (15N) and 4 to 6%
(C). The parameter Aaantl shows standard deviations between 30 and
50% of the average values.
4.4 Fluctuations of the CSA tensors 127
Table 4.4 Fluctuations of the !5N and 13C CSA parameters during the CSA trajectories.The average value with standard deviation of the anisotropy, the asymmetry, and Aoan/|for the antisymmetric part, which were calculated from the ensemble of 625 snapshots,are given for the peptide planes 10, 30, and 43.
spinpeptide
plane no.
anisotropy
Act [ppm]asymmetry r\ Ao„ [PPffl2]
15N
10 148.7 ±14.8 0.31 ±0.15 561± 239
30 142.7 ±10.1 0.39 ±0.10 595 ±148
43 144,2 ±10.5 0.39 ±0.14 569 ±157
13C
10 119.0 ±7.5 0.72 ±0.17 68 ±37
30 114.0 ±5.5 0.95 (±0.14)" 137 ±41
43 116.3 ±5.0 0.82 (±0.11)" 21 ±13
1The distribution of the CSA asymmetry values is strongly asymmetric.
100
80w*-»
c
<Den
> 60
CD
£40n
E
=j 20
a) (142.7 ±10.1) ppm100
80w
§ 600
o
&-40-Q
£3 20
b)
0.39 ±0.10
_JM1JU I uwin _J L^. . ,
^i-200 -180 -160 -140 -120 -100 -1 0 1
Ao [PPm] T|
Fig. 4.3 Fluctuation of I5N CSA parameters of peptide plane 30. The distributions of
(a) the anisotropy Ac and (b) the asymmetry r\ of the 15N CSA tensor are shown for
the ensemble of 625 snapshots of the CSA trajectory.
128 CSA Fluctuations Studied by MD simulation and DFT Calculations
4.4.2 Orientational fluctuations of the CSA tensors
The orientational fluctuations of the CSA tensor during the trajectory are
mainly given by the peptide plane reonentation Additional fluctuations of
the CSA pnncipal axes with respect to the peptide plane frame are induced
by the changing chemical and electronic environment and can be extracted
by removing the internal motion of the peptide plane from the trajectory
which has been descnbed for the "local frame averaging" approach in
Section 4 3 1 The onentational distnbutions of the principal axes (with
respect to the equilibnum peptide plane) before and after the removal of the
internal motion are depicted in Fig 4 4 and Fig 4 5 for the N and 13C
CSA tensors, respectively In all cases, the onentational fluctuation due to the
peptide plane reonentation is dominant Fluctuations due to a changing
environment (see nght sides of Fig 4 4 and Fig 4 5) are generally larger in
case of the 15N CSA tensors Here, the distnbutions of the x and y pnncipal
axes are broader than that one of the z pnncipal axis This reflects a strong
influence of the asymmetry of the tensor on the pnncipal axis onentations
The N tensor of peptide plane 10 shows, on average, the smallest
asymmetry values during the CSA trajectory (see Table 4 4) In extreme
cases of r| —> 0, the x and y pnncipal axes he arbitrarily in a plane
orthogonal to the z pnncipal axis Correlations between the asymmetry r| of
the tensor and the orientations of the principal axes are an important result
indicating that onentation and pnncipal value fluctuations must not be
treated separately when calculating their influence on CSA relaxation (see
Section 2 5 1) In case of 13C and all three peptide planes, the onentational
distributions in the "local frame" are similar for the three pnncipal axes of
the CSA tensor reflecting its high asymmetry In addition, similar
distributions are found for all peptide planes including plane 10 This
confirms that the c tensor is less sensitive on changes of the chemical
environmental which is also visible m its relatively small amsotropy
fluctuations (see Table 4 4)
4.4 Fluctuations of the CSA tensors 129
global frame local frame
Fig. 4.4 Orientation fluctuations of the N CSA principal axes during the CSA
trajectory. Results are shown in (a,b,c) for peptide planes 10, 30, and 43. The
projections of the principal axis unit vectors on a unit sphere are given with respect to
the equilibrium peptide plane frame ew e„, e .In each row the left part corresponds to
the fluctuations including the internal motion of the peptide plane, whereas the right
part corresponds to the fluctuations left after removing the internal motion by
alignment of the instantaneous peptide plane in the equilibrium orientation.
130 CSA Fluctuations Studied by MD simulation and DFT Calculations
global frame local frame
Fig. 4.5 Orientation fluctuations of the 13C CSA principal axes during the CSA
trajectory. Results are shown in (a,b,c) for peptide planes 10, 30, and 43. The
projections of the principal axis unit vectors on a unit sphere are given with respect to
the equilibrium peptide plane frame ea, e», e .In each row the left part corresponds to
the fluctuations including the internal motion of the peptide plane, whereas the right
part corresponds to the fluctuations left after removing the internal motion by
alignment of the instantaneous peptide plane in the equilibrium orientation.
4 5 15N and 13C CSA relaxation of the protein backbone 131
4.5 iaN and UC CSA relaxation of the protein back¬
bone
In Section 4 5 1, the calculated CSA trajectory is used as a model system
to estimate systematic errors on CSA relaxation rate constants which result
from simplifying assumptions on the CSA interaction strength Since it is not
possible to measure the CSA fluctuations in a time-resolved manner it has to
be tested whether CSA fluctuations have to be explicitly taken into account
for CSA relaxation or whether the use of averaged CSA interaction strengths
is sufficiently accurate In Section 4 5 2, CSA relaxation is regarded from a
different point of view, and it is asked which assumptions on the CSA tensors
are reasonable when extracting motional parameters from NMR relaxation
measurements A discussion of the role of the antisymmetric part of the CSA
is given in Section 4 5 3
4.5.1 Implications of CSA fluctuations on CSA relaxation
CSA relaxation in proteins in the liquid state is normally calculated using
a motionally averaged CSA interaction strength obtained by solid-state NMR
measurements Fluctuations of CSA tensor principal values and orientation
(with respect to the molecular fragment) due to a changing chemical
environment induced by intramolecular motion have not been taken into
account in NMR relaxation studies so far However, these fluctuations
directly enter the autocorrelation function which determines CSA relaxation
The systematic error connected with the assumption of a constant CSA
interaction strength is estimated by comparing four different approaches for
analyzing the CSA trajectory In the "exact" approach, the time-dependent
fluctuations of the CSA tensor principal values and onentations in the
molecular frame are descnbed by a 2nd rank correlation function which can
be calculated from the CSA trajectory according to Eq (2 84) Assuming
isotropic overall tumbling and independence of internal and overall tumbling
motion, the correlation function and spectral density function in the
laboratory frame are given by
132 CSA Fluctuations Studied by MD simulation and DFT Calculations
C(x) = e Cml(x) and /(co) = j C(t)cos(cox)rfx. (4.2)
The CSA contribution to T{ relaxation of 15N and 13C spins can then be
calculated
(V-l = T-AJ^x)- (4-3)
In the "global frame average" approach, the corresponding CSA tensors,
listed in Table 4.1 and Table 4.2 for the 15N and 13C backbone spins, are
fixed at the equilibrium peptide plane. Then, a CSA trajectory is calculated
by aligning the equilibrium peptide plane with each instantaneous peptide
plane of the series of 625 snapshots. The corresponding autocorrelation
function is, again, given by Eq. (2.84). This approach corresponds to the
normally applied procedure of using a constant CSA tensor determined by
solid-state NMR to calculate CSA relaxation. The "local frame average"
approach and the "3D GAF averaging" approach (in case of peptide planes
30/43) are analogous to the "global frame averaging" approach using the
"local frame average" and "3D GAF average" CSA tensors instead which
have been calculated in Section 4.3.
In Fig. 4.6, the correlation functions of the four approaches are compared
with each other for the 15N and 13C CSA tensors of the peptide planes 10,
30, and 43. Note, that the functions are not normalized and contain the
interaction strength. The resulting relaxation rate constants (Tj )CSA are
listed in Table 4.5. In all cases, the "exact" approach corresponds to the
largest rate constants, the "global frame averaging" approach to the lowest
rate constants. The ratio y of the relaxation rate constants of the "exact" and
the "global frame averaging" approach is a measure for the systematic error
when neglecting the CSA fluctuations contained in the CSA trajectory. For
peptide planes 30 and 43, which show 3D GAF motion, y takes about the
same value of approximately 1.05 in all cases. A large systematic error
results in the flexible loop region with y = 1.75 and 1.27 for the N and
13C CSA tensors of peptide plane 10, respectively. However, the correlation
functions of the "exact" and the "global frame averaging" approach show a
4 5 15N and 13C CSA relaxation of the protein backbone 133
large correlation (constant offset between the correlation functions in
Fig 4 6) Thus, the fluctuations have not to be taken into account explicitly
but the systematic error can be compensated by an upscahng of the "global
frame average" CSA tensor Using the "local frame average" CSA tensor or
the "3D GAF average" CSA tensor as reference CSA tensor is appropriate in
case of 13C and peptide planes 30/43 which show no large-amplitude
intramolecular motion For N, the "3D GAF average" CSA tensor comes
closer to the "exact" approach than the "local frame average" CSA tensor
which reflects the averaging effect in the "local frame" In the loop region,
this effect is even larger and the "local frame average" approach only partly
compensates the systematic offset between the "exact" and the "global frame
averaging" approach However, peptide plane 10 represents a "worst case"
and it is not likely that a CSA tensor of a very flexible loop region measured
by solid-state NMR would be used for NMR relaxation calculations
Table 4 5 I5N and 13C CSA relaxation rate constants at field strength corresponding to
600 MHz proton resonance frequency for the peptide planes 10, 30, and 43 (TA )CSAwere calculated according to Eqs (4 2) and (4 3) using the correlation functions in
Fig 4 6 for three different approaches namely the "exact", the "local frame average",
and the "global frame averaging" approach The overall tumbling correlation time xc
was set to 4 03 ns which corresponds to ubiqmtm dissolved in water at 300 K
spinPP
no
(ri')«A [* ']
'exact'
approach
"local frame
average" approach
"global frame
average' approach
la
15N
10 0315 0 280 0 180 175
30 0 474 0 465 0 451 105
43 0 484 0 476 0 463 105
13C.
10 0 374 0 362 0 295 127
30 0 499 0 495 0 478 104
43 0 484 0 483 0 460 105
"
y is the ratio of the relaxation rate constants (T] )c&4of the "exact" and the
"global frame averaging" approach
134 CSA Fluctuations Studied by MD simulation and DFT Calculations
500
time [ps]500
time [ps]
f)
EQ.
1 65
LL 1 55
o
500
time [ps]1000
1 45500
time [ps]
1000
1000
1000
Fig 4 6 Auto-correlation functions for CSA interaction in the molecular frame
extracted from the CSA trajectory according to Eq (2 84) for the "exact" (e), "local
frame average" (1), "3D GAF average" (f), and "global frame average" (g) approach
Results for N and peptide planes 10, 30, and 43 are given in panels (a, c, e), for C
and peptide planes 10, 30, 43 in panels (b, d, f) Note the different scales of the Cmt(t)ordmates
4 5 15N and 13C CSA relaxation of the protein backbone 135
Dipole-CSA cross-correlated relaxation is sensitive to fluctuations of CSA
pnncipal values and angles between the CSA principal axes and the dipolar
director (it is assumed that the intemuclear distance is constant) Hence, the
relaxation rate constants do also depend on fluctuations of the dipolar
director with respect to the peptide plane The JH-15N dipole- N CSA cross-
correlation was studied following the strategy applied for the auto¬
correlation case The fluctuation distributions of the angles 0^ and
®nh zbetween the NH bond and the y and z 15N CSA pnncipal axes are
shown in Fig 4 7a and b, respectively, for peptide plane 43 The standard
deviations of the angle fluctuations of 4 2° for 6^ and 2 6° for 0^ z
reflect the larger fluctuation in case of the y pnncipal axis due to the low
asymmetry of the tensor (see also Fig 4 4) The mean value of 18 3° for
®NH zcomes cl°se t0 tne result for the "local frame averaging" approach
given in Table 4 1
The cross-conelation function in the molecular frame takes the form
Cn.nhW = (ox(0)(P2(ex(0) eNH(x))))
+ (ay(0)(P2(ey(0) eNH(T)))) + (oz(0)(P2(ez(0) eNH(x)))) (4 4)
The principal value fluctuations linearly enter the correlation function On
the other hand, the onentational fluctuations appear in the Legendre
polynomial terms which leads to non-linear averaging In case of isotropic
overall tumbling of the molecule the corresponding correlation function and
spectral density function in the laboratory frame are given by
JN NH(<o) = j CNfNH(x)cos(m)dx (4 5)
The cross-correlation relaxation rate constant FN NHcan be calculated
^ NH= -^N^NH(4JN, NH®) + 3JN, NH«°n))> <4 6)
136 CSA Fluctuations Studied by MD simulation and DFT Calculations
with^ = ^^WfA'W assuming a constant NH bond length rm
The cross-correlation functions CN NH(t) were calculated according to
Eq (4 4) for all three peptide planes using, again, the "exact", the "local
frame averaging", and the "rigid averaging" approach Internal correlation
functions for the different approaches are plotted in Fig 4 7c and d for
peptide planes 10 and 43, respectively It can be seen, that even m case of
peptide plane 10 the fluctuations have not to be taken explicitly into account
and it is appropriate to use the "local frame average" CSA tensor for the
calculation of the rate constants TN NHAt field strengths corresponding to
600 MHz proton resonance the calculated ratios for the rate constants of the
"exact" and the "global frame averaging" approach are 1 33, 1 06, and 1 05
in case of peptide planes 10,30, and 43, respectively Thus, a systematic error
of at least 5% has to be considered when calculating the rate constants with
the "global frame average" CSA tensors
4.5 15N and 13C CSA relaxation of the protein backbone 137
110
200 400 600 800 1000
time [ps]
200 400 600 800
time [ps]
1000
Fig. 4.7 Fluctuation distributions of (a) the angle BNH zbetween the NH bond and die
z principal axis of the 15N CSA tensor and (b) the angle 9^ between the NH bond
and the y principal axis. Mean values and their standard deviations for the ensemble of
625 snapshots are indicated in the figure. H- N dipole- N CSA cross-correlation
functions are calculated according to Eq. (4.4) for peptide plane 10 (see panel c) and 43
(see panel d) using the "exact" (e), "local frame average" (1), and "global frame
average" (g) approach.
4.5.2 CSA relaxation based on average CSA tensors
The last chaper raises the question which average CSA tensor should be
used when motional parameters are extracted from NMR relaxation
measurements. In Table 4.6 different CSA tensors, either obtained by the
138 CSA Fluctuations Studied by MD simulation and DFT Calculations
CSA trajectory of peptide planes 30 and 43 with the "global frame
averaging" approach or by solid-state NMR experiments, are translated into
the auto- and cross-correlated relaxation rate constants (Tx N)CSA and
FN NHfor 15N, (T{ C)CSA and rc cc
for 13C assuming anisotropic 3D
GAF motion of the peptide plane. A comparison of the calculated rate
constants for the CSA tensors DFT pp 30, DFT pp 43, and Ala-Ala shows
that the cross-correlation relaxation rate constants are much more sensitive
to even slight differences in the orientation of the CSA tensors since the
relative orientation of the CSA tensor with respect to the dipolar director
enters the cross-correlation function. Hence, assumptions which are
reasonable for the analysis of auto-correlation data cannot be simply
transferred to the analysis of cross-correlation data.
For the auto-correlation rate constants (7, x)cSA' tne calculated CSA
tensors located in the a helix and a (3 strand of ubiquitin yield very similar
results for bothlN and C. This supports the assumption of uniform CSA
tensors used in Chapter 3. On the other hand, the rate constants calculated
from the two different experimental CSA tensors show quite large
differences. The rate constants calculated from N and C CSA tensors of
the peptide Ala-Ala compare well with the results from the calculated
tensors. However, the large anisotropy Aa = I64.4ppm of the N CSA
tensor of Hiyama et al. [119] (for monoclinic Boc(Gly)3OBz), most often
used in NMR relaxation studies of proteins, results in a rate constant
(Tl N)CSA which is by a factor of 1.30 larger than the one for the 15N CSA
tensor of Ala-Ala. In case of 13C, the discrepancy is smaller, but with the
CSA tensor of Stark et al. [120] a relaxation rate (Ty C)CSA by a factor of
1.13 larger than calculated for the 13C CSA ofAla-Ala is obtained. Whether
the relatively small anisotropics of the calculated CSA tensors are
appropriate for the interpretation of protein relaxation data, cannot be
decided from experimental relaxation data alone since an extraction of CSA
parameters from the relaxation data -needed for comparison- does rely on
assumptions for the dipolar interaction strength affecting both the analysis of
auto- and cross-correlated relaxation data. This illustrates that the study
presented in this chapter is only a first step on the way to an accurate
4 5 15N and 13C CSA relaxation of the protein backbone 139
interpretation of CSA relaxation data in proteins Further DFT calculations
for other peptide planes in ubiquitin or different proteins, solid-state NMR
expenments for non-glycme peptides, but also more reliable data on the
dipolar interaction strengths are necessary to gain more information about
the magmtude of the CSA tensors
Table 4 6 Calculated auto- and cross-correlated relaxation rate constants at 600 MHz
proton resonance frequency using different CSA tensors obtained either by averaging
over the CSA trajectories or by solid-state NMR The relaxation rate constants are
calculated assuming isotropic overall tumbling with correlation time %c = 4 03 ns and
intramolecular 3D GAF motion of the peptide plane with fluctuation amplitudes
°«p = 7° and Gv= 14°
15N CSA tensor"
(^1 N^CSA Is 1 *N NH *-S 1
solid-state NMR Boc(Gly)3OBz [119] 0 58 4 27
solid-state NMR AlaAla[124] 0 44 4 46
DFT pp 30 "global frame averaging" 0 43 4 08
DFT pp 43 "global frame averaging" 0 44 3 76
13C CSA tensor* (7"l c)c\SA [* ] rcc [sV
solid-state NMR GlyGlyHCl [120] 0 52 -135
solid state NMR AlaAla [124] 0 47 -1 19
DFT pp 30 'global frame averaging" 0 46 101
DFT pp 43 "global frame averaging" 0 45 -103
"For the CSA tensors obtained by solid-state NMR see also Table A 4 in the
appendixbFor the CSA tensors obtained by solid-state NMR see also Table A 5 in the
appendix
C
T/v nhls tne 15N-LH dipole 15N CSA cross-correlation relaxation rate constant
and was calculated according to Eq (2 73)dTc c
is the 13C'-13Ca dipole-13C CSA cross-correlation relaxation rate
constant and was calculated according to Eq (2 73)
In case of cross-correlated cross-relaxation, the situation is even more
difficult Most published studies are based on the simple assumption of an
140 CSA Fluctuations Studied by MD simulation and DFT Calculations
axially symmetric N CSA tensor which is approximately true for the tensor
determined by Hiyama et al [119] with r\ = 0 06 The DFT calculations
and other experimental results (see Table A 4 in the appendix) indicate that
this assumption might not be appropriate since asymmetries up to r| = 0 45
have been found This may have considerable influence on the interpretation
of cross-correlation relaxation data In addition, the assumption of uniform
CSA tensors for different peptide planes seems to be much more cntical than
in the case auto-correlated relaxation (see above)
In this section, motionally averaged CSA tensors were compared In the
last section, it was shown that an average CSA tensor which has been
upscaled to compensate for the motional scaling would be appropriate for
describing NMR relaxation In principle, one could use "local frame
average" CSA tensors extracted from CSA trajectories for which it has been
shown that the amount of intramolecular motion in the liquid state is properly
reflected in the trajectory If the relaxation calculations are based on
experimentally determined CSA tensors, the upscahng of the tensor could be
done using the "3D GAF averaging" approach described in Section 4 3 4
The magnitude of the intramolecular motion present in the solid state has
then to be estimated
4.5.3 Role of the antisymmetric part of CSA tensor
2The antisymmetric part Aoantl is not manifested in standard solid-state
NMR experiments, since it is non-secular with respect to the Zeeman
interaction However, it contributes to CSA relaxation (see Section 2 3)
although it is normally neglected when calculating relaxation parameters of
proteins in solution The DFT calculations provide data for the
antisymmetric part of the CSA tensors In Table 4 1 and Table 4 2 the average2 7 2 2
values for the parameter Aaantl = a~v + axz + a in the principal axes
system of the symmetric CSA are given for N and C, respectively The
relaxation contribution of the antisymmetric and symmetric part can be
calculated with Eqs (2 36), (2 66), and (2 74) considering only overall
tumbling motion In Fig 4 8, the relative contribution of the antisymmetric
part at 600 MHz proton resonance frequency, expressed as ratio
4 5 15N and 13C CSA relaxation of the protein backbone 141
_1 anti _j sym
(^l x)csa//(^1 x)c5A' *s pl°tte(i versus the rank 2 correlation time of the
overall tumbling motion for the 15N and 13C CSA's of peptide plane 30 and
43. The relative contribution of the antisymmetric part is much larger in case
of N. For larger molecules with tumbling correlation times tc > 4ns a
lower plateau is reached. Here, the antisymmetric CSA contributions of
about 6% for 15N and less than 2% for 13C are rather small. For smaller
molecules with tumbling correlation times xc < 300ps much larger
contributions of up to 40% and 10% in case of 15N and 13C, respectively, are
predicted.
_1 anti
(Tl,x)cSA_1 sym
(Tl,x)cSA
15N, pp 30~~-0\
0.3 N\ 15N, pp43
0.2 - \
13C, pp 30\
0.1 /
v^13C, pp43^\/
10-11 1Q-10 10-9 -i 0-8
Fig. 4 8 Ratio of relaxation rate constants of the antisymmetric and the symmetric
CSA relaxation pathways at 600 MHz proton resonance frequency as a function of the
rank 2 correlation time xc for isotropic overall tumbling motion. Results are shown for
N and 13C CSA tensors of peptide planes (pp) 30 and 43 in ubiquitin (indicated m
the figure) No intramolecular motion is included. The CSA parameters were taken
from the "local frame averaging" approach (see Table 4.1 and Table 4.2).
Relaxation due to the antisymmetric part of the 15N CSA tensor should be
measurable according to the shown results. An ideal test system would
consist of a small peptide dissolved in a deuterated solvent in order to
142 CSA Fluctuations Studied by MD simulation and DFT Calculations
suppress the dominant H- N dipolar relaxation. Tl Nmeasurements at
different field strengths, temperatures or viscosities would be required to
achieve a decomposition of the symmetric and antisymmetric CSA
relaxation contributions due to their distinct dependence on the rank 2
tumbling correlation time for the overall tumbling.
4.6 Conclusions
By combination of DFT calculations and MD simulation, a CSA trajectory
is created which can be used as a model system to estimate systematic errors
in the calculation of relaxation rate constants which result when introducing
simplifying assumptions on the CSA interaction strength. It was found that
for the studied peptide planes in ubiquitin a description with an average CSA
tensor, which does not contain the reorientational averaging of the peptide
plane, is sufficient and the principal value and orientational fluctuations of
the CSA tensor have not to be taken explicitly into account. Results for the
calculated average CSA tensors located in the a helix and a P strand of
ubiquitin support the assumption of uniform CSA tensors in different peptide
planes which has been made use of in Chapter 3 when extracting motional
parameters from NMR relaxation data.
It is difficult to decide whether the absolute values of the calculated CSA
tensor parameters are realistic or not. On the one hand, there is a good
agreement of both 15N and 13C CSA tensors with a single experimental
study ofthe peptide Ala-Ala [124] which corresponds to an anisotropy at the
lower end of all experimental studies. On the other hand, the calculated
anisotropics are significantly smaller than what is normally used in NMR
relaxation studies (see Chapter 3) or, recently, has been extracted from cross-
correlation relaxation experiments ( N anisotropies of about 170ppm, see
e.g. ref. [107]). However, such NMR relaxation studies do not rely on a self-
consistent parameter set. In particular, the assumption on the NH bond length
is very critical, and it significantly affects the results for the extracted
anisotropies. Similar problems are contained in the calculation of the CSA
trajectory since the DFT calculations rely on geometric parameters
determined by the MD force field which was not optimized for this purpose.
4.6 Conclusions 143
The DFT calculations have shown that the relaxation contribution of the
antisymmetric part with respect to the symmetric part of the CSA tensor is
less than 10% in case of proteins dissolved in water. For these systems, it
seems to be unlikely that the antisymmetric part can be accurately extracted
from standard NMR relaxation measurements. On the other hand, this
relaxation contribution might be wrongly attributed to the symmetric CSA
tensor resulting in increased scaling factors for the symmetric part of the
CSA (i. e. scaling factors XN and Xc in Chapter 3).
144 CSA Fluctuations Studied by MD simulation and DFT Calculations
5.1 Introduction 145
5 Rotational Motion of a Solute Molecule
in a Highly Viscous Liquid
5.1 Introduction
NMR measurement parameters are sensitive to an extremely wide range of
motional time scales from picoseconds to seconds [7]. The fastest
intramolecular dynamical processes in the picosecond to microsecond range
can be monitored by various types of relaxation experiments in the laboratory
frame. However, the inherent power of NMR relaxation spectroscopy for
studying mobility in the liquid phase is hampered by the fact that only the
composite effect of intramolecular and overall tumbling motion is sensed by
the relaxation parameters. It is impossible to measure intramolecular
mobility alone. This becomes apparent from the Lipari-Szabo approach
which has been introduced in Section 2.2.3.2. The relaxation-active motional
correlation function for an auto-correlated process of a rank 2 interaction
\l = v in the laboratory frame is then given by a combination of Eqs. (2.39)
and (2.42):
C (T) = YS + (1 -S )e ]e , (5.1)
where the definitions S = S^ and xint = x\„t have been used. The
rotational tumbling with the correlation time xc acts as a multiplicative
"masking" process of the intramolecular mobility with correlation time xmt
that is ofprimary interest. As a result, xtcan be determined with reasonable
accuracy only in a certain range, which defines the "observation window" of
intramolecular motion. This is a severe practical restriction in motional
studies. The size of the observation window depends on the type of
measurement, the type of internal motion, the overall tumbling correlation
time and the magnetic field strength. In Fig. 5.1, the dependence of the
C Tl of a dipolar ^C-1!! system on the internal correlation time xint is
calculated for different values of the overall tumbling correlation time. As
can be seen in Fig. 5.1a and b, the size of the observation window increases
with increasing overall tumbling correlation time. The center of the
146 Rotational Motion of a Solute Molecule in a Highly Viscous Liquid
observation window is approximately given by xint = xc in cases of
tumbling correlation times xc<lns. For xc>lns, the functional
dependence of 13C T{ on %int is no longer monotonous and the center is
given by xint < xc. The center is slightly shifted towards increasing internal
correlation times for increasing xc. However, this effect which is not linear
in xc decreases for increasing x .Outside of this observation window,
intramolecular motion leads to xt -independent averaging effects. In the
extreme narrowing regime tmf«xc, only the order parameter remains active
(see Eq. (5.1)) and the anisotropic relaxation-active interactions, such as
2
dipolar interaction or CSA interaction, are scaled by S .In the case of slow
internal motion (xint » xc) the relaxation rate constants which belong to
different conformations of the molecule are averaged.
It would be highly desirable to shift the observation window for xint by a
selective change of xc [142] while the intramolecular motion is not affected.
This might be illustrated by two examples. In case of the cyclic peptide
antamanide with an overall tumbling correlation time of approximately
150 ps for a solution in CDC13 (for details see refs. [70,142]), the internal
correlation times for discrete jumps between different conformations in the
phenylanaline sidechains of antamanide, which are much larger than xc [70],
might be extracted when increasing xc by a factor of at least 10 as shown in
Fig. 5.1a. In case of larger molecules as the protein ubiquitin with a overall
tumbling correlation time xc = 4.0 ns for a solution in water (see Section
3.5.3), the internal correlation times for very fast intramolecular motion with
xint < 50 ps (see Section 3.5.4) might be accessible when going to correlation
times xc in the range of 10 ns as shown in Fig. 5.1b.
Within the framework of classical hydrodynamic theory for a continuous
solvent medium, the rotational correlation time for an interaction of rank 2 in
a spherical molecule obeys the Stokes-Einstein-Debye (SED) relationship
(see Section 2.2.2.1):
<<" ?rwhere V is the effective spherical molecular volume, T| is the shear viscosity,
T is the absolute temperature, and k is the Boltzmann constant. This relation
5 1 Introduction 147
has frequently been used in NMR since the pioneering work of Bloembergen,
Purcell, and Pound [143] Based on Eq (5 2), it is tempting to shift xc and
the observation window of xint by varying the viscosity T) of the solvent,
while hoping that the intramolecular mobility of the solute is not significantly
affected
tint NFig 5 1 Illustration of the observation window (indicated as rectangle) for the
correlation time of intramolecular motion The dependence of C T, at 600 MHz
proton resonance frequency on the internal correlation time xmt is given for (a) the
overall tumbling correlation times xc = 150ps and 1 5 ns, and (b) for xc = 4ns and
10 ns assuming the spectral density approach of Lipan-Szabo in Eq (5 1) The C T,
was calculated for an ^C-'H bond with bond length rCH=\ 08 A considering only
dipolar relaxation
148 Rotational Motion of a Solute Molecule in a Highly Viscous Liquid
Williamson and Williams [144] used polychlorotrifluoroethene (PCTFE)
as a polymeric solvent for NMR studies. Its viscosity can be varied in a large
range by changing the temperature and by mixing fractions of different
degrees of polymerization. PCTFE was subsequently used in a number of
NMR investigations [145-148] for adjusting the viscosity in structural
studies of solute molecules measuring the nuclear Overhauser effect (NOE)
which depends on xc. However, no quantitative study of the dynamical
aspects of solute molecules in PCTFE has been reported so far.
The present experimental study explores the potential of PCTFE as a
solvent to control the overall tumbling correlation time of a simple solute
molecule by varying the viscosity. The rigid and symmetrical molecule 1,3-
dibromoadamantane (see Fig. 5.2) was selected as a probe and the relaxation
properties of its methine carbon-13 spins were studied with the goal of
developing a motional model for this solute in the accessible temperature-
dependent viscosity range between 10 cP and 10 cP.
5.2 Experimental
5.2.1 Characterization of the system and sample preparation
Voltalef 10S (Atochem, France) is a highly viscous, non-polar, aprotic oil
consisting of polychlorotrifluoroethene (PCTFE) with 6 to 7 monomer units
(see Fig. 5.2) and a molecular mass of approximately 800 Da.
Cl-
F
I-C-
Cl
-C-I-ClI
n = 6-7
PCTFEJ3C
H
Fig 5 2 Components of the studied system 1,3-Dibromoadamantane is dissolved in
polychlorotnfluoroethylene (PCTFE)
5.2 Experimental 149
Viscosity measurements were performed using a dynamic stress rheometer
(Rheometrics) with a Couette geometry (concentric cylinders). At
temperatures above 282 K, the viscosity of PCTFE is within experimental
accuracy not dependent on the shear rate in the range from 10"2 to 102 s.
The shear viscosity at a shear rate of 1 s, given in Fig. 5.3, covers more
than two orders of magnitude from 59 cP at 344.1 K to 8920 cP at 283.3 K.
In this range a small but systematic deviation from an Arrhenius temperature
dependence is found, similar to other highly viscous liquids [50,149], with
an activation energy varying from Ea ~ 87kJmol at 283 K to
Ea ~ 55kJmol at 344 K. The measured temperature dependence is shown
in Fig. 5.3. On the other hand, the temperature dependence can be fitted
within experimental accuracy by a power law [149]:
r\{T) = c(T-TQ)q, (5.3)
where c = 7.119- 1012cPK_<?, TQ = 241.8K, and q = -5.501. Below
282 K, PCTFE becomes opaque and the viscosity dependent on the shear
rate. At 281 K, for example, viscosities of r| = 1.25 • 10 cp at a shear rate
9-1 4 -1of 10 s and rj = 3.7 10 cp at a shear rate of 10 s were measured (see
Table 5.1).
It is known that fluorinated organic compounds have exceptional abilities
to dissolve oxygen [150,151], which could cause paramagnetic relaxation of
guest molecules. Therefore, the polymeric oil was degassed under high
vacuum and handled in an argon atmosphere. 1,3-Dibromoadamantane
(Aldrich) was dissolved in PCTFE with a concentration of 85 mM. Benzene-
dg, enclosed in a coaxial capillary, was used for field-frequency locking.
150 Rotational Motion of a Solute Molecule in a Highly Viscous Liquid
104
57,
& 103
'53oo
.«2 102>
'"
280 290 300 310 320 330 340 350
Temperature T [K]
Fig. 5 3 Temperature dependence of the viscosity of PCTFE. The experimental data
points measured at a shear rate of 1 s are given for a temperature range from 283 to
344 K. The curve corresponds to the fitted power law of Eq.(5.3)
5.2.2 NMR relaxation measurements
The NMR measurements were performed on Bruker DMX-400 and AMX-
600 spectrometers, operating at 400 and 600 MHz proton resonance
frequency, respectively. Tj, NOE ,Ty, and T2 have been measured for the
methine carbons in 1,3-dibromoadamantane at temperatures of 274, 281,
290, 300, 324, and 339 K and a proton resonance frequency of 600 MHz.
Additional measurements at 400 MHz proton resonance were carried out at
a temperature of 300 K.
The 13C spin-lattice relaxation times T{ were measured by the standard
inversion-recovery method, after an initial polarization transfer from protons
by refocused INEPT and with proton decoupling during mixing and
acquisition times. Eight or more mixing times %m in the range
3(is < Tm < 5Tj were used. The inversion-recovery data were fitted by a
three-parameter mono-exponential function, employing the simplex
algorithm from the toolbox of MATLAB [152], The covariance matrix of the
fitted parameters was calculated for uncorrelated measurement errors of
5.2 Experimental 151
equal variance.
The 13C NOEs were measured after presaturation of the protons by a
series of 120° pulses with 10-ms interpulse delays, applied for at least
5 Tj [153]. The peak areas were integrated and the NOEs were calculated
as ratios with and without proton saturation. The experimental error was
estimated by repeating the measurement for some of the NOEs.
The 13C spin-spin relaxation times T2 were determined by the Carr-
Purcell-Meiboom-Gill (CPMG) pulse sequence after an initial INEPTIT 1 J
transfer. C rotating-frame relaxation times T{ were measured for a C
spin-locking field of 1.2 kHz. It proved unnecessary to eliminate the cross-
correlation between dipolar and chemical shift anisotropy relaxation by
applying pulses to the protons during the mixing time. A quantitative
estimate of the cross-correlated cross-relaxation effect, using an experiment
similar to the one of ref. [71], showed that the contribution to T2 or Tl is
below 1%, indicating a negligible anisotropy of the CH carbon shift. This is
in agreement with quantum chemical calculations of the CSA tensor based
on density functional theory (DFT) which showed that Ac < lOppm. In the
T2 and Ti measurements, the carrier frequency was set 100 Hz off-
resonance. The magnetization decays were fitted by a two-parameter mono-
exponential function. It was found that at 300 K and 600 MHz T{ is
independent of the spin-locking field strength set to 0, 0.6, 1.2, 1.8, and
2.8 kHz. This shows that no slow exchange processes with Texch > 10 |xs
occur.
Decoupling of the 19F spins of the solvent by strong rf irradiation during
the mixing time of NOESY and lH-T{ inversion-recovery experiments did
not noticeably affect relaxation, verifying that there is no measurable cross-
relaxation between the 19F spins of PCTFE and the spins of the guest
molecules. However, a small contribution of 19F to dipolar T^ and T2relaxation of C, not exceeding a few percent, cannot be excluded.
A mono-exponential relaxation behavior at all temperatures (even at 281 K
and 274 K where the solvent is very viscous and opaque) was observed. The
152 Rotational Motion of a Solute Molecule in a Highly Viscous Liquid
measured C relaxation parameters are listed in Table 5 1 The errors for T{and T2 vary between 0.5 and 6%, and the standard deviations for the
heteronuclear NOEs at 290 and 300 K were found to be 1%.
Table 5.1 Experimental C relaxation data for CH groups in 1,3-dibromoadamantane
dissolved in PCTFE and viscosity r| of the solvent PCTFE at the correspondingtemperatures.
COff/27I
[MHz]
T
[K]
11
[cP]
7",
[ms]
?2
[ms]
NOE
600
274 791 ± 16rf 162
281
125000"
50000 *
37000c
771 +9 d n±2d 166
290 3928' 731 ±14 e n±i,e 1 75 + 0.02'
300 1392* 750 ± 3 e 19611' 1.82 ± 0.02e
324 208' 1058 ±5 d 784 ± 4' 2 19
339 83* \5\5±\1d 133418'' 2 43
400 300 1392' 535 ± 3 e 188 ±2 d 191+002'
"
Viscosity measured at a shear rate of 0 01 s
bViscosity measured at a shear rate of 1 s
c
Viscosity measured at a shear rate of 10 s
Error is the standard deviation obtained from fitting the time course of the peakintegrals
eError is the standard deviation obtained by repeating the experiment
5.3 Results and Discussion
5.3.1 Test of a single-correlation-time model
The relaxation of the 13CH carbon spins is governed by the 13C-'H dipolar
interaction with the standard relations for 7^ and T2 according to Eqs.
(2.65) and (2.71), respectively, the cross relaxation rate constant
a = Tc Hand the heteronuclear NOE according to Eqs. (2.67) and
5.3 Results and Discussion 153
(2.68), respectively.
For simplicity it is often assumed that the correlation function of the
random process causing relaxation is mono-exponential with the correlation
time xc. This is justified for a rigid molecule tumbling isotropically in a
homogeneous medium (see Section 2.2.2.1) which leads to the Lorentzian(2)
spectral density function/(co) = 7L(co) of Eq. (2.36):
7(q>) = c—2. (5.4)1 + (coxc)
However, an analysis of the experimental data shows that the three 13C
parameters T{, T2, and NOE contradict each other if a Lorentzian spectral
density is assumed. This is illustrated in Fig. 5.4 for the experimental values
at 300 K and 600 MHz. The measured T, and T2 values were corrected by
subtracting the ~10% contribution of intramolecular dipolar relaxation
induced by remote protons, assuming isotropic tumbling and internuclear
distances rc//=1.08 A, rcc=1.54A, and tetrahedral bond angles as
geometric parameters. The observed Tl value may be rationalized by one of
the two correlation times xc « 60 ps or xc~ 7 ns .
Both are in contradiction
with the value xc = 400 ps deduced from the measured T2 and NOE values.
Fitting of the experimental data by a non-axially symmetric rotational
diffusion tensor and by applying the Woessner equations [154] would lead to
an anisotropy of rotational tumbling of at least two orders of magnitude. This
is, however, in contrast to the moment of inertia tensor of 1,3-
dibromoadamantane whose principal values have a ratio of 1:2.6:3.1,
rendering this interpretation unlikely.
154 Rotational Motion of a Solute Molecule in a Highly Viscous Liquid
0.001
Fig 5 4 Illustration of the failure of a single-correlation-time model The theoretical
dependencies of C T{, T2, and NOE on a single correlation time xc are plotted,
assuming rCH-\ 08 A and cow/2it= 600 MHz The experimental values for the data
at 300 K corrected for the dipolar relaxation by i emote protons, are given as solid
circles It is not possible to find a single correlation time x( for which all three
experimental values he on the corresponding theoretical curves The xc "unceitainty
range" is marked in gray
5.3.2 Distribution of correlation times
It is attempted to rationalize the experimental data by a distribution of
rotational correlation times (see Section 2 2 2 2) which may be the result of
an exchange between environments with different effective microviscosities
or rotational hindering potentials The strictly exponential time course in the
relaxation experiments suggests rapid exchange on time scales much shorter
than Tn If each environment has a characteristic correlation time T and an
occupational probability p{x[), the mean relaxation rate constants are given
by
5.3 Results and Discussion 155
(T~x% = jP^fJT^c <°K = Jp(Tc)c(Tc)rflc, (5.5)
0x c
0
with x = 1, 2, or 1 p; and the mean iVOZs is given by
In (°\(NOE)T = 1 + -2- -£-. (5.6)
Because of the linear dependence of the relaxation rate constants on 7(a)), it
is possible to express them by a mean spectral density function
(7(0))), = \p{xc) X-£—1dxc, (5.7)
which is the Fourier transform of Eq. (2.37). Based on this equation, the
relaxation data of Table 5.1 measured at 300 K for 400 and 600 MHz proton
resonance were fitted by various distribution functions p(ic).
The first models to be explored are those with n= 2 or 3 discrete
environments [155,156] leading to the mean spectral density function
"2t
<J(a»>T = Xp* ^—. (5.8)
At first, n = 2 is assumed. Prior to fitting, the experimental values were
corrected as described below Eq. (5.4). The corrected values are given in
Table 5.2. The resulting fit parameters p{ = 1 - p2, xcl, and xc2, and the
back-calculated relaxation parameters T{, T2, and NOE are included in
Table 5.2 (row A). As expected, the fit is significantly better than that for a
single correlation time. But the differences between the back-calculated
values and the corrected experimental values are still significantly larger than
the estimated measurement errors. It is remarkable that the two resulting
correlation times tc1 = 37ps and xc2 = 3.3ns differ by nearly two orders
156 Rotational Motion of a Solute Molecule in a Highly Viscous Liquid
of magnitude and indicate two vastly different environments. Environment 1
with a population of 76% seems to approximate freely moving solute
molecules, while environment 2 with a population of 24% suggests strong
association with PCTFE molecules or rotational hindrance by the PCTFE
molecules. Interestingly, for certain parameters p{ = 1 -p7, xcl, and xc2
of the bimodal model the calculated NOE parameters show an anomalous
field dependence where, in contrast to the single-correlation-time model, the
NOE may increase with increasing magnetic field strength. This contradicts
the experimental findings. In a further attempt to improve the fit for all six
relaxation parameters, an adjustable rCH bond length was used. Although2
the error of the fit (not shown) is reduced to % =53, this led to a physically
unreasonable bond length of rCH = 1.28 A.
The inclusion of a third environment (n = 3 in Eq. (5.8)) with correlation
time xc3 significantly improved the fitting quality. The fit, documented in
Table 5.2 (row B), leads to three correlation times that differ from each other
by more than an order of magnitude. Such a fitting procedure should be
considered rather as an attempt of mathematically modeling the measured
data than to imply a specific physical model. Nevertheless, it shows that the
environment experienced by the solute molecules is highly inhomogeneous,
requiring more than just two discrete correlation times.
NMR relaxation times in highly viscous liquids have previously been
successfully fitted by assuming a Cole-Davidson spectral density function
[53, 157]:
(jm\-l(
sin(pCDatan(<BTCD))
2 Pcd/2(i + (cotCD) ) )
(5.9)
with the distribution parameter (3CD, 0 < Pcz) < 1, and the effective
correlation time xCD. For (3C0= 1, the spectral density merges into the one
for a single correlation time. A simple physical interpretation of this
empirical spectral density function is not apparent. Fitting the corrected
relaxation data by the Cole-Davidson spectral density function leads to the
5.3 Results and Discussion 157
results in Table 5.2 (row C). The fitting error lies between the ones for two
and three discrete correlation times (Table 5.2, rows A and B). The very low
value pC£) = 0.094 indicates a significant deviation from a single-
environment situation. The fitted Cole-Davidson spectral density function is
compared with those of the bimodal and trimodal models in Fig. 5.5. While
the values (7(0)) are nearly identical for the three spectral density
functions, their forms differ at the other relevant frequencies, in particulary
for 300 and 750 MHz.
10
i ' I ' I '
0'9
— bimodal— trimodal
- t\• - - - Cole-Davidson :
\-\ .
\'\ -
\\.
-10:
::-^:::'
I L I , I
0 100 200 300 400 500 600 700 800
00/271 [MHz]
Fig 5 5 Comparison of three model spectral density functions fitted to the
experimental C data Tt, T2, and NOE at 300 K, measured at a>H/2it= 400 and
600 MHz Solid line, bimodal spectral density function; dashed line, trimodal spectral
density function, dotted line, Cole-Davidson spectral density function
158 Rotational Motion of a Solute Molecule in a Highly Viscous Liquid
Table 5 2 Model parameters obtained by fitting various motional models to the
experimental data and back-calculated relaxation data for 1,3-dibromoadamantane at
300 K and (0H/2n = 400 and 600 MHz
Motional
Model
(0H/2n
[MHz] [ms]
T2
[ms]
NOE Model Parameters" 24X
A Bimodal
model
400 560 202 1 81p, =0 757 ±0 002
p2 = 0 243 ± 0 002
Tcl =(36 7±0 5)ps
xc2 = (3 32 ± 0 03) ns
326
600 843 217 2 10
B Tnmodal
model
400 590 201 1 90p, =0 693 ±0 012
p2 =0 181 ±0010
p3 =0 126 ±0 012
xc1 =(169±2 2)ps
xc2 = (528 ± 44) ps
Tc3 = (5 75 ± 0 30 ns
8
600 825 217 1 83
C Cole
Davidson
model
400 560 200 172 (3CD =0 094 ±0 001
xCD = (8 90 ± 0 06) ns249
600 846 218 1 70
Experimentaldata at
400 589
±3
207
±2
191
±0 02
300 K c
600 826
+ 3
216
±1
182
±0 02
"The error limits of the model parameters are determined by a Monte Carlo
procedure consisting of 100 fits with random errors corresponding to the
experimental standard deviations added to all relaxation parameters
Fitting error
2_
Tl calc
A.
-T
(Ar, exp)
i exp)2+(T2cak-T2 eJ
+{NOELalc-NOE )2
^^
(Ar2 exp) (AW^exp)standard deviations AJ, , A7"2 and ANOE
The values are experimental measurements corrected for dipolar relaxation byremote protons (see text)
5.3 Results and Discussion 159
5.3.3 Analysis of the temperature-dependent relaxation data
The temperature-dependent relaxation data of Table 5.1 were used to study
the viscosity dependence of the dynamical behavior of the solute molecules.
At 290 K and 600 MHz a Tx minimum with T{ min= 731 ms is found. A
comparison with the theoretical value of Tl min= 224 ms for a single-
correlation-time model and rCH=l.08 A provides another indication for a
multiple-site dynamics of 1,3-dibromoadamantane. Because it is difficult to
relate the deduced correlation times to a specific physical model of the
inhomogeneous polymer environment of the solute molecules, it might be
appropriate to analyze the data at a single proton-resonance frequency of
600 MHz in terms of the bimodal model. The three model parameters
Pj = 1 - p2, xcl, and xc2 were deduced from the three measurements Tl,
T2, and NOE, again corrected for the dipolar relaxation of remote protons.
They are given in Table 5.3.
The slight discrepancy between the values at 300 K in Table 5.2 and Table
5.3 is caused by the fact that additional 400 MHz measurements are used in
the fitting procedure leading to the values in Table 5.2. Throughout the
viscosity range investigated two significantly different correlation times
result which monotonously increase for increasing viscosity. It is remarkable
that at the highest measured viscosity (281 K) the environment with the
longer correlation time is populated to 78%, whereas at the lowest measured
viscosity (339 K) the environment with the shorter correlation time
dominates with 95% population. It appears that the equilibrium is shifted
toward the "free" form of the solute at high temperature.
In Fig. 5.6, the products xc]T and xc2T are plotted in a double logarithmic
form against the viscosity r\. Based on the SED relationship of Eq. (5.2), a
linear dependence with unit slope is expected. For both components the slope
is significantly smaller than 1, following a relationship
m,
Txci = const r\ ,i= 1,2. (5.10)
A least-squares fit leads to the values /n, = 0.12 for the rapidly tumbling
160 Rotational Motion of a Solute Molecule in a Highly Viscous Liquid
population and m2 = 0 45 for the slowly tumbling population The
functional form of Eq (5 10) is in agreement with earlier findings for other
systems using NMR and dielectric relaxation measurements [158,159]
Table 5 3 Parameters of the bimodal model computed from the experimental values of
r,, T2, and NOE (corrected for dipolar relaxation by remote protons) at various
viscosities and co„/2jt= 600 MHz
T
[K] [cp]Populations
* Correlation times*
281p, =0221+0023
p2 = 0 779 ± 0 023
tcl =(74 2 + 10 7)ps
Xc2 =(7 80±0 30)ns
290 3928p, =0 505 ±0 012
p9 =0495+0012
Tc, =(359±16)ps
Tc2 = (4 79 ± 0 13) ns
300 1392p, =0 70910 003
p, =0 291 ±0 003
Tcl =(276±08)ps
ic2 = (2 84 ± 0 03) ns
324 208p, =0 909 ±0 002
p, =0 091 ±0 002
xcl =(22 5+0 5) ps
xc2 =(1 11 ±003) ns
339 83p, =0 953 ±0 003
p2 = 0 047 + 0 003
Tcl =(18 2±04)ps
tc2 = (748 + 56) ps
"
Viscosity measured at a shear rate of 1 s The viscosity at 281 K depends on the
shear rate (see Table 5 1) and was not used
bThe error limits of the model parameters are determined by a Monte Carlo
procedure consisting of 100 fits with random errors corresponding to the
experimental standard deviations added to all relaxation parameters
The violation of the SED relation implies that the motion of the solute
molecules is less influenced by the solvent viscosity than expected This is
not unreasonable for the rapidly tumbling population as the small molecules
may move with little hindrance in cavities formed by the polymenc solvent
For the derivation of the SED relation, on the other hand, it was assumed that
the solute molecules move in a homogeneous environment of very small
solvent molecules which can be approximated to be a continuum (see
5 3 Results and Discussion 161
Section 2.2.2.1) where obviously the formation of long-lived cavities is
unlikely. An Arrhemus-type approximation of the correlation times xcl and
Tc2, Tcl 2= kexp(Eal 2/(i?r)), leads to the activation energies
Eal = llkJmol'
and£a2 = 33kJmor'.
5x10
2x10
TlIcP]
11 [cP]
Fig 5 6 Viscosity dependence of xcXT for the fast and xc2T for the slow tumbling
population in the bimodal model. Only data points with shear-rate-independent
viscosities are included (see Table 5 3) The linear fits corresponds to Eq (5 10) with
the fitting parameters m,
162 Rotational Motion of a Solute Molecule in a Highly Viscous Liquid
5.4 Conclusions
This study demonstrates that 1,3-dibromoadamantane experiences PCTFE
as an inhomogeneous solvent medium The observations can formally be
modeled by a rapid exchange between at least two different environments
Similar situations have been found for solutes in other highly viscous
solvents [50,51,157] For the bimodal model, the apparent correlation times
differ by two orders of magnitude, one environment corresponding to
virtually free guest molecules, and the other environment exhibiting a strong
solute-solvent interaction The temperature and viscosity dependence of the
tumbling correlation times of the two populations is much weaker than would
be expected from the Stokes-Einstein-Debye relationship The exchange of
the guest molecules between the different environments appears to be rapid,
but is difficult to quantify based on NMR measurements alone, as no rf field
strength dispersion is observed mlj experiments The results presented in
this chapter show that the tumbling behavior of small solute molecules in
solvents such as PCTFE is not as simple as desired for shifting the rotational
correlation time in view of NMR studies of intramolecular motion
6 1 Appendix to Chapter 3 163
6 Appendix
6.1 Appendix to Chapter 3
6.1.1 Analysis of molecular dynamics trajectory
6.1.1.1 Extraction of fluctuation amplitudes from MD trajectory
Table A 1 MD reonentational fluctuation amplitudes about the pnncipal axes ea, en, e
and dihedral angles co of the peptide planes in ubiquitin The values for aa, o«, and cywere extracted from the 1 5 ns MD trajectory according to Eq (2 82) Only those 57
peptide planes which can be descnbed by a 3D GAF motion are listed
peptide
plane nooa [deg] Op [deg] oY [deg] co [deg]
a
2 4 883 4 208 9 806
3 4 282 4 133 10718 167 7 ±7 3
4 4 516 3 844 8 304 179 3 + 6 7
5 3 946 3 388 10 539 176 1 ±6 8
6 4 704 3 910 8 472 175 7 ±7 2
12 4 986 5 223 14 232 172 617 8
13 4 342 5 689 12 764 1813 + 79
14 4 855 3 916 11893 176 9 + 8 7
15 4 414 6 089 14 358 176 8 + 7 9
16 5 079 5 282 10 863 178 2±8 3
17 4 753 4 954 14 817 181 1+70
18 7 115 3 789 9 800 168 8 + 7 2
20 6 019 6 519 12 496 175 1 ±7 2
21 7 093 5 495 9 348 1806±7 6
22 4 436 4 500 10 960 181 6±7 7
23 4 465 3 064 7 765 175 0 + 6 0
24 3 887 3 837 6 946 167 9 ±60
25 4 498 3 029 7 373 1764±5 8
26 4 201 3 718 8 114 172 9±60
164 Appendix
peptide
plane noCa [de§] Op [deg] Oy [deg] ro [deg]°
27 4 279 3 231 5 973 175 5 ± 5 4
28 4 379 2 921 6 964 174 1 ±5 5
29 4 069 3 638 7 557 173 6 + 5 7
30 3 700 3 146 7 322 1709 + 5 6
31 4 291 3 670 6 997 170 7161
32 4 380 5 574 8 607 1760 + 6 3
33 5 693 4 755 10 180 179417 6
34 5 091 5 078 11 612 1865 + 8 1
39 5 006 5 985 9 305 173 2169
40 5 690 4 467 11740 1780169
41 6 345 3 994 11566 176 317 8
42 6 937 3 553 9 426 161 2168
43 4 096 4 134 9211 184 516 8
44 4 097 3 357 8 888 178 816 8
45 4 641 3 672 9717 172 616 8
46 7 776 6 985 12 565 187 217 7
48 7 992 7 574 15 645 183 9184
49 6 097 4 821 11 825 176 617 7
50 4 173 5 141 8 823 179 318 8
51 4 774 4 655 10 755 1816180
52 4 888 4 450 11588 190 1 + 8 7
55 4213 3 857 11676 174017 9
56 4 363 4 612 10 955 173 5166
57 5 216 4 229 9 408 175 116 8
58 6 199 3 650 8 772 1763 + 7 2
59 5 964 4 252 9 624 175 6 + 6 5
60 4 659 5 304 15 217 184 417 5
61 4 401 7 269 13 748 1786178
6 1 Appendix to Chapter 3 165
peptide
plane noo« tdeg] Op [deg] cy [deg] to [deg]"
62 4 294 7 840 11572 1816±68
63 6 486 4 431 9 288 1792 + 78
64 4 852 6 018 8 075 178 2167
65 4 229 6 131 14 147 1785 + 7 7
66 5 317 4 406 9 896 182 3 + 72
67 4 464 4 280 10 315 174 3 ±7 0
68 4 675 3 584 11836 1765 + 75
69 4 935 3 100 8 873 176 5 + 7 1
70 3 790 3 440 11934 1779 + 77
71 5 768 3 868 10 236 1789 + 63
"
Average value and standard deviation are given for the ensemble of 1500
snapshots
166 Appendix
6.1.1.2 Analysis of dihedral angles in ubiquitin
a)
10 20 30 40 50 60 70
Residue number i
140 -
100 -
60 * H
20
-20 -
-60
-100
-140
ion
'/$**»t10 20 30 40 50 60
Residue number i
70
Fig A 1 Average dihedral angles (a) \\rt and (b) (p, in ubiquitin as function of the
residue number i The standard deviation of the dihedral angle distribution resultingfrom the ensemble of 1500 snapshots is indicated in the figure (see bars) Glycineresidues are given as open circles The only non-glycine residues in a left-handed
conformation (cp, > 0°) are Ala 46, Asn 60, and Glu 64 (see also ref [96])
6.1 Appendix to Chapter 3 167
10 20 30 40 50 60 70
Residue number i
10 20 30 40 50
Residue number i
60 70
Fig. A.2 Correlation between dihedral angles. In Panel (a) the correlation coefficients
for the dihedral angles y, _,, <p, (filled circles) of the same peptide plane and for
y,, cp; (open circles) of the same residue i are compared with each other. In Panel (b)
the correlation coefficients for the dihedral angles \|/(_|,a>; (filled circles) and for
cpj, co, (open circles) are given. The coefficients are calculated from the ensemble of
1500 snapshots.
168 Appendix
6.1.2 Experimental results
Table A 2 N and C relaxation data of the backbone of human ubiquitin measured at
400 and 600 MHz proton resonance frequency and at a temperature of 300 K The data is
listed according to the peptide plane number For error bars see Section 3 5 1
pp
no"
15N
r,
400
MHz
[s]
15N
r,
600
MHz
[s]
15N
NOE
400
MHz
15N
NOE
600
MHz
15N
T2
600
MHz
[s]
13C,
Tx
400
MHz
[s]
13C.
T\
600
MHz
[s]
13C,
T2
400
MHz
[s]
13C.
T2
600
MHz
[s]
2 0 345 0 480 0 612 0 721 0 160 1057 1227 0 257 0 149
3 0 328 0 443 0 635 0 762 0171 1074 1248 0 228 0 131
4 0 322 0 438 0 639 0 769 0 168 1 111 1321 0 232 0 141
5 0 337 0 464 0 619 0 752 0 171 1088 1308 0217 0 131
6 0 317 0 445 0 633 0 752 0 169 1067 1263 0 213 0127
7 0 331 0455 0 595 0 746 0 174 0 975 1 188 0 225 0 139
8 0 351 0 469 0517 0 672 0 179 1043 1 170 0 247 0 143
9 0 349 0510 0 483 0 635 0195 1064 1 155 0 265 0 152
10 0 372 0 488 0 471 0 607 0 195 1 193 1306 0 294 0 179
11 0 399 0518 0 465 0 608 0 205 1 177 1321 0 286 0 179
12 0 357 0 493 0 556 0 689 0 186 1 120 1253 0 244 0 142
14 0 343 0 469 0 643 0 764 0 171 1072 1268 0 222 0 136
15 0 329 0 451 0613 0 762 0 166 1 113 1303 0 226 0 134
16 0 358 0 490 0 623 0 736 0 188 1095 1271 0 222 0130
17 0 326 0 452 0 633 0 769 0 163 1062 1236 0 219 0 132
18 0 334 0 478 0 653 0 772 0 168 1049 1 256 0 217 0 133
20 0 337 0 464 0 626 0 761 0 180 1061 1215 0 221 0 129
22 0318 0 447 0 637 0 763 0 178 1080 1306 0 203 0 127
23 0317 0 430 0 641 0 786 0 149 1002 1287 0 201 0 124
25 0 314 0 441 0 665 0 776 0 126 1050 1203 0 172 0 090
26 0 325 0 436 0 628 0 768 0 164 1098 1283 0 224 0 129
27 0315 0 429 0 645 0 793 0 164 1027 1 242 0 205 0 130
6 1 Appendix to Chapter 3 169
PP
noa
,5N
r,
400
MHz
[s]
15N
7",
600
MHz
[s]
15N
NOE
400
MHz
15N
NOE
600
MHz
15N
600
MHz
[s]
13C
r.
400
MHz
[s]
13C
T{
600
MHz
[s]
13C,
T2
400
MHz
[s]
13C.
T2
600
MHz
[s]
29 0 326 0 438 0 637 0 786 0 163 1039 1226 0 231 0132
30 0 325 0 437 0 621 0 768 0164 1053 1254 0 225 0 128
32 0 330 0 438 0 644 0 776 0164 1031 1211 0 223 0130
33 0 341 0 456 0611 0 764 0 173 1 112 1348 0 234 0135
34 0 338 0 459 0 617 0 752 0 175 1050 1260 0 216 0 134
35 0 329 0 454 0 665 0 756 0 162 1059 1250 0 231 0 144
36 0 375 0516 0 669 0 779 0178 1090 1328 0 252 0152
39 0 341 0 447 0 613 0 750 0179 1031 1 194 0216 0124
40 0 337 0 444 0 634 0 766 0 167 1 131 1312 0 230 0137
41 0 334 0 450 0 603 0 756 0 177 1 127 1328 0 237 0 139
42 0 333 0 458 0 628 0 750 0174 1059 1265 0 201 0124
43 0 332 0 465 0 639 0 758 0171 1034 1268 0 207 0130
44 0 328 0 446 0 625 0 764 0 176 1057 1272 0 210 0 120
45 0 331 0446 0 627 0 762 0 164 0 992 1222 0186 0 124
46 0 323 0 459 0 603 0 741 0 171 1036 1211 0 219 0 124
47 0 354 0 473 0618 0 738 0 178 1 112 1306 0 249 0137
48 0 349 0 466 0 645 0 750 0 171 1 165 1360 0 251 0155
49 0 372 0 484 0 567 0 691 0 189 1082 1294 0 236 0139
50 0 335 0 450 0 587 0 748 0170 1 117 1310 0 228 0133
51 0 335 0 483 0615 0 762 0171 1053 1261 0 225 0136
52 0 356 0 497 0 637 0 746 0179 1030 1 163 0118 0 063
54 0 346 0 471 0 655 0 768 0 167 1 110 1348 0 218 0 134
55 0 335 0 450 0 635 0 749 0166 1 127 1322 0 221 0131
56 0315 0 430 0 627 0 792 0 172 1019 1295 0 201 0 122
57 0 328 0 445 0 620 0 782 0 164 1043 1220 0 236 0122
58 0 321 0 436 0 673 0 787 0 161 1060 1233 0 214 0 128
170 Appendix
PP
no"
15N
r,
400
MHz
[s]
15N
r,
600
MHz
[s]
15N
NOE
400
MHz
15N
NOE
600
MHz
15N
T2
600
MHz
[s]
13C.
r,
400
MHz
[s]
13C.
Tl
600
MHz
[s]
13C,
T2
400
MHz
[s]
13C.
T2
600
MHz
[s]
59 0 335 0 446 0 602 0 762 0 179 1 171 1333 0 232 0 131
60 0 331 0 439 0 643 0 756 0 176 1 132 1343 0 226 0 137
61 0 336 0 441 0 661 0 764 0 171 1 176 1385 0 242 0 141
62 0 379 0 521 0 493 0 626 0 189 1 157 1317 0 237 0139
63 0 350 0 474 0 617 0 756 0167 1083 1256 0 230 0 137
64 0 324 0 440 0 643 0 765 0 169 1092 1268 0 222 0 129
65 0 331 0 459 0 638 0 790 0 173 1 175 1345 0 248 0 135
66 0 332 0 471 0 629 0 755 0179 1 119 1349 0 220 0 142
68 0331 0 458 0 661 0 756 0 179 1080 1289 0 221 0 131
70 0 331 0 442 0 641 0 748 0 154 1084 1222 0 226 0134
71 0 355 0 465 0 600 0 710 0 180 1020 1228 0219 0 143
74 0 540 0 623 -0 25 0 107 0 355 1 114 1043 0 474 0 281
75 0 706 0 819 -0 91 -0 35 0 540 1251 1 106 0 678 0 408
76 1 163 1242 -144 -0 92 0 804 2 021 1764 1099 0 662
aThe peptide planes are labeled and numbered by the amino acid residue that
contributes the nitrogen atom The plane of (a non proline) peptide bond i
contains the atoms N,, H, , C, ,,and O, _,
6 1 Appendix to Chapter 3 171
6.1.3 Analysis of relaxation data
6.1.3.1 Calculation of the order parameter of the 3D GAF model
2The calculation of the order parameters S„v according to Eq (2 57) is the most
expensive step when fitting the fluctuation amplitudes to the relaxation data The
MATLAB code [152] for the order parameter calculation which was compiled with an
ANSI C compiler and linked with the MATLAB C library is given below
Function Squad.m
function ret=Squad(fhetam,thetan,sigmal,sigma2,sigma3),%
% This function calculates the order parameter for a certain
% auto-correlated (thetam=thetan) or cross correlated (thetan.thetam)
% spin interaction in the peptide plane exhibiting 3D GAF motion
%
% The peptide plane frame is defined as follows
% principal axis gamma colhnear to C_alpha-C_alpha axis
% principal axis alpha in plane axis, perpendicular to gamma-axis, in direction C - > O
% principal axis beta axis perpendicular to the plane%=> right-handed coordinate system with alpha (x), beta (y), gamma (z)
%
% Input parameters
%thetam polar angles (theta,phi) of the interaction m (in rad)
%thetan polar angles (theta.phi) of the interaction n (in rad)
%sigmal fluctuation amplitude sigma_alpha (in rad)
%sigma2 fluctuation amplitude sigma_beta (in rad)
%sigma3 fluctuation amplitude sigma_gamma (in rad)
%
% Output%ret order parameter
% Wigner rotation matrix elements ace to R N Zare, Angular Momentum, Table 3 1, p 89
% The matnx elements are given for an angle pi/2
rmn=[[0 25 0 5 sqrt(3/8) 0 5 0 25],[ 0 5-0500505],
[sqrt(3/8) 0-050 sqrt(3/8)],[-0 5050-0505],
[0 25 -0 5 sqrt(3/8) -0 5 0 25]],
% List of normalized spherical harmonics (see R N Zare Angular Momentum, p 10)
%see function Ym.m below
Ymm=[Ym(-2 thetam(l),thetam(2)) Ym(-1 thetam(l),thetam(2)) Ym(0,thetam(l),thetam(2))
Ym( 1 ,fhetam( 1 ),thetam(2)) Ym(2,thetam( 1 ),fhetam(2))],
Ymn=[Ym(2,thetan( 1 ),thetan(2)) -Ym( 1 ,thetan(l),thetan(2)) Ym(0,thetan(l),thetan(2))
Ym( l,thetan(l),thetan(2)) Ym( 2,thetan(l),thetan(2))],
%%%%%%%%%%%%%%%%%%
%Calculation of the order parameter
%%%%%%%%%%%%%%%%%%
sigmalqh=0 5*sigmal*2,sigma2q=sigma2A2,
sigma3qh=0 5*sigma3A2
172 Appendix
pih=pi/2,
expsig2=zeros(size(rmn)),expsig3=expsig2,
for m=-2 2
mq=mA2,for md=-2 2
expsigl(m+3,md+3)=exp(-sigmalqh*(mq+mdA2))*exp(-j*pih*(m-md)),
expsig3(m+3,md+3)=exp(-sigma3qh*(mq+mdA2))*exp(j*alpha*(md-m)),end
end
% Loop to get the Order Parameter
Squad=0,for l=-2 2
a=exp(-sigma2q*lA2),for k=l 5
forkd=15
b=a*expsig 1 (k,kd)*rran(k,l+3)*rmn(kd,l+3),
for m=l 5
for md=l 5
Squad=Squad+b*expsig3(m,md)*rmn(m,k)*rmn(md,kd)*Ymm(m)*Ymn(md),end
end
end
end
end
ret=rea](Squad)*4*pi/5,
Function Ym.m:
function ret=Ym(m,theta,phi),%
% This function calculates the normalized spherical harmonics Ym
% of rank 2 for the angle theta and phi% ace to R N Zare, Angular Momentum, p 10
%
% Input%m index
%theta, phi set of angles (in rad)
%
% Output%ret Complex number'
ifm=0
ret=sqrt(5/(16*pi))*(3*cos(theta)A2-l),elseif (m==l)
ret=-sqrt(15/(8*pi))*sin(theta)*cos(theta)*exp(j*phi),elseif (m==-l)
ret=sqrt(15/(8*pi))*sin(theta)*cos(theta)*exp(-j*phi)elseif (m==2)
ret=sqrt(15/(32*pi))*sin(theta)A2*exp(2*j*phi),elseif (m==-2)
ret=sqrt(15/(32*pi))*sm(theta)A2*exp(-2*j*phi),else
fpnntf( 'Wrong input-index"')end
6.1 Appendix to Chapter 3 173
6.1.3.2 Fit results of 3D GAF motional model
Table A.3 Results of the least-squares fits of the 3D GAF motional model to the
experimental relaxation data of Table A.2. In addition to the three parameters aaa, Cy,
xmt for each peptide plane a global scaling parameter Xc was optimized to
Xc = 1.083 ± 0.004. Only those 45 peptide planes whose motion can be described by a
3D GAF model are considered. The error limits of the fitted parameters are determined
by a Monte Carlo procedure consisting of 60 fits with random errors of the experimental
standard deviations added to all relaxation parameters.
no °a|3 tdeg] Oy [deg] Aa/aa \n, [Ps] X?'
3 8.03 ± 1.16 15.53 ±0.97 0.71 ±0.23 11.2 5.6
4 11.45 ±0.85 11.12 ±1.48 -0.03 ±0.20 8.5 5.4
5 8.24 ±0.88 16.58 ±0.80 0.76 ±0.17 12.2 9.4
6 8.24 ±1.21 14.64 ±1.29 0.62 ± 0.26 13.4 3.4
12 7.47 ±1.23 19.97 ±0.84 1.07 ±0.25 30.1 6.2
14 6.97 ±1.28 17.71 ±0.85 1.02 ±0.27 7.7 8.2
15 9.25 ±1.10 14.56 ±1.13 0.48 ± 0.22 11.8 8.5
16 4.68 ±1.73 21.03 ±0.79 1.61 ±0.43 14.3 3.6
17 6.55 ±1.13 16.06 ± 0.92 0.98 ± 0.25 8.4 7.2
18 6.99 ±1.27 17.42 + 0.93 1.00 ±0.26 3.5 8.4
20 1.00 ±1.31 19.58 ± 0.48 2.58 ±0.40 12.0 2.4
22 6.88 ±1.40 16.03+0.96 0.92 ±0.33 9.4 18.1
26 7.61 ± 1.29 14.94 ±1.10 0.73 ±0.28 9.4 11.6
27 6.25 ±1.71 14.77 ±1.05 0.94 ± 0.46 2.0 8.4
29 5.73 ±1.59 16.18 ±0.90 1.13 ±0.41 4.2 11.4
30 7.58 ±1.28 15.24 ±1.14 0.76 ± 0.28 10.5 10.4
32 1.00 ±1.43 17.62 ±0.58 2.54 ± 0.46 7.0 12.8
33 7.87 ±1.18 16.92 ±0.96 0.83 ± 0.24 10.6 15.2
34 2.78 ±1.66 19.22 ±0.69 1.99 ±0.48 14.5 8.8
39 1.00 ±0.40 19.33 ±0.37 2.58 ±0.11 17.8 16.5
40 9.39+1.03 14.73 ±1.11 0.48 ± 0.20 8.8 16.3
174 Appendix
no °ap [deg] Oy [deg] Ao/Oa
\nt M2 b
X,
41 9 82 ± 0 86 14 8910 95 0 44 + 016 15 7 86
42 5 49 11 62 18061088 1 3010 40 126 10 8
43 6 37 ± 1 42 17 45 ± 0 98 1 1010 32 92 67
44 6 16+136 17 03 10 90 1 11 10 32 10 3 16 6
45 1 00 ± 1 53 18 1310 63 2 55 10 47 11 1 218
46 1 00 ± 1 45 18 6010 54 2 5610 44 20 0 34
48 12 7310 73 12 9311 10 002 + 0 14 119 20 3
49 5 85 ± 1 61 213410 82 141+038 27 2 15 8
50 8 62 ±1 14 15 8110 99 0 6510 23 18 8 13 3
51 4 6011 77 19051073 1531047 10 3 10 8
55 10 04 10 84 14 23 11 00 0 371017 117 16 8
56 6 2211 46 15 521108 100 + 035 20 26 2
57 3 26+183 17391081 1 77 + 0 54 70 13 5
58 3 76+178 16 151091 157+051 20 83
59 8 801101 15 9810 99 0 6410 20 13 5 19 1
60 9 34 ± 0 94 14 7910 95 0 491018 11 1 14 9
61 11 08 10 67 12 981101 0161014 58 165
63 6 55 + 131 18 4410 79 1 1310 28 118 20 5
64 7 7210 87 15 4110 86 0 75+019 86 59
65 9 371088 14 89 10 94 0 49 + 017 20 45
66 10 2710 86 15 2510 95 0 421016 10 6 44
68 7 6511 12 16 67 + 0 90 0 85 10 23 88 48
70 7 341 1 42 15 40+102 0 801031 14 5 313
71 5 4911 72 19 9310 81 1 4010 40 24 7 22 2
a
Amsotropy of peptide plane fluctuation as defined in Eq (3 2)
*The overall fitting error % of the global fit is the sum of the fitting error
2contributions % of each peptide plane i
6.1 Appendix to Chapter 3 175
25
20
a15
CCL
£10o
nwMiiiinMTnnnMnMninmTiMW
A •
V •
O A A &
a a
O a
an a O
#A
*
^O
O AA°
AAA
° °°
oo
10 20 60 7030 40 50
Peptide-plane number
Fig. A.3 Results of an alternative fitting procedure. The 45 peptide planes are divided
into three categories: "a helix", "(3 sheets", and "other". Three global scaling factors
are extracted for the three categories: )£"'= 1.070 + 0.010, X*e,,Jt=1.076 ± 0.008,
and X°c er=1.096 ±0.010. The fitted fluctuation amplitudes are labeled as follows:
helix, x
helix .~.. . . sheet, i \
sheet,,,„ ,
a„R (open squares), ay (rilled squares), aaB (open circles), aY (filled
other . . , .other
Ja(5
circles), aan (open triangles), ay (filled triangles).
176 Appendix
6.2 Appendix to Chapter 4
6.2.1 15N and 13C CSA tensors of several peptides measuredby solid state NMR
*
Table A 4 Results for the principal values, the amsotropy Acs, asymmetry r\, and
orientation of the N CSA tensors of several peptides studied by different authors using
solid state NMR
model peptide[ppm] [ppm] [ppm]
A5
[ppm]
11ft a
[deg]Ref
GlyGlyHCl -55 4 -44 8 100 2 150 3 011 21 [125]
GlyGlyHCl 51 8 -49 2 1010 1515 0 03 20 [126]
AcGlyAlaNH2 -75 1 -34 6 109 7 164 6 0 37 21 [1261
AcGlyTyrNH2 -60 8 -35 7 96 5 144 8 0 26 19 [126]
AcGlyGlyNH2 -64 4 -410 105 4 158 1 0 22 21 [126]
Boc(Gly)3OBz (m) -58 3 513 109 6 164 4 0 06 22 [119]
Boc(Gly)3OBz (t) -77 0 -30 0 107 0 160 5 0 44 24 [119]
AlaAla -54 4 -415 95 9 143 9 0 13 116 [124]
[15N] alanyl3
gramicidin A
-65 -39 104 156 0 25 12-14 [127]
[15N] leucyl4-
gramicidin A
-65 -36 101 151 0 29 12-14 [127]
AlaLeu 55 4 -42 3 97 7 146 6 0 13 17 [128]
QNH is the angle between the z principal axis of the CSA and the NH bond It
is assumed that the z principal axis lies in the peptide plane and the y principalaxis is orthogonal to the plane
6 2 Appendix to Chapter 4 177
Table A 5 Results for the pnncipal values, the anisotropy Aa, asymmetry r\, and
onentation of the 13C CSA tensors of several peptides studied by different authors using
solid state NMR
model peptide°xx
[ppm]
°yy
[ppm] [ppm]
AG
[ppm]
Tl QCOy
"
[deg]
Ref
GlyGlyHCl -74 4 -7 4 818 122 7 0 82 13 [120]
GlyGlyHCl -73 7 -7 2 80 9 1214 0 82 12 [121]
AcGlyGlyNH2 -70 2 -114 816 122 4 0 72 0 [121]
AcGlyTyrNH2* -72 5 -2 3 74 8 1122 0 94 6* [121]
AcGlyAlaNH2 -69 7 -12 6 82 3 123 5 0 69 2 [121]
GlyAla -71 -13 84 126 0 0 69 10 [122]
ValGlyAla -75 -5 80 120 0 0 88 8 [122]
AlaAla -74 0 -0 9 74 9 1124 0 98 4 [124]
["Cjl-alanyb;gramicidin A
-73 4 -7 3 80 7 121 1 0 82 0 [123]
cyclofpCJ-Val-Pro-Gly-Val-Gly}
-60 6 -9 7 70 3 105 5 0 73 4 [123]
"The angle 6C0 is the angle between the CO bond and the y pnncipal axis of
the CSA tensor The z principal axis is assumed to he orthogonal to the peptide
plane*Please mind that the onentations of the x and z pnncipal axes are permutedwhen compared with the results from the other solid state NMR studies
178 Appendix
6.2.2 Orientation of the calculated CSA tensors
Table A 6 Orientation of the principal axes (in polar coordinates) of the calculated
"global frame averaged" 15N CSA tensors (see Table 4 1) of peptide planes 10, 30, and
43 of ubiquitin with respect to the peptide plane frame ea, ea, e of the corresponding
equilibnum peptide plane For comparison, the polar coordinates of some dipolar
directors in this peptide plane frame are given
interaction(9,tp)
peptide plane 10
(e,q»
peptide plane 30
0,9)
peptide plane 43
15Naxx (40 8, 149 1) (310,155 3) (32 2, 165 9)
15N oyy (107 8,80 9) (100 9,83 9) (98 0, 88 8)
15N0ZZ (125 3, 184 1) (118 7, 179 9) (1210, 183 6)
NHN (103 6, 180 9) (102 1, 177 1) (1017,179 6)
C'N (42 0, 180 4) (38 8, 179 8) (42 8,180 2)
NCa (14 6,13) (115,359 0) (17 3,0 6)
Table A 7 Orientation of the principal axes (in polar coordinates) of the calculated
"global frame averaged" 13C CSA tensors (see Table 4 2) of peptide planes 10, 30, and
43 of ubiquitin with respect to the peptide plane frame ea, e«, ey of the corresponding
equilibnum peptide plane For comparison, the polar coordinates of some dipolar
directors in this peptide plane frame are given
interaction(e,q>)
peptide plane 10
(6,9)
peptide plane 30
(6,9)
peptide plane 43
13C'<TXX (93 0, 83 9) (94 8, 85 1) (89 5, 88 0)
13C ayy (90 7, 353 8) (76 9, 356 3) (79 0, 357 0)
13C azz (3 1,70 7) (14 0,155 4) (110,180 6)
CO (817,358 6) (79 6, 0 4) (78 5, 359 0)
C'N (42 0, 180 4) (38 8, 179 8) (42 8, 180 2)
C'C (160 4, 179 6) (159 1,180 2) (160 8, 179 8)
References 179
References
1 T. E. Creighton, Proteins, structures and molecularproperties, W. H.
Freeman and Company, New York (1993).
2 L. Stryer, Biochemistry, W. H. Freeman and Company, New York
(1988).
3 K. C. Holmes and D. M. Blow, The use ofX-ray diffraction in the study
ofprotein and nucleic acid structure, Wiley Interscience (1995).
4 J. P. Glusker, M. Lewis, and M. Rossi, Crystal structure analysis for
chemists and biologists, VCH Publ., New York (1994).
5 J. D. Dunitz, X-ray analysis and the structure of organic molecules,
VCH, Basel (1995).
6 G. E. Bacon, Neutron diffraction, Oxford University Press, Oxford
(1975).
7 R. R. Ernst, G. Bodenhausen, and A. Wokaun, Principles of nuclear
magnetic resonance in one and two dimensions, Clarendon Press,
Oxford (1987).
8 K. Wuthrich, NMR ofproteins and nucleic acids, Wiley, New York
(1986).
9 A. Abragam, Principles of nuclear magnetism, Clarendon Press,
Oxford (1961).
10 R. Tycko (Ed.), NMR probes of molecular dynamics, Kluwer,
Dordrecht (1994).
11 A. G. Palmer III, J. Williams, and A. McDermott, J. Phys. Chem. 100,
13293 (1996).
12 D. E. Woessner, /. Chem. Phys. 36, 1 (1962).
180 References
13 J. N. S. Evans, Biomolecular NMR spectroscopy, Oxford University
Press, Oxford (1995).
14 G. Lipari and A. Szabo, J. Am. Chem. Soc. 104, 4546 (1982).
15 G. Lipari and A. Szabo, J. Am. Chem. Soc. 104,4559 (1982).
16 S. F. Lienin, T. Bremi, B. Brutscher, R. Briischweiler, and R. R. Ernst,
J. Am. Chem. Soc. 120, 9870 (1998).
17 B. T. M. Willis and A. W. Pryor, Thermal vibrations in
crystallography, Cambridge University Press, London (1975).
18 C. L. Brooks III, M. Karplus, and B. M. Pettitt, Proteins: A theoretical
perspective ofdynamics, structures, and thermodynamics, John Wiley
and Sons, New York (1987).
19 W. F. van Gunsteren, H. J. C. Berendsen, Angew. Chem. Int. Ed. Engl.
29, 992 (1990).
20 T. Bremi and R. Briischweiler, J. Am. Chem. Soc. 119, 6672 (1997).
21 H. A. Havel (Ed.), Spectroscopic methods for determining protein
structure in solution, VCH, Weinheim (1996).
22 M. Mehring, Principles of high resolution NMR in solids, Springer
Verlag, Berlin (1983).
23 R. G. Parr and E. Yang, Density functional theory of atoms and
molecules, Oxford University Press, Oxford (1989).
24 S. F. Lienin, R. Briischweiler, and R. R. Ernst, J. Magn. Reson. 131,
184(1998).
25 A. Messiah, Quantenmechanik, Walter de Gruyter, Berlin (1991).
26 H. Primas and U. Muller-Herold, Elementare Quantenchemie,
Teubner Verlag, Stuttgart (1990).
References 181
27 C. M. Schoenenberger, Ph. D. Thesis No. 8652, ETH Zurich (1988).
28 R. K. Wangsness and F. Bloch, Phys Rev. 89, 728 (1953).
29 F. Bloch, Phys. Rev. 102,104 (1956).
30 A. G. Redfield, IBM J. Res. Develop. 1, 19 (1957).
31 A. G. Redfield, Adv. Mag. Reson. 1, 1 (1965).
32 T. O. Levante and R. R. Ernst, Chem. Phys. Lett. 241, 73 (1995).
33 N. R. Skrynnikov, S. F. Lienin, R. Briischweiler, and R. R. Ernst, J.
Chem. Phys. 108, 7662 (1998).
34 R. Zare, Angular Momentum, Wiley Interscience, New York (1988).
35 H. W. Spiess, In NMR basic principles and progress, Vol. 15 (Eds.: P.
Diehl, E. Fluck, and E. Kosfeld), Springer Verlag, Berlin (1978).
36 M. Goldman, J. Magn. Reson. 60,437 (1984).
37 T. Bremi, R. Briischweiler, and R. R. Ernst, /. Am. Chem. Soc. 119,
4272 (1997).
38 N. G. van Kampen, Stochastic processes in physics and chemisry,
North-Holland, Amsterdam (1992).
39 P. Debye, Polar Molecules, Dover, New York (1929).
40 P. S. Hubbard, Phys. Rev. 131,1155 (1963).
41 D. Kivelson, Faraday Symp. Chem. Soc. 11, 7 (1977).
42 G. Stokes, Trans. Cambridge Philos. Soc. 9, 5 (1856).
43 C. M. Hu and R. Zwanzig, J. Chem. Phys. 60,4354 (1974).
44 L. D. Favro, Phys. Rev. 119, 53 (1960).
45 J. H. Freed, J. Chem. Phys. 41, 2077 (1944).
182 References
46 E. N. Ivanov, Soviet. Phys. JETP 18, 1041 (1964).
47 W. T. Huntress, Jr., J. Chem. Phys. 48, 3524 (1968).
48 W. T. Huntress, Jr., Adv. Magn. Res. 4,1 (1970).
49 R. Bruschweiler and D. A. Case, Progr. in NMR Spectr. 26,27 (1994).
50 E. Rossler and H. Sillescu, In Materials science and technology, Vol.
9: Glasses and amorphous materials (Ed.: J. Zarzycki), VCH,
Weinheim(1991).
51 C.J. F. Bottcher and P. Bordewijk, Theory ofelectric polarization, Vol.
2: Dielectrics in time-dependent fields, Elsevier, Amsterdam (1978).
52 M. D. Ediger, C. A. Angell, and S. R. Nagel, J. Phys. Chem. 100,
13200 (1996).
53 P. A. Beckmann, Phys. Rep. Ill (3), 85 (1988).
54 W. Kauzmann, Rev. Mod. Phys. 14, 12 (1942).
55 K. L. Ngai and R. W. Rendell, In Relaxation in complex systems and
related topics (Ed.: I. A. Campbell and C. Giovannella), Plenum Press,
New York (1990).
56 H. Sillescu, J. Chem. Phys. 104,4877 (1996).
57 R. Bruschweiler, J. Am. Chem. Soc. 114, 5341 (1992).
58 R. Bruschweiler and D. A. Case, Phys. Rev. Lett. 72, 940 (1994).
59 R. Bruschweiler, / Chem. Phys. 102, 3396 (1995).
60 R. M. Levy and M. Karplus, Biopolymers 18, 2465 (1979).
61 B. Roux and M. Karplus, Biophys. J. 53, 297 (1988).
62 T. Bremi, Ph. D. Thesis No. 12240, ETH Zurich (1997).
63 R. Bruschweiler and P. E. Wright, J. Am. Chem. Soc. 116,8426 (1994).
References 183
64 B. Brutscher, T. Bremi, N. R. Skrynnikov, R. Bruschweiler, and R. R.
Ernst, J. Magn. Reson. 130, 346 (1998).
65 E. R. Henry and A. Szabo, J. Chem. Phys. 82,4753 (1985).
66 D. Wallach, J. Chem. Phys. 47, 5258 (1967).
67 S. Chandrasekhar, Rev. Mod. Phys. 15, 1 (1943).
68 A. Szabo, /. Chem. Phys. 37, 150 (1984).
69 L. G. Werbelow and D. M. Grant, Adv. Magn. Reson. 49, 107 (1982).
70 T. Bremi, M. Ernst, and R. R. Ernst, J. Phys. Chem. 98, 9322 (1994).
71 N. Tjandra, A. Szabo, and A. Bax, J. Am. Chem. Soc. 118, 6986
(1996).
72 D. Yang, R. Konrat, and L. E. Kay, J. Am. Chem. Soc. 119, 11938
(1997).
73 B. Brutscher, R. Bruschweiler, and R. R. Ernst, Biochemistry 36,
13043 (1997).
74 R. Bruschweiler and R. R. Ernst, J. Chem. Phys. 96,1758 (1992).
75 H. M. McConnell, J. Chem. Phys. 28,430 (1958).
76 C. Deverell, R. E. Morgan, and J. H. Strange, Mol. Phys. 18, 553
(1970).
77 R. B. Brooks, R. E. Bruccoleri, B. D. Olafson, D. J. States, S.
Swaminathan, and M. Karplus, J. Comput. Chem. 4,187 (1983).
78 A. D. MacKerell, Jr., D. Bashford, M. Bellot, R. L. Dunbrack, M. J.
Field, S. Fischer, J. Gao, H. Guo, D. Joseph, S. Ha, L. Kuchnir, K.
Kuczera, F. T. K. Lau, C. Mattos, S. Michnick, D. T. Nguyen, T. Ngo,
B. Prodhom, B. Roux, B. Schlenkrich, J. Smith, R. Stote, J. Straub, J.
Wiorkiewicz-Kuczera, and M. Karplus, Biophys. J. 61, A134 (1992).
184 References
79 W. F. van Gunsteren and H. J. C. Berendsen, Groningen molecular
simulation package and manual (GROMOS), Biomos, University of
Groningen, Groningen (1987).
80 W.F. van Gunsteren, S. R. Billeter, A. A. Eising, P. H. Hunenberger, P.
Kriiger, A. E. Mark, W. R. P. Scott, and I. G. Tironi, Biomolecular
simulation: The GROMOS96 manual and user guide, vdf
Hochschulverlag AG of ETH Zurich and Biomos b.v., Zurich,
Groningen, 1996.
81 H. Le, J. G. Pearson, A. C. de Dios, and E. Oldfield, J. Am. Chem. Soc.
117, 3800 (1995).
82 T. B. Woolf, V. G. Malkin, O. L. Malkina, D. R. Salahub, and B. Roux,
Chem. Phys. Lett. 239,186 (1995).
83 A. C. de Dios, Prog. NMR Spectrosc. 29, 229 (1996).
84 D. Sitkoff and D. A. Case, J. Am. Chem. Soc. 119, 12262 (1997).
85 D. Sitkoff and D. A. Case, Prog. NMR Spectrosc. 32, 165 (1998).
86 T. Ziegler, Chem. Rev. 91, 651 (1991).
87 S. F. Lienin and R. Bruschweiler, unpublished work.
88 R. Bruschweiler and D. A. Case, J. Am. Chem. Soc. 116,11199 (1994).
89 J. Engelke and H. Riiterjans, J. Biomol. NMR 9, 63 (1997).
90 P. Allard and T. Hard, J. Magn. Reson. 126, 48 (1997).
91 K. T. Dayie and G. Wagner, J. Magn. Reson. Series B109,105 (1995).
92 K. T. Dayie and G. Wagner, J. Am. Chem. Soc. 119, 7797 (1997).
93 M. W. F. Fischer, L. Zeng, Y. Pang, W. Hu, A. Majumdar, and E. R. P.
Zuiderweg, J. Am. Chem. Soc. 119, 12629 (1997).
References 185
94 K. D. Wilkinson, In Heat shock proteins in the nervous system (Eds.:
R. J. Mayer and I. R. Brown), Academic Press, London (1994).
95 K. D. Wilkinson, In Ubiquitin (Ed.: M. Rechsteiner), Plenum Press,
New York (1988).
96 S. Vijay-Kumar, C. E. Bugg, and W. J. Cook, J. Mol. Biol. 194, 531
(1987).
97 P. J. Kraulis, /. Appl. Crystallogr. 24, 946 (1991).
98 D. L. Di Stefano and A. J. Wand, Biochemistry 26,7272 (1987).
99 P. L. Weber, S. C. Brown, and L. Mueller, Biochemistry 26, 7282
(1987).
100 P. L. Weber, D. J. Ecker, J. Marsh, S. T. Crooke, and L. Mueller, Trans.
Am. Cryst. Assoc. 24, 91 (1988).
101 R. E. Lenkinsiki, D. M. Chen, J. D. Glickson, and G. Goldstein,
Biochim. Biophys. Acta 494, 126 (1977).
102 M. S. Briggs and H. Roder, Proc. Natl. Acad. Sci. U. S. A. 89, 2017
(1992).
103 A. C. Wang, S. Grzesiek, R. Tschudin, P. J. Lodi, and A. Bax, J.
Biomol. NMR 5, 376 (1995).
104 A. J. Wand, J. L. Urbauer, R. P. McEvoy, and R. J. Bieber,
Biochemistry 35, 6116 (1996).
105 D. M. Schneider, M. J. Dellwo, and A. J. Wand, Biochemistry 31,3645
(1992).
106 N. Tjandra, S. E. Feller, R. W Pastor, and A. Bax, J. Am. Chem. Soc.
117, 12562 (1995).
107 M. Ottiger, N. Tjandra, and A. Bax, /. Am. Chem. Soc. 119, 9825
186 References
(1997).
108 J. A. Braatz, M. D. Paulsen, and R. L. Ornstein, /. Biomol. Struct.
Dynam. 9, 935 (1992).
109 R. Abseher, S. Liidemann, H. Schreiber, and O. Steinhauser, J. Mol.
Biol. 249, 604 (1995).
110 V. Sklenar, M. Piotto, R. Leppik, and V. Saudek, J. Magn. Reson.
Series A 102, 241 (1993).
111 S. Grzesiek and A. Bax, J. Am. Chem. Soc. 115, 12593 (1993).
112 A. Bax and S. Pochapsky, J. Magn. Reson. 99, 638 (1992).
113 D. Marion, M. Ikura, R. Tschudin, and A. Bax, /. Magn. Reson. 85,
393 (1989).
114 N. J. Skelton, A. G. Palmer III, M. Akke, J. Kordel, M. Ranee, and W.
J. Chazin, J. Magn. Reson. Series B 102, 253 (1993).
115 L. Zeng, M. W. F. Fischer, and E. R. P. Zuiderweg, J. Biomol. NMR 7,
157 (1996).
116 G. Zhu and A. Bax, /. Magn. Reson. 90,405 (1990).
117 J. P. Ryckaert, G. Cicotti, and H. J. C. Berendsen, J. Comput. Phys. 23,
327 (1977).
118 LA. Goichuk, V. V. Kukhtin, and E. G. Petrov, J. Biol. Phys. 17, 95
(1989).
119 Y. Hiyama, C. Niu, J. V. Silverton, A. Bavoso, and D. A. Torchia, /.
Am. Chem. Soc. 110, 2378 (1988).
120 R. E. Stark, L. W. Jelinski, D. J. Ruben, D. A. Torchia, and R. G.
Griffin, J. Magn. Reson. 55, 266 (1983).
121 T. G. Oas, C. J. Hartzell, T. J. McMahon, G. P. Drobny, and F. W.
References 187
Dahlquist, J. Am. Chem. Soc. 109, 5956 (1987).
122 F. Separovic, R. Smith, C. S. Yannoni, andB. A. Cornell, J. Am. Chem.
Soc. 112, 8324 (1990).
123 Q. Teng, M. Iqbal, and T. A. Cross, J. Am. Chem. Soc. 114, 5312
(1992).
124 C. J. Hartzell, M. Whitfield, T. G. Oas, and G. P. Drobny, /. Am. Chem.
Soc. 109, 5966 (1987).
125 G. S. Harbison, L. W. Jelinski, R. E. Stark, D. A. Torchia, J. Herzfeld,
and R. G. Griffin, J. Magn. Reson. 60, 79 (1984).
126 T. G. Oas, C. J. Hartzell, F. W. Dahlquist, and G. P. Drobny, J. Am.
Chem. Soc. 109, 5962 (1987).
127 Q. Teng and T. A. Cross, /. Magn. Reson. 85,439 (1989).
128 C. H. Wu, A. Ramamoorthy, L. M. Gierasch, and S. J. Opella, J. Am.
Chem. Soc. 117, 6148 (1995).
129 N. Tjandra, P. Wingfield, S. Stahl, and A. Bax, J. Biomol. NMR 8, 273
(1996).
130 C. Scheurer, N. R. Skrynnikov, S. F. Lienin, S. Strauss, R.
Briischweiler, and R. R. Ernst, submitted to / Am. Chem. Soc.
131 V. G. Malkin, O. L. Malkina, M. E. Casida, and D. R. Salahub, J. Am.
Chem. Soc. 116, 5898 (1994).
132 V. G. Malkin, O. L. Malkina, L. A. Eriksson, and D. R. Salahub, In
Modern densityJunctional theory: A toolfor chemists, Vol. 2 (Eds.: J.
M. Seminario and P. Politzer), p 77, Elsevier, Amsterdam (1995).
133 V. G. Malkin, O. L. Malkina, and D. R. Salahub, Chem. Phys. Lett.
261, 335 (1996).
188 References
134 D. R. Salahub, R. Foumier, P. Mlynarski, I. Papai, A. St-Amant, and J.
Ushio, In Density functional methods in chemistry (Eds.: A.
Labanowski and J. Andzelm), Springer Press, New York (1991).
135 A. St-Amant and D. R. Salahub, Chem. Phys. Lett. 169, 387 (1990).
136 N. F. Ramsey, Phys. Rev. 78, 699 (1950).
137 S. Foster and S. Boys, Rev. Mod. Phys. 32, 303 (1960).
138 J. P. Perdew and Y. Wang, Phys. Rev. B 45,13244 (1992).
139 W. Kutzelnigg, U. Fleischer, and M. Schindler, NMR basis principles
and progress 23,165 (1991).
140 N. Godbout, D. R. Salahub, J. Andzelm, and E. Wimmer, Can. J.
Chem. 70, 560 (1992).
141 C. Sosa, J. Andzelm, B. C. Elkin, E. Wimmer, K. D. Dobbs, and D. A.
Dixon, /. Phys. Chem. 96,6630 (1992).
142 R. R. Ernst, M. J. Blackledge, T. Bremi, R. Briischweiler, M. Ernst, C.
Griesinger, Z. L. Madi, J. W. Peng, J. M. Schmidt, and P. Xu, In NMR
as a structural tool for macromolecules: Current status and future
directions (Eds.: B. D. N. Rao and M. D. Kemple), Plenum Press, New
York (1996).
143 N. Bloembergen, E. M. Purcell, and R. V. Pound, Phys. Rev. 73, 679
(1948).
144 M. P. Williamson and D. H. Williams, /. Chem. Soc, Chem. Commun.,
165 (1981).
145 E. Hyde, J. R. Kalman, D. H. Williams, D. G. Reid, and R. K. Olsen,
/ Chem. Soc, Perkin Trans. 1, 1041 (1982).
146 D. Neuhaus and J. Keeler, J. Magn. Reson. 68, 568 (1986).
References 189
147 M. P. Williamson and D. Neuhaus, J. Magn. Reson. 72, 369 (1987).
148 L. A. Luck and C. R. Landis, Organometallics 11, 1003 (1992).
149 E. Rossler, Ber. Bunsenges. Phys. Chem. 94, 392 (1990).
150 J. J. Delpuech, M. A. Hamza, G. Serratrice, and M. J. Stebe, J. Chem.
Phys. 70, 2680 (1979).
151 M. A. Hamza, G. Serratrice, M. J. Stebe, and J. J. Delpuech, J. Magn.
Reson. 42, 227 (1981).
152 MATLAB Reference Guide, The Math Works Inc., Natick, MA
(1995).
153 L. E. Kay, D. A. Torchia, and A. Bax, Biochemistry 28, 8972 (1989).
154 D. E. Woessner, J. Chem. Phys. 37, 647 (1962).
155 D. Beckert and H. Pfeifer, Ann. Phys. 16,262 (1965).
156 H. Sillescu, J. Chem. Phys. 54, 2110 (1971).
157 E. Rossler, J. Tauchert, and P. Eiermann, / Phys. Chem. 98, 8173
(1994).
158 T. K. Hitchens and R. G. Bryant, J. Phys. Chem. 99, 5612 (1995).
159 O. F. Kalman and C. P. Smyth, / Am. Chem. Soc. 82, 783 (1959).
190 References
191
Acknowledgements
During three fascinating years at the ETH Zurich I had the opportunity
to collaborate with excellent scientists and to meet numerous interesting
people who supported me with their advice, knowledge, and friendship.
They all created a very special and exciting atmosphere which will be
missed. I would like to express my vivid thanks to
- Prof. Dr. Richard R. Ernst for providing excellent research conditions,
for teaching me how to think thoroughly about scientific problems, for his
very inspiring guidance in many discussions and his valuable advice in
various situations. His continuous support of my work has been very
motivating for me.
- Prof. Dr. Rafael Briischweiler for introducing me to the exciting field
of NMR, for sharing his deep understanding of science with me, for his
most stimulating advice in numerous scientific discussions and other
questions of life, and for the very pleasant atmosphere during our
collaboration over the last years.
- Dr. Tobias Bremi for introducing me to many different techniques, for
the exciting collaboration in the 3D GAF project, for being a great office-
mate in F34, and for being a friend.
- Dr. Christoph Scheurer for providing the DFT calculations on
ubiquitin, for first-class computer management, and for his friendship
during the last months of my work.
- Dr. Nikolai Skrynnikov for the fruitful collaboration in different
projects and for always helping when help was needed.
- Dr. Bernhard Brutscher for introducing me to advanced experimental
NMR techniques.
- Dr. Thomas Schulte-Herbriiggen for his support in the maintenance of
the AMX-600 and for valuable discussions about many questions of life.
192
- all other members of the group of Prof. Ernst for creating a very
stimulating and friendly atmosphere: Ute Drexler, Prof. Dr. Edward B.
Fel'dman, Dr. Zhehong Gan, Dr. Sabine Hediger, Dr. Hongbiao Le, Dr. Tilo
Levante, Dr. Zoltan Madi, PD Dr. Rolf Meyer, Dr. Pierre Robyr, Dr. Martin
Schick, Patrick Sommer, Dr. Suzana Straus, Dr. Marco Tomaseili, Dr.
Marcel Utz, and Dr. Rico Wiedenbruch.
- Dr. Serge Boentges, Dr. Gerald Hinze, and Dr. Bettina Wolff for their
technical support and helpful discussions in the "viscosity project".
- Prof. Dr. Beat H. Meier and Dr. Matthias Ernst for their kind support
during the last months of my work.
- our secretary Irene Miiller and many other people in the LPC for
providing an excellent infrastructure.
- Dr. Alexander Ernst, Thomas Gilbert, Dr. Felix Graf, and Dr. Michael
Wilier for many interesting discussions and their friendship during the last
years.
- my parents for their constant encouragement and support which made
my stay in Zurich possible.
193
Curriculum Vitae
Personal data
Name Stephan Frank Lienin
Date of birth March 23,1970
Place of birth Rheinfelden (Baden), Germany
Citizenship German
Parents Rosemarie and Heinz Lienin
Education
1976-1980 Primary school in Lorrach (Germany).1980-1989 Hans-Thoma-Gymnasium, Lorrach.
1989 Abitur
Studies
1990-1992
1992-1995
Aug 1995
Sept 1998
Undergraduate studies in chemistry at the university of
Mainz (Germany). Best leaving exam (Vordiplom) of the
year.
Undergraduate studies in chemistry at the EidgenossischeTechnische Hochschule (ETH) in Zurich. Degree in
chemistry in 1995. Diploma thesis in nuclear magneticresonance in the group of Prof. Dr. Richard R. Ernst.
Graduate studies in physical chemistry in the research
group of Prof. Dr. Richard R. Ernst at ETH Zurich.
Professional and Teaching Experience
1995-1998 Teaching assistant for various lectures at the Laboratoryof Physical Chemistry.
1996-1998 Responsible for maintenance of a Bruker AMX-600
high-resolution NMR spectrometer.
Scholarships
1992-1995 Studienstiftung des Deutschen Volkes
1992-1995 Deutscher Akademischer Austauschdienst (DAAD)
194