Glycyl Cα chemical shielding in tripeptides: Measurement by solid-state NMR and correlation with X-ray structure and theory

Article abstract of DOI:10.1021/ja026700g

We report here ¹³C^α chemical shielding parameters for central Gly residues in tripeptides adopting α-helix, β-strand, polyglycine II, and fully extended 2° structures. To assess experimental uncertainties in the shielding parameters

Full text of DOI:10.1021/ja026700g

Published on Web 09/13/2002
Glycyl Cr Chemical Shielding in Tripeptides: Measurement by
Solid-State NMR and Correlation with X-ray Structure and
Theory
Eduard Y. Chekmenev, Ray Z. Xu, Mark S. Mashuta, and Richard J. Wittebort*
Contribution from the Department of Chemistry, UniVersity of LouisVille,
LouisVille, Kentucky 40292
Received April 26, 2002
Abstract: We report here 13CR chemical shielding parameters for central Gly residues in tripeptides adopting
R-helix, â-strand, polyglycine II, and fully extended 2° structures. To assess experimental uncertainties in
the shielding parameters and the effects of 14N-13CR or 15N-13CR dipolar coupling, stationary and magic
angle spinning (MAS) spectra with and without 15N decoupling were obtained from natural abundance and
double-labeled samples containing [2-13C, 15N]Gly. We find that accurate (<1 ppm uncertainty) shielding
parameters are measured with good sensitivity and resolution in 15N decoupled 1D or 2D MAS spectra of
double-labeled samples. Compared to variations of isotropic shifts with peptide angles, those of 13CR shielding
anisotropy and asymmetry are greater. Trends relating shielding parameters to the 2° structure are apparent,
and the correlation of the experimental values with unscaled ab initio shielding calculations has an rms
error of 3 ppm. Using the experimental data and the ab initio shielding values, the empirical trends relating
the 2° structure to shielding are extended to the larger range of torsion angles found in proteins.
compared with calculations.8 The correlation between theory
and experiment is good if the ab initio calculations are scaled.5
Introduction
Chemical shielding is at the heart of most NMR experiments.
At the most rudimentary level, it is the source of resolution
which separates lines from chemically different groups. In turn,
the correlation of isotropic shifts with chemical connectivity
and environment provides numerous insights about chemical
properties. With the development of ab initio quantum chem-
istry, empirical correlations are augmented by a direct com-
parison of experiment with calculation. For example, the large
database of 13CR isotropic chemical shifts from solution NMR
protein structures indicates a good correlation with the protein
2° structure,1 and calculations suggest that 13CR shielding tensor
principal components are largely dependent on the 2° structure,
that is, backbone torsion angles (φ,ψ).2,3 On the basis of this
idea, a simple strategy for determining 2° structure in randomly
oriented polypeptides has been described.4 Measured shielding
tensor principal components for Ala, Val, and Leu residues in
small peptides of known structure5-7 have been determined and
Development of this approach and experimental strategies for
exploiting shielding anisotropies,9 thus, requires an additional
experimental determination of 13CR shielding tensors in peptides
of known structure. The modest experimental database of known
13CR shielding tensors has been recently summarized6 and
includes no systematic study of glycyl residues.
Herein, we report such a study of tripeptides containing
central glycyl residues. Glycine contains no side chain, eliminat-
ing any possible effects of side-chain conformation, and it is a
frequently occurring residue in proteins (6.8% of residues in
known proteins). Tripeptides studied here have central Gly
residues with torsion angles characteristic of R-helices, â-strands,
the polyproline/polglycine II collagen structure, and fully
extended conformations. We discuss the variation of the
shielding tensor parameters (δ11 - δ33) and (δ22 - δ33) with
torsion angles and the correlation with published calculations.
A noteworthy feature of these results is that the shielding
parameters vary over larger ranges than the isotropic shifts. This
greater sensitivity is of practical utility in solid-state NMR,
where the resolution of the experiment is substantially less than
in solution NMR. In addition, it is less demanding of the
theoretical shielding calculation.
* Address correspondence to this author. E-mail: rjwitt01@
athena.louisville.edu.
(1) Wishart, D. S.; Sykes, B. D. J. Biomol. NMR 1994, 4(2), 171-180.
(2) De Dios, A. C.; Pearson, J. G.; Oldfield, E. Science 1993, 260, 1491-
1496.
(3) De Dios, A. C.; Oldfield, E. J. Am. Chem. Soc. 1994, 116, 11485-11488.
(4) Heller, J.; Laws, D. D.; Tomaselli, M.; King, D. S.; Wemmer, D. E.; Pines,
A.; Havlin, R. H.; Oldfield, E. J. Am. Chem. Soc. 1997, 119, 7827-7831.
(5) Havlin, R. H.; Laws, D. D.; Bitter, H.-M. L.; Sanders, L. K.; Sun, J. S.;
Wemmer, A.; Pines, A.; Oldfield, E. J. Am. Chem. Soc. 2001, 123, 10362-
10369.
Several experimental questions regarding accurate measure-
ments of (δ11 - δ33) and (δ22 - δ33) under conditions
appropriate for examining larger biomolecules are explicitly
(6) Wei, Y.; Lee, D. K.; Ramamoorthy, A. J. Am. Chem. Soc. 2001, 123, 6118-
6126.
(7) Havlin, R. H.; Le, H.; Laws, D.; deDios, A. C.; Oldfield, E. J. Am. Chem.
Soc. 1997, 119, 11951-11958.
(8) Yao, X.; Hong, M.; J. Am. Chem. Soc. 2002, 124, 2730-2738.
(9) Hong, M. J. Am. Chem. Soc. 2000, 122, 3762-3770.
9
11894
J. AM. CHEM. SOC. 2002, 124, 11894-11899
10.1021/ja026700g CCC: $22.00 © 2002 American Chemical Society

R
Glycyl C Chemical Shielding in Tripeptides
A R T I C L E S
addressed by using triple resonance (1H/13C/15N) spectroscopy
of stationary and spinning samples with double labeling ([2-
X-ray Crystallography. The unit cell dimensions and space groups
were determined for all of the peptides studied here by X-ray
crystallography and conformed to the published values. The complete
X-ray structure of one GGG polymorph with cell dimensions and space
group equivalent to those of CSD (Cambridge structural database) code
TGLYCY10 was determined using a Bruker SMART APEX CCD area
detector system at 100 K and standard software.22-24 A thin, colorless
needle (0.03 × 0.04 × 0.31 mm3) was determined to be a single crystal.
For all 3516 unique reflections (R(int) ) 0.032), the final anisotropic
full-matrix least-squares refinement on F2 for 323 variables converged
at R1 ) 0.046 and wR2 ) 0.093 with a GOF of 1.05.
1
13C, 15N]Gly) or at natural abundance. Comparison of H and
15N decoupled stationary and 1-D MAS spectra allows us to
assess the accuracy of determining principal components from
spinning sideband intensities. The effect of 14N coupling10 to
CR is studied by comparing 15N decoupled spectra of double-
labeled samples with spectra at natural abundance. The latter
are sufficiently crowded that a 2-D technique placing isotropic
shifts and spinning sidebands in separate dimensions is used.
When given the 1.1% natural abundance of 13C, successful
application of this technique suggests the feasibility of perform-
ing these experiments on proteins containing a limited number
of labeled sites.
NMR Spectra. 13C NMR spectra were obtained on a home-built
11.7-T instrument (13C Larmor frequency of 124.59 MHz) with 5-mm
(1H/13C) double-resonance or 4-mm (1H/13C/15N) triple-resonance MAS
probes built on a design previously described.25 Sample spinning speeds
were controlled to within (3 Hz (Doty Scientific, Columbia, S.C.).
13C spectra were excited by cross-polarization from abundant protons
using a 2-ms Hartman-Hahn contact (γBC1 /2π ) γB1H/2π ) 42 kHz)
and accumulated with high power (γBH1 /2π ) 125 kHz) two pulse
phase modulated (tppm) decoupling26 with recycle times sufficiently
long to give equilibrium signal intensities. The tppm phase shift was
Experimental Methods
Peptide Synthesis. Peptides studied at natural abundance were
obtained from Bachem (King of Prussia, PA). Seven tripeptides with
*G ) [2-13C, 15N]Gly were prepared (G*GV, G*GG, A*GG, P*GG,
F*GG, Y*GG, and V*GG) by solid-phase synthesis. [2-13C, 15N]Glycine
was purchased from CIL (Andover, MA) and converted to t-Boc-*Gly
by reaction with tert-butyloxycarbonyl anhydride in t-BuOH/NaOH.
The product was acidified with HCl/CH3COOH, extracted with ethyl
acetate, and recrystallized.11 The course of coupling reactions were
monitored by reacting a few of the nascent peptide resin beads with
ninhydrin and examining their color under a microscope. Reactions
were deemed complete when all color (initially purple and later yellow)
was absent. An ∼1.5-fold excess of t-Boc-*Gly was coupled to Gly
resin (Chem-Impex International, Wood Dale, IL) using the coupling
reagent 1-hydroxybenzotriazole (HOBt) and Castro’s reagent (BOP)
in DMF. The N-terminal residue was coupled using a 6-fold excess of
the t-Boc amino acid. Peptides were cleaved from the Merrifield resin
with anhydrous HF.11 The product was extracted with 25% acetic acid
and lyophilized. Initial purification was by cation exchange chroma-
tography. A 100-200-mg amount of crude peptide dissolved in 50%
acetic acid was applied to a Pasteur pipet containing ∼1 g of Dowex
50WX8-100. The column was washed with several volumes of water,
and the peptide eluted with 1% NH4OH. Final purification was by size
exclusion chromatography on a 1-m column of Sephadex LH-20
(Amersham Pharmacia Biotech AB, Uppsala, Sweden) in 50% acetic
acid. When prepared in this fashion, peptides yielded single spots in
silica gel thin-layer chromatography (solvent system H2O (30%)/
C2H5OH (70%)), had the expected molecular weights as determined
by MALDITOFF mass spectroscopy, and readily crystallized.
NMR Samples. Crystalline samples of the peptides for solid-state
NMR (containing 8-25 mg of labeled material) were grown according
to the procedures described in the original X-ray literature: G*GG,12,13
A*GG,14 P*GG,15 F*GG,16 Y*GG,17 and V*GG,18 G*GV,19 WGG.20
Cambridge Crystal Data Base21 reference codes for these structures
are listed in Table 2.
1
22.5°, and line widths were minimized by adjusting the H flip angle
(∼150°). From spectra obtained in this way, we arrived at the following
criteria for an acceptable sample: CR line widths in the range 50-100
Hz (0.4-0.8 ppm). Samples of WGG and F*GG were doped with
CuSO4 to shorten the 1H T1. Triple resonance experiments used
unmodulated 15N decoupling (γBN1 /2π ) 40 kHz). The 2-D PASS
experiment was implemented on our instrument as described by
Anzutkin et al.27 All 13C spectra were referenced to external adamantane
using the more intense, downfield line at 38.6 ppm.
NMR Spectrum Analysis. Principal shielding components were
obtained from 1-D and 2-D MAS spectra by the Berger-Herzfeld28
procedure using computer programs (i) kindly provided by Professor
Malcolm Levitt or (ii) written by us. The latter provides a surface of
ø2 ) (1/σ2noise)∑(Ijexp - Ijcalc)2 as a function of the two fitting param-
eters, (δ11 - δ33) and (δ22 - δ33), the covariance matrix, and a Monte
Carlo error analysis. The fitting parameters used here were chosen,
since they vary continuously for all possible values of δii. Visual
inspection of the ø2 surface confirms the optimum fit (global minimum
in ø2) and, if present, shows local minima. With the assumption that
experimental errors are dominated by spectrum signal-to-noise (σn2oise
is the mean square noise amplitude measured directly from the NMR
spectrum), standard errors and confidence intervals are determined in
three ways:29,30 (i) from the covariance matrix, (ii) from ∆ø2 ) ø2 -
ø2min contours, and (iii) by Monte Carlo simulation. In a Monte Carlo
trial, exact sideband intensities corresponding to the best-fit parameters
are added to Gaussian random noise with the experimental σ2noise and
then refitted. On the basis of a large number of trials (∼103), the range
of parameters found provides a good estimate of parameter error limits
when the relation between data and parameters is either linear or, as is
the case here, nonlinear.29,30 Source code (Fortran), look-up table and
i/o files are available upon request.
(10) Wang, C.; Teng, Q.; Cross, T. A. Biophys. J. 1992, 61(6), 1550-1556.
(11) Stewart, J. M.; Young, J. D. Solid-Phase Peptide Synthesis, 2nd ed.; Pierce
Chemical Co.: Rockford, IL, 1984; Chapter 2.
(12) Srikrishnan, T.; Winiewicz, N.; Parthasarathy, R. Int. J. Pept. Protein Res.
1982, 19, 103-113.
(21) Allen, F. H.; Kennard, O. Chemical Design Automation News 1993, 8(1),
1 and 31-37.
(22) SHELXTL, version 6.12; Bruker AXS, Inc.: Madison, WI, 2001.
(23) Sheldrick, G. M. SHELXS-90. Acta Crystallogr. 1990, A46, 467.
(24) Sheldrick, G. M. SHELXL-97: Program for the Refinement of Crystal
Structures; University of Go¨ttingen: Go¨ttingen, Germany, 1997.
(25) Zhang, Q. W.; Zhang, H.; Lakshmi, K. L.; Lee, D. K.; Bradley, C. H.;
Wittebort, R. J. J. Magn. Reson. 1998, 132, 167-171.
(13) Lalitha, V.; Subramanian, E. Cryst. Struct. Commun. 1982, 11, 561-564.
(14) Subramanian, E.; Lalitha, V. Biopolymers 1983, 22, 833-838.
(15) Lalitha, V.; Murali, R.; Subramanian, E. Int. J. Pept. Protein Res. 1986,
27, 472-477.
(16) Subramanian, E.; Sahayamary, J. Int. J. Pept. Protein Res. 1989, 34, 211-
214.
(26) Bennett, A. E.; Rienstra, C. M.; Auger, M.; Lakshmi, K. V.; Griffin, R. G.
J. Chem. Phys. 1995, 103, 6951-6958.
(27) Antzutkin, O. N.; Shekar, S. C.; Levitt, M. H. J. Magn. Reson., Ser. A
1994, 115, 7-19.
(28) Herzfeld, J.; Berger, A. E. J. Chem. Phys. 1980, 73, 6021-6030.
(29) Press, W. H.; Flannery, B. P.; Teukolsky, S. A.; Vetterling, W. T. Numerical
Recipes; Cambridge University Press: Cambridge, U.K., 1988.
(30) Bevington, P. R.; Robinson, D. K. Data Reduction and Error Analysis for
the Physical Sciences, 2nd ed.; McGraw-Hill: New York, 1992.
(17) Carson, W. M.; Hackert, M. L. Acta Crystallogr., Sect. B: 1978, 34, 1275-
1280.
(18) Lalitha, V.; Subramanian, E.; Parthasarathy, R. Int. J. Pept. Protein Res.
1986, 27, 223-228.
(19) Lalitha, V.; Subramanian, E.; Bordner, J. Int. J. Pept. Protein Res. 1984,
24, 437-441.
(20) Subramanian, E.; Sahayamary, J. Int. J. Pept. Protein Res. 1989, 34, 134-
138.
9
J. AM. CHEM. SOC. VOL. 124, NO. 40, 2002 11895

A R T I C L E S
Chekmenev et al.
Table 1. Comparison of (δ11 - δ33) and (δ22 - δ33) Determined by
Different Methods
No. of
sb’s
a
sample
method
decoupling S/N
(δ11 − δ33)b,c (δ22 − δ33)b,c
A*GG powder
A*GG MAS
G*GV powder
G*GV MAS
P*GG powder
P*GG MAS
A*GG MAS
1H/15N
1H/15N
1H/15N
1H/15N
1H/15N
1H/15N
1H
166
137
120
90
150
127
130
39
29
100
45
110
43
51.5(1.5)
53.1(0.4)
48.5(1.5)
48.9(0.6)
50.5(1.5)
50.7(0.7)
52.6(0.4)
51.1(1.0)
50.4(1.2)
49.3(0.8)
48.3(1.1)
49.1(0.6)
49.7(1.6)
23.7(1.5)
24.8(0.5)
18.3(1.5)
20.3(0.8)
21.8(1.5)
22.0(0.7)
26.0(0.5)
25.6(1.7)
25.8(1.8)
20.9(0.8)
22.8(1.3)
22.8(0.9)
23.9(1.8)
9
7
9
9
7
7
8
7
7
7
AGG
AGG
MAS
2D-PASS
1H
1H
G*GV 2D-PASS
GGV 2D-PASS
F*GG MAS
FGG 2D-PASS
1H/15N
1H
Figure 1. Powder and MAS (νr ) 1.1 kHz) 1H, 15N decoupled 13C spectra
of P*GG. Spectra were obtained with 2048 and 720 transients, respectively.
1H/15N
1H
a Spectrum signal-to-noise ratio, S/N, is for the centerband. b Joint 95%
confidence limits are in parentheses. c Shielding parameters are in ppm
relative to TMS with the convention that δ11 > δ22 > δ33.
Table 2. Shielding Tensor Data for Central Gly Residues
CSD
code
∼2°
structureb
(φ,ψ)
δiso
δ11 − δ33
δ22 − δ33
a,c
a,c
a,c
peptide
G*GG TGLYCY10 extended
178, -172 42.4(0.2) 43.3(0.4) 19.7(0.4)
-165, 175
G*GG BIBRUZ anti-â
-153, 160
-155, 155
-71, 167
-83, 169
43.3(0.4) 35.2(0.7) 26.0(0.7)
43.0(0.2) 34.7(0.6) 25.0(0.6)
43.0(0.2) 50.7(0.5) 22.0(0.5)
42.9(0.2) 53.0(0.7) 24.9(0.7)
V*GG COPBIS10 anti-â
P*GG FABXUB10 31-helix
A*GG CALXES20 31-helix
F*GG FIZWIU01 RR-helix -90, -29 44.5(0.4) 49.1(0.6) 22.8(0.9)
G*GV CUWRUH RR-helix -77, -22 44.8(0.2) 48.9(0.6) 20.3(0.8)
Figure 2. ∆ø2 contours obtained from the least-squares fitting of the P*GG
MAS spectrum and parameter values (gray dots) from 103 Monte Carlo
trials simulating the effect of experimental noise on the fitted parameters.
WGG FIZWOA01 RL-helix
Y*GG LTYRGG10 RL-helix
88, 10
81, 12
44.3(0.2) 50.3(5)
44.2(0.2) 41.3(0.5) 16.3(0.6)
25.0(5)
a Numbers in parentheses are 95% joint confidence intervals. b RR-helix
and RL-helix are right and left-handed R-helices. c Shielding parameters are
in ppm relative to TMS.
accuracies are tested by standard error analysis of the values
determined by both methods.
Figure 1 shows the 15N/1H decoupled 13C powder and MAS
spectra of P*GG, wherein the observed spectral patterns are
determined by the CR chemical shielding alone.
Results
With the goal of measuring chemical shielding parameters
in proteins, essential practical considerations are sensitivity,
resolution, and the reliability of the shielding values. We
investigate these by summarizing results obtained with several
peptides using stationary and spinning samples and different
decoupling schemes with labeled and unlabeled samples, Table
1. Experiments, for example, with and without 13C, 15N labeling
of the central Gly residue allow a direct assessment of the effects
of dipolar coupling between CR and bonded 14N or 15N.
Shielding parameters are listed in terms of (δ11 - δ33) and (δ22
- δ33) with the convention that δ11 and δ33 are, respectively,
the most downfield and upfield components. Note that these
two parameters with the isotropic shift, δiso (Table 2), determine
the three shielding tensor principal components. The value of
(δ11 - δ33) gives the overall breadth of chemical shielding, while
(δ22 - δ33) is closely related to the asymmetry parameter, η.
When (δ22 - δ33) vanishes or equals (δ11 - δ33), the tensor is
Best fit values of the shielding parameters, S/N values, and
confidence intervals are listed in Table 1. Importantly, the ∆ø2
surface shows a single minimum, indicating that the shielding
parameters are uniquely determined by the MAS spectrum. To
specify confidence limits in the fitted shielding parameters, three
∆ø2 contours for the P*GG MAS data and the results of the
Monte Carlo simulation are superposed in Figure 2.
Shown are frequently used contours for 68.4% confidence
for individual (∆ø2 ) 1) or joint (∆ø2 ) 2.3) parameter values
as well as the 95% joint confidence level (∆ø2 ) 6.17).29,30
Individual parameter confidence intervals determined by the
limits of the ∆ø2 ) 1 contour are in good agreement with
standard errors from the covariance matrix and the individual
parameter distributions found in the Monte Carlo simulation.
However, the latter shows that parameter values consistent with
the experimental noise level frequently vary over a larger range
when viewed in the two parameter space shown in Figure 2.
Thus, confidence limits listed in Tables 1 and 2 for MAS spectra
are at the 95% joint confidence level (∆ø2 ) 6.17). Parameter
accuracy from stationary powder spectra is limited by spectral
resolution; the reciprocal of the spectrum acquisition time31
which is limited in practice by the time in the acquisition period
1
axial, η ) 0, or when (δ22 - δ33)/(δ11 - δ33) ) /2, η ) 1.
In simple cases (nonoverlapping powder patterns and absence
of dipolar couplings), shielding tensor principal components are
determined from powder patterns by inspection of the turning
points or by line shape fitting. Alternatively, Berger-Herzfeld
analysis of MAS sideband intensities28 provides a substantial
resolution advantage, albeit at the possible expense of determin-
ing principal component frequencies via line intensities. The
(31) Farrar, T. C.; Becker, E. D. Pulse and Fourier transform NMR; Introduction
to Theory and Methods; Academic Press: New York, 1976.
9
11896 J. AM. CHEM. SOC. VOL. 124, NO. 40, 2002

R
Glycyl C Chemical Shielding in Tripeptides
A R T I C L E S
(1 for 14N).34 D is the traceless, axially symmetric dipole tensor
with the unique component νD along the C-N bond and
expressed in the same frame as σ. Consequently, the observed
MAS spectrum is a sum of center and sideband intensities
resulting from two or three “effective” tensors. While the effect
of dipolar coupling is apparent in simulations with νr , νD,
with νr ) 1.2 kHz, simulated changes in sideband intensities
are a few percent as experimentally observed, Figure 3.
When compared to 15N, the effect of dipolar coupling to 14
N
is potentially larger (νD for 14N is smaller but ms takes on values
2-fold larger). This probably has no observable effect in larger,
hydrated peptides where efficient 14N relaxation results in self-
decoupling.10 For tripeptides, we have observed self-decoupling
Figure 3. 15N decoupled (256 transients) and coupled (132 transients) 13
MAS (νr ) 1.2 kHz) spectra of A*GG. The small signal at 20 ppm is a
natural abundance Ala methyl group signal.
C
2
of the amide H-14N dipolar coupling (νD ) 1200 Hz) in one
case (GAL) but not in GGV where the Gly2 amide coupling is
at which the free induction signal disappears into the noise, ∼2
ms in these experiments. This corresponds to a spectral
resolution of (1.5 ppm for the powder patterns. For the three
cases where both powder and MAS spectra of double-labeled
samples were obtained (P*GG, A*GG and G*GV), the shielding
parameters agree within the stated confidence limits. However,
in all cases, the MAS parameters are larger than those from the
stationary experiment. This is likely due to the lower spectral
resolution of the stationary sample experiment, wherein broad-
ening of the powder pattern shoulders results in overestimating
seen.35 Thus, Gly2 amide 13CR-14N dipolar coupling likely
explains the systematically larger (1.9 ppm) value of (δ22
-
δ33) measured in GGV as compared to the 15N decoupled MAS
value observed in G*GV.
Regarding the measurement of shielding parameters, we
conclude by noting that parameters measured by the 2D
experiment are equivalent to those measured in the 1D experi-
ment (Table 1; A*GG, F*GG, and G*GV examples). Further-
more, since it is a constant time 2D experiment,27 the only
increase in total spectrum acquisition time relative to the that
of the 1D experiment is from the modest reduction in signal
(∼20%) that we observe in practice from the sequence used to
encode sideband dependent phase. Thus, the resolution advan-
tage of the 2D experiment comes at no cost in accuracy or
acquisition time. The 2D spectra with sufficient S/N are obtained
from tripeptides at natural abundance, indicating that the
approach can be used for labeled proteins. Shielding parameters
are most reliably determined using triple resonance 1D or 2D
MAS with double-labeled samples, wherein error limits in the
shielding parameters are determined by spectrum signal-to-noise.
For example, if the center band S/N ratio is in excess of 100:1
and the spinning speed is sufficiently slow, νr < 4(δ11 - δ33)ν0,
δ33 and underestimating δ11. In the data reported here, with S/N
≈ 125:1 and 7 or more sidebands, the MAS experiment is
superior to the stationary sample experiment, wherein accuracy
is limited by lower spectral resolution.
Spectra of CR shielding are affected by dipolar coupling with
the bonded 15N or 14N.6,10 We now consider dipolar effects in
MAS experiments. The effect of the quadrupole mixing of 14
N
1
Zeeman states on a dipolar coupled spin /2 spectrum in MAS
NMR has been quantitatively explained.32 MAS lines appear
as 2:1 doublets with a maximum splitting of 9νD(e2qQ/h)/20ν0.
Using representative values (N-CR bond length of 1.47 Å,
amide 14N quadrupole coupling of 3.2 MHz,33 and NMR
frequency of 35.8 MHz), this splitting is ∼28 Hz, a value 2- or
3-fold less than the line widths observed here. This expectation
is experimentally confirmed, that is, no such splittings were
observed in numerous spectra of unlabeled peptides.
shielding parameters are determined within 1 ppm. Without 15
N
decoupling and double-labeled samples, only (δ22 - δ33) appears
to be altered and by a small amount (1-2 ppm). An alternative
to 15N labeling and decoupling is to use a higher magnetic field
(>18 T) and spinning speed such that dipolar coupling is
insignificant relative to both shielding and the spinning speed.
The main results presented here are summarized in Table 2.
Shielding parameters were obtained with 15N decoupled MAS
spectra of samples with [2-13C, 15N]Gly in the central residue
with one exception, WGG. Because of limited S/N in the WGG
spectrum, error limits are larger than dipolar effects and we
anticipate that the parameters are reliable within these limits.
The (φ,ψ) angles for the central residues, on the basis of the
X-ray coordinates, are listed, and the 2° conformation is labeled
according to the standard structure with which it most closely
corresponds. Commonly accepted (φ,ψ) values are left-handed
R-helix (-62°, -41°), right-handed R-helix (57°, 47°), parallel
â-strands (-119°, 113°), antiparallel â-strands (-139°, 135°),
and 31 helix/polyglycine II (-80°, 150°).36 Table 2 provides
examples reasonably close to each of these. The most repre-
Consequently, 13CR sideband intensities at the field used here
(11.7 T) are determined by a combination of first-order chemical
shielding (Hcs ) σIz) and dipolar coupling (Hdipole ) DIzSz).
For the standard N-CR bond length 1.47 Å, the dipolar coupling,
νD, is 690 Hz or 5.5 ppm for 14N and 960 Hz or 7.7 ppm for
15N. Although these couplings are 10-15% of (δ11 - δ33) at
11.7 T, MAS spectra at νr ) 1.2 kHz with and without 15N
decoupling, Figure 3, and the shielding parameters obtained by
neglecting dipolar coupling, Table 1, show only a small effect.
Only (δ22 - δ33) is systematically different (larger by 1.2
ppm) in the absence of 15N decoupling when 15N coupling is
neglected in the fitting procedure. This observation is quanti-
tatively simulated as follows: compared to the MAS experiment
involving only chemical shielding, the shielding tensor, σ, is
replaced by “effective” tensors, σeff(mS) ) σ + mSD, for the
two or three nitrogen spin states (mS ) (1/2 for 15N or mS ) 0,
(32) Oliveri, A. C.; Frydman, L.; Diaz, L. E. J. Magn. Reson. 1987, 75, 50-
(34) Tycko, R.; Weliky, D. P.; Berger, A. E. J. Chem. Phys. 1996, 105, 7915-
7930.
62.
(33) Rabbani, S. R.; Edmonds, D. T.; Gosling, P.; Palmer, M. H. J. Magn. Reson.
1987, 72, 230-237.
(35) Pometun, M. S.; Usha, M. G.; Richardson, J. F.; Wittebort, R. J. J. Am.
Chem. Soc. 2002, 124, 2345-2351.
9
J. AM. CHEM. SOC. VOL. 124, NO. 40, 2002 11897

A R T I C L E S
Chekmenev et al.
sentative are the â-strand (V*GG and G*GG) and 31 helix/
polyglycine II (A*GG and P*GG) examples, since both torsion
angles fall within 20° of standard values. G*GG is also studied
in a fully extended polymorph (CSD code TGLYCY10), and
the shielding parameters are quite different from those of the
â-strand form. Thus, conformation can have a strong effect on
chemical shielding. This sample did not, as might be expected,
show distinct isotropic 13CR shifts for the central Gly residues
of the two molecules in the asymmetric unit. Because of this
observation and the difficulties described in obtaining the
published structure,12 the X-ray structure was reinvestigated.
An excellent, low temperature (100 K) structure (R1 ) 0.046)
was obtained that confirmed the published structure (see
Experimental Methods). Moreover, to confirm the equivalence
of the materials used for the NMR and X-ray experiments, we
reproduced the NMR results using a separately recrystallized
sample of G*GG and a natural abundance sample (GGG) which
shows, in addition to the 13CR signal of the labeled sample, four
additional 13CR lines confirming the presence of two nonequiva-
lent molecules (at the terminal Gly residues) in the NMR
samples. For R-helices, examples of both left- and right-handed
helices were studied. Since Gly is not chiral, shielding is
independent of the helix handedness.7 The torsion angles
conform less closely to the ideal values but are within 35° of
ideal R-helix values. This point is further addressed in the
Discussion.
Figure 4. Ramachandran plot showing torsion angles for Gly residues in
the peptides studied here (open circles) and Gly residues found in proteins
with R-helical (red), â-sheet (brown and green), and the 31-helix/polyglycine
II (blue) conformations.
When given the small size of the NMR data set, an important
question is whether these observations extend beyond the data
presented here. In Figure 4, torsion angles on a Ramachandran
plot for the peptides studied here (open circles) are compared
with those observed for Gly residues from a randomly selected
set of proteins in the Protein Data Bank.37
Figure 4 includes 208 examples of Gly residues found in
R-helices (red) and a similar number found in â-strands (brown
and green). Due to a limited number of collagen-like examples
in the protein database, the 31-helix/polyglycine II set (blue) is
smaller (53 examples). We note that an initial set of examples
half as large covered the same overall space on the plot.
The tripeptide examples of R-helix and 31-helix/polyGly II
conformation are representative of the torsion angles found for
Gly residues in proteins. Moreover, the regions of a Ramachan-
dran plot occupied by Gly in R-helices and 31-helices are well
localized. In contrast, the well-known conformational variability
of Gly in proteins is displayed only in â-strands. In addition to
examples in a somewhat expanded normal range for â-strands
(brown), there are numerous examples outside the normal space
(green) in which the sign of φ is positive. Thus, while the
experimental rules concluded above are useful guidelines for
identifying R- and 31-helix/polyGly II residues, either a much
larger set of shielding data or reliable calculations are necessary
to identify â-strands by shielding parameters.
To circumvent the need for a large data set, we compare our
experimental values, Table 2, with the calculations of Oldfield
and co-workers for a tripeptide-like fragment, N-formyl glycyl
amide.7 This type of comparison has been made previously for
Ala, Val, and Leu residues.5 When comparing this nonionic
fragment and a tripeptide, two obvious questions arise: the
effects of charges at the peptide termini and different flanking
residues. First, even though most of the peptides studied here
are zwitterionic, we anticipate that the comparison is relevant,
since termini charges appear to have only a small effect on CR
shifts at the adjacent residues. In the case of isotropic shifts,
this was noted some time ago38 in peptides of the form GGXGG
Table 2 yields three trends relating glycyl shielding param-
eters to the 2° structure. (i) Without exception, isotropic shifts,
δiso, for R-helix residues are downfield from â-strand residues
with an average shift of 1.5 ppm. This is in agreement with the
chemical shift index.1 Additionally, we observe that 31-helix/
polyGly II isotropic shifts are in the range observed for
â-strands. (ii) The range of anisotropies, (δ11 - δ33), vary from
34.7 ppm (VGG, â-strand) to 53.0 ppm (PGG, 31 helix), an
order of magnitude larger than the range of isotropic shifts. In
terms of 2° structures, 31 helices are largest (50.7 and 53.0 ppm),
R-helices are intermediate (41.3 to 50.3 ppm), and â-strands
are the smallest (34.7 and 35.2 ppm). (iii) A third parameter
distinguishing 2° structures is the ratio (δ22 - δ33)/(δ11 - δ33),
1
which is less than /2 for both R-helix and 31-helix/polyGly II
1
residues but greater than /2 for â-strands.
Discussion
The data presented here sample glycyl residues in a variety
of backbone conformations close to R-helix, â-strand, 31-helix/
polyGly II, and fully extended 2° structures. Using MAS triple
resonance spectroscopy, accurate shielding parameters have been
measured. The observed chemical shielding parameters show
consistent results as follows: the â-strand examples have upfield
δiso ≈ 43 ppm, small (δ11 - δ33) ≈ 35 ppm, and (δ22 - δ33)/
(δ11 - δ33) > 1/2; the 31-helix/polyGly II examples have upfield
δiso ≈ 43 ppm, large (δ11 - δ33) ≈ 52 ppm, and (δ22 - δ33)/
(δ11 - δ33) < 1/2; while right- and left-handed R-helix examples
have downfield δiso ≈ 44.5 ppm, intermediate (δ11 - δ33) ≈
41-50 ppm, and (δ22 - δ33)/(δ11 - δ33) < 1/2. The downfield
isotropic shifts for R-helix relative to â-strand residues are in
the range observed previously in numerous cases, while the
experimental observation of trends for (δ11 - δ33) and (δ22
δ33) in glycyl residues is new.
-
(37) Berman, H. M.; Westbrook, J.; Feng, Z.; Gilliland, G.; Bhat, T. N.; Weissig,
H.; Shindyalov, I. N. Bourne, P. E. Nucleic Acids Res. 2000, 28, 235-
242. Available at www.rcsb.org/pdb/.
(38) Keim, P.; Vigna, R. V.; Marshall, R. C.; Gurd, F. R. N. J. Biol. Chem.
1973, 248, 6104-6113.
(36) Creighton, T. E. Proteins: Structures and Molecular Properties, 2nd ed.;
W. H. Freeman: New York, 1993; Chapter 5, p 183.
9
11898 J. AM. CHEM. SOC. VOL. 124, NO. 40, 2002

R
Glycyl C Chemical Shielding in Tripeptides
A R T I C L E S
0.5 ppm) and requires scaling of the theoretical shifts by 0.82.
We conclude that in most cases calculated values of (δ11
δ33) and (δ22 - δ33) are reliable within several ppm.
-
The single outlier in the above correlation is the GGG
polymorph (CSD code TGLYCY10), which contains two
molecules in similar, fully extended conformations. For these
two conformations, a single isotropic shift is observed with (δ11
- δ33) ) 43.3(0.4) ppm, a value substantially different from
values interpolated from the ab initio surface,38 34.2 and 36.0
ppm. As described in the Results section, both the NMR results
and the X-ray structure were confirmed. We note that for nearby
torsion angles in the area of (-150, 175), shielding parameters
vary substantially and include those experimentally observed.
Also, sparseness of ab initio calculations near the fully extended
conformation7,38 is a potential source of the disagreement.
In summary, the following generalizations are consistent with
this experimental and the published theoretical7 studies of Gly
CR chemical shielding. In most cases, (δ11 - δ33) values
distinguish between R-helix (∼32 ppm to ∼50 ppm) and 31-
helix/polyGly II (∼45 ppm to ∼55 ppm), but neither is separated
from â-strands which displays a wide range of values (∼24 ppm
to ∼58 ppm) because of the wide range of torsion angles. In
terms of (δ22 - δ33)/(δ11 - δ33), R-helix (∼0.25 to ∼0.4) and
31-helix/polyGly II (∼0.35 to ∼0.55) are again easily distin-
guished, as is, in most cases, R from â (∼0.35 to ∼0.90). Finally,
if (δ22 - δ33)/(δ11 - δ33) > 0.5, the structure is â.
The results presented here show that measurements of Gly
13CR shielding parameters by MAS spectroscopy are a viable
experimental technique for qualitative investigations of 2°
structures in proteins. Attractive cases include proteins not
amenable to crystallization or solubilization in their native state.
Examples are collagens, elastins, and plaques associated with
a variety of connective tissue and neurodegenerative diseases.
In all of these cases, determination of the 2° structure is
important and currently difficult. Also, the 13CR shielding
parameters reported here should serve as experimental tests for
developing improved ab initio calculations. The use of triple
resonance and double-labeled samples not only place dipolar
coupling to the bonded nitrogen under experimental control but
also makes 15N/13C resolved spectroscopy possible and 15N
shielding parameters available. These would further improve
resolution, make the approach more robust, and resolve ambi-
guities in cases where 13C shielding parameters do not cor-
respond to a unique set of torsion angles or 2° structure. Results
of 15N shielding in these peptides will be reported separately.
Figure 5. Correlation of experimental, Table 2, and theoretical7 shielding
parameters, δ22 - δ33 (circles) and δ11 - δ33 (squares), respectively. Points
are color coded as follows: R-helix (red), â-sheet (green), and the 31-helix/
polyglycine II (blue) conformations.
where CR shifts of the central three residues are unchanged in
the low pH cationic, neutral pH zwitterionic, and high pH
anionic forms. Second, isotropic shifts of the Gly residues
flanking the variable X residue vary by less than 0.2 ppm for
all residues except X ) pro, in which case the shift is 0.8 ppm
at the N-terminal side. These observations are also indicated in
the anisotropic shielding parameters in two peptides studied here.
The hydrogen chloride salts of GGG (BIBRUZ) and zwitterionic
VGG have different charges at the termini and different
sequences but the same conformation at the central Gly residue,
and we observe equivalent shielding parameters. In Figure 5,
Table 2 values of (δ11 - δ33) and (δ22 - δ33) are correlated
respectively with (σ33 - σ11) and (σ33 - σ22) values interpolated
from the published Gly shielding surface.7,39
By correlating shielding differences, (δ11 - δ33) and (δ22
-
δ33), we eliminate the need to establish a common reference
for theory and experiment. The largest deviations are for R-helix
residues, and outliers in the data from the fully extended GGG
polymorph (CSD code TGLYCY10) have been eliminated.
Unlike the correlation for Ala and Leu where an empirical
scaling (0.72-0.84) of the ab initio values was required to bring
theory and experiment within reasonable agreement,5 no scaling
is needed or would improve the agreement between theory and
experiment for Gly. The rmsd error, 3.0 ppm, is modest in
comparison to the properties correlated, indicating a good
correlation between theory and experiment (F2 ) 0.97). Cor-
relating isotropic shifts is less successful (F2 ) 0.74, rmsd )
Acknowledgment. This research was supported by NIH Grant
AR41751-07. We thank the Kentucky Research Challenge Trust
Fund for the purchase of the CCD X-ray instrument.
(39) Shift Calculator, Eric Oldfield web page. http://feh.scs.uiuc.edu/.
JA026700G
9
J. AM. CHEM. SOC. VOL. 124, NO. 40, 2002 11899