Macromolecular NMR Spectroscopy BCH 5886

T.M. Logan Spring, 2005

Lecture 10. Assignment and Structure Determination in Proteins.

We have presented several experiments over the past few lectures, and haven’t spent any
time really discussing how they are used in chemical shift assignment and structure
determination. We’ll correct that error in this lecture. I’ll break this up into three categories: 1.
natural products, 2. peptides, and 3. proteins. Nucleic acids will be discussed separately, but
brief comments about how nucleic acids would be treated differently than proteins will be made
where appropriate.

10.1. Assignment and Structure Determination of Natural Products.

The objective of assigning natural (and non-natural) products is to obtain some sort of
picture of the molecule you are studying. In most cases, an accurate three-dimensional structure
will not be determined. Instead, you will be interested in determining a “two dimensional
structure”, similar to what can be drawn on paper. In more sophisticated examples, we will need
three dimensional information to shed light on a particular aspect of the molecules chemistry.
First we’ll cover chemical shift assignments, then briefly discuss “structure determination”. Many
of the rules for chemical shift assignment in natural products are the same for amino and nucleic
acids, and the common rules/strategies will only be covered here. As I am not an expert in
natural products, much of what I say in the next few paragraphs is anecdotal, but since that
hasn’t stopped me in the past, here goes.
Chemical shift assignment consists of three conceptual steps: 1. identifying unique spin
systems; 2. assigning the “structure”, or topology of the individual spin systems (or networks);
and 3. connecting the different spin systems together to identify the molecule. This is, in general,
a non-trivial process, as you might have gathered from the exam problem you had to solve, and
these steps are called conceptual because you generally work on steps 1 and 2 simultaneously.
The first step, identifying individual spin systems, is based in part on identifying coupled nuclei
through COSY-type experiments. The problem can be approached by finding “starting” points.
Singlets, of course, aren’t scalar coupled to anything (otherwise they wouldn’t be singlets), and
can be assigned by inspection. Assigning singlets doesn’t help with identifying spin system
topologies, but these resonances must be assigned eventually, and they make convenient
starting positions. Other good starting points are up-field or down-field shifted resonances.
These arise from 1Hs next to electron donating or withdrawing functional groups, respectively. If
these resonances are well separated from the others, then their couplings and connectivities are
easily identified. The scalar-coupled resonances are identified by:
1. identifying collapsed couplings in selective decoupling experiments, as was done in your
exam,
2. identifying cross peaks in a 2D COSY experiment. Of course, other experiments can, and
should, be used to confirm the correlations identified in the COSY (what might those be?), and
3. measuring the scalar coupling constants observed in the sell-resolved resonances. This
is because the J values must be symmetric; scalar coupling of 1Hs A and B gives the same
interaction whether you look at nucleus A or B. A trivial example is shown in Figure 10.1; it is
often not this simple to identify the couplings when the coupling patterns become complicated

10.1
Macromolecular NMR Spectroscopy BCH 5886
T.M. Logan Spring, 2005

Figure 10.1. Stick

A B C D
spectrum showing
four sets of doublets.

due to multiple coupling partners. In the example shown, the couplings between resonances A
and C, and B and D, are the same, demonstrating that nuclei A and C, and B and D, are scalar-
coupled, but that A is not coupled to B or D, etc. Specific problems to be aware of in identifying
spin coupling networks:
1. chemical shift degeneracy (overlap). Several 1Hs can have similar magnetic
environments that contribute to “accidental” degeneracy; other 1Hs can be related by symmetry.
This degeneracy can lead to fewer than expected resonances. Accidental degeneracy can
generally be identified by complicated multiplet structures (different multiplets are overlapped)
and by the use of relayed-COSY and TOCSY experiments because different correlation peaks
will be identified. These overlapped resonances might not be assigned initially as overlapped;
many overlapped peaks are identified only after you have enough information to say that it
doesn’t make sense that this many cross peaks can be associated with a single 1H. Symmetry-
related degeneracy can be distinguished by simplified multiplet structures, and by higher
intensity (the resonance contains signals from more than one 1H). A rough estimate of the
number of 1Hs involved in a single resonance can be obtained by integrating the spectrum. This
is only approximate because of relaxation differences between the different nuclei.
2. odd local geometry. Scalar coupling constants are dependent upon the dihedral angle
between the two 1Hs; it is largest for anti conformations; smallest for eclipsed. In some
structures, there will be only weak coupling between two nuclei that are part of the same spin
coupling network because of this. In extreme cases, the entire spin system won’t be identified
without NOESY data. On the other hand, you can attempt to assign these 1Hs to the same spin
network using heteronuclear correlations. Assume you’ve assigned the majority of 1H spins. If
you perform an HMQC to correlate directly-bound 1H-13C spin pairs, then you can run an HMBC
experiment to identify correlations between 1H and other 13Cs. This experiment will be
discussed again below.
3. Poly aromatic spin systems. These are particularly difficult to assign because they are
especially poor in 1Hs, and the 1Hs that are present are often overlapped. In this case, there can
be too many carbons between the 1Hs to use the HMQC / HMBC trick I just mentioned, and
NOESY experiments don’t work because the 1Hs are too far away.
4. 1H chemical shifts are strongly affected by neighboring functional groups. For instance,
the CH3 resonance frequency shifts when bound to an oxygen or nitrogen; it also shifts when
bound to sp3 vs. sp2 carbons. The same is true for CH2 and CH groups. There are published
tables that indicate commonly observed chemical shift ranges for 1Hs near important organic
functional groups; these should be consulted when assigning natural products. However, I can

10.2
Macromolecular NMR Spectroscopy BCH 5886
T.M. Logan Spring, 2005

guarantee that just about each interesting system will have one or two 1Hs that don’t “fit” into
these categories and will require more thought in assignment.
5. The types of coupling networks encountered are, generally, not known at the beginning of
the assignment process. This is in contrast to proteins and nucleic acids, where the basic spin
systems are known ahead of time.
Connecting assigned spin coupling
OH
networks in natural products can be difficult;
many times the different spin networks are HO
H H
separated by several bonds lacking 1Hs, and / O
H
or by functional groups. There are two
H OH
approaches to bridging these gaps: using NOE
interactions, or using heteronuclear couplings. H HO
The NOE is limited to 1Hs that are about 5-7 Å
apart (the exact distances will depend on the H
molecular tumbling rate, and the local “proton Figure 10.2. Isolated spin systems can sometimes
density”). Another way to correlate the be connected using HMBC correlations, as in the
glycosidic linkage of carbohydrates.
different spin systems is to use HMQC / HMBC
experiments. This can be accomplished in the
following manner. Consider the spin system indicated in Figure 10.2, which indicates the 1H and
13C nuclei around the glycosidic bond in polysaccharides. From the COSY / DQF-COSY, and

TOCSY experiments, the 1H spin systems of each individual sugar can be assigned; furthermore,
the specific carbohydrate can be identified through scalar coupling constants between different
1Hs (all hexoses are epimers of each other). The problem is in connecting the individual sugar
spin systems. This is done using a combination of HMQC and HMBC experiments. Recall that
the HMQC is used to provide 1H-13C correlations, and provides the C1 and C4 carbon chemical
shifts. Once these are obtained, an HMBC experiment is performed that correlates H4 of sugar A
with C1 of sugar B (arrows in the figure). In the same experiment, you get a correlation between
H1 of sugar B and C4 of sugar A to confirm the assignments.
This particular example was chosen to represent a general solution to connecting spin
systems separated by an ether linkage; similar approaches can be taken to correlate spin
systems separated by a limited number of bonds; the limitation arises from the size of the 1H-X
nucleus coupling constant relative to the T2 relaxation time of the molecule.
To summarize, COSY-type experiments are used to identify and assign individual scalar
coupling networks. The different networks are correlated using either NOE information or
heteronuclear correlations. In general, assignment and structure determination in these
molecules is complicated by the variety of spin systems that can be encountered, by the relative
lack of 1Hs, and by an absence of a defined, three dimensional structure.

10.2. Assignment of Peptides and Small Proteins. Homonuclear Methods.

In this section, I will discuss assignment and structure determination of peptides and small
proteins (< 100 amino acids). These methods discussed in this section are based on
homonuclear, 1H-1H correlations. The approach and limitations will be discussed. In the next

10.3
Macromolecular NMR Spectroscopy BCH 5886
T.M. Logan Spring, 2005

section, I’ll discuss specific methods for dealing with larger proteins.
Assignments of these molecules follow the same conceptualized steps identified for natural
products; many of the steps are the same, some are different. Peptides and proteins offer some
simplifications over natural products, and some additional complications. These are:
1. The individual spin systems are well-known, and the relative chemical shifts and coupling
topologies are documented (Wüthrich’s book). This is also true of nucleic acids. The fact that the
spin systems are known in proteins and nucleic acids represents a substantial simplification over
natural products where the spin systems are not known a priori.
2. Peptides are generally flexible and don’t have a fixed secondary or tertiary structure in
solution. Assignments are simplified in peptides because T2 is generally long, which results in
nearly maximal cross peak intensity. On the other hand, the flexibility increases the likelihood of
accidental chemical shift overlap, e.g., all valines have similar chemical shifts for the Hα, Hβ and
Hγ's. On the other hand, short DNA/RNA fragments can have a defined secondary and tertiary
structure through intramolecular hydrogen bonding.
3. As peptides get longer, they tumble more slowly and adopt a more compact, well-defined
structure. This has the following implications: as T2 decreases, cross peak intensities decrease
because the time needed to evolve antiphase states approaches T2. This is true in nucleic acids,
as well. Second, as the structure becomes more compact, there are many more unique chemical
environments, which removes the accidental overlap problem, but complicates the spectral
appearance.
In the following paragraphs, I’ll discuss the analysis of 2D spectra and assignment of amino
acid spin systems. A complete and thorough discussion of this can be found in the book by
Wüthrich. This book covers peptide and nucleic acid assignment strategies. it is a bit out-date in
that there is no discussion of heteronuclear 3D methods, but it nevertheless is a solid book that
belongs on the shelf of any serious biomolecular NMR spectroscopist.
Peptide assignment strategy: Example. I previously showed you TOCSY and NOESY
spectra from a 17 residue peptide that we are studying (Figure 7.19 and 20). In general, the
TOCSY and NOESY spectra can be broadly classified in two different “parts” - the amide part
and the sidechain part. There are two approaches to assigning the spectrum: one is to start with
the amide 1Hs; the other is to start with the sidechain 1Hs. There are pros and cons to either
approach.
Starting with the HN. This method generally allows you to identify the individual amino
acids with ease because each amino acid contains a single backbone amide 1H (the sidechain
amides of glutamine and asparagine are easily identified and distinct from the backbone amides).
Figure 10.X shows an expansion of the TOCSY spectrum shown in Figure 7.20, emphasizing the
ability to identify discrete amino acid spin systems from the correlations detected between
aliphatic sidechain and amide protons Using the standard chemical shift ranges and known
coupling topologies for the different amino acids, the data in this TOCSY spectrum allowed us to
1. identify the individual spin systems (indicated by the lines drawn in the figure) and 2. assign
them to amino acid type in a relatively short amount of time. In many cases (e.g., when a
particular amino acid occurred once in the peptide), this information was sufficient to obtain the
sequential assignment; in cases where there were two amino acids of the same kind, we couldn’t

10.4
Macromolecular NMR Spectroscopy BCH 5886
T.M. Logan Spring, 2005

tell specifically which amino acid it was.

.In my opinion, this is the easiest and most direct
method for assignment. The limitations to this 0.6
tocsy amide region
approach, though, are that 1. only HN-Hα correlations
are observed in a COSY, which doesn’t provide
enough information to assign the amino acid type; 2. 1.1

since we need to observed correlations with HN, we

need to use H2O solvent, which presents severe
1.6
problems with solvent suppression; and 3. if the HN
chemical shift dispersion is not high, then there can be
significant problems due to chemical shift overlap. 2.1

H1, ppm
Starting from the Sidechain 1H. An alternative
approach is to start from the sidechain resonances
2.6
(see Figure 7.20). The advantages of this approach
are that 1. there are no exchangeable resonances so
the spectra can be run in D2O solvent and solvent 3.1

suppression becomes trivial; 2. many correlations are

used to assign the individual amino acids, so overlap is
3.6
not as much of a problem; 3. the cross-peak patterns
and chemical shifts of the different amino acids are
well-known, and easily assigned in 2D spectra of 4.1
peptides. The disadvantages of this approach are 1.
certain regions of the spectrum exhibit significant
chemical shift degeneracy (e.g., methyls); 2. it is 4.6

sometimes hard to identify individual amino acids 8.9 8.7 8.5 8.3

because of overlapping or missing correlations; 3. in H2, ppm

D2O there are no correlations with backbone amides. Figure 10.3. TOCSY spectrum indicating
how this data can be used to identify
This last point is important because the primary way to
individual amino acid spin systems and
identify secondary structure involves backbone amide then assign them to amino acid type. This
resonances. peptide contains 17 residues; 13 spin
In fact, you generally need to assign all cross systems are easily identified. Subsequent
analysis showed two different spin
peaks in every spectrum you collect. The approach I
systems were overlapped and one was
use is to 1. identify individual spin systems from the outside of this spectral region.
amides; 2. use the correlations from the HN protons to
assign amino acid type where possible; 3. confirm and extend these assignments by assigning
the sidechain correlations. Using this approach, overlapping resonances in one region of the
spectrum can usually be worked around because there may be no (or at least different) overlap
in the other region. Also, the assignments are obtained in H2O, which will facilitate assigning the
NOE spectra used in structure characterization.
Sequential assignments. This approach doesn’t provide complete, sequence-specific
chemical shift assignments; we’ve only gone through the first two conceptual assignment steps.

10.5
Macromolecular NMR Spectroscopy BCH 5886
T.M. Logan Spring, 2005

Figure 10.4. Sche H NH O H NH

matic indicating O
NOEs involved in
sequential
assignments of
peptides.
NH O H NH O H

We still need to sequentially link the individual amino acids. This is invariably done using a 2D
NOESY experiment (Figure 10.5). The amino acids are linked together in what’s called a
NOESY walk. The NOESY experiment, of course, contains correlations between 1Hs nearby in
space. This always includes an intra-residue HN-Hα correlation (fixed by bonded geometry to be
less than ~3.5 Å away, independent of secondary structure). The NOESY experiment also
contains other HN-Hα correlations, and, in small peptides, these are generally “sequential” NOEs
to the preceding amino acid residue (Figure 10.4, 10.5). Therefore, from one amide resonance,
we get information on (at least) two spin systems: the intraresidue spin system, and the
preceding spin system. In the absence of assignments, it would be nearly impossible to
determine which resonances belong to the intra- versus sequential residues from the NOESY

Figure 10.5. Fingerprint

region of a NOESY
spectrum indicating
I8
sequential assignments
identified in a NOESY walk.
The NOESY spectrum is
shown in black; the TOCSY T12β
spectrum is shown in red. I13
Assignments from
Takahashi and Logan, K3 K4 A11
unpublished.
R16
T12α
R2 L5
R14

F9 S15

H7 F6, H10

10.6
Macromolecular NMR Spectroscopy BCH 5886
T.M. Logan Spring, 2005

Figure 10.6.
δ ε H H
δ H H
H H H H
H H H H
H H H
HO NH
H Hδ
NH NH N
H
δ
Hδ
H H H H H

spectrum alone. This distinction is easily made by comparing the data in the TOCSY and
NOESY spectra because the TOCSY experiment only provides correlations with 1Hs of the intra-
residue amino acid (Figure 10.5).
This assignment strategy (using TOCSY and NOESY spectra) provides unambiguous
sidechain assignments, especially for aliphatic resonances, but there are two problematic groups
of residues: aromatics and prolines. Aromatic sidechain 1Hs are difficult to assign because the
aromatic and aliphatic protons are not scalar coupled. Aromatic resonances are assigned by 1.
chemical shift patterns in the TOCSY (to distinguish Phe, Tyr, and Trp residues); correlations
observed in the TOCSY; and Hδ ↔ Hβ crosspeaks in NOESY spectra to connect the aromatic
and aliphatic portions (Figure 10.6). Also, assigning Trp aromatic sidechains generally requires
NOE correlations between the indole HN and the other aromatic 1Hs. Prolines are difficult to
assign because they don’t have an amide 1H. The correlations expected in a proline residue are
well-known, and distinct from any other residue. Sequential assignments are made via NOE
interactions, but from the HN on the residue preceding the proline (Figure 10.7) to either Hα or
Hδ of the proline for trans and cis conformations, respectively.

10.3. Assigning Proteins using Heteronuclear Methods.

The homonuclear methods described in the previous section become useless when proteins
approach sizes > 100 amino acid residues. There are two reasons for this: first, the number of
resonances becomes larger and overlap becomes increasingly severe; but, second, the tumbling
rate slows and T2 relaxation rates increase, preventing the buildup of correlations through the
relatively small 1H-1H coupling constants. Therefore, for proteins larger than ~100 amino acid
residues, the double- and triple-resonance 3D methods discussed in the previous lecture are
used for making chemical shift assignments. Again, the same three conceptual steps are

Figure 10.7. Hg Hb Hb Trans form

Hg Hb Hg
Hb O
Hg
O
Hd
Hd
Hd H
N
Hd H
N H
H N
N H O
Cis form
O
O
H

10.7
Macromolecular NMR Spectroscopy BCH 5886
T.M. Logan Spring, 2005

involved in making chemical shift assignments, but a completely different set of experiments are
used.
Identifying Individual Spin Systems. There are two experiments used to identify the
individual spin systems (e.g., count the number of amide resonances): the HSQC and HNCO.
The HSQC provides a high resolution 2D spectrum correlating 1H bound to 15N; the HNCO is a
very sensitive, (relatively) high resolution 3D triple resonance experiment that correlates COi-1,
Ni, and HNi, where i and i-1 refer to the nuclei being of the intra-residue and preceding residue,
respectively. A single correlation is expected for each residue in these two experiments
(glutamine and asparagine also give amide resonances, but, again, these are distinct in chemical
shift and spectral appearance from the backbone resonances, and are readily identified). A
protein of n residues, m of which are prolines, should have n-m-1 correlations in these spectra
(often a few less than that are observed due to chemical shift overlap or protein dynamics).
Figure 10.8 shows a TROSY spectrum of arginine kinase kindly provided by Omar Davulcu, Jack
Skalicky and Michael Chapman.
Assigning Spin Systems to Amino Acid Type. There are several basic types of amino
acids that are classified according to their topology: These groups are 1. glycine (1 methylene);
2. proline (five membered ring); 3. AMX (asn, asp, cys, his, ser, phe, tyr, trp; AMX refers to
relative 1H chemical shifts of the one Hα and two Hβ signals); 4. long spin systems (gln, glu, lys,
met, & arg); 5. branched chain (val, ile, & leu); and 6. other (ala, thr). Most of these can be
identified based on the 13C chemical shift of the α and β nuclei (see Spera and Bax reference,
end of Lecture 2). For instance, Val Cα and Cβ have chemical shifts of 60 & 33 ppm,

Figure 10.8. TROSY spectrum or

arginine kinase. For larger proteins,
TROSY spectra replace the HSQC
due to the more favorable
relaxation properties of the TROSY
correlation compared to the HSQC.
From this spectrum one obtains
information about the total number
of individual spin systems.
Davulcu, Skalicky, and Chapman,
unpublished work.

10.8
Macromolecular NMR Spectroscopy BCH 5886
T.M. Logan Spring, 2005

Figure 10.9.
HN(i) HN(i) (Ω3)
HNCACB N(i) N(i) (Ω2)
Cα/Cβ(i, i-1) (Ω1)

Hα/Hβ(i-1) HN(i) (Ω3)

CBCA(CO)NH
Cα/Cβ(i-1) (Ω1) N(i) (Ω2)
CO(i-1)

respectively, while leu has 56 & 42 ppm. The values for glu, gln, & met are distinct enough to
allow a strong tentative assignment based on the Cβ chemical shift alone; the aromatic amino
acids can generally be distinguished from the asn/asp/cys spin systems in a similar fashion.
Thus, the individual amino acid spin systems are assigned to type based on their 13C chemical
shifts. These are generally obtained from the HNCACB experiment (see section 9.10 for a
discussion and sample spectra).
Linking Amino Acids into the Amino Acid Sequence. Of course, the final stage in
backbone assignments is to link together the individual amino acids to map the amino acid
sequence of the protein being studied. This process provides the sequence-specific chemical
shift assignments. Again, this is generally done using the HNCACB in combination with the
CBCA(CO)NH experiment (see Lecture 9 for description of these experiments). As discussed in
section 9.10, one often detects correlations that indicate the Cα/Cβ chemical shift of the residue
preceding the amide on which you are detecting. This information can be used to make the
sequential assignments by finding the 13Cα/13Cβ frequencies that match best. Sample data
indicating the use of the HNCACB and CBCA(CO)NH spectra for sequential assignments are
shown in Figure 10.10. In this specific example of arginine kinase, the high molecular weight
required sidechain deuteration which precludes using the CBCA(CO)NH experiment (because
the polarization transfer starts on sidechain protons). In this case, the HN(CO)CACB experiment
was used.
A summary of the triple resonance experiments commonly used in assigning protein
backbone resonances is shown in Figure 10.12. This list is by no means exhaustive. For a more
complete list, see Edison et al. 1994 Methods Enzymol. 239, 3-79.
Sidechain assignments. In the final step, the side chain 1H and 13C resonances must be
assigned; this is done using the HCCH-TOCSY experiment, as discussed in Lecture 9, although
there are other experiments that provide very specific information. Among the more important of
these is the HNHA experiment, providing correlations between HN, N, and Hα atoms and the
15N-separated TOCSY-HSQC, which frequently only provides correlations between HN and Ha,
but sometimes with sidechain methyl signals in larger proteins.
In summary, the assignment of proteins is accomplished as follows: 1. identify the individual
spin systems in HSQC / HNCO spectra; 2. assign to type using HNCACB; 3. link sequentially

10.9
Macromolecular NMR Spectroscopy BCH 5886
T.M. Logan Spring, 2005

using CBCA(CO)NH; assign all sidechains using HCCH-TOCSY. In general, one collects many
experiments that provide redundant correlations to make sure that the chemical shift
assignments are as complete as possible. In the end, your structures are only as good as your
assignments!.

Figure 10.10. Strip plot indicating

combined analysis of HNCACB /
CBCA(CO)NH type spectra. The
figure shows side-by-side strips from
an HNCACB and an HN(CO)CACB
spectrum collected on arginine
kinase prepared with uniform
15N,13C, and sidechain 2H labeling.
THe HN(CO)CACB experiment was
run rather than the CBCA(CO)NH
because the latter experiment
requires polarization transfer from
sidechain 1H to 13C, which is
impossible when the sidechains are
deuterated. The analysis starts by
identifying the intramolecular and
sequential (i-1) correlations in the
right-most panel. The residue that
identically matches the Cαi-1 and
Cβi-1 resonances is selected as the
next residue in the polypeptide chain
and the process is repeated
(Davulcu, Skalicky, and Chapman,
unpublished).

10.10
Macromolecular NMR Spectroscopy BCH 5886
T.M. Logan Spring, 2005

Experiment Correlations Observed Magnetization Flow

C H C H

HNCA H(i), N(i), Cα(i, i-1) N C C N C C

H H O H H O

C H C H

HN(CO)CA H(i), N(i), Cα(i-1) N C C N C C

H H O H H O

C H C H

HNCO H(i), N(i), CO(i-1) N C C N C C

H H O H H O

C H C H

HNCACB HN(i), N(i), Cα/Cβ(i, i-1) N C C N C C

H H O H H O

C H C H

CBCA(CO)NH HN(i), N(i), Cα/Cβ(i-1) N C C N C C

H H O H H O
Figure 10.11.

Determining Structure in Proteins.

In X-ray crystallography, one observes scattering reflections, which is FT’d to present
regions of electron density. The structure of the protein is determined by fitting a polypeptide
chain into this electron density. Protein structure determination by NMR is not as direct in that
there is no structure into which nuclei are fit. Instead, NMR structures are determined by
providing a different set of structural constraints; NOE interactions provide pair-wise distance
information, and scalar couplings are used to provide pair-wise dihedral angle information. Other
information, such as 13C chemical shift and amide hydrogen exchange rates, are used to identify
secondary structures. All of this data is used to calculate structures that are consistent with the

10.11
Macromolecular NMR Spectroscopy BCH 5886
T.M. Logan Spring, 2005

observed structural constraints. We’ll discuss the calculation and evaluation of structures in the
next lecture; here we’ll discuss experiments to collect and provide structural constraints.

10.4. NOE and Protein Structure.

Again, the best reference for this section is the Wüthrich book, chapters 7-9. Recall that the
NOE interaction between two nuclei can be represented as

1
NOE ∝ 〈 ----6〉 f ( τ c )p A [10.1]
r
where f ( τ c ) represents a function describing the local motional properties of the spin and p A
represents the population of the molecule in a given conformation (typically assumed to be
100%). While in principle, the NOE can provide precise internuclear distances; in practice, the
data are much less precise. There are three reasons for this:
1. The NOE equations are only valid for two isolated spins. This situation is not
obtained in proteins where there are on the order of 1000 or more 1Hs.
2. Spin diffusion effects cause problems in quantifying NOEs (spin diffusion is non-
direct dipolar interactions between two protons; it makes two 1Hs appear closer in
space than they really are).
3. Motion during the NOE mixing period (~ 80-100 ms) averages the distance of the
detected interactions.
For these reasons, the NOE is used in a semi-quantitative or qualitative manner, depending
on the particular application. In the semi-quantitative mode, we evaluate NOE intensity as
indicating distance within a certain range; in the qualitative mode, we simply determine whether
there is an interaction between two nuclei. One application of the qualitative mode is secondary
structure identification. This is discussed below.

10.5. Identifying Secondary Structure from NOEs.

There are four basic types of secondary structure in proteins: helix (α or 310), strand
(parallel or anti-parallel sheets), turn (several types), and random coil, each with its own unique
NOE signature. These are presented in the following paragraphs, but first, some nomenclature,
Figure 10.12. The standard nomenclature is to indicate the two nuclei participating in the NOE;

Figure 10.12. αβ(i, i+3)

αN(i, i+2)

CH2 CH2
CH2
O H H
H H O H H

N N N
N N

O O H NN H O H
H H βN
αN CH2
CH2

αN(i, i+3)

10.12
Macromolecular NMR Spectroscopy BCH 5886
T.M. Logan Spring, 2005

Figure 10.13. H O H H O

N N
N

H H O H

H H
H
H N O O
O
O H N H N
H N
H H H
H
N H O O
O
O N H N H
N H
H H H
H
H N O O
O
O H N H N
H N
H

Anti-parallel Parallel

the N-terminal nucleus is indicated first, and the relative residue positions are indicated in
parenthesis, e.g., aN(i,j). If the NOE is between atoms on sequential residues, then there is no
numerical index. For instance, an NOE between Hα of residue i and the HN of residue j (j=i+2) is
indicated as αN (i,i+2), say, or αβ (i, i+3) for an NOE between the Hα and Hβ protons of two
residues separated by two intervening amino acid residues. NOEs are classified as sequential if
|i-j| = 1; medium-range NOEs are for 1< |i-j| ð 5, and long range for |i-j| > 5.
β-Strands and β-Sheets.. These are the easiest secondary structures to identify. In the
extended conformation, the characteristic NOEs are strong NOEs between HN of residue i and
Hα of residue i-1 (denoted as αN(i,i+1) or αN), and weak NOEs between HN on adjacent
residues (NN (i, i+1) or NN), Figure 10.13. In β-sheets, individual strands can be oriented in a
parallel or anti-parallel fashion; the distinguishing feature here being strong NH-NH and NH-Hα

Figure 10.14. H
H H
O
N H
αN(i,i+2) N H

O H
H N O H N O
O N
H O H N

H H
H H

H O
O
NN(i,i+2) H

Type II Type I

10.13
Macromolecular NMR Spectroscopy BCH 5886
T.M. Logan Spring, 2005

NOEs between residues that are distant in the amino acid sequence. Parallel vs. anti-parallel are
distinguished by the N → C sense of the strands, and by the presence of Hα-Hα NOEs in the
antiparallel sheets.
Turns. There are several types of turns found in proteins; I will concentrate only on hairpin
turns of type I and type II, Figure 10.14 These are characterized by short NN distances between
residues 2,3 and 3,4, and by a short αN (i, i+2) distance between residues 2 and 4 for Type I
turns.
Helices. There are two commonly
encountered helices in proteins: α-helix and
310-helix, which differ in their characteristic
NOE patterns. In α-helices, the characteristic
NOEs are strong NN, weak αN, strong
αN(i,i+3) and αN(i, i+4), with some αN(i, i+2)
observed, but being weaker than the “i+4”
NOEs (Figure 10.15). In 310 helices, there are
no αN(i, i+4) NOEs and the αN(i, i+2) are
much more intense. Another characteristic
NOE observed in α-helices is the αβ(i, i+3).

10.6. Scalar Couplings and Protein

Structure.
Scalar couplings between two nuclei Figure 10.15. Stereo drawing of a typical α-helix
separated by 3 bonds are important indicators segment found in proteins. Atoms colors are: Hα -
of dihedral angles in proteins, and currently an gold; HN - cyan; CO - green; N - blue; Cα - red;
Cb - grey; Hβ- also blue (oops). The NOEs that
area of active interest in the development of
characterize α-helix are indicated and consist of
methods to provide a range of precise J short αN(i,i+2, αN(i,i+3), αN(i,i+4) and αβ(i,i+3).
coupling measurements. The idea is shown in The short NN distances are not shown be easily
Figure 10.16. The angle between HN & Hα visualized. The helix runs in the N to C terminal
direction from bottom to top of the figure.
(as shown in the Figure) provides a measure
of the φ angle; χ angles are obtained by
measuring either 1H-1H, 1H-13C, or 13C-13C couplings at various points along the sidechains.
From the plot shown in this figure, we can identify scalar couplings consistent with β-strand (φ = -
150° → -90°) and α-helix (φ = -60° → -30°) for 3JHNHα > 8 Hz and 3JHNHα < 4 Hz, respectively.
However, it is also obvious that the Karplus curve does not give a unique φ value for 3JHNHα
between 4 and 9 Hz, and this is the limitation of scalar coupling constants.
The most commonly measured coupling constant is the 3JHNHα, and the experiment most
commonly used to measure it is the HNHA-J experiment. The original HNHA sequence was
published by Vuister and Bax (1993 JACS 115, 7772-7777) with a significant improvement
published by Kuboniwa (also in Bax’s group; 1994 JBNMR 4, 871-878). The experiment is
shown in Figure 10.17, and a brief product operator analysis is presented after that. Essentially
this experiment is an HMQC correlation between 1H and 15N, coupled with a COSY-type HN-Hα
coherence transfer (t2 period) in the middle of the experiment. This results in two peaks per spin

10.14
Macromolecular NMR Spectroscopy BCH 5886
T.M. Logan Spring, 2005

H 3JHNHa = 6.71cosΘ + 1.76cosΘ + 1.60

Figure 10.16.
C
O

φ N

O H

system: the HN “diagonal” peak and the Ha “cross peak”. Each has intensity given by the
following expressions immediately prior to detection:
2
diagonal peaks = cos ( 2πJδ 2 ) , [10.2]

2
cross peaks = – sin ( 2πJδ 2 ) [10.3]

where J is the 3JHNHα and δ2 = 13 ms. The value of J is obtained from the ratio of the cross
peak and diagonal peak intensity,
2
J = – tan ( 2πJδ 2 ) . [10.4]

Similar experiments are used to measure coupling constants between other nuclei, and Karplus
curves are used to analyze them, as well, but with different coefficients.

10.7. Amide Hydrogen Exchange and Protein Structure.

A third experimental measure of protein structure the determine the amide hydrogen
exchange rates; the exchange rate is slowed when the amide hydrogen is involved in H-bonded
secondary structure. Note that this data does not tell us what type of secondary structure a
residue is involved in, but can be used to identify hydrogen bonds. The use of amide hydrogen
exchange to identify hydrogen bonds has been supplanted to some extent by the ability to
directly measure hydrogen bonds via scalar couplings, as described in Lecture 2.

Figure 10.17.

H δ1 δ2 δ2 t2 δ2 δ2 δ1

N δ3 + t1/4 δ3 + t1/4 δ3 + t1/4 δ3 + t1/4 dec

10.15
Macromolecular NMR Spectroscopy BCH 5886
T.M. Logan Spring, 2005

10.8. Chemical Shift and Protein Structure.

This discussion was presented some time ago, in the second lecture. Again, the idea is that
the chemical shift of a particular amino acid type is shifted upfield or downfield if it is involved in
secondary structure formation. For α-helices, Cα resonances are shifted downfield and upfield,
respectively; while the opposite shift is observed in β-strand. Secondary chemical shifts of CO
and Hα are also useful.

10.9. Structure Refinement.

10.16