3D NMR and Beyond

What do you do when resonances are overlapped in 2D? => 3D NMR

The concept of 3D NMR is essentially the same as 2D, except now there are 2 indirect
detect dimensions, so 2 times (now t1 and t2) that get incremented to give a second and
third dimension FID

Of course, what do you do when resonances are overlapped in a 3D experiment, you

acquire a 4D experiment.

As far as I know, nobody has ever acquired a 5D data set though, 4D is the practical
limit.

The main application of 3D and 4D NMR is structural biology, using NMR to determine
3D structures of macromolecules, primarily proteins and nucleic acids. High-resolution
3D structures of molecules of ~50,000 MW have been determined with NMR.

Why should an organic chemist care about 3D NMR?

1) NMR and Drug Design- ligand-macromolecule interactions. Structural biology meets

organic chemistry. The limit of the largest structure of a macromolecule structure
determined by NMR is always increasing. Even when a high-resolution structure is not
realistic, NMR assignments can be made on the protein backbone or the ligand. 3D NMR
and 4D NMR is required for the protein, 3D NMR can also be very useful although not
necessarily required on the ligand.

2) Larger small molecules can have overlapped protons, even in 2D NMR, especially
molecules with somewhat repetitive units like a sugar.
Why was 3D NMR not worth the trouble for any molecule that did not
absolutely need three-dimensions in the past; and how have NMR
spectroscopists worked around the difficulties?

1) Difficult to display. Since 2D data is really three-dimensional (remember that the 2D

data has varying intensities or contours), 3D data is really four-dimensional, and 4D data,
five-dimensional.

People have gotten around this by displaying planes of data, rather than a 3D plot.
Actually can be very simple to interpret.

2) Data sets are disk space intensive (>100 mb - 1 gb)

Hard disks are cheap now.

3) Data processing is excruciating.

Computers are much faster, so 3D processing can take as little as a minute or so, not
hours. Better software macros allow for easier processing, no more difficult than 2D data.

4) Data acquisition takes several days of NMR time.

Gradients have reduced phase cycles and improved sensitivity; probes and higher fields
have improved sensitivity. Linear prediction has allowed for reduction in the amount of
points needed to be acquired in the second and third dimensions (the indirect
dimensions). Some 3D experiments can be acquired in a few hours (some still take days
though). Rarely is even a 4D data set acquired for more than 3 days, and usually 3 days
on an instrument is not that difficult to get (even our 500 can be reserved from Friday
night to Monday morning). Considering that 2D's used to be acquired over days to weeks;
three days for a 4D is not that bad.

5) Pulse sequences written have limited use and are somewhat difficult to optimize; often,
users would have to write and optimize their own sequences.

Numerous experiments now exist, with the optimal way to acquire them already
determined. Experiment application is no longer trial-and-error. 3D acquisition is as easy
as 2D acquisition.
6) Experiments depend upon full isotope labeling of 13C and/or 15N.

Improved instrument sensitivity allows many experiments to be done in natural

abundance for carbon and with a cryo probe even nitrogen, although 3D data sets with
15
N will still require higher concentration (20 mM?).

7) More dimensions on an experiment, longer pulse sequence, T2 relaxation kills you

Methods have been developed to shorten pulse sequences, so less magnetization is lost to
T2. Also, T2 not a big problem for smaller molecules, but is a problem for ligand-
macromolecule complexes. Also, TROSY can help reduce T2 effects.
 What experiments exist in 3D?

2D NMR can be broken down into through-bond and through-space

3D NMR can be broken down into through-bond, through-space, through-bond-through-

space, through-space-through-bond

Through-Space

NOESY-NOESY

One of the first, may have been the first 3D experiment, but no real practical relevance
today. Disperses NOEs over 3 dimensions.

Through-Space-Through-Bond

NOESY-HSQC (first NOESY then an HSQC) (there are also ROESY-HSQC or NOESY-
HMQC ...)

Basically a proton-proton NOESY with one of the proton dimensions correlated to a

Carbon or Nitrogen.

Often referred to as a 3D NOESY or as a 13C-edited NOESY (a NOESY edited by

spreading proton NOEs over a third dimension 13C or 15N)

First, start with magnetization on 1H, in the first indirect dimension, the magnetization is
detected, then there is a mixing time, where magnetization is transferred through-space to
other protons- in a normal NOESY, there would be a 90º pulse and magnetization would
be detected- in a NOESY-HSQC, after the mixing time, the magnetization is transferred
to the 13C (or 15N) as in an HSQC, then the magnetization is detected so the second
indirect detect dimension is 13C (or 15N), then the magnetization is transferred back to 1H
and detected. Thus, there are 2 1H dimensions and one 13C (or 15N) dimension. NOEs are
observed from one proton to another, and the proton being detected is correlated to a 13C
(or 15N).
 NOESY-HSQC

Pulse Sequence Comparison of NOESY, HSQC, and NOESY-HSQC:

2D NOESY vs. 3D NOESY-HSQC (15N) on a protein.

2D NOESY

3D NOESY-HSQC (15N)
Through-Bond-Through-Space

1) HSQC-NOESY

Approximately the same as the NOESY-HSQC. The HSQC dimension is first rather than
the NOESY, so 1st indirect dimension is 1H, then the 13C (or 15N), then there is an NOE
mixing time then detect.

2) TOCSY-NOESY

First 1H-1H TOCSY, then NOESY

For example- suppose proton H5 has identical chemical shift to H5*, but H6 has a
different chemical shift than H6*, in 2D NOESY it was established that H5 or H5* has
NOE to HA.

First, TOCSY:
H6 has TOCSY crosspeak with H5, H6* with H5*.

then NOESY, H5 has NOE to HA, H5* does not. The NOE has been discriminating based
upon the chemical shift of H6 and H6*

At chemical shift of H6, H5, HA: there will be a crosspeak but not at H6*, H5*, HA
Through-Bond

Through-Bond can be broken into two categories: experiments with a diagonal (like a
COSY or TOCSY in 2D) and experiments without a diagonal (like an HSQC in 2D)

Experiments with a diagonal:

1) HSQC-TOCSY

First, an HSQC correlating the proton and nitrogen or carbon, then a TOCSY. Similar to
a NOESY-HSQC

When there are resonances with coincidental chemical shifts in proton but not carbon,
you can acquire a 3D HSQC-TOCSY and determine which resonance is which. This can
be acquired as 2D carbon/proton as well.

First, HSQC, record 13C shift

Then, TOCSY, detect proton

Overlap of TOCSY crosspeaks of Neomycin

If the dispersion in 13C chemical shift is used, then the TOCSY crosspeaks would be
resolvable.
13
C HSQC Neomycin
2) HCCH TOCSY (a 13C-13C TOCSY)

Similar to HSQC-TOCSY in that it is an HSQC first then a TOCSY, except it uses the
large 13C-13C coupling for the TOCSY. Must be 13C labeled though. 13C-13C coupling is
40-80 Hz so signal-noise is much better than in standard proton TOCSY, also there is no
torsion angle dependence, so correlations can be observed even when proton-proton three
bond scalar coupling is 0 Hz, e.g. when the torsion angle between 2 protons is ~90°.
Common experiment on labeled nucleic acids to assign the ribose/deoxyribose and on
proteins to assign the side chain.

First, transfer magnetization from proton to carbon as in HSQC

Then, do 13C-13C TOCSY

Then, transfer magnetization back to proton and detect

Through-Bond Experiments without a diagonal

Too numerous to discuss.

Essentially, these experiments are developed to have one or two correlations per
repeating unit to assign the proton spectrum. Many of these can be acquired as 2D
experiments. Many require either 13C or 15N labeling, but some can certainly done in
natural abundance especially with a cryo probe.

DNA/RNA

1) HCNCH/HCN

2) Exchangeable Correlations of RNA/DNA base

3) Backbone HCP Correlation

Proteins

Backbone correlations
 Assignment of NMR Spectra of Macromolecules

Proteins:
Protein assignments depend primarily upon 3D experiments correlating 1H, 13C and 15N:

Normally, one starts with the amide proton-nitrogen correlation (HSQC), as the amide
proton and nitrogen are common to all amino acids except proline and well dispersed in
proteins.
Protein Backbone Correlations- Protein Assignments
HNCA
 Nucleic Acids

Usually, you start assignments with the protons on the base of the nucleotide and the
anomeric proton (H1') as those are the most dispersed in chemical shift.

HCN

Magnetization is transferred from H8 to C8 to N9 and from H1' to C1' to N9. Correlations

are observed between H8 and N9 and H1' and N9 in 2D (in 3D H1', C1', N9 and H8, C8,
N9). There are many different versions of this sequence and the base to sugar correlation.

HCCH TOCSY

Magnetization is transferred from H1' to C1' (also H2' to C2' ...) then C1' to C2', C3' C4',
C5' then from C2' to H2', C3' to H3' ... Correlations are observed
H1'/C1'/H2',H3',H4',H5',H5"; H2'/C2'/H1',H3',H4',H5',H5" ...
 Nucleic Acid Backbone Correlations (1H/13C/31P)

3D HCP
Two (1H/31P) or three dimensional (1H/13C/31P) experiment similar to HNCA for proteins

Correlates H4' - C4' - P (i)

H4' - C4' - P (i+1)
H5' - C5' - P (i)
H5" - C5' - P (i)
H3' - C3' - P (i+1)
H2' - C2' - P (i+1)
2D 1H/31P Heteronuclear COSY (or HP-COSY)

Correlates
H5' - P (i)
H5" - P (i)
H3' - P (i+1)
(H4' - P (i))
 Measurement of Coupling Constant in 2D

Coupling constants can be measured in 2D experiments in several ways, the most

common being measuring the frequency difference between 2 resonances. Another
method is to measure the intensity of the crosspeak of interest and compare that intensity
to the diagonal peak in a specific spectrum or similarly measure the intensity of a
resonance and compare to the intensity of the resonance in a reference spectrum. Only the
standard measuring the frequency difference method will be discussed.

Previously, the measurement of 1H-1H coupling constants with DQF-COSY was

discussed in the section of DQF-COSY.

Measuring Heteronuclear 1JCH/1JNH with an HSQC:

 Dipole-Dipole Coupling

Dipole-Dipole Coupling = dipolar coupling

Basically, this direct interaction of two spins with the magnetic field

∆B = 3µ(3cos2θ - 1) r-3(µ0/4π)

∆B is the splitting induced by the magnetic field or dipolar coupling

µ is the magnetic moment of the nucleus (proton)
θ is the angle between the vector connecting the two nuclei (protons) and the magnetic
field
r is the distance between the two nuclei
Dipole-Dipole couplings can be 1000s of Hz (normal scalar couplings are
10s or 100s of Hz):

Methylene Chloride
 Why do you not normally observe dipole-dipole couplings?

∆B = 3µ(3cos2θ - 1) r-3(µ0/4π)

In solution, molecules normally are tumbling freely, or isotropically.

In solids, molecules are not freely tumbling, and dipolar couplings are observed. They
can be very useful or very annoying as there is much information in them- both a distance
and an angle term, but they are huge, and can easily get in the way, so 1H decoupling is
often turned on to decouple the spectra.

Is there a way to observe them in solution?

By inducing anisotropic tumbling of the molecule in solution.

One way (including the example above) is with liquid crystals.

In Biological NMR, Bax (NIH) developed a method of putting protein solutions in

detergent micelles, inducing partial orientation on the sample. Pardi (U. Colorado)
accomplished the same on nucleic acid samples with rod like phage particles.

What use are dipolar couplings?

More constraints for structure calculation of macromolecules:

 Scalar Coupling Constants Across a Hydrogen Bond

Hydrogen Bonds can be directly detected.

Small scalar coupling constants have been observed and measured for nuclei involved in
Hydrogen bonds.

Essentially, the experiments used to detect Hydrogen bonds are COSY type experiments,
either heteronuclear or homonuclear.

Primarily these experiments have been applied on nucleic acids, proteins and protein-
nucleic acid complexes, but in principal are applicable to any molecule with Hydrogen
bonds.

The experiments are much more sensitive for N-H•••N Hydrogen bonds using 15N/1H
correlations; but could use 15N/15N correlation or 1H/1H correlation for detection. Also,
N-H•••O(-CH) or O-H•••O(-CH) using 1H/1H or 1H/13C correlations are possible, but
very weak.

Cytosine Guanine

JHN COSY-
Transfers magnetization from 1H to 15N (across the Hydrogen bond, uses 1H - 15N
coupling constant across Hydrogen bond) and back the 1H; in this case from the imino 1H
(H1) of Guanine to N3 of Cytosine (and N1 of Guanine) and then back. What is observed
is an 1H/15N HSQC with peaks at the 1H chemical shift of the Guanine H1 and the
nitrogen chemical shift of both the Cytosine N3 and Guanine N1.
 Uracil Adenine

JNN COSY

Transfers magnetization from 1H to 15N to 15N (across the Hydrogen bond, uses 15N -15N
coupling constant across Hydrogen bond) and back the 1H; in this case from the H2 of
Adenine to the N1 of Adenine to the N3 of uracil (across the Hydrogen bond) and back to
the N1 of Adenine to H2 of Adenine (an alternative would to detect the imino (H3)
proton of uracil rather than going back to the Adenine). What is observed is an 1H/15N
HSQC with peaks at the 1H chemical shift of the Adenine H2 and the nitrogen chemical
shift of all of the Adenine N3, Adenine N1 and Uracil N3.
 TROSY

TROSY = Transverse Relaxation Optimized SpectroscopY

Transverse Relaxation = T2

As molecular weight increases, T2 relaxation times get shorter. As T2 gets shorter, the
efficiency of through-bond correlation experiments gets worse. The problem with
studying large proteins (>30,000 MW) is not just the amount of resonances and the
broadening of the resonances dues to slow tumbling (which is correlated to T2), but also
the loss of signal in correlation experiments from T2 relaxation (not only are there alot
resonances that are overlapped but they are broad making them more overlapped and the
intensity is less because signal is lost during the magnetization transfer steps due to T2).

The goal of TROSY is to limit the effect of T2.

Essentially, TROSY is another way to transfer magnetization from one nucleus to another
(proton to carbon, proton to nitrogen), and can be incorporated into any of the
heteronuclear correlation experiments (HSQC, HNCA...)

Normal HSQC:
Normal HSQC (without decoupling small molecule):

Normal HSQC (without decoupling large molecule):

TROSY (large molecule):

Applications of TROSY:

The TROSY effect has now been applied to all pulse sequences relevant to
macromolecules.

The TROSY effect is greater for larger molecules.

The TROSY effect is enhanced at very high field for 15N, and sometimes for 13C;
theoretically, 1 GHz would produce the maximal TROSY effect for 15N/1H correlations.

TROSY is thus most useful on high molecular weight molecules at high field for 1H/15N
correlations. An important use is thus a quick examination of protein structure. Acquire
1
H /15N-TROSY-HSQC on large protein, then bind small molecule drugs to see possible
changes in spectra upon binding. Drugs that don't bind should cause no changes in
spectra, those that bind specifically should cause dramatic changes of a few residues.
 Structure Calculation
NMR parameters can be used to determine three-dimensional structures of small
molecules up to moderate molecular weight proteins, nucleic acids, sugars, protein-DNA
complexes, protein-RNA complexes, etc.

Covalent structural information (Two-dimensional) can be inferred from correlations

(through-bond experiments) of protons to each other and other NMR active nuclei
(including 13C, 15N, etc.). Generally, this information is contained in TOCSY, DQF-
COSY, COSY, Relay, HSQC, HMQC, HMBC, as well as the specialized experiments for
macromolecules- HNCA, HNCO... (for proteins), HCP, HCN... (for nucleic acids).

Three-dimensional structure information comes from both through-bond and through-

space experiments.

Through-Bond:

3-bond scalar coupling constants (mostly proton-proton, but sometimes proton-carbon or

nitrogen or phosphorus etc.) imply torsion angle using the Karplus relationship. In
general, cis and trans result in maximal coupling constant.

Why are coupling constants rarely enough information by themselves to determine

three-dimensional structure of even higher molecular weight small molecules?

1) You would need at least 2 measurements for every three-bond torsion angle in the
whole molecule (you cannot distinguish cis and trans very well).

2) Precise and accurate measurements are difficult due to a number of factors including
overlap of resonances and dynamics.

Dynamics: If there is motion, the coupling constant will assume average value

3) The Karplus curve is calibrated based upon model compounds so you would need a
well calibrated curve for each different torsion angle you are measuring to be precise.

4) A few degrees of error of each measurement leads to large error over the whole
molecule.
Through-Space

NOEs/ROEs between protons are distance dependent.

In small molecules, NOEs are often used for distinction between isomers etc. that are
similar or the same through-bond but different through-space. In large molecules, NOEs
are used as distance meters.

Since NOE Intensity is proportional to r-6, NOE intensity can be used for three-
dimensional structure determination. Torsion angle only yield local structural
information, NOEs yield global structural information. The two measurable parameters
are complementary to each other.

Why can't the NOE be used as a precise measure of distance?

1) Precision of measurement

2) Spin Diffusion

3) Overlap of NOEs prevent accurate measurement of all NOEs

4) Strongly coupled protons have distorted NOE intensity

5) Dynamics. NOE intensity is biased toward overly intense NOEs because intensity is
proportional to r-6. Fast Dynamics usually leads to overly intense NOEs; intermediate
dynamics can lead to overly weak NOEs; slow dynamics leads multiple sets of NOEs.
If neither torsion angles nor NOEs are precise enough, then how do you use the
information for structure determination of a molecule by NMR?

A range of acceptable values is inputted to a computer, NOEs are usually classified as

strong, medium or weak based upon intensity at various NOESY mixing times. Torsion
angles are also given a range of acceptable values in accordance with the measured scalar
coupling constant.

e.g.:

Strong NOE: 1.8 - 3.0 Å

Medium NOE: 2.5 - 4.0 Å
Weak NOE: 3.5 - 6.0 Å

Torsion angle with small 3JHH = 90° ± 30°

Torsion angle with large 3JHH = 180° ± 30° or 0° ± 30°

These are referred to as constraints (or distance constraints and dihedral or torsion angle
constraints)

A computer program uses these constraints and other constraints to output structures- a
structure calculation

Other constraints:

1) Covalent structure

2) Van Der Waals interactions to prevent atoms from being to close and sometimes from
being too far away (repulsive and attractive)

3) Bond lengths/bond angles- C-H, N-H etc. bond lengths/angles are defined

4) Electrostatics
Other experimentally determined constraints:

1) Hydrogen Bonds
Scalar Coupling Constants can be observed across a Hydrogen-bond; these can then be
used as a constraint in the calculation

2) Dipolar Coupling Constants

Dipolar coupling constants have both a distance and an angle term, and are potentially
very powerful for structure determination

3) Chemical Shift
Chemical shift is a very sensitive indicator of local structure, but difficult to interpret the
meaning of small changes

What does the computer do with the constraints?

The computer attempts to satisfy all constraints inputted into the program. Any constraint
that is not satisfied will cause a violation of the constraint, which is assessed a penalty,
normally expressed in terms of energy. The computer attempts to minimize the energy of
the structure. At the end of the calculation the computer outputs a three-dimensional
structure. Normally, the same calculation is accomplished many (~100) times yielding a
family of structures that satisfies the constraints equally well.
NMR Solution Structure of RNA-drug Complex
 Macromolecule-Ligand Interactions

NMR can be a very effective technique to study macromolecule-ligand interactions,

studying either the macromolecule or ligand or both

Macromolecule

Chemical Shift:
Use chemical shift to identify ligands that bind and NOE to determine what part of the
ligand is interacting with the specific residues of the macromolecule. This is often
referred to as chemical shift mapping.

Chemical shift (1H, 13C or 15N) is a very sensitive indicator of local structure

The most efficient way of scanning the effect of ligand binding on the structure of a
macromolecule is through a 1D if there is sufficient dispersion else an HSQC.
SAR by NMR (Structure Activity Relationship by NMR- Fesik et al.)
15
N/1H HSQC of FKBP (FK506 binding protein bound to tethered ligand)

This protein, which is called the FK506 binding protein (FKBP), inhibits calcineurin (a
serine-threonine phosphatase) and blocks T cell activation (4) when it is complexed to
FK506
NOE:

NOE experiments are used to identify which protons of the ligand are close to which
protons of the macromolecule, but proton assignments are thus required

Since resonance assignment of small molecules is trivial compared to macromolecule,

simply assigning the small molecule ligand in combination with NOE experiments is
enough to identify what portion on the ligand is interacting with the macromolecule

How can you separate NOEs into ligand-ligand, macromolecule-macromolecule, ligand-

macromolecule?

1) Chemical shift-

Macromolecules and ligands do not have resonances in all regions of the NMR spectrum
so NOEs in those regions are easily sorted in intramolecular and intermolecular
(nucleic acids have few resonances < 4.0ppm; proteins have few resonances between
5.5 ppm and 7.0 ppm).

Section of 2D NOESY of RNA oligonucleotide-paromomycin complex

RNA has only a few resonances upfield of 4.00 ppm. Paromomycin has a number of
resonances between 3.0 and 4.0 ppm, and only 3 resonances between 5.0 and 6.0 ppm.
2) With 13C or 15N Labeled macromolecule (or ligand) and non-labeled (so 12C and 14N)
ligand (or macromolecule), acquire a NOESY that filters out 12C (or 14N) bound protons,
approximately a NOESY-HSQC or HSQC-NOESY-HSQC.

NOESY-HSQC (3D NOESY or could be acquired as 2D NOESY with HSQC filter

12
C-HA HB-13C
12
C-HC HD-13C

Assume that all protons are close enough to observe an NOE, HB and HD are from
macromolecule, HA and HC are from small molecule ligand. In a NOESY-HSQC (13C),
only the HA-HC crosspeak (so ligand-ligand) will not be observed, since neither are
attached to 13C.

2D Plane at ~76 ppm of 3D NOESY-HSQC (13C) of RNA oligonucleotide-paromomycin

complex

Only crosspeaks from 13C labeled RNA are possible. In 3D at ~76 ppm, only crosspeaks
from H2' and H3' are possible as only C2' and C3' on ribose resonate between 70 and 80
ppm.
HSQC-NOESY-HSQC

In an HSQC-NOESY-HSQC there are two HSQC steps so the magnetization is

transferred from 1H (detect 1H) to 13C (detect 13C) then back to 1H, then NOESY, then 1H
to 13C (detect 13C) then back to 1H then acquire. That would be a 4D NOESY,
alternatively either or both 13C detection periods could be left out to make a 3D or 2D
spectrum; whether it is a 2D, 3D, or 4D, the only magnetization detected will be 13C
bound 1H to 13C bound 1H.

In the Example above, the only crosspeak detected would be HB-HD.

Alternatively, the 13C could be used as a filter, meaning that instead of detecting 13C
bound 1H to 13C bound 1H, one of the HSQC steps could be used to selectively NOT
detect the 13C bound 1H; that way, you would observe 13C bound 1H to 12C bound 1H
ONLY. In the example above, that would mean that HA-HB, HA-HD, HC-HB, and HC-HD
crosspeaks would be observed. This is referred to as an X-Filtered NOESY or 13C-filtered
NOESY.

Similarly, both HSQC periods could be used to filter out 13C bound protons, that way
only the 12C bound 1H to 12C bound 1H NOEs would be observed or in the case above,
HA-HC.
3) Linewidth

If the ligand is much smaller than the macromolecule (>10X MW) and the ligand is in
fast exchange (affinity is weak; binding constant is ~10-3M or weaker), then the
resonances of the ligand will be sharp compared to the macromolecule (Affinity and
Exchange will be discussed later). Thus, ligand-ligand NOEs will be sharp in both
dimensions of a 2D NOESY, macromolecule-macromolecule NOEs will be broad, and
ligand-macromolecule will be broad.

The ligand-ligand NOEs will be obvious; since the chemical shifts of the ligand will also
be obvious, the ligand-macromolecule NOEs will be moderately easy to pick out.

4) Ligand MW ~800 or less

If the ligand is much smaller than the macromolecule (>10X MW) and the ligand is in
fast exchange (affinity is weak; binding constant is ~10-3 or weaker), then the resonances
of the ligand will be sharp compared to the macromolecule Ligand-ligand NOEs will be
opposite in sign to the diagonal. Macromolecule-macromolecule NOEs are the same sign
as the diagonal.
 Transfer NOE

Transfer NOE is a specific NOE experiment used to determine the conformation of the
small molecule ligand in the bound conformation. The binding of the ligand to the
macromolecule must be weak, because within some hundreds of milliseconds, the ligand
has to bind to the macromolecule, remain bound for some time (at least tens of
milliseconds) and then dissociate again to yield a transfer NOE signal. Through this
technique the observed signals are as narrow as the ligand signals, but the NOESY
spectrum contains NOE cross peaks for only the bound state that build up during the time
when the ligand is bound to the protein.

Whether transfer NOEs can be observed has to be tested for each case, but the binding
constant should normally be somewhere around ~ 1 mM.

What can thus be determined with the Transfer NOE experiment is not where the small
molecule binds on the macromolecule or even what groups on the small molecule are
interacting with the macromolecule, instead what the Transfer NOE experiment gives you
is the three dimensional structure of the ligand in the bound conformation.
 Chemical Exchange

Chemical Exchange rate of a macromolecule-ligand complex is correlated to the affinity

of the complex:

Macromolecule + Drug ↔ Macromolecule-Drug

KA = [Macromolecule-Drug]/[ Macromolecule]*[ Drug]

KD = 1/KA

In the NMR tube, the macromolecule concentration will be ~1mM. Thus, there is
virtually no free macromolecule when KD ~1 µM. However, if KD ~1 mM, then there
will be a mixture of free and bound ligand with equal concentration of macromolecule
and ligand.

KD can be correlated to rate by:

KD = R-1/R1

If the binding is diffusion limited, then k1 ~ 107M-1sec-1

Thus, a complex with KD ~1 µM would have a K-1 = 10 sec-1 (or a time constant of 100
ms). A lower affinity complex of KD = 1 mM would thus have a time constant of 100 µs.

Chemical Shift Difference and Chemical Exchange:

Ligand binds to the macromolecule, induces a shift of one proton 0.2 ppm (as stated
earlier, chemical shift is very sensitive to local environment. Chemical shift changes of
several ppm or 0 ppm are possible, but for the example we are using 0.2 ppm). On a 500
MHz spectrometer 0.2 ppm = 100 Hz = 100 sec-1, results in a time constant of 10 ms. A
complex with KD = 1 mM, would have a time constant of exchange of 100 µs, which
would be fast in comparison to the chemical shift difference of free and bound. A
complex with KD = 1 µM, would have a time constant of exchange of 100 ms, which
would be slow in comparison to the chemical shift difference of free and bound. A
complex with KD = 10 µM, would have a time constant of exchange of 10 ms, which
would be intermediate in comparison to the chemical shift difference of free and bound.
 What do you see in the NMR tube for different affinity complexes?

Paromomycin Binds an RNA Oligonucleotide in Slow Exchange

 What do you see in the NMR tube for different affinity complexes?

Ribostamycin Binds an RNA Oligonucleotide in Fast/Intermediate Exchange

G1491 and U1490 each shift as in fast exchange, but are slightly broader in
the complex as in intermediate exchange. U1406 and U1495 both broaden as
in intermediate exchange.
 What do you see in the NMR tube for different affinity complexes?

Neomycin Binds an RNA Oligonucleotide in Slow/Intermediate Exchange

G1491 and U1490 each shift as in slow exchange with two resonances
observed, but are slightly broader in the complex as in intermediate
exchange. U1406 broadens as in intermediate exchange. U1495 shifts as in
fast exchange.
 Structure Calculation of Macromolecule-Ligand Complex

Structure Calculation of Macromolecule-Ligand Complex is accomplished in the

essentially the same way as for a standard 3D structure calculation with NMR. The only
major difference is the need for constraints to put the two structures together. The
intermolecular constraints have to be through-space, so that means NOEs between the
ligand and the macromolecule primarily (usually referred to as intermolecular NOEs),
and potentially Hydrogen bonds (intermolecular Hydrogen bonds). Dipolar coupling
constraints potentially are useful as well, but have not as yet been applied to both the
macromolecule and ligand.

Intermolecular NOEs

Ligand-Macromolecule Complex Exchange: Intermolecular NOEs are normally hard to

observe in complexes that are in intermediate exchange as the lines are broader, so there
is more overlap, the intensity of the NOE is less (even in the NOE volume is equivalent)
and usually the volume is also less as there is loss of signal due to T2 relaxation. In fast
exchange, the lines are sharp, but the molecules are bound for only a fraction of the time.
To get around this problem, large amounts of ligand are added to force the complex to be
bound most of the time (so instead of a 1:1 ratio of ligand:macromolecule, a 3:1 to 5:1
ratio are used).

NOE Volume: The NOE volumes of intermolecular NOEs are often weaker than
corresponding intramolecular NOEs of protons separated by equivalent distance due to
the exchange of the complex. Thus, when inputting intermolecular constraints, wider
acceptable ranges (for example instead of 1.8-3.0 Ǻ, 1.8-4.0 Ǻ for the strongest
intermolecular NOEs) are often used.

3D Structure Precision: The precision (or root mean square deviation from the average
structure) of the complex will depend heavily upon how many intermolecular constraints
there are. The individual structures (the structure of the macromolecule in the complex
and the structure of the ligand in the complex) can be precise, but without many
intermolecular constraints the whole complex will not be precise.