Transcript Lecture Slides (Powerpoint)

The NMR Fingerprints of Proteins:
What you can see in simple spectra
• typical chemical shifts observed in proteins
• interpreting simple 1D spectra
• interpreting 15N-1H 2D correlation (HSQC)
spectra
• using changes in HSQC spectra to measure
binding events
Chemical shift
not all nuclei in a protein will have the same resonance frequency
(the spectrum would be pretty uninformative if they did!)
a reference frequency of
0 ppm is defined by the signal
from an internal standard
such as DSS or TMSP
x axis is the chemical
shift d, in units of
parts per million (ppm)
of the B0 field strength
chemical shift dispersion is very small (~12 ppm)
compared to the B0 field strength (like ripples on an
ocean surface). At 500 MHz, 12 ppm is a 6 KHz
range. This makes it easy to pulse in the center of the
1H spectrum, around 4.5-5 ppm, and excite all
resonances nearly evenly.
so why don’t
all the nuclei
have the
same frequency?
Shielding
electrons surrounding
the nucleus create a
magnetic field which
opposes the B0 field
and reduces the
effective field felt by
the nucleus. This is
called diamagnetic
shielding.
fr. Macomber p. 69
Simple shielding effects--electronegativity
The amount of shielding the nucleus experiences will
vary with the density of the surrounding electron cloud
If a 1H nucleus is bound to a more electronegative atom
e.g. N or O as opposed to C, the density of the electron
cloud will be lower and it will be less shielded or
“deshielded”. These considerations extend beyond what
is directly bonded to the H atom as well.
N
H
more electron
withdrawing-less shielded
C
H
less electron
withdrawing-more shielded
Simple shielding effects--electronegativity
less shielded
more shielded
higher resonance frequency lower resonance frequency
amides (HN)
aliphatic/alpha/beta etc.(HC)
most HN nuclei come between 6-11 ppm while most
HC nuclei come between -1 and 6 ppm.
More complex shielding effects
Groups such as carbonyls and aromatic rings have associated
anisotropic fields or “shielding cones” that will either shield or
deshield nearby 1H nuclei depending upon where the 1H nuclei are
located relative to the shielding cone.
shielding
shielding
attached
protons
are deshielded
deshielding
deshielding
deshielding
deshielding
shielding
shielding
More complex shielding effects:
Aromatic protons
aromatic region (6-8 ppm)
amide region (7-10 ppm)
One consequence of these effects is that aromatic protons, which are
attached to aromatic rings, are deshielded relative to other HC protons.
In fact, aromatic ring protons overlap with the amide (HN) region.
It should now be apparent to you that different types of proton in
a protein will resonate at different frequencies based on simple
chemical considerations. For instance, Ha protons will resonate
in a region centered around the relatively high shift of 4.4 ppm,
based on the fact that they are adjacent to a carbonyl and an
amine group, both of which withdraw electron density. But not all
Ha protons resonate at 4.4 ppm: They are dispersed as low as
~3 and as high as ~5.5. Why?
“Ha region”
“Average” or “random coil”
chemical shifts in proteins
One reason for this dispersion is
that the side chains of the 20 amino
acids are different, and these
differences will have some effect on the
Ha shift.
The table at right shows “typical” values
observed for different protons in the 20
amino acids. These were measured in
unstructured peptides to mimic the
environment experienced by the proton
averaged over essentially all possible
conformations. These are sometimes
called “random coil” shift values.
Note that the Ha shifts range from ~44.8, but Ha shifts in proteins range from
~3 to 5.5. So this cannot entirely explain
the observed dispersion.
Amino acid structures and chemical shifts
note: the shifts are somewhat different from the
previous page because they are measured on the free amino
acids, not on amino acids within peptides
A simple reason for the increased shift dispersion is that the
environment experienced by 1H nuclei in a folded protein (B) is not
the same as in a unfolded, extended protein or “random coil” (A).
shift of particular proton in
unfolded protein is averaged
over many fluctuating structures
will be near
random coil
value
shift of particular proton in folded
protein influenced by
groups nearby in space,
conformation of the backbone,
etc. Not averaged among many
structures because there is only
one folded structure.
So, some protons in folded
proteins will experience very
particular environments and will
stray far from the average.
Example: shielding by aromatic side
chains in folded proteins
Picture shows the side
chain packing in the
hydrophobic core of a
protein--the side chains are
packed in a very specific
manner, somewhat like a
jigsaw puzzle
shielded methyl
group
methyl region
of protein spectrum
+
+
a consequence of this packing is
that some protons may be
positioned within the shielding
cone of an aromatic ring such as
Phe 51. Such protons will exhibit
unusually low resonance
frequencies (see picture at left).
Note that such effects depend
upon precise positioning of side
chains within folded proteins
so you can tell if your protein is folded or not by looking at the 1D spectrum...
poorly
poorly
dispersed amides
dispersed alphas
poorly
dispersed aromatics
unfolded
poorly
dispersed methyls
ubiquitin
folded
ubiquitin
very shielded
methyl
What specifically to look for in a nicely
folded protein
notice
notice alpha protons
aromatic/amide
with shifts above 5
protons with
shifts above 9
notice all these methyl
peaks with
and below 7
chemical shifts around
zero or even
negative
Linewidths in 1D spectra: aggregation and
conformational flexibility
Linewidths get broader with larger
particle size, due to faster
transverse relaxation rates. We’ll
learn the physical basis for the faster
relaxation later. Broader than
expected linewidths can indicate that
the protein is aggregated. It can also
indicate that the protein has
conformational flexibility, i.e. that its
structure is fluctuating between
several slightly different forms. We’ll
learn why this is when we cover the
effect of protein dynamics on NMR
spectra. Conformational flexibility
also tends to reduce dispersion by
averaging the environment
experienced by a nucleus.
An example of analyzing
linewidths and dispersion:
Hill & DeGrado used
measurements of chemical
shift dispersion and line
broadening in the methyl
region of 1D spectra to
gauge the effect of
mutations at position 7 on
the conformational flexibility
of a2D protein
leucine and valine mutants have poor
dispersion and broad lines, despite being very stably folded
and not aggregated (circular dichroism and analytical ultracentrifugation measurements). These mutants are folded
but flexible.
Hill & DeGrado (2000) Structure 8: 471-9.
Limitations of 1D NMR
In general, 1D NMR provides only qualitative information about
your protein:
Does it have a stable, specific folded structure under the NMR
conditions?
Does it seem to be aggregated?
Spectra of even small proteins (e.g. 6 kD), unlike the spectra of
small organic molecules and short peptides, are just too
complex to be studied by 1D methods.
Adding a second dimension: Isotopic labelling
There are many overlapping resonances in 1D protein spectra. One
way to remove this overlap is to label your protein with 15N and/or 13C
and correlate the chemical shift of each 1H nucleus with the chemical
shift of the 15N or 13C atom to which it is directly attached.
This is done by transferring the magnetization between the two
atoms using the large one-bond 13C-1H or 15N- 1H scalar coupling
15N---1H
J = ~89-95 Hz
13C---1H
J = ~110-160 Hz
HSQC: Simple 2D Heteronuclear Correlation Spectrum
Heteronuclear
Single
Quantum
Coherence
1D 1H spectrum of same region
peaks occur where the
chemical shift of a 1H and the
chemical shift of the attached
15N atom intersect
**
note less overlap
in 2D!
the ability to spread the
spectrum out into a second
dimension, thus achieving
better resolution of the
resonances, is a major
reason for isotopic labelling
aromatics in 1D e.g. * not visible in HSQC due to
isotope editing--they aren’t bound to 15N!
Appearance of 15N-1H Correlation Spectra (HSQC)
glutamine and
asparagine peaks
good shift
Since each amino acid
dispersion in both
residue (except proline)
dimensions--folded protein
has an amide proton, we
might expect about the
same number of peaks
as residues. This is roughly
true, but some side chains,
like glutamine, asparagine
and arginine have amides too,
so there will be more peaks
than residues.
The glutamine and asparagine
peaks are especially
recognizable--they are pairs of
1H shifts correlated to a single
nitrogen in the upper right
portion of the spectrum
This is about a 60-residue protein-->small
15N-1H
HSQC spectra as “protein fingerprints”
because there is
about one peak per
residue, a 15N-1H HSQC is
something of an NMR
fingerprint of a protein.
HSQCs are very commonly
used to detect ligand
binding--if the fingerprint
changes (peaks move) it
indicates that binding is
occurring. This is the basis
of “SAR by NMR”, about
which you read in Shuker et
al.
some peaks
move, some don’t-binding to specific region
HSQC in absence (magenta) and
presence (black) of ligand
Notice that SAR by NMR uses an
approach whereby two ligands are sought
which bind to two different places with
micromolar affinity. They are then linked
together to produce a single ligand with
nanomolar/picomolar affinity. One can tell
that the two ligands bind in different places
because different peaks move, or because
one will bind in the presence of the other
(no competitive inhibition).
Why don’t they just screen for a single
ligand with nanomolar/picomolar affinity in
the first place?
Structural interpretation of changes in HSQC
upon binding requires resonance assignment
Notice that the peaks which move upon binding have been labelled with
the residue names and numbers--if they want to know where on the protein
the ligand is bound they need to know which peak corresponds to which
amino acid residue. How is this determined? This is the fundamental
problem of resonance assignment, which we will cover a few lectures later.
compound binds
in this region