Transcript No Slide Title

Protein NMR III - Other types of structural data
• In the last two classes we looked at how we get structural
information from basically two sources - 3J couplings and
NOEs enhancements (correlations).
• NOEs gave us approximate distance information, and 3J
couplings could be transformed into dihedral constrains.
• NMR spectra have a lot more information than that, which we
usually dump. First, some of the information was originally
fudged to make it work better with current MM programs of
those days (couplings into dihedrals…).
• Today we’ll see how we can employ some of the NMR data
in a better fashion, as well as use other information obtained
from NMR. As we said before, as long as we can get a
relationship between the NMR derived parameter (S) and the
geometry of the atoms involved, we can use it in MM:
Scalc = f(xyz)
ES = KS * f [ (Scalc - Sobs) ]
• A physicist can tell you that ALL NMR observations depend
entirely on the geometry of the molecular system, so there is
an equation for everyone. The problem is to find them and
parametrize them.
Direct use of coupling constants
• Couplings are perhaps the easiest ones to start with. They
were not included as they where originally because the MM
programs did dihedral constraints easier.
• As we saw last time, this had the disadvantage of creating
ambiguities on the number of possible dihedrals for a certain
coupling constant.
• The assumption that only certain angles are allowed is fine in
globular proteins (for which the X-ray trends were found), but
it is a big no-no if we are dealing with small flexible peptides
or peptides containing unnatural amino acids.
• The best thing to do would be to include directly the coupling
constant as part of an energy term of our MM force field. This
is what we do, and it works like a charm…
EJ = KJ * ( Jcalc - Jobs )2
Jcalc = A * cos( f )2 + B cos( f ) + C
• The computer back-calculates the 3J coupling using the
current dihedral angle and compares it to the observed value.
Since we don’t choose any particular angle, we can use a
single value instead of a range (simple quadratic function)
Use of chemical shifts
• What about chemical shifts? After all, we have chemical shifts
because we have different conformations for different amino
acids in the peptide.
• However, nobody really cared about them until recently. The
main problem is that, as opposed to couplings, rules or
parameters for chemical shifts can only be used in regular
structures.
• Since nobody looked at proteins by NMR until the mid ‘80s,
there were no good parametrizations or good reference data.
• The idea is that we can assign a random coil chemical shift
value to all the protons in an amino acid. Any deviation from
it, or secondary shift, arises from different effects:
a) Peptide group anisotropy. The local magnetic field of the
peptide group (CO-NH) will make protons lying above or to
the side be shifted up- or down-field.
H
N
r
C
C
O
f
spga = CCO * r-3 * [ 1 - 3 * cos ( f )2 ]
Use of chemical shifts (continued)
b) Ring current effects. The local magnetic field created by the
e- current of aromatic rings will cause protons lying above or
to its the side be shifted up- or down-field. This example is
archetypal and you’ll find it in every organic chemistry book.
H
r
f
src = Cring * r-3 * [ 1 - 3 * cos ( f )2 ]
c) Polarization of C-H bonds by polar/charged groups. The
electron cloud of the s bond goes back or forth the C-H bond
depending of the presence of groups of different polarity
aligned with them:
qi
r
C
-
H
qi
f
r
C
Upfield shift
+
H
Downfield shift
selec = C * r-2 * qi * cos ( f )
f
Use of chemical shifts (...)
• So, since we have equations for each effect, we can calculate
it to a certain degree of accuracy in the computer. If we know
both the random coil and the experimental value we can tell
the MM program to make the calculated mach the observed
values or else put an energy penalty:
Es = Ks * [ ( dobs - drandom ) - ( spga + src + selec ) ]2
dobs - drandom is the secondary shift
• This works great in some cases. Those of you who endured
my talk in 9/98 probably remember some of this…
Without d constraints
With d constraints
Finding Bonds, H-bonds…
• A hydrogen bond (HB) allows chunks of peptide relatively far
away from each other to come close together. They are all
over the place in globular proteins, so if we could identify
were they are (donor and acceptor atoms), we have a huge
constraint in the structure.
• In a protein the most interesting HBs are those formed
between the peptide backbone amide protons and carbonyls,
as the ones we see in a-helixes and b-sheets. We can also
see some from side the chains (Asn, Asp, Gln, Glu) to the
backbone amides or carbonyls:
• To find them we study the exchange rates of amide protons.
The idea is that labile protons (NHs) that are protected from
the solvent will not exchange as fast with solvent protons as
things that are solvent exposed.
Amide exchange rates
• Therefore, if we add D2O to our H2O solution and take
spectra at different times, we’ll see that signals from different
amide protons will decrease in size at different rates.
• Since the amide region of a 1D is way too crowded in
proteins, we normally use a quick 2D experiment, as a DQFCOSY. We look at the NH to Ha fingerprint at different times.
4.0
t = 0 - No D2O
Add D2O
4.0
(Has)
t = t1
4.0
t = t2
8.0
(NHs)
7.0
Amide exchange rates
• From this data we can tell which which amide is H-bonded
strongly, which one weakly, and which ones not at all. Since
we also have NOE and 3J coupling data, we can try to see
if these hydrogen bonded amides match with regions that
we identified previously as a-helices, b-sheets, or b-turns.
• If we can do this, then, and ONLY then, we can use a H-bond
constraint during the generation of our 3D model.
• Why the ONLY? We only now the H-bond donor, but there is
(or there was until a month ago) no way we can tell who the
acceptor atom is (the C=O oxygen). If we miss-place one of
these we screw up big time - Since we are basically cyclizing
the peptide, there is no way we can get the right structure.
• If we decide that it’s reasonable to use a H-bonding energy
penalty, we can put it into the force field more or less as a
distance constraint:
EHB = KHB * ( ri - rHB-ideal )2
• rHB-ideal is ~2.5 Å (depending on the reference). Since Hbonds also have an angular requirement (the N-H…O angle
has to be between 135 and -135), we can make a more
complex term to reflect this.
Amide temperature gradients
• Studying exchange rates works OK in proteins, because the
time in which the amides turnover is long (globular). In small
peptides this ain’t true.
• Since we have a lot more flexibility in a peptide (a lot more
contact with solvent), everything usually exchanges in
relatively short times (minutes as opposed to hours). By the
time you put some D2O in the tube, brought it to the NMR lab,
placed it in the magnet, and shimmed the sample, there are
no amide protons…
• For peptides, instead of studying the exchange rates, we
analyze the change in chemical shift of the amide protons
with change in sample temperature (temperature gradients).
• This is because the more the proton is exposed, the more it’ll
interact with solvent as we increase temperature, moving it
upfield towards water…
• We measure T gradients in parts per billion (ppb). Values
below -2 indicate shielding from the solvent (water), between
4 and 5 ppb we have partial shielding (some of the time we
have a H-bond, some of the time we don’t), and above 6 or 7
ppb indicates complete exposure.
• As with proteins, we cannot tell which one is the acceptor (the
oxygen). Therefore, we have to be really really careful using
these data…
Using ERs and TGs
• Knowing that you have a H-bond and not being able to use it
as a constraint in the model is painful.
• If we want to be safe, we can just do the whole calculation of
structures with NOEs and 3Js as we saw last time, and then
discard structures in which the NH 1H we know is H-bonded
does not appear H-bonded (use it as a check).
• The other way is to have some other sources to corroborate
that the H-bond exists (NOEs and couplings). This works
better in proteins because we have sizable a-helices and bsheets. In peptides we may have a b-turn, which is very tiny,
and may not have decent NOEs and 3J couplings.
• Or, we may do it the hard way - If we have 3 or 4 possible
H-bond acceptors, we can try each one of them in different
simulations and see at the end which one gives us the
lowest energy structures:
O
O
H
O
O
O
H
E1
O
O
O
H
O
E2
O
O
H
O
E3
Isotopic labeling
• The only nuclei that we can look at in a protein is usually the
1H. In small proteins (up to 10 KDa, ~ 80 amino acids) this is
OK. We can identify all residues and study all NOEs, and
measure most of the 3J couplings.
• As we go to larger proteins (> 10 KDa), things start getting
more and more crowded. We start loosing to many residues
to overlap, and we cannot assign the whole chain.
• What we need is more NMR sensitive nuclei in the sample.
That way, we can edit the spectra by looking at them, or,
for example, add a third (and maybe fourth) dimension.
• To do this we need several things:
a) We need to know the gene (DNA chunk) that is responsible
for the synthesis of our proteins,
b) We need a molecular biologist to get a plasmid that
overexpresses this gene, possibly in an E. coli vector
system in minimal media, so that even Joe Blow, the new
(and clueless) undergrad in the lab, can grow lots of it.
c) The overexpressed protein has to be functional (85% of the
time you get a beautiful band in the SDS-gel that are all
inclusion bodies…).
d) A good/simple purification procedure.
Isotopic labeling (continued)
• 10 to 1 that your particular protein will fail one of these
requirements in real life. But most of the time, we can work
around either overexpression, activity and purification
problems. Getting the gene is the toughest one to overcome.
• In any case, now that we have the plasmid, we grow it in
isotopically enriched media. This usually means M9 (minimal
media), which only has NH4Ac and glucose as sources of N
and C. No cell homogenates or yeast extracts.
• So, if we want a 15N labeled protein, we use 15NH4Ac (that
is dirt-cheap). Glucose-U-13C is a lot more expensive, but it is
sometimes necessary.
• In that way we get partially- or fully-labeled protein, in which
all nuclei are NMR-sensitive (13C=O, 13Ca, and 15Ns). All the
protein backbone is NMR-sensitive.
H
15
O
13
N
13
C
AA1
AA2
13
C
15
N
H
H
15
C
13
C
O
O
13
N
13
C
C
AA3
• Another cool thing is that we now we have new 3J-couplings
to use for dihedral angles: H-C-C-15N, H-N-C-13C. These have
their own Karplus parameters.
Isotopic labeling (…)
• One of the most common experiments performed in 15Nlabeled proteins is a 15N-1H hetero-correlation. Instead of
doing the normal HETCOR which detects 15N (low sensitivity),
we do an HSQC or HMQC, which gives us the same data but
using 1H for detection.
• This experiment is great, because we can spread the signals
using the chemical shift range of 15N:
7.0
1H
d
8.0
“0”
185.0
15N
d
165.0
• It is ideal for several things. One of them is measurement of
amide exchange rates.
• It is also good to do spin system identification: If we don’t
have good resolution in the COSY or TOCSY and some
signals are overlapped, we can use the 1H-15N correlation to
spread the TOCSY correlations in a third dimension…
Summary
• If we have a reliable physical description of an NMR
observable in which we can include the geometry of the
molecular system, we can fudge an energy penalty into the
MM calculations to get results consistent with the NMR
measurements. 3J, 1H ds, 13C ds (in the making…).
• We can determine if an amide is involved in hydrogen bonds
by studying the exchange rates with D2O (in proteins) or by
measuring their temperature gradients (peptides)
• Irrespective of how we find that the amide proton is hydrogen
bonded, we have to be really careful if we want to incorporate
them into the refinement step as constraints.
• If we have an engineered organism that overexpresses our
protein we can prepare it fully labeled with 13C and 15N.
This allows for the detection of almost all the nuclei of the
molecule, use of more dimensions, measurement of other
couplings and parameters, etc., etc…
Next class
• We’ll see a brief introduction to 3D-NMR (very brief).
• Information we can get from 3D experiments (read the stuff
from M. P. Williamson…).