NMR - Sample Preparation & Miscellaneous Information
A typical sample volume for the new Varian Unibody probes is ca.
0.75 ml for a 5 mm tube. For macromolecules, 650 to 700 mL is acceptable. Thin-walled
Shigemi tubes require 0.35 ml (5 mm) or 1.2 ml (8 mm).
Amounts of material needed for small molecules (e.g. metabolite mixtures)
With the cold probe, 1D spectra of aqueous (TCA) extract can be easily obtained
with 1-2 mg dry weight. For lipoid extracts or organic solvents, concentrations
of one microM can be recorded, which is equivalent to about 0.3-0.4 nmole
material (or < 1 microgram of material having a molecular weight of 500).
2D really needs concentrations in the 100 + microM range (>35 nmole).
Even low (10%) 13C enrichment can make heteronuclear experiments possible.
Amounts of material needed for macromolecules
Signal is proportional to the number of molecules in the receiver coil, so
concentration is important. As a rule of thumb, about 0.5 to 1 µmole
solute is needed for high resolution work (2 and 3D spectroscopy). For 1D NMR
for screening purposes (e.g. to verify that a protein is folded), much less
sample can be used, e.g. 0.75 ml @ 100 µM (0.075 µmole).
For a 10 kDa protein, 1 mM is 10 mg/ml, so for 1D NMR you would need
ca. 0.7-1 mg. For a 20 kDa protein, you would need at least twice that amount
of mass. For 2D, 5-10 mg of a 10 kDa protein would be needed.
These numbers can be substantially reduced where cold probes are used. For
non-lossy solvents (e.g. salt-free water or methanol), the cold probe
is >3X more sensitive than the corresponding RT probe, i.e. A saving of
around 10 X in time (or concentration for the same spectrometer time).
solubility
The protein should be soluble to ca. 1 mM without significant aggregation,
and also stable for days under the condition of the experiment. This
may require preliminary work to establish appropriate conditions of temperature,
salts and buffer ions (but see below). Often this can be done with small amounts
of material on the bench (e.g. by functional assays or CD etc.) Nucleic
acids are generally readily soluble in aqueous buffers.
The solubility of most globular proteins increases with increasing ionic
strength between 0 and ca. 300 mM (salting in). On the other hand, 300
mM KCl or NaCl degrade the performance of the detection system (dielectric
losses). This becomes worse the higher the magnetic field strength. A reasonable
compromise is 100 mM KCl. Neutral osmolytes such as glycine at 0.5 M or so
may be an alternative way of increasing solubility without changing the ionic
strength (at least between pH 5 and pH 8). Use d2Gly!
pH: avoid the isoelectric pH.
Influence of size and temperature
The line-width of a resonance is proportional to the correlation time, which
is in turn proportional to the molecular mass, the viscosity and 1/T. In addition
to decreasing resolution, broad lines mean lower sensitivity (they have lower
intensity), and decreased efficiency in nD experiments owing to loss of coherence
during the t1 period. The viscosity of water varies approximately exponentially
with temperature. At 20°C, h(H2O)=0.01 Poise, and at any other T,
h =0.01exp(2200(1/T-1/293). The viscosity of D2O is ca. 1.23 that of H2O.
25°C, H2O, c=15 mg/ml
| Mass |
tc |
NH width |
c |
relative sensitivity |
| kDa |
ns |
Hz |
mM |
|
| 5 |
2.1 |
3 |
3 |
1 |
| 10 |
4.2 |
6 |
1.5 |
0.25 |
| 15 |
6.3 |
9 |
1 |
0.11 |
| 20 |
8.4 |
12 |
0.75 |
0.063 |
| 30 |
12.6 |
18 |
0.5 |
0.028 |
| *50 |
21 |
30 |
0.3 |
0.01 |
Thus aggregation is very bad news.
*Could not be done at 14.1 T.
Note that in order to reach a given signal-to-noise ratio (SNR), signal
averaging is required. SNR increases with the square root of measurement
time. So, all other things being equal, a 20 kDa protein requires 16 fold more
spectrometer time than a 10 kDa protein at the same mass per unit volume.
The cold probe increases sensitivity ca. 2.5 fold (6x decrease in time). There
is a premium on concentration!
5 kDa, 15 mg/ml
| T/K |
tc |
NH width |
relative sensitivity |
| 278 |
3.8 |
5.5 |
0.55 |
| 288 |
2.8 |
4 |
0.75 |
| 298 |
2.1 |
3 |
1.0 |
| 308 |
1.6 |
2.3 |
1.3 |
| 318 |
1.2 |
1.8 |
1.7 |
Effects of pH
pH: avoid the isoelectric pH. Don't use buffer that have CH groups, unless
you can deuterate (very strong signals). Remove glycerol. Cysteine
is bad, because it oxidises, so DTT+EDTA may be needed. Either mutate to
Ser, react with IAA, or as a last resort use low concentration of DTT and
EDTA (<100 µM) and degas the buffers. Note that a small molecule
has a very sharp line. 1 mM EDTA will have an intensity ca. 50 times that
of a 1 mM protein of M=10 kDa. This leads to noise, spectral overlap and
dynamic range problems on the detection system. If such additives must be
used, consider buying deuteriated versions (very expensive). Even these are
typically only 99% deuteriated, so that the residual signal will still be
comparable to that of the protein.
Acceptable buffers :
K or Na phosphate (pK ca. 6.8) + KCl ( to ca. 0.1 M), d3acetate (pK≈4.6)
(+KCl to ca. 0.1 M). But consider using low conductivity buffer species on
the cold probe (e.g. deuterated Tris)
Very high pH is bad, because the amide protons that are accessible to solvent
exchange rapidly with solvent and become broad (leading to their disappearance
spectroscopically).
Isotopic substitution
Small proteins (< 10 kDa) that are monomeric can be studied in detail without
isotopic substitution. Increasing number of residues increases both the size
and the complexity of the NMR spectrum (overlap). Minimal substation is desirable.
15N
15N ammonium sulfate or chloride is cheap (ca. $35/g) and E.coli or P.
pastoris can be grown on a defined minimal medium containing controlled
N and C sources. For an M9 medium, 1.5 g/L ammonium sulfate is sufficient.
For good expression (cf. 5-10 mg/L in log phase growth) a couple of L will
give enough protein for an NMR sample (Link to Expression Laboratory).
15N substitution allows 2D and 3D edited experiments to be carried out, with
greatly increased resolution, and reliability of spectral analysis. It can
also be used to measure dynamic properties of the protein that cannot be obtained
by X-ray crystallography. So even for 10 kDa protein, this is highly
recommended.
13C
Detailed analysis of proteins in the size range 10-25 kDa requires additional
labeling with13C. Again, M9 medium supplemented with 13C6 glucose is a simple
means of achieving this. Current (June 2004) bulk price of glucose is
ca. $100/g. Minimal quantity is ca. 2 g/L, and maybe up to 4 L culture are
required for this. An NMR sample doubly labeled will cost ca. $1000. The
double label allows triple resonance experiments to be carried out, which gives
unambiguous spectral analysis.
Note that homodimers of small subunits also require full labeling and the formation
of mixed isotopic dimers for analysis the dimerisation interface.
2H
Larger proteins (M>25 kDa) suffer from broad lines in the proton spectrum. This can be alleviated
by isotopic dilution with 2H, which is a much less active NMR nucleus. This
requires growing the cells in a deuterated medium. For some experiments, the
cells have to be trained to grow on >95% D2O, with correspondingly increased
growth times and reduced cell yield/expression. 4 L D2O costs ca. $1200, so
a fully labeled sample will cost ca. $2200 just in isotopes.
Spectrometer time needed
A 1D spectrum can be acquired very quickly. The main amount of time needed
is to equilibrate the temperature, tune the probe, shim and calibrate pulse
power.
A 2D spectrum is a series of 1D spectra. Typically 600 1D spectra are recorded,
so it takes a few hours. At 1 mM sample, one can expect a reasonably
good moderate resolution spectrum in about 10-12 hours. A high resolution spectrum
will require up to 24 h.
Several 2D spectra are needed for full analysis. For the 10 kDa case, without isotopes, one needs
a good COSY, TOCSY and a couple of NOESY spectra as a minimum. So the 2D work
needs at least 2 days, and probably closer to 3 days spectrometer time. A pH
or ligand titration would probably take about 12 hours to complete.
3D spectra are series of 2D spectra, and the data sizes can become rather large.
A moderate 3D experiment would occupy about 32 Mwords of storage (raw
data). The experiment takes 2-3 days to record. Several 3D spectra are
needed for full analysis. For 15N only, TOCSY HSQC and 2 NOESY-HSQC spectra
amount to around 8 days of time. A few 2D spectra would also be needed.
For the triple resonance cases, several 3D spectra are needed, e.g. HNCA, HNCO,
CBCA(CO)NH, HCCH-TOCSY and X-edited NOESY, amounting to up to around 20 d spectrometer
time.
Data analysis can be very time consuming: first assign, then create lists of
restraints and then calculate the structure.