Philipp Selenko *, Gerhard Wagner
Journal of Structural Biology 158 (2007) 244–253
NOTES (excerpts from the journal spliced together, along with my own comments) - I'm here hitting the highlights of the article, particular the parts I found most interesting, with comments.
Labeling of proteins with 15N and 13C, over-expression in bacteria, and NMR of the crude extract. Harder to do in EU cells because of more expensive growth media, difficulty of over-producing the bacteria, poor incorporation of the labeled atoms….
Micro-injection vs Protein Over-production…. with in-cell NMR Advantages and Disadvantages:
Advantage of protein over-production
- Does not require purification of the labeled polypeptide, which can be beneficial when a protein
Disadvantage of protein over-production
- On the other hand, in-cell protein production cannot be easily quantified or reproduced under exactly identical
- At low levels of recombinant protein over-expression, other cellular components become increasingly labeled, which results in the generation of signal artefacts that have to be actively suppressed (Rajagopalan et al., 2004).
Advantage of micro-injection
- Defined concentrations and less background noise - These adverse effects are absent when a labeled compound is introduced into the cellular environment at defined concentrations, as can be achieved by microinjection, for example.
Disadvantages of micro-injection
- Limited to certain cell-types
- Requires protein to be concentratable
This approach, however, is restricted to a few eukaryotic cell types, which can be manipulated in such a way, and requires the labeled protein samples to be sufficiently soluble at high concentrations, since the maximum injection volume per single cell is typically on the order of nano-liters.
Alternatively, one could envisage intracellular sample
delivery by means of cell-permeable synthetic vectors. We
are particularly intrigued by the potential application of
‘Trojan’ peptide tags, which confer efficient cell membrane
transduction activities to a wide range of fused protein substrates
(Derossi et al., 1998; Dietz and Bahr, 2004). These
internalization peptides are composed of short, positively
charged amino acid residues, which can be genetically engineered
to be part of virtually any recombinant polypeptide
(Li et al., 2002). Upon labeled expression and purification
of tagged fusion proteins, these substrates are simply added
to the growth medium of a variety of cultured laboratory
cell lines and readily internalized. In theory, this method
should be generally applicable to a wide range of eukaryotic
cells and quantitatively accomplishable for a large
number of cells.
The method of choice for eukaryotic in cell NMR measurements will hence be dictated by both the suitability of the protein of interest for either approach and by the overall biological question to be addressed.
Difficulties of in-cell NMR in EU cells
The major obstacles for the selective labeling of recombinant proteins with NMR-active isotopes in eukaryotic cells
have been …
- the difficulty to achieve adequate levels of protein over-expression
- sufficient isotope incorporation and (due to complexity of EU metabolisms)
- the costs of isotope-enriched growth media
- induction can take days rather than hours for bacteria (which can increase background artifacts)
Growth media for NMR labeling in E. coli are simple in their composition, easily prepared and, depending on the type of labeling, relatively cheap. Bacteria will also incorporate isotopes with high efficiency (_98%). Labeling media for eukaryotic cells are sophisticated, they must often be obtained commercially for satisfactory results, and they are expensive.
Tumbling rate and line widths gives information about cellular binding partners
In general, small proteins display large tumbling rates, which
lead to slow overall relaxation and narrow NMR line
widths (Fig. 3a). Molecules of larger size tumble more
slowly, relax faster and exhibit broader resonance signals (Fig. 3b).
Any molecule in a cellular environment exhibits
a reduced tumbling rate due to intracellular viscosity
and hence displays broader NMR line widths (Fig. 3c).
In the absence of protein binding to endogenous cellular
factors, a direct comparison of individual protein line
widths in buffer and in in-cell NMR experiments will
readily yield a qualitative estimate about intracellular viscosity.
Upon sample binding to cellular components, the
resulting protein complexes can either display tumbling
rates that correspond to the sum of their individual
masses (Fig. 3d), or individual contributions to a mixed
set of rates, when the interaction is restricted to a subset
of residues (Fig. 3e).
The latter results in residue specific line broadening, which yields
information on the dynamics and localization of the cellular interaction.
Binding to quasi-static cellular structures like organelles or membranes
results in severe line broadening (Fig. 3f), which can serve as a
qualitative indicator for the kind of interaction.
Many biological binding events are dynamic and
modulated by cellular signaling, which often leads to
transient and interpretable changes of NMR line widths.
Getting around the complicity of the cellular environment and binding partners
It is apparent that a biologically active protein can experience
any of the aforementioned conditions, and superpositions
thereof, in a cellular environment. Complicated or poor quality in-cell
NMR spectra are the likely result.
In such cases, the researcher needs to reduce the complexity
of the system under investigation, which can be
achieved either by ‘chopping up’ full-length proteins into
individual domains in order to selectively probe differential
biological activities or by introducing site-directed
mutations that abolish certain functional characteristics
and similarly enable one to discriminate between specific
Example: Indeed, given the multitude of cellular binding partners for
ubiquitin in Xenopus oocytes, in-cell NMR experiments for
the wild-type protein yielded poor quality NMR spectra.
Only upon mutating conserved binding sites of known
interaction surfaces, did ubiquitin provide interpretable
in-cell NMR data. This work thus serves a fine example
for what kind of biological studies will be amenable to
in-cell NMR approaches.
Note that: The necessary concentration of injected protein for the minimally sufficient experimental NMR signal has to be evaluated empirically for each sample.
Other factors affecting the minimal amount of labeled compound needed for a minimal NMR signal…
- the protein’s size
- expected cellular activity
- the type of labeling, will determine the minimally
Injection concentrations are typically in the range of 0.5–3.0 mM.
- These quantities might not be achievable for all recombinant protein samples nor generally suffice for a satisfying experimental readout
It is evident that a compromise between the
experimentally achievable signal-to-noise, the duration of
individual NMR experiments, the cellular concentration
of labeled proteins and its physiological relevance, will
have to be found if in-cell NMR measurements in X. laevis
oocytes are to yield biologically meaningful results.
1H, 15N correlation pulse-sequence,
isotope-edited correlation techniques
Solvent suppression was achieved by a Watergate NMR pulse-sequence
2d scan was described as follows:
Fig. 5 (b) Identical samples as in (a) but recorded as 1D 1H(15N) hetero-nuclear single quantum coherence (HSQC) correlation spectra, using a Watergate version of a standard, sensitivity-enhanced HSQC pulse-sequence with the same number of scans as in (a).
1d NMR provides too much noise, but 2d isotope edited technique provides a clearer picture
Fig. 5a shows the characteristic appearance of 1-dimensional
proton-only NMR spectra (no isotope-edited correlation
‘filter’) of a 15N-labeled protein sample in its pure
form (top panel), resuspended in crude Xenopus egg extract
(middle panel) or upon oocyte injection (bottom panel). It
is apparent that by conventionally recording all NMR signals
from 1H nuclei only, these spectra do not allow to discriminate
between resonances from endogenous, cellular
components and the ones from the labeled compound.
Moreover, it is evident that the quality of in-cell NMR
experiments, recorded in this mode, is too poor to provide
any information on the injected protein sample. When,
however, these same samples are measured with the application
of a 1H, 15N correlation pulse-sequence, similar 1-
dimensional traces selectively display NMR resonances
from the labeled protein only.
The good quality of these background-suppressed spectra
enables the selective detection of protein NMR signals under
extract and in-cell conditions (Fig. 5b). Changes in NMR line widths of
individual resonance signals in these different aqueous solutions
are readily visible (Fig. 3).
What will cause in-cell NMR to fail
Once cellular proteins engage
the labeled specimen in too many generic interactions, or
scavenge the NMR-active protein into complexes of molecular
weights too large for detection by conventional solution-
state NMR methods, the envisaged in-cell NMR
approach is likely to fail. In most instances, such unfavorable
binding behaviors are easily pre-assessed by NMR
experiments in cellular extracts (Serber et al., 2006).
This future application sounds FASCINATING to me
In-cell NMR analyses of intrinsically disordered proteins
Intrinsically disordered proteins (IDPs) represent a
growing class of gene products (Dyson and Wright,
2005), which are characterized by lack of secondary and/
or tertiary structure in their pure forms and at physiological
pH (Uversky, 2002). IDPs are estimated to account for
_20% of all human proteins (Dunker et al., 2000) and exert
important functions in key cellular processes (Dunker
et al., 2002). A significant number of IDPs are implicated
in human protein deposition diseases, in which a normally
soluble polypeptide forms insoluble aggregates in a subset
of cells and precipitates in the form of amyolid fibrils
(Fink, 2005). Little is known about the general 3-dimensional
conformation of IDPs in cellular environments and
it is tempting to speculate whether unfolded protein conformations
are preserved under native, intracellular conditions.
Could it be possible that some IDPs are not per se
structurally deficient protein entities but that their
unfolded state results as a consequence of the isolated
in vitro experimental setup employed for their characterization?
In-cell NMR experiments on the natively unfolded
FlgM protein indeed suggest a more folded conformation
in the cellular environment of live bacteria (Dedmon
et al., 2002). Are there cell type specific differences in the
conformation of unfolded proteins and to what extent do
different cellular environments modulate the pathological
conversion of IDPs during amyloid formation? In-cell
NMR spectroscopy appears to be a most appropriate tool
to address these questions in live cells.
6.2. In situ observation of post-translational protein
A limitation of in-cell NMR spectroscopy in bacteria is
the inability to study post-translational protein modifications.
Whereas the function of most eukaryotic proteins
is regulated by a variety of sometimes transient, covalent
modifications, mammalian proteins expressed in bacterial
organisms are typically not modified. Eukaryotic protein
modifications, with the exception of post-translational glycosylation,
involve the covalent attachment of small chemical
entities, acetyl-, methyl-, or phosphate-groups, which
do not greatly alter the overall molecular weight of the
modified substrates. This is particularly favorable for biomolecular
NMR analyses as the spectral quality of a
unmodified protein is likely to be preserved upon covalent
modification. In addition, post-translational protein modifications
greatly change the chemical environment of targeted
residues, which translates into large chemical shift
changes. In theory, transferring an unmodified protein substrate
into a eukaryotic cellular environment should result
in its post-translational modification by endogenous
enzymes and according to a specific, biologically relevant
pattern. In-cell NMR measurements should then enable
the in situ observation of the establishment of these modifications
in a time-dependent and residue-specific manner.
We have recently confirmed that these in vivo approaches
can indeed resolve cellular phosphorylation reactions in
X. laevis oocytes (Selenko et al., submitted). They point
to a plethora of possible in-cell NMR applications in
eukaryotic post-translational protein modification
In discussing micro-injection…
About 200 manipulated oocytes are required for an in-cell NMR sample, which corresponds to >250 lL in sample volume and, inside a Shigemi_NMR tube, will suffice to span transmitter- and receivercoil extensions of most commercial NMR spectrometer probes (Fig. 4e).
Considering the small injection volume per cell (50 nL), an oocyte sample requires only about 10 lL of labeled protein.
Question: in a sample NMR tube, one can be reasonably sure that each molecule is only influenced by itself and/or average interactions in an inert environment. But here, each protein or even, each protein group in each of the 200 cells could be being acted on differently and giving different resonances. How does one sort out all this noise?
Question: Water is not inert to H1 resonances. Cells are full of water. How does one get around the interactions on water on these cellular proteins or is that something that one wants to see? I'm assuming from this article that in a 2d NMR map, only the water directly interacting with the protein would be visible indirectly on the spectra, via it's influence on the N or C under observation.
Question: I need more information about 2d NMRs. I’ve only worked on the 1d kind.
Question: In traditional NMR, you have the reference compound TMS. What is the equivalent for in-cell NMR or is there one? If there is not one, how does one determine a reference point? Note that figure 5 description seems to imply a standard, or, I could just be reading that completely wrong.
Question: While I can see the advantages of micro-injection into X. laevis oocytes, I do wonder, since the injected protein is not native to that cell, I presume the interactions within that cell would be different from its host cell, thus not giving the true picture of how it may structurally relate to certain molecules.
ANS: Most biological reactions are executed with comparable activities in Xenopus cell extracts and can easily be modulated by the addition of small molecule inhibitors or activators. These extracts can also be selectively depleted of certain cellular components, and replenished with labeled, NMR-active substitutes, which enables the
isolated investigation of biological processes without compromising the qualitative nature of the experimental readout by residual endogenous activities.