I.
Specific Aims
In the recent years we have developed protocols and designed throughput processes for
the production of large amounts of a diverse variety of membrane proteins with cell-free
expression systems. A particular focus has been on the preparative scale production of
larger polytopic membrane proteins. We have furthermore established different modes of
cell-free expression of membrane proteins and we could identify parameters and
approaches for the significant quality improvement of cell-free produced membrane
proteins. We propose to further develop our methods for the direct expression of
membrane proteins into artificial hydrophobic environments as a completely new way for
the rapid generation of samples suitable fir structural and functional analysis. In particular
we want to
1) Develop expression protocols for the direct expression of membrane proteins into
defined lipid environments like liposomes, bicelles and nano discs.
2) Evaluate, document and improve the quality of the expressed proteins in view of
structural homogeneity and functional folding. Specific guidelines will be defined
II.
Background and Significance
One of the main obstacles in determining the three-dimensional structure of membrane
proteins is the production of the required large amount of material. Both bacteria as well as
eukaryotic cells (yeast, insect cells, mammalian cells) have been used for the expression of
membrane proteins, however, often with limited success. In particular eukaryotic membrane
proteins are difficult to obtain in the amounts necessary for structural studies, which is also
reflected by the fact that only approximately 10 structures of eukaryotic membrane proteins
have been determined so far. Expression of membrane proteins in cellular systems suffers
from several problems: targeting and translocation of the synthesized protein to the correct
membrane destination, overloading of the transport system and toxic effects upon insertion
into the membrane.
Cell-free (CF) expression systems offer a completely new principle for the preparative
scale production of membrane proteins. These systems do not only overcome the above
mentioned problems of cellular based expression systems, they also provide a considerable
number of additional benefits. A highly valuable characteristic is the open nature of the
reaction. This allows the addition of many different compounds, such as protease and RNAse
inhibitors, ligands or chaperones directly into the reaction. For the production of membrane
proteins for NMR-based structure determination cell-free expression systems offer even more
advantages. Due to the lack of metabolic scrambling amino acid type specific labeling is
possible in almost any combination, enabling the development of efficient labeling protocols
for the assignment of the protein’s resonances despite the limited chemical shift dispersion
that is characteristic for a-helical membrane proteins (Klammt et al., 2004; Trbovic et al.,
2005; Reckel et al., 2008). Samples of membrane proteins for the structural analysis by NMR
spectroscopy can therefore be generated in less than two days and this perspective offers an
enormous potential for new NMR applications.
A unique and fascinating option is the CF production of soluble membrane proteins into
defined hydrophobic environments. Several detergents are tolerated by the system at
concentrations above their critical micellar concentration (CMC) without significant decrease
in protein yield. Addition of such detergents produces micelle solubilized membrane proteins
directly. These membrane proteins have never formed a precipitate nor had to be extracted
from a a membrane with harsh detergents. In particular detergents of the Brij family have
proven to be very useful for the production of large amounts of micelle-solubilized membrane
proteins. Unfortunately these detergents cannot be used for NMR investigations since the
protein-micelle complex is too large, leading to very fast relaxation resulting in broad and
overlapping peaks. Detergents, however, can be exchanged, for example by immobilizing the
membrane protein to a column and washing it with a solution containing the detergent of
choice.
The experience with membrane proteins in different detergents and liposomes has suggested
that their structure in micelles is less well defined than in lipid bilayers which could have
consequences for their function. Liposomes would be the ideal “native-like” surrounding for
membrane proteins since many membrane proteins – including membrane proteins expressed
n cell-free expression systems - have proven to be functional in this environment.
Unfortunately liposomes are too large for liquid state NMR spectroscopy, necessitating the
search for alternative “native like environments”. Recently bicelles and nano discs have been
suggested as hydrophobic environments that are small enough for liquid state NMR
spectroscopy but at the same time mimic better the environment of a lipid bilayer. In this
grant we propose to establish protocols that enable us to directly express membrane proteins
into such hydrophobic surroundings using our cell free expression system.
III
Preliminary work by this research group
During the last 5 years we have optimized a cell-free expression system that is based
on an E. coli S30 extract for the production of large amounts of membrane proteins and were
the first to achieve the preparative scale production of functionally folded membrane proteins
(Klammt et al., 2004). We further developed protocols for the cell-free production of
membrane proteins in different modes (Fig. 1, Schwarz et al., 2006). In the P-CF mode, the
proteins precipitate after translation. This precipitate, however, is very different from
inclusion bodies formed in E. coli and the membrane proteins can be resolubilized by addition
of detergent without prior denaturation and refolding steps (Klammt et al., 2004). The P-CF
synthesis followed by subsequent solubilization and reconstitution can principally result in
functionally folded MPs as shown by us and others with EmrE, the mechanosensitive channel
MscL or the light harvesting protein LH1 (Elbaz et al., 2004; Klammt et al., 2004; Berrier et
al., 2004; Shimada et al., 2004). In the D-CF mode, the membrane proteins are synthesized in
the presence of detergents and remain soluble by insertion into micelles (Klammt et al.,
2005). This approach is only possible with cell-free systems and avoids the denaturation and
refolding of inclusion bodies or the extraction from a cellular membrane by harsh detergents.
For this soluble expression of membrane proteins we have identified a diverse number of
detergents that are tolerated by the cell-free systems (Berrier et al., 2005; Klammt et al.,
2005). In addition, the cell-free synthesis of functionally folded MPs directly into defined
preformed liposomes has recently been demonstrated for the first time (Kalmbach et al.,
2007). However, the yield of this L-CF mode of cell-free synthesis is currently very low and
needs significant further improvement.
Fig. 1. CF expression modes for
high
level
MP
production.
Schematic view of CF approaches for
the production of MPs. P-CF: MPs
are synthesized in absence of supplied
hydrophobic
compartments
and
precipitate after translation. D-CF:
MPs are kept in soluble form by
insertion into provided detergent
micelles. L-CF: MPs can become
inserted into preformed liposomes of
defined composition.
1 Development of cell-free expression systems for the high-level production of
functionally folded membrane proteins. We established in our group the high level
expression of MPs in CF systems by modifying Escherichia coli S30 extract preparation
protocols and by optimizing reaction processes (Klammt et al., 2004). Our most productive
set-ups are continuous exchange cell-free (CECF) expression systems having a reaction
mixture (RM) separated by a semipermeable membrane from a feeding mixture (FM) (Alakov
et al., 1995). As reaction devices, we have developed specifically designed Plexiglas
containers. We have further optimized essential reaction components like the energy system,
concentrations of distinct ions and precursors and the supplementation of beneficial additives.
Techniques for the preparation of key compounds like E. coli cell extracts and purified T7polymerase are established procedures in our lab (Schwarz et al., 2006). We can yield up to 6
mg of recombinant MP per ml RM in 12 hours of incubation and numerous MPs of
prokaryotic and eukaryotic origin, with either -helical or ß-barrel type secondary structures
and with up to 14 transmembrane segments (TMSs) have already successfully been produced.
Most of those targets like the tellurite transporter TehA, the cysteine exporter YfiK or
eukaryotic proteins like G-protein coupled receptors (GPCRs), aquaporins or the ion channels
rOCT1 and rOCT2 could not be produced in significant amounts in conventional in vivo
expression systems based on living cells.
For P-CF expressions, we have screened detergents that are suitable for the efficient
solubilization of the MP precipitates (Klammt et al., 2004; Klammt et al., 2005). Especially 1myristoyl-2-hydroxy-sn-glycero-3-[phospho-RAC-(1-glycerol)]
(LMPG)
and
related
derivatives from the detergent family of lyso-phospholipids were found to have optimal
properties in the solubilization of structurally different MPs. LMPG solubilized MPs can then
be transferred into other hydrophobic environments by buffer exchange. Spectroscopic
analysis and reconstitution followed by transport assays demonstrated the folding in
functional conformations of P-CF produced transporters (Elbaz et al., 2004; Klammt et al.,
2004). In addition, our recorded NMR spectra of solubilized EmrE, SugE, TehA and YfiK
further indicate structurally folded conformations of the proteins. We use the P-CF mode for
rapid expression screening of new MP targets and for the development of protocols for high
yield production.
If we observe aggregation or problems in the functional folding of P-CF expressed
MPs, we switch to D-CF expression into defined detergent micelles. We have analyzed a
representative variety of detergents including non-ionic detergents (alkyl-glycosides (-OG,
DDM, DM), polyoxyethylene-alkyl-ethers (Brij derivatives, Thesit, Genapol), polyethyleneglycol derivatives (Triton, Tween, NP40), steroid-derivatives (CHAPS, Digitonin),
zwitterionic detergents (DPC, DHPC, diC6PC, diC8PC), and long-chain ionic detergents
(LMPG, LPPG) for their suitability for the production of soluble MPs (Klammt et al., 2005).
In this screen, we identified the group of polyoxyethylene-alkyl-ethers to be best suited for
soluble expression of MPs. In summary, P-CF and D-CF expression modes have been
established in our group and we have already demonstrated the efficiency of these techniques
for the high level production of functionally folded MPs in several cases.
2 Development of throughput processes for the optimization of cell-free expression
protocols. Screening parameters in cellular-based throughput expression approaches is
usually restricted to screening of only few variables due to the complex nature of cellular
environments and degradation and conversion problems. In contrast, the open nature of cellfree systems enables the development of expression protocols for individual target proteins by
screening of a large number of expression conditions. This strategy is optimal for robotic
devices and we have established processes for the automatic cell-free throughput expression
screening based on a Tecan freedom Evo200 platform (Fig. 2). Reactions are set up in
microplate formats by high performance nanoscale pipetting roboters and all subsequent steps
like incubation, centrifugation, purification and quantification can be operated as integrated
processes in order to determine ideal expression conditions. In a first approach, we have
analyzed the cell-free expression of a comprehensive subset of the E. coli inner membrane
proteome containing C-terminal fusions of GFP for fast monitoring (Daley et al., 2005). In a
subset of 128 different MPs, > 80 % of the targets could be synthesized after two optimization
screens in D-CF and P-CF expression modes.
Fig. 2. Throughput design for the optimization of cell-free membrane protein production. An integrated process
for pipetting, incubation, separation, purification, quantification and parameter selection of CF MP expression
conditions is shown. The expression screening is demonstrated with an example of Mg 2+/K+ ion concentration
optimization.
The class of electrochemical potential-dependent transporters of E. coli includes
several hundred proteins belonging to at least 50 different families (www. tcdb.org). Exact
numbers are impossible to define as many MPs cannot be functionally assigned so far. In our
cell-free expression optimization screen, 39 transporters of this class belonging to 19 different
families could already be successfully produced (Fig. 3, Table 1).
Fig. 3: CF expression of
EPD transporter. Samples
containing CF produced transporter were separated by 17.5
% SDS-PAGE and stained
with Coomassie-Blue. 2.5 µl
of CF RM was analyzed.
For most of them, we have optimized the expression protocols to achieve preparative scale
amounts.
This expression screen has demonstrated that we can produce electrochemical potentialdependent transporters ranging in size from 110 to 1041 amino acids with 4 to 14 TMSs,
while those numbers do not define limits yet. These preliminary results indicate that (I) there
is no evident bias in the size of CF synthesized MPs and even very large proteins can be
synthesized in mg amounts. (II) Topologies and predicted number of TMSs do not seem to
play a significant role in the expressability of MPs by CF systems at least in the P-CF and DCF modes. (III) For the majority of MPs, the optimization of basic screening parameters like
the concentrations of ions, detergents and precursors are sufficient in order to obtain
expression levels sufficient for structural analysis.
Cell-free synthesis is not confined to prokaryotic targets. In cooperation with the
group of Hermann Koepsell from the University of Würzburg we have analyzed the
expression of the cation transporter rOCT1 and the anion transporter rOAT1 from rat. Both
proteins have a molecular mass of approx. 60 kDa and 12 predicted TMSs. After protocol
optimization, both proteins could be produced in levels of several mg per ml of RM in the PCF mode. With reconstituted CF produced rOCT1, the specific transport of methyl-phenyl
phosphonium (MPP+) was measured with a KM (MPP+) of 35 µM and a minimal turnover
rate of 19 sec-1 (Fig. 4). These results agree well with data previously obtained from rOCT1
isolated from insect cells with a KM (MPP+) of 30 µM and a minimal turnover rate of 20 sec-1
quinine inhibited MPP uptake
[nmol MPP  mg-1 sec-1]
(Keller et al., 2005).
Fig. 4. CF expression and
functional analysis of eukaryotic
electrochemical potential-dependent
transporters.
Transport
activity of CF produced and
reconstituted rOCT1. Reconstituted
proteoliposomes were analyzed
with radiolabeled MPP+. The
measured transport activity could
be inhibited by the specific
inhibitor quinine. Figure is courtesy
of T. Keller and H. Koepsell.
300
200
100
0
0
50
100
150
200
250
MPP-conc. [µM]
This approach demonstrates that even complex eukaryotic electrochemical potentialdependent transporters with 12 TMSs can be obtained in a functional conformation and at
high levels by CF expression. It should further be mentioned that rOCT1 as well as rOAT1
samples of similar quality can be obtained from conventional insect cell expression systems
only in µg amounts from several litres of culture after weeks of work and at several times
higher costs.
3 CF expression and functional analysis of the human endothelin B receptor.
We have shown that cell-free expression is an interesting and promising alternative for the
production of G-protein coupled receptors (GPCRs) as well (Klammt et al., 2007a; Klammt et
al., 2007b). The receptors for human endothelin B (ETB), human and porcine vasopressin
type 2 and rat corticotropin releasing factor are synthesized in amounts of up to 6 mg per ml
reaction mixture in P-CF and D-CF modes. Single particle analysis revealed homogenous
dispersion of dimer particles in cell-free produced GPCR samples (Klammt et al., 2007b). By
various complementary techniques, we have confined the dimer interface to the N-terminal
end of ETB involving TMS 1 (Klammt et al., 2007a).
4
Development of labeling strategies for the structural investigation of membrane
proteins by liquid state NMR spectroscopy.
Cell-free expression solves a major problem for the structure determination of MPs by
providing sufficient material, but further technical challenges remain. NMR spectra of helical MPs suffer from significant peak overlaps due to the limited 1H chemical shift
dispersion and from fast relaxation. Consequently, only the most basic NMR experiments
yield sufficient sensitivity for detailed investigations of larger MPs. We have developed
NMR-based techniques that allow us to efficiently obtain backbone assignments of
solubilized MPs as the first step in structural determinations (Trbovic et al., 2005; Koglin et
al., 2006). Our methods take advantage of the high flexibility of CF systems in combination
with almost completely suppressed metabolic scrambling of added isotope labeled
compounds. Our assignment strategy is based on a two step approach in which we first use
standard triple resonance experiments to obtain as many assignments as possible and then use
a double labeling scheme to select amino acid pairs with a two-dimensional version of the
HNCO experiment to identify sequence-specifically additional amino acids. CF expression
allows obtaining amino acid type selective labeled samples in basically any combination of
combinatorial labeling schemes. In addition, we have recently developed a new labeling
protocol (TMS-labeling for transmembrane segment labeling) that takes advantage of the fact
that roughly 60% of the transmembrane helices of MPs consist of the six amino acid types G,
A, L, F, I and V. Producing MPs that are double labeled only with these six amino acid types
significantly reduces peak overlap in HNCA and HNCOCA spectra, while providing many
sequential connectivities due to the clustering of these six amino acid types in the TMSs.
We have started to assign the backbone resonances and to investigate the structure of
MPs. In particular we have focused on TehA, a 36 kDa transporter with 10 TMSs involved in
detoxifying tellurite compounds in bacteria (Turner et al., 1997; Taylor, 1999). We have
structurally analyzed a truncated derivative of TehA (∆TehA) that is still functionally active
and covers the first seven TMSs (219 amino acids) (Fig. 5A). Following sequential
assignments of more than 90% of the backbone resonances we have started to analyze the
three-dimensional structure of the protein. Secondary structural analysis was mainly based on
the
13
C and
13
Cß chemical shifts as well as on the pattern of sequential NOEs in the
15
N-
edited [1H,1H]-NOESY spectrum. We have further started to use paramagnetic relaxation
enhancement (PRE) to obtain distance constraints for the structure determination of ∆TehA.
For this purpose we have mutated all three naturally occurring cysteines to alanines and have
reintroduced a cysteine residue at specific locations. Due to the significant peak overlap in the
[15N,1H]-TROSY spectrum we cannot use a fully
15
N-labeled sample for these investigations
but have to rely on amino acid type selectively labeled samples. Using this approach we have
obtained ~100 PRE constraints so far which we have combined with angle constraints derived
from 13C-chemical shifts and sequential NOEs in structure calculations. Figure 5B shows the
current structural model of TehA. This model will be refined with additional NOEs and PREs
in the future.
A
B
Fig. 5. Structural evaluation of ∆TehA. A:
[15N, 1H]-TROSY spectrum of 15N/2H labeled
∆TehA containing the first seven TMSs. B
Structural model of ∆TehA. All seven predicted helices are present, however, several
of them are still distorted.
IV Research Plan
1) Development of expression protocols for the direct expression of membrane proteins
into liposomes, bicelles and nano discs.
Based on our established P-CF and D-CF expression protocols we want to develop procedures
to directly express membrane proteins into liposomes, bicelles and nano discs. Expression
into liposomes is of interest for functional studies, in particular for transporters that need two
different compartments that are separated by a membrane for the any functional studies.
Liposomes are however not suitable for high resolution studies such as liquid state NMR
spectroscopy or x-ray crystallography. To be able to investigate the structures of membrane
proteins in hydrophobic environments that mimic the natural environment better than
detergent micelles we will develop protocols for the direct incorporation into bicelles and
nano discs.
1.1 Incorporation into liposomes.
Direct incorporation of a membrane protein into a lipid bilayer is significantly more difficult
than incorporation into detergent micelles but can be achieved by destabilizing the liposomes
with detergents. We will use different types of lipids such as DMPC, DMPG and E. coli total
lipid mixtures. If available we will use ether-linked versions of these lipids which are not
sensitive to hydrolization. Before using these lipids in cell-free reactions we will determine
the amount of detergent necessary to destabilize the liposomes without completely dissolving
them. This can best be achieved by measuring the optical density at 580 nm. Dissolving the
liposomes results in a decreased optical density and we will determine the detergent
concentration of the mid-point of this transition. We will use lipid : detergent ratios around
this mid-point concentration in the cell-free expression experiments. These expression
experiments will be carried out with our robotic platform in a 96 well plate format by
changing in addition to the detergent and lipid concentrations the concentrations of different
salts and amino acids. Liposomes will be created with a diameter of 4 m by dissolving the
lipids in a suitable buffer and using an extruder to create a homogeneous particle size. At the
beginning we will use those detergents that are well tolerated by the cell-free system.
However, since destabilization of the liposomes can be achieved at concentrations lower than
the CMC of the detergent we can use some detergents that decrease the expression yield when
added at concentrations above their CMC. The effect of low concentrations of such detergents
will be tested in expression experiments without added lipids.
After the cell-free expression experiments, the liposomes will be harvested by centrifugation
as the insoluble fraction. SDS-PAGE and western blot analysis will be used to test if this
fraction contains the expressed protein. The successful incorporation into liposomes will be
tested by obtaining freeze-fracture images by electron microscopy. The necessary equipment
and expertise for these types to obtain these images is available at the Max-Planck Institute
for Biophysics that is located next to our institute on the campus of the University of
Frankfurt and we have frequently used this MPI facility in the past. If incorportation into
liposomes can be confirmed in these images the liposomes will be purified from other
components of the insoluble fraction by sucrose density gradient centrifugation.
1.2 Soluble expression into bicelles
Bicelles are formed by mixing long-chain and short-chain lipids. If the lipid : detergent ratio
is smaller than 1, isotropic bicelles can form that are small enough to allow high resolution
liquid state NMR spectroscopy but provide at the same time a disk-shaped form that mimics
better the natural flat bilayer of membranes than the curved detergent micelles. Above a ratio
of 1, also depending on the temperature as well as ionic strength and presence of additional
charged detergent molecules, the solution will form a lyotropic liquid crystal in the presence
of strong magnetic fields that have been used for the measurement of residual dipolar
couplings. For the structural investigation of membrane proteins, however, incorporation into
isotropic tumbling bicelles is required and we will, therefore keep the ratio in a range that
allows the creation of such bicelles. Incorporation into bicelles can in principle be achieved in
different ways: 1) by reconstituting the membrane proteins in DMPC liposomes and
subsequent formation of bicelles by adding detergent (DHPC), 2) by solubilizing the
membrane protein in DHPC micelles and subsequent formation of bicelles by addition of
lipids (DMPC) or 3) by dissolving the membrane protein in organic solvents followed by
adding both bicelles components and removal of the organic solvent. All three different ways
to produce membrane proteins incorporated into bicelles can in principle be used for the
creation of bicelles samples based on membrane proteins expressed in a cell-free system with
the third option most likely being restricted to only a few smaller membrane proteins.
Obtaining micelle solubilized membrane proteins, however, is an established technique in our
laboratory and detergents can be routinely exchanged to obtain a sample dissolved in a
suitable detergent for bicelles formation (DHPC). As mentioned before, direct incorporation
into liposomes has been shown to be in principle possible and protocols for obtaining
preparative amounts of liposome incorporated membrane proteins will be developed in
specific aim 1.2 (see above). The open nature of the cell-free reaction, however, opens an
alternative avenue towards obtaining bicelle-solubilized membrane proteins by formation of
bicelles in the reaction mixture and direct expression into these bicelles. Since the size and the
behavior of the bicelles depend on many parameters (lipid : detergent ratio, salt concentration,
additional detergent components) we will screen many different parameters using our robotic
platform. One advantage of bicelles is the possibility to purify the protein / bicelles complex
by ion chelate chromatography when the membrane proteins contain a His-tag. If reaction
conditions can be identified that suggest that a large percentage of the expressed protein is
contained in the soluble phase we will purify the protein. To investigate if the protein is
incorporated in a bicelles (in contrast to a micelle formed by the detergent component) we
will measure 31P NMR spectra. If these spectra indicate the existence of bicelles we will use
the cell-free expression system to label the protein with 15N-labeled amino acids. At the
beginning we will use the amino acids glycine, alanine, serine and tryptophan since they
cover a wide chemical shift range and are useful to investigate the homogeneity of the sample.
The homogeneity and stability of the bicelles can be improved by doping the bicelles with
negatively charged dihexanoyl- / dimyristoylphosphatidylserine or other components. After
optimization of the spectral quality based on the use of a limited number of amino acids we
will prepare a triple labeled sample 13C/15N/2H by using the adequate amino acid mixtures
(in deuterated extract to avoid back exchange of deuterated a-positions) and investigate the
possibility to measure multi-dimensional triple resonance experiments. Depending on the
protein we will also investigate the biological function by performing binding assays if
binding partners / inhibitors are known and available.
1.3 Soluble expression into nano discs
The recently developed nano disc technology allows the in vitro generation of preformed
planar membranes of defined shapes and sizes that are suitable to accommodate only defined
numbers of membrane proteins (Bayburt et al., 2002). Nanoparticulate phospholipid bilayer
disks can be assembled from phospholipids and specific membrane scaffold proteins. The
scaffold proteins can be overproduced by conventional expression systems in E. coli cells and
detailed protocols for the in vitro nano disc formation have been published (Bayburt et al.,
2002). We have established the expression of membrane scaffold proteins in our lab and they
will be used in mixtures with different lipids to produced a variety of nano discs in vitro. The
nano discs will be analysed for the insertion of a representative selection of membrane
proteins. Protocols will be established for the efficient insertion of membrane proteins
produced in the P-CF mode. Alternatively, nano discs will be supplied directly into CF
reactions for the co-translationally insertion of membrane proteins. This technology would
provide the advantage of uniform membrane protein samples inserted into a natural
membrane-like environment. The geometry of nano discs can be manipulated as such that
only a single protein can be accommodated in one nano disc, resulting in a homogenous
sample probably also suitable for structural approaches even by NMR spectroscopy. This
approach would therefore provide a further alternative option for the incorporation of
membrane proteins into a hydrophobic environment. Recent reports have already
demonstrated that CF expression in presence of supplied preformed nano discs can be a
promising tool for the soluble expression of complex membrane proteins including GPCRs
(Katzen et al., 2008). Optimization of expression conditions and fine-tuning of protocols will
make use of our robotic platform and build on the experience obtained with the expression
into micelles, liposomes and bicelles.