2D-crystallization and 3D-structures of membrane channels and transporters

Membrane proteins are responsible for a broad spectrum of biological functions
such as signal transduction, structural functions, energy conversion or transport
of matter across the membranes. Around 30% of the protein-sequences encode
membrane-proteins [88, 80]: Most of them are not directly involved in cell-housekeeping
functions but are indispensable for multicellular life. Therefore, many of
these proteins are of high importance in various medical relevant areas, such as
neurobiology, cell-cycle controlling or immune response, just to mention a few of
them. Malfunctions of these proteins can result in severe diseases such as Cystic
Fibrosis. Other membrane-proteins, such as the multidrug resistance-pump,
are responsible for major medical problems if they are present: Examples are
the resistance of microbes against antibiotics or ineffective treatments of cancer
patients due to resistance of the cancer cell against the chemotherapeutic agents.
The medical relevance of these proteins was and is accompanied by a lack
of structural information. However, significant progress was made in the last
three years. Several leading structures were solved for some protein-families,
mostly using large automated screening approaches. But others are still in
the dark. Furthermore, the current structure exploration still fails in solving
routinely the structure of specific membrane-proteins. More effort has to be
put into the development of these methods for specifically crystallizing (in 2D
or 3D) medically relevant proteins with the goal of bringing membrane protein
structures to structural biology and medicine.
2D-crystallization is a promising approach for these difficult projects since
the protein is reconstituted in its natural environment soon after purification.
The protein can be kept under physiological conditions for the structural exploration.
Some of the obtained 2D-crystals were reported to be still functional
active such as the AQP1 crystals [92].
However, as with 3D-crystallization for X-ray crystallography, the 2Dcrystallization
is still the bottle-neck in electron crystallography. Chapter 2
describes the attempts to 2D-crystallize a highly flexible multidrug-resistance
protein LmrA from Lactococcus lactis. This ABC-transporter was subjected to
a broad crystallization pre-screen to find stabilizing conditions for subsequent
crystallization experiments: First, various detergents were tested for their ability
to preserve LmrA in a healthy state. Since no functionality test for solubilized
LmrA was known, indirect methods to test the integrity of the protein
had to be used, such as solubilization tests, electron microscopy of solubilized LmrA or sucrose gradients. In the second part of the pre-screen the selected
detergents were tested against various lipids for LmrA reconstitution. These
reconstitution tests were done with the monolayer technique [41]. From these
experiments we learned that the detergents C12E8, Triton X-100 and (with some
restriction) DDM are most suitable for further crystallization tests. The lipidscreen
revealed that synthetic lipids mimicking the membrane of Lactococcus
lactis are better suitable for LmrA-reconstitution than other synthetic lipids
(except DMPC). Especially the results for a POPG:cardiolipin mixture seemed
to be promising. From the natural lipid-isolates, Escherichia coli -lipid, also
containing cardiolipin has the ability to incorporate LmrA in lipid bilayers (see
next paragraph).
In parallel, reconstitution and crystallization experiments were performed: A
major problem was the proper reconstitution of LmrA. Tests of different detergent
removal methods revealed, that LmrA reconstitutes nicely in Escherichia
coli -lipid by a stepwise detergent-removal with bio-beads [65, 64]. In these experiments,
fungi-like structures were visible as spikes sticking out for 6 nm of
thick double-membranes. These structures were interpreted as the soluble part
of LmrA. The measured dimensions are compatible with the findings of Chang et
al., 2001 [10] on MsbA, a LmrA homologue of Escherichia coli. Taken together,
these results are a good starting point for further crystallization experiments,
if the found conditions can be combined and the protein can be trapped in a
specific confirmation using LmrA-inhibitors.
In chapter 3 the successful 2D-crystallization of GlpF, the glycerol facilitator
protein of Escherichia coli, is described. To assess the GlpF structure, recombinant
histidine tagged protein was overexpressed, solubilized in octylglucoside
(OG) and purified to homogeneity. Negative stain electron microscopy of solubilized
GlpF protein revealed a tetrameric structure of approximately 80 °A
sidelength. Scanning transmission electron microscopy (STEM) yielded a mass
of 170 kDa corroborating the tetrameric nature of GlpF. These results are contradictory
to previous speculations that GlpF is a monomer [9, 39]. However,
the tetrameric architecture has been confirmed by the atomic structure of GlpF
[23].
Reconstitution of GlpF in the presence of lipids produced highly ordered
two dimensional crystals, which diffracted electrons to 3.6 °A resolution. Cryo
electron microscopy provided a 3.7°Aprojection map exhibiting a unit cell comprised
of two tetramers. In projection, GlpF is similar to AQP1, the erythrocyte
water channel. However, the major density minimum within each monomer is
distinctly larger in GlpF than in AQP1. This finding was confirmed with the
comparison of the refined AQP1 model [14] and the x-ray structure of GlpF
[23], see also figure 1.6 p. 12.
To obtain the three dimensional (3D) structure of GlpF, the two-dimensional
crystals were tilted up to 62� in the electron microscope to get the side-views
of the protein. The resulting 6.9 °A density map showed the GlpF helices to
be similar to those of AQP1, the erythrocyte water channel. While the helix
arrangement of GlpF does not reflect the larger pore diameter as seen in the
projection map, additional peripheral densities observed in GlpF are compatible
with the 31 additional residues in loops C and E, which accordingly do not
interfere with the inner channel construction. Therefore, the atomic structure
of AQP1 was used as a basis for homology modeling of the GlpF channel, which
was predicted to be free of bends, wider, and more vertically oriented than the AQP1 channel. Furthermore, the residues facing the GlpF channel exhibited an
amphiphilic nature, being hydrophobic on one side and hydrophilic on the other
side. This property was speculated to partially explain the contradiction of
glycerol diffusion but limited water capacity. Both, the additional densities and
the greasy slide similar to maltoporin [72] have also been observed in the atomic
structure of GlpF [23]. The importance of the amphiphilic channel architecture
is also corroborated by molecular dynamic calculations [35].