Detergent-protein and detergent-lipid interactions : implications for two-dimensional crystallization of membrane proteins and development of tools for high throughput crystallography

2.1 Scope of this Thesis
This thesis represents an attempt to enlighten the role of the detergent in reconstitution and more specifically
in two-dimensional (2D) crystallogenesis of membrane proteins. The construction of a tool for precise
and routine measurements of detergent concentrations provided a valuable tool for better understanding
and controlling the detergent issue. Additionally, a novel approach for detergent removal in 2D crystallization,
i.e. the use of cyclodextrins was explored and a nanoliter dispensing high throughput tool was
developed allowing for profound and sophisticated screening of optimal conditions for protein reconstitution
and crystallization.
2.2 Combining Electron Microscopy
and Atomic Force
Microscopy
Although electron crystallography has proven to be
a powerful approach to structure determination of
membrane proteins (for a recent example see (Gonen
et al., 2005)) successes are somehow restricted
to certain classes of membrane proteins (e.g., outer
membrane porins, aquaporins, naturally occurring
crystalline proteins). This is mainly due to the stability
of these proteins with respect to biochemical
manipulation. One can not exclude however, that
these are simply more amenable to crystallization
due to the nature of their molecular surfaces.
2D crystallization exhibits several advantages
compared to 3D crystallization of membrane proteins:
The simple fact that the proteins are allowed
to reside in a native-like environment, i.e.,
the membrane and that their function is not impaired
by the lateral crystal contacts is of considerable
interest. If structural investigations shall not
be restricted to static snapshots of different conformations
and moreover structure-function relationships
shall be established, then electron microscopy
(EM) in combination with atomic force microscopy
(AFM) surely represent a valuable approach.
In Chapter 2 the combination of such data has
been successfully applied to the ammonium transporter
AmtB from Escherichia coli. The aim was to
determine the crystal packing of the double-layered
2D crystals of AmtB by AFM in order to process the
cryo EM data. Additionally, the AFM images, due
to their outstanding signal-to-noise ratio, enabled
the direct visualization of trimers in the reconstituted
membranes. The topographical data from
the AFM allowed the assessment of a single layer
within the double layered crystals.
2.3 Investigating the Role of the
Detergent
In Chapter 3 the development of a fast and precise
method for detergent concentration determination
is presented. The robustness and wide application
range of this method has been demonstrated
by comparing concentrations of radioactively
labeled dodecyl-[beta],D-maltoside (DDM) with
measured contact angles, by measuring the amount
of DDM bound to the proton/galactose symporter
GalP from E. coli, by measuring the effects
of 100 mM NaCl on the cmc of dodecyl-N,Ndimethylamine-
N-oxide, by characterizing the surface
energy of Parafilm, and finally by revealing
the stoichiometry of complex formation between
methyl-[beta]-cyclodextrin (MBCD) and different de-
tergents. The possibility of performing such measurements
routinely in membrane biochemistry is
unique compared to all other methods available to
date.
Chapter 4 addresses the major aspects of detergent
use in membrane protein purification and
crystallization. First, the stability of GalP in different
detergents is assessed, unveiling profound
differences in the capacity of detergents to keep
the protein in solution. Second, it is demonstrated,
that the amount of a detergent, i.e., dodecyl-�,Dmaltoside,
bound to a protein can be controlled
during purification. At last the amount of different
detergents for solubilization of E. coli lipids is
determined, showing differences in the mechanisms
by which detergents promote solubilization.
Banerjee et al. (Banerjee et al., 1995) examined
the preferential affinity of detergents for different
lipids in mixed membranes (such as biological membranes).
They showed that different detergents extract
the serotonin 5-HT1A receptor from native
membranes along with different lipids. The effect
is considerable and might explain why different detergents
exhibit such a different ability to keep a
protein in its native state, because some might simply
not be able to co-solubilize native lipids essential
for the stability (and function) of the protein.
The amount of detergent bound to a protein is
of special interest when using dialysis or dilution
for detergent removal. Furthermore, in most cases
the protein must not be exposed to excess detergent
which anyway fails to satisfactorily mimic the
native bilayer. As pointed out in the discussion of
Chapter 4, protein reconstitution is facilitated when
the detergent collar that is present around the hydrophobic
region of membrane proteins in solution
is near its solubility limit (Psol).
The same is true for the lipid: Reconstitution is
likely to happen when liposomes are forming, therefore
an excess of detergent is not desirable either.
Additionally, even detergents known to have adverse
effects on protein stability can be used for
lipid solubilization, given that they are present at a
minimal concentration. The use of detergent mixtures
in crystallization can also have the effect of
reducing the size of the detergent collar around the
protein. Moreover, the free detergent concentration
in detergent mixtures is altered by the presence of
the second species and can be crucial to the formation
of crystals in some cases (Koning, 2003).
When using minimal amounts of detergent in a
crystallization mixture, special care should be taken
with respect to the formation of ternary micelles.
Ideally, equilibration of the ternary mixtures prior
to detergent removal needs to be completed.
2.4 The Use of Cyclodextrins for
High Thorughput 2D Crystallization
of Membrane Proteins
Chapter 5 demonstrates the feasibility of the
cyclodextrin-based detergent removal for twodimensional
crystallization. The possibility of
choosing different kinetics, simply by adding different
amounts of cyclodextrin at various time intervals
is one of the major advantages of this method.
By implementing optical spectroscopy, it would be
possible to slow down the detergent removal rate at
the onset of proteoliposome and 2D crystal formation.
As pointed out by Lichtenberg et al. (Lichtenberg
et al., 2000) the rate of detergent removal has
to be slow enough to allow for detergent-induced
vesicle size growth, a process which is usually quite
slow. This aspect is important to keep in mind as
one defines the rate of detergent neutralization (in
contrast to dialysis). At a first glance one might
think that in this respect the cyclodextrin approach
bears no advantage compared to dialysis. However,
the rate of low-cmc detergent removal using dialysis
can be too slow, thereby keeping the protein out
from its native environment for too long, ultimately
promoting its precipitation.
In Chapter 6 we present an apparatus for parallel
quantitative reconstitution and 2D crystallization
of membrane proteins. Cyclodextrin provides a
unique opportunity for high throughput implementation
compared to other methods available today.
Protein concentrating through controlled evaporation
with concomitant detergent neutralization (to
prevent detergent concentrating) is advantageous
compared to commercially available protein concentrating
devices which very often concentrate detergent
micelles too. Moreover, the possibility of using
one protein preparation for wide screening ensures
that inconsistencies in results arising from preparative
differences are excluded. Often, the detergent
and lipid concentration of the purified protein are
ill characterized, and this variability may be a cause
for much of the irreproducibility and failure in crystallization
(Wiener, 2004).
So far the use of wide screening matrices (sparse
matrix design) in 2D crystallography was restricted
by the enormous number of experiments and
amount of protein needed for a rigorous screening.
The presented machine makes it possible to partially
compensate for the first bottleneck in protein
structure elucidation, which is the over-expression
of membrane proteins.
Fig. 2.1 summarizes the screening strategy based
on the criteria discussed in Chapter 6 and above.
Screening efficiency is provided by the subdivision
of the problem into multiple subproblems and by
their sequential screening.
With the high throughput approach however, a
new bottleneck arises as one will produce a large
number of crystallization trials, which have to be
screened for their outcome. Therefore –in analogy
to the x-ray community– the development of automated
sample preparation and automated electron
microscopic analysis would provide substantial support
to the 2D crystallographer.
Combining step-by-step identification of key values
necessary for crystallization (and/or efficient reconstitution)
together with high throughput screening
matrices opens up new prospects in the en
deavor to membrane protein structure and function
determination. Now it is possible to apply a
semi-rational screening strategy and this might contribute
to transform 2D crystallization from art to
science (Jap et al., 1992).