research papers
1062 doi:10.1107/S0907444909029163 Acta Cryst. (2009). D65, 1062–1073
Acta Crystallographica Section D
Biological
Crystallography
ISSN 0907-4449
Effects of impurities on membrane-protein
crystallization in different systems
Christopher A. Kors,a Ellen
Wallace,b Douglas R. Davies,b
Liang Li,c Philip D. Laiblea and
Peter Nollertb*
aBiosciences Division, Argonne National
Laboratory, 9700 South Cass Avenue, Argonne,
IL 60439, USA, bdeCODE biostructures,
7869 NE Day Road West, Bainbridge Island,
WA 98110, USA, and cDepartment of Chemistry
and Institute for Biophysical Dynamics,
University of Chicago, 929 East 57th Street,
Chicago, IL 60637, USA
Correspondence e-mail: pnollert@decode.com
When starting a protein-crystallization project, scientists are
faced with several unknowns. Amongst them are these
questions: (i) is the purity of the starting material sufficient?
and (ii) which type of crystallization experiment is the most
promising to conduct? The difficulty in purifying active
membrane-protein samples for crystallization trials and the
high costs associated with producing such samples require an
extremely pragmatic approach. Additionally, practical guide-
lines are needed to increase the efficiency of membrane-
protein crystallization. In order to address these conundrums,
the effects of commonly encountered impurities on various
membrane-protein crystallization regimes have been investi-
gated and it was found that the lipidic cubic phase (LCP)
based crystallization methodology is more robust than
crystallization in detergent environments using vapor diffu-
sion or microbatch approaches in its ability to tolerate
contamination in the forms of protein, lipid or other general
membrane components. LCP-based crystallizations produced
crystals of the photosynthetic reaction center (RC) of
Rhodobacter sphaeroides from samples with substantial levels
of residual impurities. Crystals were obtained with protein
contamination levels of up to 50% and the addition of lipid
material and membrane fragments to pure samples of RC had
little effect on the number or on the quality of crystals
obtained in LCP-based crystallization screens. If generally
applicable, this tolerance for impurities may avoid the need for
samples of ultrahigh purity when undertaking initial crystal-
lization screening trials to determine preliminary crystal-
lization conditions that can be optimized for a given target
protein.
Received 11 March 2009
Accepted 22 July 2009
1. Introduction
The relatively low number of currently available X-ray crys-
tallographic membrane-protein structures compared with
those of soluble proteins points to the need for better methods
for membrane-protein crystallization (von Heijne, 2007). Most
of the atomic level details can be directly accredited to the
generation of protein crystals suitable for structure determi-
nation by X-ray crystallography.
Membrane proteins are comprised of hydrophobic and
hydrophilic regions, rendering them soluble in the cell mem-
brane and in artificial amphiphile structures such as lipid
bilayers and detergent micelles. For purification and subse-
quent biochemical and biophysical analyses, the extraction of
membrane proteins from their native membrane is achieved
through solubilization with a detergent, forming protein–
detergent complexes (PDCs; le Maire et al., 2000; Moller & le
Maire, 1993). The properties of a particular solubilizing

detergent affect the extraction yield, protein purity, type and
quantity of co-extracted lipids, as well as the functional and
structural integrity of the protein (Garavito & Ferguson-
Miller, 2001; Prive´, 2007). It is important that the artificial
microenvironment is suitable to support the native confor-
mation of the membrane protein and allows productive
protein–protein interactions during crystallization and
prevents protein precipitation and phase separation. Current
standard laboratory practice calls for the purification of a
target protein sample to usually better than 90% as judged
by SDS–PAGE prior to submitting samples to crystallization
trials. However, it is unknown whether a protein sample is
truly ‘pure enough’ to produce crystals. With regard to the
crystallization of soluble proteins, a study (Geerlof et al., 2006)
backed this laboratory practice, showing that proteins of
unknown structure with purity levels of >95% (as determined
by SDS–PAGE) yielded crystals in 59% of all instances,
whereas samples that were <95% pure yielded crystals with a
success rate of only 37%. A different study investigating the
effects of macromolecular impurities on the crystallization of
eubacterial aspartyl-tRNA synthetase, either by vapor diffu-
sion or in capillaries, called for purity-level requirements of
99% for the production of high-quality crystals (Moreno et
al., 2005). Findings such as these point towards protein purity
being a dominant contributor to failed crystallization attempts
and have formed more-or-less unsubstantiated guidelines
for purity requirements in membrane-protein crystallization
trials.
With respect to lipid content, the guidelines for membrane-
protein crystallization have changed over time. At the dawn of
membrane-protein crystallization, it was believed to be of the
utmost importance to remove as many endogenous lipids from
the extracted protein molecules as possible. However, results
over the years have changed this perception (De Foresta et al.,
1994; Garavito & Ferguson-Miller, 2001; Garavito et al., 1996;
Haneskog et al., 1996; Kragh-Hansen et al., 1998; Lund et al.,
1989). The proper functioning of membrane proteins in the
complex environment of a biological lipid bilayer (White &
Wimley, 1999) may require specific lipid–membrane protein
interactions and indeed lipids have been directly observed in
several membrane-protein crystal structures (Ferguson et al.,
2000; Jones, 2007; Camara-Artigas et al., 2002; McAuley et al.,
1999; Roszak et al., 2007; Zhang et al., 2003; Belrhali et al.,
1999; Ferguson et al., 1998; Harrenga & Michel, 1999; Nollert,
2005; Tsukihara et al., 1996; Jordan et al., 2001). The supple-
mentation of solubilized membrane-protein samples with
lipids after their purification has improved or was required in
the procedure to crystallize several membrane proteins (Guan
et al., 2006; Newman et al., 1981; Toyoshima et al., 2000).
For the growth of membrane-protein crystals of sufficient
quality for X-ray diffraction experiments, protein particles
must associate in an amphiphilic environment that leads to the
formation of an ordered crystalline structure, as opposed to a
non-ordered aggregated material (Nollert, 2005). This is
practically pursued by testing crystallization conditions in
crystallization trials, the latter of which are usually conducted
in time-consuming trial-and-error experiments in which
hundreds or even thousands of crystallization cocktails are
sampled (Chang et al., 1998; Dahl et al., 2004). In these trials
most experiments produce aggregated material, rarely pro-
viding clues for further experiments. The initial discovery of a
crystallization ‘hit’ for a given membrane protein, however, is
an important milestone since these crystals demonstrate that
many of the experimental parameters required for crystal
growth have been identified and optimized parameters may be
within short reach. Even the appearance of tiny or poorly
diffracting crystals confirms that the protein under investiga-
tion can indeed form crystals (is ‘crystallizable’) and, provided
that the protein sample and crystallization parameters can be
improved, better crystals may eventually form. In addition, the
crystalline material itself can be utilized as seeds in further
crystallization experiments (Bergfors, 1999).
Conversely, the formation of unproductive aggregates may
arise for a number of reasons including poor quality or purity
of the protein sample, insufficient chemical composition of the
crystallization cocktail or inadequate crystallization kinetics.
Most researchers eventually call into question the purity of
their sample or their choice of purification detergent. Indeed,
protein-sample purity is important for successful nucleation,
growth, and affects crystal quality since impurities cause
undesirable interactions on the surface of growing crystals
(Anderson et al., 1988; Caylor et al., 1999; Kurihara et al., 1999;
Plomp et al., 2003; Van der Laan et al., 1989; Vekilov &
Rosenberger, 1996). Impurities are often associated with ‘step
pinning’, where they adsorb to the surface of a growing crystal
and impede the addition of desired components (Land et al.,
1999; McPherson et al., 1996; Plomp et al., 2003; Sangwal, 1996;
van Enckevort et al., 1996).
The sitting-drop vapour-diffusion technique produced the
crystals used to solve the first known structure of a membrane
protein (Deisenhofer et al., 1984) and has long been most
frequently employed for membrane-protein crystallization.
The preparation of such crystallization experiments involves
combining solutions of salts and/or polyethylene glycol (PEG)
with a protein sample (small amphiphiles such as heptane-
1,2,3-triol can also be supplemented), causing the protein to
become supersaturated, which is aided by concomitant con-
trolled dehydration. If conditions are favorable, the growth of
structured and highly ordered protein crystals ensues after the
formation of stable nuclei (Caffrey, 2003; Wiener, 2004).
Over a decade later, the crystallization of bacterio-
rhodopsin within a lipidic cubic phase (LCP) matrix was first
described (Landau & Rosenbusch, 1996). This crystallization
methodology involves two simple steps. At first, the protein
solution is mixed with a lipid, for example monoolein. In this
material, the lipid self-assembles into a continuous bilayer, for
example a lipidic cubic phase, containing the membrane
protein. In a second step, crystallization is initiated by adding a
crystallization-inducing reagent to the lipid material. The
mechanism of crystallization from LCP is not fully understood
(Nollert et al., 2001), but is likely to involve diffusion within
the bilayer and local concentration of the protein and
restructuring of the lipid mesophase to a lamellar arrangement
where crystal growth occurs through stacked bilayers (Caffrey,
research papers
Acta Cryst. (2009). D65, 1062–1073 Kors et al.  Impurities and membrane-protein crystallization 1063

2008; Grabe et al., 2003). While the actual format has changed
(Cherezov & Caffrey, 2006; Cherezov et al., 2004; Nollert,
2002), this methodology has been key to the crystallization
and subsequent structure determination of a number of
membrane proteins, most prominently that of the -adrene-
nergic receptor (Cherezov et al., 2007).
Interestingly, the crystallization of the bacterial photo-
synthetic reaction center from Rhodopseudomonas viridis in a
monoolein-based matrix (Katona et al., 2003) starts out in a
bona fide LCP and, depending on the duration of the crys-
tallization experiment, the lipid matrix loses viscosity but
remains transparent and nonbirefringent. Lipidic cubic phases
are typically highly viscous, however, and in the absence of
further characterization of the exact nature of the materials
when crystals form, such host matrices will be referred to as
‘lipid mesophases’.
In order to minimize protein-sample consumption in crys-
tallization trials, a plug-based crystallization technique akin to
miniaturized microbatch setups (Chayen, 1992) has been
developed. Unlike in vapor diffusion, the concentration of
each component in the crystallization experiment, once set up,
remains constant. Porin from Rhodobacter capsulatus and the
photosynthetic reaction center (RC) from Rhodopseudo-
monas viridis have been crystallized using this microfluidic
approach (Li et al., 2006). Crystallization-inducing agent(s),
buffer and protein are combined in microchannels made
from polydimethylsiloxane (PDMS) and simultaneously form
10–20 nl crystallization experiments in the form of ‘plugs’
carried by immiscible fluorinated oil (Zheng et al., 2005).
The goal of this study was to compare (i) vapor-diffusion,
(ii) plug-based and (iii) lipidic cubic phase-based crystal-
lization approaches with regard to the sensitivity of crystal-
lization success to sample purity. These experiments were
designed in order to develop best practices in membrane-
protein crystallization projects. The bacterial photosynthetic
reaction center (RC; Allen et al., 1987; Arnoux et al., 1995;
Chang et al., 1991; Deisenhofer et al., 1985; Ermler et al., 1994;
McAuley-Hecht et al., 1998; Stowell et al., 1997) was employed
as a model membrane protein.
This study shows that the LCP crystallization method
produces crystals from RC samples with substantial impurity
levels and in this respect outperforms all other tested methods.
If this finding is applicable to many other membrane proteins,
then its tolerance for impurities predestines the LCP crystal-
lization method as an effective tool for initial crystallization
screening trials.
2. Materials and methods
2.1. Preparation of RC samples
Rhodobacter sphaeroides strains expressing recombinant
polyhistidine-tagged RCs [C-terminal tag (7 CAC) on the M
subunit; GenBank Accession No. K00827; Pokkuluri et al.,
2002] were cultured in YCC medium (Taguchi et al., 1992) for
2–3 d in 2.8 l Fernbach flasks (2 l per flask). Cells were
harvested at 12 500g. The pellets were combined and washed
in buffer 1 (10 mM Tris pH 7.8, 10 mM NaCl). The cell pellets
were resuspended in buffer 1 and lysed by sonication and
three serial passages through a microfluidizer (Model M-110L,
Microfluidics, Newton, Massachusetts, USA). Unbroken cells
and debris were removed by centrifugation at 22 000g for
15 min at 277 K. Membranes were pelleted by ultra-
centrifugation of the supernatant at 245 000g for 120 min at
277 K. Membrane pellets were weighed and resuspended
in buffer 1 at 12.5 ml g1 using a tissue homogenizer. The
proteins embedded within these homogenized membranes
were then solubilized by incubation for 2–3 min at 310 K (with
stirring, in darkness) in 1%(w/v) N,N-dimethyldodecylamine-
N-oxide (LDAO; Sigma–Aldrich, St Louis, Missouri, USA;
CMC 0.023% as per Hermann, 1962).
Membrane debris was removed by ultracentrifugation of
the suspension at 245 000g for 120 min at 277 K. The super-
natant was then filtered (0.45 mm) prior to protein purification
via one of the two following methods.
(i) The sample of lowest purity level (A280/A800 = 2.4) was
prepared using customized automated scripts (adapted from
Kirmaier et al., 2005) on an A¨KTA FPLC (GE Healthcare,
Piscataway, New Jersey, USA), employing incomplete column-
washing steps. In this method, the supernatants were passed
twice over a 5 ml HiTrap Chelating HP Column (GE Health-
care) prepared with 0.1 M NiSO4. The column was then
washed partially ( three column volumes) with 10 mM Tris
pH 7.8, 0.05%(w/v) LDAO to remove a portion of the loosely
bound components. Proteins were eluted with 10 mM Tris,
0.05%(w/v) LDAO, 100 mM imidazole pH 7.8 and subse-
quently desalted using a HiPrep 26/10 column (GE Heath-
care). Collected fractions were combined and concentrated in
a centrifugal filter (Amicon Ultra, 30 000 molecular-weight
cutoff; Millipore, Billerica, Massachusetts, USA). The protein
concentration and purity level were determined by UV–Vis–
near-IR spectroscopy. The A280/A800 ratio was used to monitor
the purity by assessing the amount of bacteriochlorophyll-
containing RCs relative to the total protein content of the
sample.
(ii) The sample of highest purity level (A280/A800 = 1.4) was
prepared manually by passing the supernatant twice over a
column composed of 10 ml Ni–NTA Superflow resin (Qiagen,
Valencia, California, USA). The column was washed exten-
sively with ten column volumes of 10 mM Tris pH 7.8,
0.05%(w/v) LDAO to remove loosely bound components.
Proteins were eluted with 10 mM Tris, 0.05%(w/v) LDAO,
100 mM imidazole pH 7.8 and were subsequently additionally
purified on a column of DEAE Sephacel resin (Sigma–
Aldrich, St Louis, Missouri, USA). The column was washed
with more than five column volumes of 10 mM Tris pH 7.8,
0.05%(w/v) LDAO until the A280 of the eluate was less than
0.1. Proteins were eluted with 10 mM Tris, 0.05%(w/v) LDAO,
280 mM NaCl pH 7.8 with manual collection of fractions.
Fractions were combined and concentrated with a centrifugal
filter as described above. The protein concentration was
determined by UV–Vis–near-IR spectroscopy.
Samples of intermediate purity were prepared by simple
mixing of the above two extremes and were monitored spec-
research papers
1064 Kors et al.  Impurities and membrane-protein crystallization Acta Cryst. (2009). D65, 1062–1073

troscopically. All RC samples were concentrated to an A800 of
18 (6 mg ml1).
2.2. Purity assessment of RC samples
The purity levels of R. sphaeroides RC samples were also
assessed by SDS–PAGE to determine the nature of the protein
contaminants. The gels were PAGEr Gold Precast 8–16%
acrylamide (Lonza, Walkersville, Maryland, USA) or Novex
NuPAGE bis-tris 4–12% acrylamide (Invitrogen, Carlsbad,
California, USA) and were stained with Bio-Safe Coomassie
Stain (Bio-Rad, Hercules, California, USA) or Coomassie
Brilliant Blue R-250 according to the manufacturers’ proto-
cols.
In addition, the lipid content of each sample was deter-
mined by thin-layer chromatography (TLC) as described
previously (Eriks et al., 2003; Kors et al., 2009). In brief, a TLC
tank lined with 3 mm CHR pure cellulose chromatography
paper (Whatman, Florham Park, New Jersey, USA) and filled
with solvent (chloroform:methanol:ammonium hydroxide,
63:35:5 by volume; Fisher Scientific, Waltham, Massachusetts,
USA) was sealed and equilibrated for 1 h. Samples (5 ml) were
spotted on 10  20 cm Silica Gel 60 TLC plates (EMD
Chemicals, Gibbstown, New Jersey, USA). Spotted plates
were allowed to dry and were then placed into the sealed TLC
chamber. Once the solvent had migrated to 1–2 cm from
the top of the plates, they were removed and allowed to dry
thoroughly.
For lipid visualization, a large desiccator was heated to
333 K in a hot-water bath. Resublimed iodine crystals (Fisher
Scientific) were placed at the bottom of the desiccator and the
TLC plates were stained in the desiccator for no longer than
15 min. Plates were imaged immediately in order to record the
maximum intensity of the short-lived iodine signal.
Destained gels and stained TLC plates were scanned and
ImageJ v.1.36 (Abramoff et al., 2004) was used to quantify the
signals from the images.
2.3. Methods for crystallization of RC samples
Crystallization experiments were performed in parallel
using LCP, plug-based microbatch and sitting-drop vapor-
diffusion crystallization techniques and were conducted within
3 d of protein purification and initial characterization.
2.3.1. Vapor diffusion. Sitting-drop vapor-diffusion crys-
tallization trials of RC samples with varying A280/A800 ratios
were set up similarly to previously described conditions
(Chang et al., 1985) at room temperature. Droplets consisted
of 0.9–2.0 mg ml1 RCs, 16–21%(w/v) PEG 4000, 0.28 M
NaCl, 3%(w/v) 1,2,3-heptanetriol (high-melting point isomer;
Sigma–Aldrich) and 0.05%(w/v) LDAO. Typical droplet
volumes were 12.5–25 ml (for microlitre trials) or 400 nl (for
nanolitre trials). This RC mixture was equilibrated against
0.75–1 ml reservoirs (for microlitre trials) or 150 ml reservoirs
(for nanolitre trials) of 10 mM Tris pH 7.8, 0.56 M NaCl,
25%(w/v) PEG 4000. Plates were stored at room temperature
in the dark. The microlitre-volume trials were set up manually,
whereas the nanolitre-volume trials were set up using a robot
(Mosquito, TTP LabTech, Melbourn, England). Independent
control crystals of RC grown under the conditions referenced
above grew up to 4 mm in size in large setups. Such crystal
sizes agree with those reported in the literature (Chang et al.,
1985).
2.3.2. Microbatch plugs. The batch-mode experimental
setup has been described previously (Li et al., 2006). On a
microfluidic chip, the protein sample was combined with
buffer and crystallization-inducing agent streams and 10–15 nl
plugs were formed in the Teflon tubing (200 mm inner
diameter, 250 mm outer diameter). The plugs were carried
by a mixture of perfluoro-tri-n-butylamine and perfluoro-
di-n-butylmethylamine (FC-40). The buffer was 0.15%(w/v)
LDAO, 9.8%(w/v) 1,2,3-heptanetriol, 10 mM Tris pH 7.8; the
crystallization-inducing agent was 0.15%(w/v) LDAO,
9.8%(w/v) 1,2,3-heptanetriol, 1.1 M NaCl, 50%(w/v) PEG
4000, 50 mM Tris pH 7.8. The flow rate of the protein sample
was kept constant at 0.6 ml min1 and the stream of the crys-
tallization-inducing agent ranged from 0.4 to 0.7 ml min1 with
0.1 ml min1 increments; the buffer stream ranged from 0.4 to
0.1 ml min1 accordingly to maintain the total aqueous flow
rate at 1.4 ml min1 and the carrier fluid, FC-40, ranged from
1.4 to 2.6 ml min1 with 0.3 ml min1 increments in phase with
the increase in flow rate of the crystallization-inducing agent.
By changing the flow rate of the carrier fluid, the size of the
plugs changed, a parameter that can be used to index the
concentration of crystallization-inducing agent such that
larger plugs contain less of it (Li et al., 2006). With this setup,
the protein, LDAO and heptanetriol were kept at the same
concentrations of 2.6 mg ml1, 0.1%(w/v) and 5.6%(w/v),
respectively; the NaCl and PEG 4000 concentrations ranged
from 0.55 M and 25%(w/v) to 0.31 M and 14%(w/v), respec-
tively. In order to prevent evaporation, the Teflon tubing
housing the crystallization trials was stored in additional glass
tubing (1 mm inner diameter, 2 mm outer diameter) which was
pre-filled with perfluorotripentylamine (FC-70) and sealed
with wax at both ends. The trials were incubated at 296 K.
2.3.3. LCP. Proteo-LCP was prepared using a 40:60(w:w)
ratio of protein solution:monoolein (Nu-Chek, Elysian,
Minnesota, USA) by the microcrystallization cubic phase
method (Nollert, 2004). Briefly, one 250 ml syringe containing
pre-weighed solid monoolein (MO) was connected to a second
250 ml syringe containing protein solution and the solutions
were passed back and forth via a syringe coupler (Cubic LCP
kit; Emerald BioSystems, Bainbridge Island, Washington,
USA) until the material became transparent and uniform. A
10 ml syringe mounted in a repeating dispenser (Hamilton,
Reno, Nevada, USA) was loaded with proteo-LCP by way of
the syringe coupler. Proteo-LCP slugs of 0.4 ml volume were
delivered into the drop chamber of a Compact Jr plate
(Emerald BioSystems) which was pre-dispensed with 2 ml
crystallization-inducing reagent solution from the reservoir.
After sealing with tape, the drop chamber was equilibrated
with a reservoir of 80 ml in volume. The plates were wrapped
in foil and stored in a dark cabinet at 289 K. Samples in LCP
were incubated in a series of conditions (comprising a
48-condition grid matrix) in which the concentration of
research papers
Acta Cryst. (2009). D65, 1062–1073 Kors et al.  Impurities and membrane-protein crystallization 1065

Jeffamine M-600 (Sigma–Aldrich) was varied from 7 to
18%(v/v), the ammonium sulfate (Sigma–Aldrich) concen-
tration ranged from 0.7 to 1.15 M and the buffer (HEPES pH
7.5) was held constant at 1 M. A significant variation was
observed in the pH of Jeffamine M-600 obtained from
different suppliers and in different lots obtained from the
same supplier; care was therefore taken to make certain that
the crystallization mixtures had a final
measured pH of 7.2. Statistics for LCP
crystallization-trial success were
computed as a percentage of hits from a
48-condition screen.
2.4. X-ray diffraction of RC crystals
In order to minimize the potential
damage to crystals arising from hand-
ling and cryoprotection, crystals from
LCP and vapor-diffusion trials were
examined at room temperature without
transfer from the bulk crystallization
solution. As expected, the X-ray
diffraction limits of RC crystals tested at
room temperature were poor (10 A˚).
Crystals of RCs grown for control
purposes and tested for X-ray diffrac-
tion at liquid-nitrogen temperature
routinely diffracted to better than 3.5 A˚
resolution using synchrotron radiation.
Large portions of the crystallization
drop were drawn into glass capillaries
by gentle aspiration of bulk crystal-
lization-inducing reagent. Capillaries
containing crystals were mounted on a
goniostat and aligned in the X-ray beam
using a microscope. In the case of
microbatch trials, in situ diffraction
experiments were conducted. X-ray
diffraction experiments were conducted
on beamline 19BM of the Advanced
Photon Source. Crystals were exposed
to the unattenuated beam for 5 s and
diffraction data were collected on a
CCD detector to assess the intrinsic
diffraction limit.
2.5. Lipids from R. sphaeroides, brain
and E. coli
For the preparation of R. sphaeroides
lipid samples, approximately 4 g of the
membrane-debris pellet from x2.1 was
homogenized in a final volume of 8 ml
10 mM Tris pH 7.5. This membrane
solution was used as the starting mate-
rial in an organic lipid-extraction pro-
cedure (Bligh & Dyer, 1959) comprising
serial additions, with vortexing, to the
homogenate of 30 ml 1:2 CHCl3:MeOH, 10 ml CHCl3 and
10 ml deionized water. The solution was centrifuged for 5 min
at 1000g at room temperature to separate the phases. The
lower organic phase was transferred to a clean glass vial, dried
under vacuum and stored at 253 K under inert gas. Polar brain
(porcine) lipid extract and E. coli polar lipid extract were
purchased from Avanti Polar Lipids (Alabaster, AL, USA).
research papers
1066 Kors et al.  Impurities and membrane-protein crystallization Acta Cryst. (2009). D65, 1062–1073
Figure 1
Characterization of protein and lipid impurities found within partially purified RC samples used for
crystallization experiments. Protein analyses (a, c) are depicted as digitized intensities from
Coomassie-stained SDS–PAGE gels. Lipid analyses (b, d) comprise iodine-stained TLC plates. For
the partially purified samples (a) and (b), the RC content was kept constant using 12 and 30 mg in
each gel and TLC lane, respectively. The numbers above the lanes indicate the A280/A800 ratio of the
sample. Assignments of spots on the TLC plate in (b) signify the detergent and lipid components
present in the samples that are resolved by this solvent system: Pigments, a mixture comprised of
bacteriochlorophylls, bacteriopheophytins, carotenoids and quinones; LDAO, N,N-dimethyl-
dodecylamine-N-oxide; MGDG, monogalactosyldiacylglycerol; CL, cardiolipin; PE, phosphatidyl-
ethanolamine; PC, phosphatidylcholine; PG, phosphatidylglycerol. Increases in LDAO intensities
observed on the TLC plate in (b) can be attributed to an overall increase in total protein in samples
with decreasing purity (increasing levels of impurities), resulting in a higher overall level of PDCs
(daCosta & Baenziger, 2003). For samples with lipids or membranes added (c, d), the exact contents
of the LCP-based trials (0.5 mg of each) in the absence of MO were loaded for analysis. The samples
contained (1) RCs only, (2) RCs plus 12% polar brain lipids, (3) RCs plus 18% polar E. coli lipids,
(4) RCs plus 1.2% extracted R. sphaeroides lipids, (5) RCs plus 12% R. sphaeroides whole
membranes. For both sets of samples, bands corresponding to the three protein subunits (L, M and
H; 25–30 kDa) of the R. sphaeroides RC complex are marked with a bracket and the lanes
containing molecular-weight standards [ProSieve Protein Markers from Lonza in (a) and Full-
Range Rainbow Marker from GE Healthcare in (c)] are indicated (lane L).

2.6. LCP crystallization in the presence of lipids
Polar brain, polar E. coli and extracted R. sphaeroides
lipids were reconstituted independently to a concentration of
100 mg ml1 in a MeOH–CHCl3 mixture (1:2 ratio). Parallel
experiments were conducted by replacing 0.5–30% of the mass
of MO (0.3–18% of the total LCP mass) in LCP trials with one
of the above lipid mixtures and setting up trials as described in
x2.3.3. The MO–lipid material was preformed by melting at
310 K to facilitate mixing. The protein sample used was the
R. sphaeroides RC sample of highest purity (A280/A800 = 1.4).
The resultant LCP preparations were screened for crystal-
lization against the same 48-condition crystallization screen
employed previously (x2.3.3).
2.7. Preparation of R. sphaeroides membrane samples
Pellets containing R. sphaeroides membrane vesicles (x2.1)
were resuspended in 12.5 ml buffer 1 per gram and were
treated with 0.03%(v/v) LDAO for 15 min with stirring in the
dark. This low concentration of LDAO preserved the
embedded membrane proteins but permeabilized the inside-
out vesicles, releasing the trapped soluble contents (peri-
plasmic proteins). The detergent was removed by ultra-
centrifugation at 245 000g for 120 min at 277 K. The pelleted
membranes were homogenized in buffer 1 and were treated in
the same manner two more times; they were then resuspended
in buffer 1.
2.8. LCP crystallization in the presence of membrane
fragments
Purified membranes from R. sphaeroides (x2.7) were added
to the LCP-based crystallization trials by diluting membranes
into the aqueous protein fraction prior to mixing it with solid
MO. Membrane paste was added by replacing 1–30%(w/v) of
the purified LDAO-solubilized R. sphaeroides RCs (0.4–12%
of the total mass of the LCP trial). The protein sample used
was an R. sphaeroides RC sample of intermediate purity (A280/
A800 = 2.0). The resulting LCP preparations were screened for
crystallization against the same 48-condition crystallization
screen employed previously (x2.3.3).
2.9. Characterization of protein samples
The purest preparations of R. sphaeroidesRCs that retained
all aspects of their light-driven charge-separation function
were characterized by A280/A800 ratios (total protein/bound
monomeric bacteriochlorophyll) of 1.2. SDS–PAGE gels of
these ultrapure RCs (stained with Coomassie Blue or Silver)
revealed few if any impurities and such a sample was defined
in this study as being 100% pure. Purified RC samples having
absorption ratios of0.4 (85% pure) are known to be highly
crystallizable (Pokkuluri et al., 2002). To explore the limits of
such samples with various crystallization approaches, two
types of RC samples were purified from membranes of
R. sphaeroides (A280/A800 = 1.4 and 2.4). Samples of inter-
mediate purity (A280/A800 = 1.5, 1.6, 2.0 and 2.2) resulted by
simple linear mixing. Analysis by UV–Vis–near-IR absorption
spectroscopy, complemented by SDS–PAGE (Fig. 1a), where
band intensities were quantified by ImageJ, suggested that
these RC samples ranged in purity from 86 to 50%, respec-
tively. This implies that 14–50% of the samples were
contaminating proteins. As expected, a gradual increase in
background staining was observed on gels for samples with
increasing A280/A800 ratios and was the direct result of rising
contamination levels of proteins of various sizes.
Gradual increases in protein contamination were also
accompanied by increased levels of lipids that were remnants
of the suboptimal purification process. TLC analysis was used
to visualize the general lipid content (Fig. 1b). Not surpris-
ingly, background staining at specific Rf values attributable to
lipids in the sample increased with increasing A280/A800 ratio.
The RC sample of least purity (A280/A800 = 2.4) appeared to
contain 12 times more iodine-staining material as quantified
using ImageJ in comparison to the sample with the highest
purity level (A280/A800 = 1.4). The most intensely staining lipid
species were identified based upon their known relative
mobilities in TLC, their presence in purified RC samples and
their ability to be removed during purification if certain
chromatographic schemes were utilized (Albuquerque et al.,
2002; Catucci et al., 2004; Dezi et al., 2007; Ventrella et al.,
2007; Camara-Artigas et al., 2002).
The impurities that were encountered in these samples
would be similar to the impurities encountered in any purified
membrane-protein sample, especially for a protein that has
not been characterized and that has possibly been purified for
the first time. No purification method is immune to the
presence of contaminants. Even with recent advances in affi-
nity chromatography, the purification of membrane-protein
samples to homogeneity is frequently an overwhelming
obstacle in structural and functional studies. Difficulties in
membrane-protein purification are in many ways inherent to
the requirement for using a detergent to solubilize the
macromolecules, thereby making them amenable to aqueous-
based chromatographies. Thus, when purifying membrane
proteins, one is actually purifying a protein–detergent
complex, with which endogenous lipids frequently associate
and copurify. The detergent micelle that encompasses the
hydrophobic regions of the proteins limits access to tags in
affinity chromatography, masks interactions with charged
resins in ion-exchange chromatography and modulates size
and limits separation in size-exclusion chromatography. It is
for these reasons that we sought a crystallization approach
research papers
Acta Cryst. (2009). D65, 1062–1073 Kors et al.  Impurities and membrane-protein crystallization 1067
Table 1
The abilities of the various crystallization methods to tolerate increased
levels of impurities.
Crystallization method
Maximum A280/A800
ratio of sample
producing crystals
RC
content
(%)
Impurity
content
(%)
LCP 2.4 50 50
Microcapillary 2.0 60 40
Sitting-drop vapor diffusion
(nanolitre trials)
1.6 75 25
Sitting-drop vapor diffusion
(microlitre trials)
1.6 75 25

that was tolerant of moieties that are encountered commonly
in membrane-protein samples.
2.10. Characterization of crystallization additives
The RC samples with varying amounts of added lipid (polar
brain lipids, polar E. coli lipids or extracted R. sphaeroides
lipids) were analyzed by SDS–PAGE (Fig. 1c). From these
analyses, it was determined that none of the lipid extracts
utilized contained extraneous protein, as reflected by the
relatively constant background:RC ratios, even when large
quantities (18% of the total weight of the droplet) of lipids
were added.
Likewise, the quality of the lipid extract added to each RC
sample was assessed on TLC plates (Fig. 1d). Background
staining attributable to a mixture of lipids in the samples
increased with increasing amounts of added lipid. The samples
shown represent those with the highest percentage for each
lipid additive utilized in these experiments. Although indivi-
dual species cannot be resolved, it is clear that there are
research papers
1068 Kors et al.  Impurities and membrane-protein crystallization Acta Cryst. (2009). D65, 1062–1073
Figure 2
Images of crystallization experiments with RC samples of increasing purity (as reflected by the A280/A800 ratio) using LCP, microfluidics or sitting-drop
vapor-diffusion techniques. (a) Holistic view of crystallization trials and (b) enhanced magnification, to a uniform scale, for comparison of crystal size and
quality.

differences in the numbers and types of lipids. The samples
containing polar brain lipids appear to be the most diverse, the
samples containing the extracted R. sphaeroides lipids are the
least complex and the polar E. coli lipid extract may share
species with the polar brain lipids.
3. Results and discussion
3.1. Effect of protein purity on crystallization
When compared with sitting-drop vapor diffusion and
crystallization in the plug-based microbatch system, our
results clearly show that the LCP method produced RC
crystals from samples containing the highest contamination
levels (Table 1, Fig. 2). Importantly, the contaminants faith-
fully represent native protein and lipid impurities that typi-
cally arise from incomplete purification processes. Should this
finding hold true for many membrane proteins, it would be
advisable for membrane-protein researchers to focus on LCP-
based crystallization experiments early on in membrane-
protein purification and crystallization trials.
RC crystals grew best in the LCP trials and are almost
unaffected by impurities, as shown in Fig. 2. Indeed, the LCP
method tolerated levels of impurities that were equal to the
amount of target protein present (up to 50% impurities;
Table 1, Fig. 1). Crystals appeared in setups with RC A280/A800
absorption ranging from 1.4 to 2.4 (Fig. 2) and were visible
after 48 h; larger crystals grew to full size after 5 d. In subse-
quent experiments with an RC sample of A280/A800 = 2.8
(43% RC) no crystals were obtained (data not shown).
For vapor-diffusion trials the formation of crystals was
limited to an A280/A800 ratio of 1.6 or below (75% RC), while
microbatch trials produced few crystals at an A280/A800 ratio of
2.0 (60% RC; Table 1, Fig. 1).
For nanolitre sitting-drop vapor-diffusion trials, samples
with lower impurity levels (A280/A800 ratios up to 1.6) yielded
crystals within 2–3 d. Samples with A280/A800 ratios of 1.4 and
1.5 yielded crystals within 2–3 d for microlitre trials, while
crystals started to appear in a sample with an A280/A800 ratio of
1.6 after 10 d (Fig. 2). Crystals were never observed with
samples having A280/A800 ratios of 2.0, 2.2 or 2.4 using vapor
diffusion. The tolerance for protein impurities using vapor-
diffusion approaches was independent of the volume of the
experiment.
For microfluidic plug-based batch trials, samples with A280/
A800 ratios of up to 1.6 yielded crystals within 2–3 d. After
10 d, the sample with an A280/A800 ratio of 2.0 also yielded
crystals. Crystals were never obtained from samples with A280/
A800 ratios of 2.2 or 2.4 using this approach (Fig. 2).
For vapor-diffusion and microbatch plug-based crystal-
lizations that employed PEG 4000 as the crystallization-
inducing agent, the numbers and sizes of crystals scaled with
the volume of the crystallization droplet (Fig. 3). In addition,
the number of crystals obtained decreased rapidly as the
impurity level increased. For microlitre-volume sitting-drop
vapor-diffusion trials crystal size decreased rapidly with the
addition of impurities, while crystal size became maximal at
intermediate impurity levels (presumably owing to decreased
nucleation events) for nanolitre-volume vapor-diffusion and
nanolitre-volume microbatch trials (Fig. 3).
These results demonstrate that crystal nucleation, not
crystal growth, is affected by the presence of protein impu-
rities. Crystal size was limited in the samples of highest purity
presumably because of the large number of nuclei that
formed, which ultimately limited the amount of protein
research papers
Acta Cryst. (2009). D65, 1062–1073 Kors et al.  Impurities and membrane-protein crystallization 1069
Figure 3
The effect of impurities (variations in the A280/A800 ratio) on (a) the average number of crystals obtained and (b) the average crystal length in LCP (black
diamonds), microfluidics (green circles), nanolitre vapor-diffusion (red triangles) and microlitre vapor-diffusion (blue squares) trials. The inset in (a)
represents a magnified view of the portion of the graph in which the data for LCP and microfluidic trials reside. Error bars represent standard deviations
from three or more independent crystallization experiments.

available for crystal growth. Although the number of nuclei
decreased in samples of lesser purity, crystal growth was not
impeded as these samples produced larger crystals than those
experiments in which showers of crystals appeared.
For the LCP-based crystallizations utilizing Jeffamine and
ammonium sulfate as the crystallization-inducing agents, the
number of crystals obtained remained constant across the
protein purity levels (Fig. 3). Not only did nuclei formation
seem to be less impeded for the LCP method, but the sizes of
the crystals obtained using this approach actually increased
with decreasing protein purity. This surprising result violates
the understanding of typical crystal growth in aqueous solu-
tion, where impurities disrupt the uniform assembly of layers
of the crystal lattice (Yip & Ward, 1996). The microscopic
distribution of impurities in the lipid mesophase matrix may
allow more efficient diffusion of monodispersed RCs to the
area of crystal growth and efficient diffusion of impurities
away from the zone of crystal growth. Crystal growth may
have ultimately been limited by diffusion and the complex
nature of the lipid mesophase matrix (Fig. 4).
We also noted that increasing sample-impurity levels
resulted in a greater amount of precipitate in microbatch trials
as well as in microlitre-volume and
nanolitre-volume vapor-diffusion trials
(Fig. 2). This gross aggregation formed
with kinetics similar to the formation of
crystals and was not observed in LCP
crystallization trials (Fig. 2). The
presence of contaminating proteins and
lipids obviously contributed to this
unproductive end product, impeding
structured crystal formation from input
proteins that are crystallizable in purer
form. Although the lack of this aggre-
gation in the lipid mesophase may be a
consequence of the differences in crys-
tallization-inducing agents, its absence
in the lipid mesophase may help to
explain the improved tolerance of this
method towards the impurities that
were introduced (Fig. 4). It is possible
that the lipid bilayer in the lipid meso-
phase may prevent aggregation or
crystal-growth poisoning by solubilizing
impurities.
3.2. Effect of protein purity on crystal
quality
RC crystals grew in the LCP trials
and diffracted X-rays to a resolution of
3.5 A˚ regardless of the purity level in
the crystallization setups (Fig. 5).
Hence, the crystallization behavior and
crystal quality are tolerant to impurity
levels typically found after only one or
two purification steps. The latter is a
highly desirable attribute of this crystallization approach. This
finding is reminiscent of the crystallization of bacter-
iorhodopsin from purple membrane fractions that have
undergone no chromatographic purification (Nollert et al.,
1999) and may indicate a general feature of the LCP-based
membrane-protein crystallization method to yield crystals of
sufficient X-ray diffraction quality from relatively impure
starting material.
Conversely, crystals grown by vapor diffusion and micro-
batch techniques did not diffract beyond 10 A˚ resolution with
synchrotron radiation. A factor in this very low diffraction
resolution may have been that the needles that formed had
only one dimension larger than 10 mm compared with the
100 mm2 size of the X-ray beam. Larger cryopreserved
crystals grown previously using slight variations in crystal-
lization conditions have diffracted to 3.2 A˚ resolution (Chang
et al., 1991; Marone et al., 1999) and we presume that the
quality of the crystals that formed in these experiments is
similar.
In this study, the X-ray diffraction limits of the small crystals
formed from impure membrane-protein samples equalled the
diffraction limits of larger crystals formed by the same method
research papers
1070 Kors et al.  Impurities and membrane-protein crystallization Acta Cryst. (2009). D65, 1062–1073
Figure 4
Illustration depicting why the LCP methodology has a high tolerance for contaminating protein
species (black and green), based on the quantitative mechanistic framework for crystal growth
(Grabe et al., 2003; Nollert et al., 2001). Here, impurities are excluded from the crystal-growth
process, essentially providing a microenvironment enriched in the crystallizing species (gray and
yellow), thus favoring crystal growth. Along the lines of a ‘kinetic exclusion mechanism’,
contaminating protein species with large hydrophilic or hydrophobic moieties face an energy
penalty for diffusion in curved membranes with small channels, resulting in less unproductive
aggregation and hence less interference with the desired crystallization process. Similarly, lipidic
cubic phases form substantial diffusion barriers for soluble proteins (Razumas et al., 1996), trapping
soluble contaminating proteins within the small hydrophilic channels of the LCP matrix, where they
are excluded from poisoning the crystal-growth surface. The local absence of contaminating species
allows crystals to grow as they would in solution-based crystallization approaches (batch and vapor
diffusion) using samples of higher purity.

but starting from pure protein. If this observation can be
generalized, the diffraction limits of crystals obtained from
initial LCP-based trials should predict the quality of larger
crystals that one could expect to obtain from samples gener-
ated from optimized purification protocols.
3.3. Effect of lipids and membranes on LCP-based
crystallization
Since the formation of RC crystals in LCP-based crystal-
lization trials was robust when challenged with non-RC
protein, we tested whether the same holds for lipids. In this
regard, the data show that the amount and source (polar brain,
polar E. coli or extracted R. sphaeroides lipids) of the lipids
added to the LCP-based crystallization experiments had no
negative effect on crystallization success (Fig. 6a). The number
of crystal-producing conditions was similar for all three lipid
sources and slight increases were observed in the number of
crystal-producing conditions containing more than 5% lipid,
with an average of two additional crystal conditions relative to
those experiments containing less than 5% lipid (Fig. 6a).
Surprisingly, adding extra lipids to LCP trials generally
increased the total number of conditions that produced crys-
tals. However, these experiments were dominated by condi-
tions producing crystals of relatively low quality (Fig. 6b). The
number of conditions producing medium-quality crystals
remained constant, while the number of conditions reporting
at least one high-quality crystal, even though relatively low,
increased significantly with the addition of lipid. This increase
was most pronounced at concentrations of 5–20% added lipids
(Fig. 6b).
Hence, none of the lipid mixtures inhibited crystal growth.
On the contrary, the use of endogenous R. sphaeroides lipids
in the formation of LCP led to slightly more hits per screen on
average (Fig. 6).
research papers
Acta Cryst. (2009). D65, 1062–1073 Kors et al.  Impurities and membrane-protein crystallization 1071
Figure 6
The effect of the addition of lipids on LCP crystallization of RCs. (a) The number of conditions producing crystals out of a total of 48 from the
ammonium sulfate and Jeffamine grid screen, as a function of increasing amounts of extracted R. sphaeroides lipids (squares, dotted line), polar E. coli
lipids (triangles, unbroken line) and polar brain lipid (circles, dashed line). (b) The quality of the RC crystals produced in LCP trials as a function of
increasing amounts of lipid additives. Data for extracted R. sphaeroides, polar E. coli and polar brain lipids were averaged. Values plotted represent
either the total number of crystals observed (diamonds, unbroken line) or a high-quality (triangles, dotted line), medium-quality (circles, dotted line) or
low-quality (squares, dotted line) visual score based on the size, degree of symmetry and edge definition of the crystals obtained.
Figure 5
Diffraction images of crystals produced via LCP approaches using RC samples of varying purity, as reflected by the A280/A800 ratio. The diffraction limit
at room temperature of the three crystals shown with A280/A800 ratios of (a) 1.4, (b) 1.6 and (c) 2.4 were determined to be 3.56, 3.58 and 3.48 A˚,
respectively, using routines built into the software package HKL-2000 (HKL Research Inc., Charlottesville, Virginia, USA).

Since the replacement of up to 30% of the MO in LCP
preparations (18% of the total weight of the droplet) with lipid
extracts did not have a negative impact on the crystallization
of R. sphaeroides RCs, the effect of adding RC-depleted
membranes to LCP preparations prior to crystallization was
examined (Fig. 7). Again, these additions had a minimal
impact on crystallization, with the number of crystal-
producing conditions decreasing only slightly as intact
membranes were introduced (Fig. 7). Unexpectedly, a signifi-
cant number of conditions still produced crystals even when
30%(w/v) of the sample consisted of R. sphaeroides
membranes, showing that additions of intact membranes
(lipids plus protein) were also tolerated well by the LCP-based
crystallization approach.
The observed robustness of the LCP-based crystallizations
are in line with a report on the crystallization of bacterio-
rhodopsin in the absence of detergent starting out from the
enriched purple membranes of Halobacterium salinarum
(Nollert et al., 1999). We may rationalize these findings by
proposing that the doping of LCP-based crystallization
experiments with detergents, protein, lipids or membranes
may encourage lipid membrane restructuring or enhance
membrane transitions during the crystallization process
(Nollert et al., 2001; Fig. 4).
4. Concluding remarks and perspectives
In order to provide guidance for the experimental design
of initial membrane-protein crystallization trials, we have
undertaken a series of comparative experiments to explore the
effect of impurities in the form of proteins and lipids on the
crystallization of membrane proteins in vapor-diffusion, in
plug-based microbatch and in LCP-based crystallization trials.
Employing the bacterial photosynthetic RC from R. sphaer-
oides as a model system, we found that its crystallization using
the LCP approach tolerated the highest levels of both protein
and lipid contaminants. If this finding holds true for many
membrane proteins then it would be advisable to include the
LCP methodology in the first set of crystallization trials.
Alternatively, LCP-based crystallization trials should be
employed with membrane-protein samples that show sub-
stantial levels of impurities. Hence, as a practical matter,
membrane-protein purification development efforts may be
best accompanied by LCP-based crystallizations and approa-
ches which allow relatively impure fractions to be used in
structural studies, as these samples have been shown to yield
good starting points for the optimization of purification and
crystallization experiments.
The authors would like to thank Scott Lovell for assistance
with the crystal diffraction data collection, Hui Li for assis-
tance with operation and programming of the Mosquito
crystallization robot and Donna Mielke and Deborah Hanson
for critical reading of the manuscript. This work was funded by
the NIH Roadmap grant P01 GM075913 and a University of
Chicago/Argonne National Laboratory (ANL) collaborative
research award. This work was also supported by the
University of Chicago and the Department of Energy under
section H.35 of Department of Energy Contract No. DE-
AC02-06CH11357 to UChicago Argonne LLC to manage
Argonne National Laboratory.
References
Abramoff, M. D., Magelhaes, P. J. & Ram, S. J. (2004). Biophotonics
Int. 11, 36–42.
Albuquerque, L., Santos, J., Travassos, P., Nobre, M. F., Rainey, F. A.,
Wait, R., Empadinhas, N., Silva, M. T. & da Costa, M. S. (2002).
Appl. Environ. Microbiol. 68, 4266–4273.
Allen, J. P., Feher, G., Yeates, T. O., Komiya, H. & Rees, D. C. (1987).
Proc. Natl Acad. Sci. USA, 84, 5730–5734.
Anderson, W. F., Boodhoo, A. & Mol, C. D. (1988). J. Cryst. Growth,
90, 153–159.
Arnoux, B., Gaucher, J.-F., Ducruix, A. & Reiss-Husson, F. (1995).
Acta Cryst. D51, 368–379.
Belrhali, H., Nollert, P., Royant, A., Menzel, C., Rosenbusch, J. P.,
Landau, E. M. & Pebay-Peyroula, E. (1999). Structure, 7, 909–917.
Bergfors, T. E. (1999). Protein Crystallization: Techniques, Strategies,
and Tips. A Laboratory Manual. La Jolla: International University
Line.
Bligh, E. G. & Dyer, W. J. (1959). Can. J. Biochem. Physiol. 37,
911–917.
Caffrey, M. (2003). J. Struct. Biol. 142, 108–132.
Caffrey, M. (2008). Cryst. Growth Des. 8, 4244–4254.
Camara-Artigas, A., Brune, D. & Allen, J. P. (2002). Proc. Natl Acad.
Sci. USA, 99, 11055–11060.
Catucci, L., Depalo, N., Lattanzio, V. M., Agostiano, A. & Corcelli, A.
(2004). Biochemistry, 43, 15066–15072.
Caylor, C. L., Dobrianov, I., Lemay, S. G., Kimmer, C., Kriminski, S.,
Finkelstein, K. D., Zipfel, W., Webb, W. W., Thomas, B. R., Chernov,
A. A. & Thorne, R. E. (1999). Proteins, 36, 270–281.
Chang, C.-H., El-Kabbani, O., Tiede, D. M., Norris, J. R. & Schiffer,
M. (1991). Biochemistry, 30, 5352–5360.
Chang, C.-H., Schiffer, M., Tiede, D., Smith, U. & Norris, J. (1985). J.
Mol. Biol. 186, 201–203.
research papers
1072 Kors et al.  Impurities and membrane-protein crystallization Acta Cryst. (2009). D65, 1062–1073
Figure 7
The effect of membrane additives (intact vesicles from engineered RC-
deletion strains of R. sphaeroides) on the crystallization of RC samples
using LCP approaches. The data plotted represent the number of
conditions producing crystals out of a total of 48 from the ammonium
sulfate and Jeffamine grid screen.

Chang, G., Spencer, R. H., Lee, A. T., Barclay, M. T. & Rees, D. C.
(1998). Science, 282, 2220–2226.
Chayen, N. (1992). J. Cryst. Growth, 122, 176–180.
Cherezov, V. & Caffrey, M. (2006). J. Appl. Cryst. 39, 604–606.
Cherezov, V., Peddi, A., Muthusubramaniam, L., Zheng, Y. F. &
Caffrey, M. (2004). Acta Cryst. D60, 1795–1807.
Cherezov, V., Rosenbaum, D. M., Hanson, M. A., Rasmussen, S. G.,
Thian, F. S., Kobilka, T. S., Choi, H. J., Kuhn, P., Weis, W. I., Kobilka,
B. K. & Stevens, R. C. (2007). Science, 318, 1258–1265.
daCosta, C. J. B. & Baenziger, J. E. (2003). Acta Cryst. D59, 77–83.
Dahl, S. G., Sylte, I. & Ravna, A. W. (2004). J. Pharmacol. Exp. Ther.
309, 853–860.
De Foresta, B., Henao, F. & Champeil, P. (1994). Eur. J. Biochem. 223,
359–369.
Deisenhofer, J., Epp, O., Miki, K., Huber, R. & Michel, H. (1984). J.
Mol. Biol. 180, 385–398.
Deisenhofer, J., Epp, O., Miki, K., Huber, R. & Michel, H. (1985).
Nature (London), 318, 618–624.
Dezi, M., Francia, F., Mallardi, A., Colafemmina, G., Palazzo, G. &
Venturoli, G. (2007). Biochim. Biophys. Acta, 1767, 1041–1056.
Enckevort, W. J. P. van, van der Berg, A. C. J. F., Kreuwel, K. B. G.,
Derksen, A. J. & Couto, M. S. (1996). J. Cryst. Growth, 166,
156–161.
Eriks, L. R., Mayor, J. A. & Kaplan, R. S. (2003). Anal. Biochem. 323,
234–241.
Ermler, U., Fritsch, G., Buchanan, S. K. & Michel, H. (1994).
Structure, 2, 925–936.
Ferguson, A. D., Hofmann, E., Coulton, J. W., Diederichs, K. & Welte,
W. (1998). Science, 282, 2215–2220.
Ferguson, A. D., Welte, W., Hofmann, E., Lindner, B., Holst, O.,
Coulton, J. W. & Diederichs, K. (2000). Structure, 8, 585–592.
Garavito, R. M. & Ferguson-Miller, S. (2001). J. Biol. Chem. 276,
32403–32406.
Garavito, R. M., Picot, D. & Loll, P. J. (1996). J. Bioenerg. Biomembr.
28, 13–27.
Geerlof, A. et al. (2006). Acta Cryst. D62, 1125–1136.
Grabe, M., Neu, J., Oster, G. & Nollert, P. (2003). Biophys. J. 84,
854–868.
Guan, L., Smirnova, I. N., Verner, G., Nagamori, S. & Kaback, H. R.
(2006). Proc. Natl Acad. Sci. USA, 103, 1723–1726.
Haneskog, L., Andersson, L., Brekkan, E., Englund, A. K., Kame-
yama, K., Liljas, L., Greijer, E., Fischbarg, J. & Lundahl, P. (1996).
Biochim. Biophys. Acta, 1282, 39–47.
Harrenga, A. & Michel, H. (1999). J. Biol. Chem. 274, 33296–33299.
Heijne, G. von (2007). J. Intern. Med. 261, 543–557.
Hermann, K. W. (1962). J. Phys. Chem. 66, 295–300.
Jones, M. R. (2007). Prog. Lipid Res. 46, 56–87.
Jordan, P., Fromme, P., Witt, H. T., Klukas, O., Saenger, W. & Krauss,
N. (2001). Nature (London), 411, 909–917.
Katona, G., Andreasson, U., Landau, E. M., Andreasson, L. E. &
Neutze, R. (2003). J. Mol. Biol. 331, 681–692.
Kirmaier, C., Bautista, J. A., Laible, P. D., Hanson, D. K. & Holten, D.
(2005). J. Phys. Chem. B, 109, 24160–24172.
Kors, C., Impellitteri, N. & Laible, P. (2009). Submitted.
Kragh-Hansen, U., le Maire, M. & Moller, J. V. (1998). Biophys. J. 75,
2932–2946.
Kurihara, K., Miyashita, S., Sazaki, G., Nakada, T., Durbin, S. D.,
Komatsu, H., Ohda, T. & Ohki, K. (1999). J. Cryst. Growth, 196,
285–290.
Land, T. A., Martin, T. L., Potapenko, S., Palmore, G. T. & De Yoreo,
J. J. (1999). Nature (London), 399, 442–445.
Landau, E. M. & Rosenbusch, J. P. (1996). Proc. Natl Acad. Sci. USA,
93, 14532–14535.
Li, L., Mustafi, D., Fu, Q., Tereshko, V., Chen, D. L., Tice, J. D. &
Ismagilov, R. F. (2006). Proc. Natl Acad. Sci. USA, 103, 19243–
19248.
Lund, S., Orlowski, S., de Foresta, B., Champeil, P., le Maire, M. &
Moller, J. V. (1989). J. Biol. Chem. 264, 4907–4915.
Maire, M. le, Champeil, P. & Moller, J. V. (2000). Biochim. Biophys.
Acta, 1508, 86–111.
Marone, P., Thiyagarajan, P., Wagner, A. & Tiede, D. (1999). J. Cryst.
Growth, 207, 214–225.
McAuley, K. E., Fyfe, P. K., Ridge, J. R., Isaacs, N. W., Cogdell, R. J. &
Jones, M. R. (1999). Proc. Natl Acad. Sci. USA, 96, 14706–
14711.
McAuley-Hecht, K. E., Fyfe, P. K., Ridge, J. P., Prince, S. M., Hunter,
C. N., Isaacs, N. W., Cogdell, R. J. & Jones, M. R. (1998).
Biochemistry, 37, 4740–4750.
McPherson, A., Malkin, A. J., Kuznetsov, Y. G. & Koszelak, S. (1996).
J. Cryst. Growth, 168, 74–92.
Moller, J. V. & le Maire, M. (1993). J. Biol. Chem. 268, 18659–18672.
Moreno, A., The´obald-Dietrich, A., Lorber, B., Sauter, C. & Giege´, R.
(2005). Acta Cryst. D61, 789–792.
Newman, M. J., Foster, D. L., Wilson, T. H. & Kaback, H. R. (1981). J.
Biol. Chem. 256, 11804–11808.
Nollert, P. (2002). J. Appl. Cryst. 35, 637–640.
Nollert, P. (2004). Methods, 34, 348–353.
Nollert, P. (2005). Prog. Biophys. Mol. Biol. 88, 339–357.
Nollert, P., Qiu, H., Caffrey, M., Rosenbusch, J. P. & Landau, E. M.
(2001). FEBS Lett. 504, 179–186.
Nollert, P., Royant, A., Pebay-Peyroula, E. & Landau, E. M. (1999).
FEBS Lett. 457, 205–208.
Plomp, M., McPherson, A. & Malkin, A. J. (2003). Proteins, 50,
486–495.
Pokkuluri, P. R., Laible, P. D., Deng, Y. L., Wong, T. N., Hanson, D. K.
& Schiffer, M. (2002). Biochemistry, 41, 5998–6007.
Prive´, G. G. (2007). Methods, 41, 388–397.
Razumas, V., Talaikyte, Z., Barauskas, J., Larsson, K., Miezis, Y. &
Nylander, T. (1996). Chem. Phys. Lipids, 84, 123–188.
Roszak, A. W., Gardiner, A. T., Isaacs, N. W. & Cogdell, R. J. (2007).
Biochemistry, 46, 2909–2916.
Sangwal, K. (1996). Prog. Cryst. Growth Charact. 32, 3–43.
Stowell, M. H. B., McPhillips, T. M., Rees, D. C., Soltis, S. M., Abresch,
E. & Feher, G. (1997). Science, 276, 812–816.
Taguchi, A. K. W., Stocker, J. W., Alden, R. G., Causgrove, T. P.,
Peloquin, J. M., Boxer, S. G. & Woodbury, N. W. (1992).
Biochemistry, 31, 10345–10355.
Toyoshima, C., Nakasako, M., Nomura, H. & Ogawa, H. (2000).
Nature (London), 405, 647–655.
Tsukihara, T., Aoyama, H., Yamashita, E., Tomizaki, T., Yamaguchi,
H., Shinzawa-Itoh, K., Nakashima, R., Yaono, R. & Yoshikawa, S.
(1996). Science, 272, 1136–1144.
Van der Laan, J. M., Swarte, M. B., Groendijk, H., Hol, W. G. &
Drenth, J. (1989). Eur. J. Biochem. 179, 715–724.
Vekilov, P. G. & Rosenberger, F. (1996). J. Cryst. Growth, 158,
540–551.
Ventrella, A., Catucci, L., Mascolo, G., Corcelli, A. & Agostiano, A.
(2007). Biochim. Biophys. Acta, 1768, 1620–1627.
White, S. H. & Wimley, W. C. (1999). Annu. Rev. Biophys. Biomol.
Struct. 28, 319–365.
Wiener, M. C. (2004). Methods, 34, 364–372.
Yip, C. M. & Ward, M. D. (1996). Biophys. J. 71, 1071–1078.
Zhang, H., Kurisu, G., Smith, J. L. & Cramer, W. A. (2003). Proc. Natl
Acad. Sci. USA, 100, 5160–5163.
Zheng, B., Gerdts, C. J. & Ismagilov, R. F. (2005). Curr. Opin. Struct.
Biol. 15, 548–555.
research papers
Acta Cryst. (2009). D65, 1062–1073 Kors et al.  Impurities and membrane-protein crystallization 1073