Published: May 25, 2011
r 2011 American Chemical Society 3432 dx.doi.org/10.1021/cg2001442 | Cryst. Growth Des. 2011, 11, 3432–3441
ARTICLE
pubs.acs.org/crystal
Random Microseeding: A Theoretical and Practical Exploration of
Seed Stability and Seeding Techniques for Successful Protein
Crystallization
Published as part of the Crystal Growth & Design virtual special issue on the 13th International Conference on
the Crystallization of Biological Macromolecules (ICCBM13)
Patrick D. Shaw Stewart,*,† Stefan A. Kolek,† Richard A. Briggs,†Naomi E. Chayen,‡ and Peter F. M. Baldock†
†Douglas Instruments, Douglas House, East Garston, Hungerford, Berkshire, RG17 7HD, U.K.
‡Biomolecular Medicine, Department of Surgery and Cancer, Faculty of Medicine, Imperial College London,
Sir Alexander Fleming Building, London SW7 2AZ, U.K.
bS Supporting Information
’ INTRODUCTION
Traditionally, microseeding has been used as an optimization
step, where seed crystals are transferred into conditions that are
similar to previously known crystallization conditions.1,2 Ireton
and Stoddard3 introduced a novel, more systematic approach,
referred to as microseed matrix-screening. This method was
automated and further improved by D’Arcy et al.,4 who were the
first to report the use of seeding with random screening kits.
Experience has confirmed that “randomMMS” (rMMS) not only
produces extra hits4,5 but also generates better-diffracting crystals.6
As shown in Figure 1, seeding in screening experiments picks
up viable crystallization conditions in the metastable zone that
without seeds would be overlooked as clear drops. The meta-
stable zone may be much larger than the labile zone, so many
additional hits may be found. Indeed, the labile zone may appear
not to exist, although the metastable zone does; see the work of
Ireton and Stoddard3 for an example.
In spite of its effectiveness, several aspects of rMMS are poorly
understood. We started this project with the objective of using
seeding and other nucleation methods to generate better-
diffracting crystals. We had eight specific questions in mind:
(1) how canwe carry out rMMS in away thatwill give themaximum
number of crystal hits? (2) How can we produce crystals with
different space groups? (3) How can we reduce the number of
salt crystals that arise in rMMS experiments? (4) How can we use
microseeding for crystallizing protein complexes? (5) How can
we stabilize seed stocks? (6) Is it helpful to “preseed” the protein
stock? (7) Can we harvest seed crystals frommicrofluidic devices
and capillaries? And, (8) can we encourage crystal nucleation if
we have not yet obtained our first crystals?
One of the innovations in D’Arcy’s 2007 paper4 was to
suspend seed crystals in the solution that was in the reservoir
of the well that the seed crystals were taken from, saving protein
and simplifying the procedure. We will refer to this solution as
the “Hit Solution”. We wanted to investigate the stability of the
seed crystals in the Hit Solution and in other precipitants because
the seed stock generally contains very little protein. In a typical
experiment, the contents of a well with crystals (with a volume of
say 0.6 μL) are suspended in 50 μL of Hit Solution. This means
that the protein concentration in the seed stock is about 1.2% of
the concentration that gave crystals in the original hit, say 0.06
mg/mL. This is likely to be near the point labeled “SHS” in
Received: January 28, 2011
Revised: May 20, 2011
ABSTRACT: Microseed matrix-screening combined with ran-
dom screens (rMMS) is a significant recent breakthrough in
protein crystallization. In this study, a very reproducible assay for
crystal seeds was set up that allowed the following recommenda-
tions to be made: (1) the suitability of a solution for suspending
seed crystals can be predicted by incubating (uncrushed) crystals
in it for one day and observing crystal stability. (2) For routine
rMMS, seed crystals should be suspended in the crystallization
cocktail that gave the original crystals. (3) Seed crystals can be suspended in PEG or NaCl solutions to reduce the prevalence of salt
crystals. (4) Protein complexes can be seeded with seed crystals suspended in PEG. If necessary, seed crystals can also be suspended in
the original crystallization cocktail with any individual ingredients that destabilize the complex removed. (5) “Preseeding” of the protein
stock should not be used if rMMS is available, because it is less effective. (6) Seed crystals can be harvested from microfluidic devices.
(7) Heterogeneous nucleants and cross-seeding are less effective than rMMS, but they can be used if seed crystals cannot be obtained.
A theoretical case and practical suggestions are also put forward for producing crystals with different space groups.
3433 dx.doi.org/10.1021/cg2001442 |Cryst. Growth Des. 2011, 11, 3432–3441
Crystal Growth & Design ARTICLE
Figure 1, that is, well below the metastable zone; therefore, the
crystals are likely to be thermodynamically unstable and to
dissolve. To put it another way, if a protein molecule happens
to fall off a microcrystal in the seed stock, it is unlikely that this
molecule or any other protein molecule will subsequently attach
itself to replace the loss. However, it is common experience that
(macro-) crystals that have been harvested into the reservoir
solution (containing no protein) may appear to be stable for
limited periods. This implies that, as long as the precipitant
concentration is relatively high, the rate of dissolution may be so
low that it is not normally noticed. A second objective of the
project was therefore to find out whether, in the conditions
typically used, instability of the seed stock was an important
consideration. This was of interest because some groups have
reported that the rMMSmethod does not work in their hands, and
we wondered if they were handling their seed stocks inappropri-
ately. Moreover, other groups have reported that seed stocks that
are left on the bench for a few hours may become inactive.
Another area of concern was the tendency for rMMS experi-
ments to generate salt crystals, which may waste investigators’
time and energy. This is a particular problem when the main
precipitant in the Hit Solution is a salt. For example, if a seed
stock containing ammonium sulfate is added to all conditions in a
typical screen, one would expect to find several wells containing
crystals of calcium sulfate (gypsum), because Ca2þ is common in
crystallization screens. Moreover, Mg, Ca, Zn, and Cd frequently
form phosphate crystals. We therefore investigated methods of
reducing the number of salt crystals in rMMS experiments.
Radaev and Sun8 showed that crystallization conditions for
protein complexes heavily favor (71% versus 27%) polyethylene
glycols rather than ammonium sulfate or other high-salt crystal-
lization conditions. Seed stocks obtained from high-salt condi-
tions may introduce harmful concentrations of salts that can
interfere with the crystallization of protein�protein, protein�pep-
tide, and protein�small molecule complexes and with the forma-
tion of heave atom derivatives. A method that replaces high-salt
precipitants with for example PEG would therefore be helpful.
An obvious disadvantage of the rMMS method is that it can
only be used when one has found at least one crystal, in order to
make the first seed stock. We had several ideas for tackling
this problem: first, we conjectured that seed crystals could be
harvested from microfluidic devices. Second, we noted the
approach of Habel & Hung,9 who collected precipitated protein
from the wells of unsuccessful screening experiments and added
it to random screens. It is likely that, even when crystals are not
visible, some of the precipitates obtained are nevertheless crystal-
line (the crystals may be smaller than the wavelength of light,
therefore undetectable with optical microscopes). Third, we
were interested in cross-seeding with crystals of unrelated
proteins (see Results and Discussion). Fourth, we noted the
investigations of Chayen & Saridakis10 of diverse nonprotein
materials for protein nucleation. A nanoporous biocompatible
glass (“bioglass”) was the most successful of these.11 This
material has cavities that are comparable in size to protein
molecules, and it has previously been shown to induce crystal-
lization of at least 14 proteins, several of which could not be
crystallized without bioglass.12 Sugahara et al.13 proposed the use
of synthetic zeolite to induce nucleation. By including these het-
erogeneous materials in our study, we planned to compare directly
their effectiveness with regular microseeding.
We emphasize that we were not seeking to prove here that
rMMS microseeding is an effective way to crystallize proteins.
This has already been clearly demonstrated by several publica-
tions4�6 and a multitude of PBD entries (e.g., 3CL0, 3CKZ,
3CL2, 3I0M from the National Institute for Medical Research,
UK). Instead, we investigated the relative effectiveness of different
methods of inducing nucleation, with enough data to give good
statistics, in order to improve nucleation techniques.
’EXPERIMENTAL SECTION
Choice of Test Proteins. Table 1 shows the proteins used in this
study. We selected proteins that could be obtained in large quantities
and which did not typically crystallize in a large number of conditions in
a screen. Other proteins, including lysozyme, concanavalin A, and
ferritin, were not used because reliable Receptive Conditions could
not be found (see below).
Reagents and Screens Used. The commercial crystallization
screens used were Molecular Dimensions’ Structure Screen 1 and
JCSGþ, Jena Bioscience’s JBScreen Membrane 3, and Qiagen’s PEG
Suite. In the experiments of Figures 4 and 8 (Supporting Information)
and Table 4, we looked at the effect on crystallization of suspending seed
crystals in the following solutions: Hit Sol.; 100% isopropanol (“IPA”);
100% polyethylene glycol, molecular weight 600 (“PEG 600”); 4 M
(NH4)2SO4; a 50�50 mixture of Hit Sol. and 4 M (NH4)2SO4; 5.8 M
NaCl; a 50�50mixture ofHit Sol. and 5.8MNaCl, and the stock solution
of the protein being tested (see Table 1 for concentrations and buffers).
Automation and Volumes Dispensed. We used an Oryx814
robot (Douglas Instruments) throughout the study.We dispensed vapor
diffusion sitting-drop experiments to 96-well two- or three-drop plates
(SwissCI). Drop volumes were 0.3 μL of protein plus 0.3 μL of screen
for all nonseeding experiments, while we used 0.3 μL of protein, 0.29 μL
of screen and 0.01 μL of seed stock for all seeding experiments except
where we used 0.1 μL of seed stock (with 0.3 μL of protein, 0.2 μL of
screen) as noted below. Optimization was carried out with the Oryx8
robot using “2-D Grid” designs (see Supporting Information).
Figure 1. A screening experiment can be thought of as a set of points
that land randomly on the phase diagram of a protein. (These “points”
are shown as arrows because a vapor diffusion setup is assumed.) This
schematic phase diagram7 has four main areas: an unsaturated zone
where drops remain clear and no crystals can grow (labeled “under-
saturated”); a zone where precipitation takes place (“precipitate”); a
zone where crystal nuclei form and grow into visible crystals (“labile”);
finally, there is a zone just below the labile zone, often called the
“metastable” zone. Here, crystals do not form spontaneously, but if you
take a crystal, for example from the labile zone, and put it into this zone, it
will grow. The line that divides the metastable zone from the under-
saturated zone is the solubility curve of the protein. In a normal
screening experiment, crystals will only appear in wells where the arrows
end up in or pass though the labile zone (thick arrows). If, however, you
add seed crystals to the screen, you will obtain a set of additional hits
where conditions end up in the metastable zone (dashed arrows). SHS
and SPS indicate the rough positions of seed stocks made up in Hit
Solution and protein stock, respectively (see text).
3434 dx.doi.org/10.1021/cg2001442 |Cryst. Growth Des. 2011, 11, 3432–3441
Crystal Growth & Design ARTICLE
Identification of Receptive Conditions. We identified Recep-
tive Conditions (see the definition belowTable 2) as follows: first, we set
up three plates (two drops per well) with regular screens (i.e., no additive
or seeds added) in order to eliminate conditions that crystallized at least
once in the absence of seeds. We then set up a plate (again two drops per
well) with the addition of crushed seed crystals suspended in IPA and
another with seed crystals suspended in 100% PEG 600. We identified
Receptive Conditions if (1) none of the six wells that were set up without
the addition of seed crystals produced crystals, but (2) three out of the
four wells with added seed crystals (suspended in IPA and PEG)
produced crystals. Hemoglobin proved to be difficult to crystallize,
and microseeding experiments were set up using the conventional
method of suspending seed crystals in the Hit Solution, with nine wells
per condition. We identified Receptive Conditions if none of the wells in
the regular screens contained crystals, but all nine of the seeded wells did.
Some proteins crystallized more quickly than others; for each protein, a
cutoff period for observation of 1�7 days was used that gave reliable
Receptive Conditions (see “Crystal Observation Period” of Table 1).
The Receptive Conditions are shown in Table 2.
Identification of Protein Crystals.We used a battery of tests to
distinguish protein crystals from salt crystals during the experiments
where Receptive Conditions were identified (see above). All crystals
grown were well formed with straight edges, with the exception of
hemoglobin, where crystals could be identified by color and clarity. We
photographed putative crystals of the colorless proteins in a darkroom
using UV light (wavelength 280 nm) with a microscope with normal
glass optics and a Panasonic DMC-FX12 compact camera, using a 30 s
“Starry Sky” setting. UV was provided by a UV Pen-280 (Douglas
Instruments) with a 2 mm thick UG11 filter (Schott) to cut out visible
light. Crystals that fluoresced with emission in the visible spectrum were
identified as protein.15 When preparing seed stock, we crushed all large
crystals with a glass probe, thereby subjecting them to the “crunch test”:
crystals that produced a click that could be heard and felt were
eliminated as salt crystals. The identification of protein crystals was
confirmed with brightly colored (Jena Biosciences) and fluorescent16
dyes, and the cross-linking agent glutaraldehyde (Fluka). Glutaralde-
hyde in high concentrations turns protein crystals brown or golden (see
below).
In the cases of crystals of the proteins concanavalin A, trypsin, and
thaumatin, we used an interesting novel method of making the distinc-
tion, which is a modification of the method of Pusey et al.17 We
covalently labeled 50 μL aliquots of the proteins with the fluorescent
dye DyLight 350 NHS Ester (from Thermo), following the manufac-
turer’s instructions except that we used higher protein concentrations
(30mg/mL for trypsin and concanavalin A, 36mg/mL for xylanase).We
added 20 nL samples of labeled protein to wells containing putative
protein crystals after the crystals had grown. We photographed crystals in
a darkroom by illuminating with theUVPen-280 or with an FL4BLBUV
lamp (Luxina), which has a peak wavelength of 370 nm. As shown in
Figure 2, crystals fluoresced brightly and were unambiguously identified
as protein rather than salt. (TheDyLight kits are very easy to use because
all resins, columns, etc. are provided. We chose the label that is excited at
350 nm because it is not necessary to use a filter since most cameras have
built-in UV filters.) The advantages of the method are (1) since it allows
protein to be seen directly, it does not give false positives or negatives
(except when the drop contains a lot of precipitate, see below). (2) It
cannot interfere with the crystallization process. (3) Labeled protein
need only be prepared if crystal identification by other methods fails;
(4) even needles and small crystals can be identified. The method does
not work well when the drop contains a lot of protein precipitate, which
may absorb the labeled protein before it can reach the crystals. Note
also that protein sometimes coats salt crystals in crystallization experi-
ments, giving a superficially similar appearance. Such cases can, however,
easily be distinguished by comparing UV images with visible light images
because the protein coating is outside the salt crystal.
Once Receptive Conditions were established, we identified protein
crystals using the UV Pen-280 only.
Preparation of Seed Stocks.We harvested all seed crystals from
sitting drop plates except for xylanase, where we harvested the initial batch
of seed crystals from microbatch-under-oil. All seed crystals (after the
initial batch) were obtained from the conditions shown in bold and
outlined in blue in Table 2 (i.e., crystals from microseeding experiments
were used to prepare seed stocks for subsequent rounds ofmicroseeding).
Seed stock suspensions were generated as follows: (1) a Seed Bead18
(Hampton Research) was placed in an Eppendorf tube on ice. (2) The
well containing the crystals to be harvested was opened by cutting the
tape, and 20 μL of reservoir solution from the well (or other solution)
was placed in the tube. (3) Using a glass probe (with a 0.25 mm bead
melted on the end) and observing with a microscope, the crystals were
thoroughly crushed in the well. (4) 2 μL of reservoir solution (or other
stabilizing solution) was transferred from the tube to the drop with seed
crystals, crystals were suspended by withdrawing and dispensing from
the pipet tip several times, and the mixture was transferred back to the
tube using a slightly higher volume setting on the pipet to ensure all
solution was transferred. (5) Step 4 was repeated to ensure that all
crystals were picked up. (6) The tube and Seed Bead were vortexed for 2
min, the tube was placed back on ice, the bead was removed, and the
stock was used immediately.
Since the Oryx8 crystallization system uses contact dispensing (the
tip touches the well during each dispensing operation), it was not
necessary to centrifuge the seed stock to remove the larger seed crystals,
which may cause blockages in noncontact systems. During experiments,
the seed stock is held in the Oryx robot in one channel of the dispensing
tip at ambient temperature. When not in use, all seed stocks were kept
frozen at �15 �C.
We also prepared a seed stock for cross-seeding with crystals of 15
proteins that were unrelated to any of the test proteins. These proteins
(PDB codes) comprised N1 (3CL2) andN2 neuraminidase, a mutant of
N1 (3CKZ), erythrocyte binding antigen, H3 hemagglutinin, Ser/Thr
protein kinase, polycomb EED protein (3IJC), a post synaptic density
protein (2RJI), the flavoprotein soxF, the cell envelope protein MtrF,
rabbit hemorrhagic disease virus, and other mutations of these proteins.
The seed stock was prepared as follows: (1) crystals of the first protein
were crushed in their drops; (2) 3.5 μL of 100% PEG was added to each
Table 1. Proteins Used in This Study
protein supplier product code
conc used
(mg/mL) protein buffer
“receptive” conditions
found
crystal observation
period (days)
glucose isomerase Hampton Research HR7-102 33 10 mM tris, 1 mM MgCl2 2 1
hemoglobin (bovine) Sigma Aldrich H2500 60 50 mM Na acetate 4 6
thaumatin Sigma Aldrich T7638 30 50 mM Na acetate 6 10
thermolysin Sigma Aldrich T7902 15 50 mM Na acetate, 14 mM NaOH 6 4
trypsin (porcine) Sigma Aldrich T7418 30 2% (w/v) benzamidine 2 3
xylanase Macro Crystal 36 43% glycerol, 0.2 M phosphate 5 7
3435 dx.doi.org/10.1021/cg2001442 |Cryst. Growth Des. 2011, 11, 3432–3441
Crystal Growth & Design ARTICLE
drop, then a