Sie sind auf Seite 1von 36

Protein Crystallization:

Theory and Practice


Excerpts from:

Structure and Dynamics of E. coli Adenylate Kinase


A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF
THE REQUIREMENTS FOR THE DEGREE DOCTOR OF
PHILOSOPHY
Michael B. Berry
W. M. Keck Center for Computational Biology
Rice University
Houston, TX
9/17/95
Copyright 1995

------------------------------------------------------------------------

Disclaimer
This is my version of a basic-to-advanced guide to the
crystallization of proteins. Actually, it's just excerpts from

Chapters 2 and 3 of my dissertation compiled together under a


single heading with a few new sections. The methodology covered
herein varies from basic to voodoo black magic (hair of dog), but
is by no means complete. This guide is intended mostly to provide
background in crystallization theory and to help in the
crystallization of unknown or otherwise resistant proteins.
Hopefully this will be helpful. Please pay attention to the warnings
in the text when dealing with some of the more toxic substances
used in the crystallization of proteins, especially cadmium,
thiocyanide, and azide. Use these at you own risk!
------------------------------------------------------------------------

Crystallization Theory
Crystallization is one of several means (including nonspecific
aggregation/precipitation) by which a metastable supersaturated
solution can reach a stable lower energy state by reduction of
solute concentration (Weber, 1991). The general processes by
which substances crystallize are similar for molecules of both
microscopic (salts and small organics) and macroscopic (proteins,
DNA, RNA) dimensions. There are three stages of crystallization
common to all systems: nucleation, growth, and cessation of
growth.
Nucleation is the process by which molecules or noncrystalline
aggregates (dimers, trimers, etc.) which are free in solution come
together in such a way as to produce a thermodynamically stable
aggregate with a repeating lattice. Crystallization is known to
lower the free energy of proteins by ~3-6 kcal/mole relative to the
solution state (Drenth and Haas, 1992). The formation of
crystalline aggregates from supersaturated solutions does not
however necessitate the formation of macroscopic crystals.
Instead, the aggregate must first exceed a specific size (the critical
size) defined by the competition of the ratio of the surface area of
the aggregate to its volume (Feher and Kam, 1985; Boistelle and

Astier, 1988). Once the critical size is exceeded, the aggregate


becomes a supercritical nucleus capable of further growth. If the
nucleus decreases in size so that it is smaller than the critical size,
spontaneous dissolution will occur. The process of formation of
nonspecific aggregates and noncrystalline precipitation from a
supersaturated solution does not involve the competition between
surface area and volume (n-mers add to the aggregate chain in a
head to tail fashion forming a linear arrangement), and thus
generally occurs on a much faster time scale than crystallization.
The degree to which nucleation occurs is determined by the degree
of supersaturation of the solutes in the solution. The extent of
supersaturation is in turn related to the overall solubility of the
potentially crystallizing molecule. Higher solubility allows for a
greater number of diffusional collisions. Thus, higher degrees of
supersaturation produce more stable aggregates (due to higher
probability of collision of diffusing molecules) and therefore
increase the likelihood of the formation of stable nuclei. In the case
of a finite number of solute molecules, this condition generally
results in the production of a large number of small crystals. At
lower solute concentrations the formation of individual stable
nuclei increases in rarity, thus favoring the formation of single
crystals.
Crystal growth generally starts at solute concentrations sufficient
for nucleation to occur, and continues at concentrations beneath the
nucleation threshold. The rate of growth is determined by a
combination of the nature of the growing crystal surface and the
diffusional rate. Addition of molecules to a rough surface requires
less energy than addition to a smooth surface, where surface
nucleation is required for addition.
According to Periodic Bond Chain theory, (Boistelle and Astier,
1988), three different types of growth faces exist: flat faces,
stepped faces, and kinked faces. Flat faces require two dimensional

nucleation (the formation of growing sheets of molecules) in order


to induce growth, and thus grow the slowest. Stepped faces grow
as columns of molecules, which requires only one dimensional
nucleation, and thus have intermediate growth rates. Stepped faces
typically occur as a result of a crystallographic screw axis causing
spiral growth patters to occur at the surface of the crystal. Finally,
kinked faces are growth sites which do not require nucleation to
promote further growth, and therefore grow faster than the other
two face types. Thus, the type of the growing crystal face (flat,
stepped, or kinked) strongly influences the rate at which crystal
growth occurs.
The growth of crystals from nuclei is also strongly influenced by
diffusional and convection effects. As with nucleation, increased
solubility results in increased growth rates. Again, this is a function
of the rate at which protein molecules reach the growing surface of
the protein crystal. Feher and Kam, through the use of ultraviolet
microscopy, have been able to demonstrate the regions
surrounding growing crystals to be lowered in protein
concentration relative to the surrounding solution (Feher and Kam,
1985). The rate of diffusion of proteins in and out of these halos
around the growing crystal provides a growth limiting factor.
The formation of halos due to lowered solute concentration around
growing crystals also has the effect of producing density gradients
in these areas. These in turn (under the effects of gravity) result in
the formation of convection currents which may dominate the rate
of simple diffusion and adversely effect crystal growth
(Rosenberger, 1986). As the formation of concentration gradients
around growing crystals is directly proportional to the rate at which
molecules add to the surface (the crystal growth rate), slower
growth results in decreased convection currents. This may be
accomplished by growing crystals in porous gel medias (Robert
and Lefaucheux, 1988). Crystal growth in zero (effective) gravity

may also be used to remove convective and sedimentary effects


(Littke and John, 1986; DeLucas et al., 1986).
Cessation of growth of crystals can occur for a multitude of
reasons. The most obvious is the decrease in concentration of the
crystallizing solute tothe point where the solid and solution phases
reach exchange equilibrium. In this case, the addition of more
solute can result in continued crystal growth. However, some
crystals reach a certain size beyond which growth does not precede
irrespective of solute concentration. This may be a result either of
cumulative lattice strain effects or poisoning of the growth surface.
Lattice strain effects in tetragonal lysozyme crystals have been
demonstrated by Feher and Kam and coworkers (Feher and Kam,
1978). Halved crystals of hen egg white lysozyme, when placed in
fresh crystallization solutions, grew to the exact same size as the
original crystal. This suggests that the long range propagation of
strain in the lattice effectively prevents addition of molecules to the
surface once a certain critical volume is reached. Crystals affected
by lattice strain are therefore inexorably size limited.
Poisoning of growing faces occurs when foreign or damaged
molecules are incorporated into the growing crystal face resulting
in successive defects which interrupt the crystal lattice. An
example of this might be the incorporation of a proteolytically
knicked protein onto the face of an otherwise perfect protein
crystal. If the knicked molecule is unable to form the same lattice
contacts with newly added molecules as would the perfect protein,
then its incorporation will cause local defects in the growing
lattice. Since the growth of crystal lattices typically selects for
perfect over damaged or incorrect molecules, the concentration of
these defective molecules relative to perfect molecules tends to
increase as growth proceeds. Thus, as crystals grow larger the
likelihood of incorporation of defective molecules into the lattice
increases (also the increase in surface area contributes). Sato and

coworkers have used laser scattering tomography to visualize


lattice defects in large crystals of orthogonal hen egg white
lysozyme (Sato et al., 1992). Their results demonstrated the
occurrence of both point and inclusion defects not only at the
surface, but within the bulk of the crystal itself.
To increase the size of surface poisoned crystals, the poisoned
surfaces must be removed by partially melting the crystal. This
provides an unpoisoned surface which can be used for further
growth in the presence of more solute. A method of applying this
technique to protein crystals has been designed by Thaller and
coworkers (Thaller et al., 1981).
------------------------------------------------------------------------

Crystallization Methodology
Solution properties influencing crystallization Crystallization of
macromolecules is a paradigm on par with protein folding as to the
number of possible substates which exist in excess of the global
minimum where perfect crystals form (Feher and Kam, 1985). The
reasons behind the high degree variability in the crystallization of
macromolecules are manifold and include: 1) the high degree of
mobility at the surface and as a whole for large molecules, 2) the
electrostatic nature of macromolecules (i.e. their composition as a
random assortment of mobile point charges embedded in a flexible
low dielectric medium), 3) the relatively high chemical and
physical instability of macromolecules (unfolding, hydration
requirements, temperature sensitivity), and 4) molecule specific
factors (such as prosthetic groups and ligands). These factors
contribute greatly to the difficulty and crudity in current attempts
to predict crystallization behavior by de novo calculations even for
relatively small and stable proteins like lysozyme (Durbin and
Feher, 1991). The above factors are also reflected in the general
trend towards greater crystal disorder (and lower resolution of

diffraction data) as molecular size increases due to the increased


flexibility and disorder in these crystals, although this is by no
means universally applicable (see Abrahams et al., 1994)
By and far the most important factor in crystallization is the purity
of the sample to be crystallized. A 1 ppm contaminant in a typical
10-20 mg/ml protein solution amounts to ~109 molecules (Carter,
1988). Although crystal growth, by nature, tends to exclude
impurities (such as in the case of rabbit muscle aldolase which
crystallizes out of homogenized muscle preps at 52% ammonium
sulfate; Taylor et al., 1948), the presence of high concentrations of
impurities in small volumes, as are present in vapor diffusion and
dialysis experiments, will undoubtedly lead to contamination of the
crystal lattice, and ultimately poorer crystals. Lin and coworkers
recommend the use of FPLC (or HPLC) for general purification of
proteins and to assure homogeneity on both macroscopic and
microscopic levels (Lin et al., 1992). Lin notes several cases where
the use of FPLC techniques has improved the reproducibility of
crystallization as well as the maximum resolution to which protein
crystals diffract.
As noted earlier, molecules crystallize from metastable
supersaturated solutions as a means of lowering the overall
solution free energy. Chemical precipitants are by and far the most
widely used method of achieving supersaturation of
macromolecules in order to induce crystallization. In general, the
main influence of these compounds is on the solvent (e.g. bulk
water) rather than on the solute (the protein), with the notable
exception of dye precipitants. For crystallization of proteins, the
major classes of precipitants may be divided into six categories:
salts, high molecular weight straight chain polymers (e.g. PEG),
MPD, organic solvents, sulfonic dyes, and deionized water
(McPherson, 1990; Arakawa and Timasheff, 1985). Although the
following discussions of the individual traits of these six categories
use proteins as examples, most of these precipitants are applicable

to other macromolecules including RNA, DNA, and


polysaccharides.
Salts Salts are by far the most common precipitant type used to
crystallize macromolecules. Historically they have been the most
effective precipitants tried, although, until recently, there were few
other options (McPherson, 1991). Unfortunately, salts generally
have the drawback of increasing the mean electron density of the
crystallization solution which decreases the signal to noise ratio for
crystallographic data. Salts also have a tendency to interact
strongly with heavy atom compounds, making crystal
derivitization for M.I.R. phasing difficult.
The efficacy of a particular salt as a precipitant is proportional to
the square of the valences of the cations and anions which make up
that salt. Typically, the anion is the more important species than
the cation, as in general, innocuous monovalent cations are used to
avoid forming strong cation/protein complexes which tend to occur
with polyvalent cations (especially transition metal ions). The
ability of a salt to precipitate proteins can be generally described
by the Hofmeister series:
PO43->HPO42-= SO42->citrate>CH3CO2->Cl- >Br-> NO3->ClO4>SCNand
NH4+>K+>Na+>Li+ .
Anions and cations which are weakest in the Hofmeister series
typically have the effect of salting in rather than salting out
proteins. In practice the PO43- anion does not exist in solution
within the range of pH typically used for crystallization trials, and
is present instead as HPO42- and H2PO4-. Also, the NH4+ cation

generally loses H+ above pH 8.0 and boils off as NH3, making


NH4+ salts difficult to use at high pH, as well as making the pH
highly unstable. An incomplete list of effective salt precipitants (in
descending order of efficacy), their maximum concentrations, and
generally effective ranges are shown in Table 1.1. (It should be
noted, however, that there is a general trend for larger molecules
[up to viruses; see Sehnke et al., 1988] to crystallize at increasingly
lower precipitant concentration as the molecular weight of the
species to be crystallized increases. Thus the values given in this
table are probably applicable for 15-50 kDa Mw proteins.).
------------------------------------------------------------------------

Table 1.1
Effective precipitant salts

Precipitant
Effective Concentration

Max Concentration

-----------------------------------------------------------------------(NH4+/Na+/Li+)2 or Mg2+
generally 50-80%
SO42saturation
NH4+/Na+/K+ PO43generally 50-80%
saturation
NH4+/K+/Na+/Li+ citrate
1.2-1.8 M
NH4+/K+/Na+/Li+ acetate
generally 50-80%

4.0 / 1.5 / 2.1 / 2.5 M

3.0 / 4.0 / 4.0 M

all ~1.8 M
~3.0 M

saturation
NH4+/K+/Na+/Li+ Clgenerally 50-80%
saturation
NH4+NO3generally ~4.0 M
KSCN
~200 mM (?)

5.2 / 9.8 / 4.2 / 5.4 M

~8.0 M
?

-----------------------------------------------------------------------Two general categories of salts exist, those which mainly interact


with water (non-chaotropic salts), and those which mainly interact
with the protein (chaotropic salts). Non-chaotropic salts are
preferentially hydrated with respect to protein solutes. The effect
of this is to increase in surface tension of the solvent surrounding
the macromolecule, thus dehydrating the protein's surface and
creating excluded volume effects, which force solute molecules to
form close interactions with each other. Solutions of these salts
charge shield proteins from one another by increasing the solvent
dielectric, which in turn further facilitates close protein-protein
interactions. Binding of non-chaotropic salts by the protein does
not generally play a role in crystallization, although in some cases
salt ions (particularly sulfates and phosphates) are seen bound by
macromolecules in crystal structures, and occasionally act as
lattice contacts between molecules in the crystal. Typically, these
salts increase the stability of macromolecules in solution (Scopes,
1994).
Chaotropic salts are generally not used for crystallization due to
their tendency to salt in macromolecules and to induce unfolding
through interactions with secondary structural elements of proteins.
However, Ries-Krautt and Ducruix have discovered that, at

reasonable concentrations (~200 mM) and low pH (~4.0-5.0),


potassium thiocyanate (a chaotropic salt) acts as precipitants by
interacting with and neutralizing positively charged residues on the
surface of proteins (lysine, arginine, histidine, and the amino
terminus; Ris-Kautt and Ducruix, 1991). It was further noted that
the addition of small amounts of KSCN (~10 mM) to nonchaotropic salt precipitants (phosphate) increased the rate of
crystallization substantially (WARNING: SCN- ENDUCES
PSYCHOSIS, don't eat it!). Other chaotropic salts (such as KI,
urea, and guanididium-HCl) may be equally effective as
precipitants or additives at concentration low enough to prevent
denaturation.
Polymers The use of high molecular weight linear polymers as
precipitating agents was pioneered by Polson and coworkers who
tried a variety of polymers including polyethylene glycol, dextran,
polyvinyl alcohol, and polyvinyl pyrrolidone (Polson et al., 1964).
Of these, polyethylene glycol (PEG) was found to be the most
effective both by way of precipitating ability and cost
effectiveness. PEG's are produced in a variety of molecular
weights, ranging from 200 (~3 monomers) to in excess of 1 million
(~15000 monomers), and as mono- and di- methyl ethers. Like
salts, PEG's compete with protein solutes for water and exert
excluded volume effects (which vary according to the length of the
polymer). However, unlike salts, PEG's decrease the effective
dielectric of the solution, which increases the effective distance
over which protein electrostatic effects occur. Solutions of
polyethylene glycols have mean electron densities roughly
equivalent to water and do not generally interact in a deleterious
manner with heavy atom compounds, thus making them
particularly well suited for macromolecular crystallization. PEG's
with molecular weights less than 1000 are typically liquids and are
generally used at concentrations above 40% v/v. PEG's with
molecular weights above 1000 are generally solids and are used in

the 5-50% w/v concentration range. All PEG solutions should be


made with the inclusion of ~0.l% Na azide (which, incidentally, is
also highly toxic to humans) to prevent bacterial growth. Also,
buffering of high concentration (40%) PEG solution with Na
citrate at concentrations above 100 mM tends to cause the
formation of phase transitions and color changes in the PEG
solution, which suggests some form of reaction (probably cross
linking by the citrate), and thus should be avoided.
MPD MPD is a small polyalcohol (2-methyl-2,4 pentane diol)
which has properties midway between those of low molecular
weight PEG's and organic solvents. MPD functions as a precipitant
by a combination of activities, including competition for water,
hydrophobic exclusion of protein solutes, lowering the solution
dielectric, and detergent-like effects. It is generally used in
concentrations in excess or 40% v/v with water/buffer, and tends to
cause phase transitions in the form of coacervate droplets which
are enriched in protein concentration (synonymous with those in
Ray and Bracker, 1986). As with PEG, the use of MPD as a
precipitant produces a low electron density solution which does not
generally interact with heavy atom compounds.
Organic solvents Historically, organic solvents have typically been
used as precipitants for protein crystallization as much by chance
as by design (McPherson, 1988). Crystallization by exposure to
organic solvents is occasionally seen during protein purification,
typically in the presence of common solvents such as ethanol,
methanol, acetone, isopropanol, DMSO, or tert-butanol
(McPherson, 1990). Due to their hydrophobic nature, organic
solvents cause phase transitions similar to those formed in the
presence of MPD and lower the bulk dielectric of the solvent.
Organic solvents also tend to cause protein denaturation unless
they are used at temperatures at or below 0 degrees C.

Sulfonic dyes Sulfonic dyes are typically large planar polycycles


with attached sulfonate groups. Initial studies by Lovrien and
coworkers suggest that these molecules may specifically interact
with the surfaces of proteins by way of the dyes sulfonate groups
(Conroy and Lovrien, 1992; Lovrien et al., 1993). The planar
polycycle of protein/dye complex greatly alters the solubility of the
protein triggering precipitation and occasionally crystallization of
the protein/dye complex. Lovrien suggests that a ratio of ~1-5 dye
molecules per protein molecule is sufficient for
precipitation/crystallization. NOTE, many sulfonic dyes are
extreme carcinogens and/or teratogens due to their tendency to
intercalate intoDNA.
Deionized water Dialysis of protein solutions against deionized
water as a method of crystallization takes advantage of the
requirement of proteins to be surrounded by a cloud of positively
and negatively ions in order to be soluble (the Donnan effect). If
this cloud of anions and cations is removed by dialysis, the protein
will attempt to surround itself with whatever charged species are
present, which, in this case, will be other protein molecules. The
formation of three dimensional lattice structures (e.g.
crystallization) is energetically favored as these arrangements
generally entail a more complete ionic atmosphere around each
molecule in the lattice.
Although all of the chemical precipitants discussed above have
worked for one system or another, oftentimes their individual
effects are too severe for crystallization to occur and instead the
protein solute leaves the supersaturated state by way of
precipitation rather than crystallization or may form poor one or
two dimensional crystals. In such cases, combinations of

precipitants (i.e. salts and PEG's, MPD and salts, salts and
organics, etc.) may produce larger, better diffracting crystals, and
occasionally new space groups. Some of the best combinations
(based on results of crystallization trials) are outlined in Table 1.2.
------------------------------------------------------------------------

Table 1.2
Proven combinations of precipitants

Major Precipitant
[Additive]

Additive

[Major Precipitant]

-----------------------------------------------------------------------(NH4)2SO4
6%-0.5%

PEG 400-2000,

2.0-4.0 M

MPD, ethanol,
methanol
Na citrate
6%-0.5%

PEG 400-2000,

PEG 1000-20000
0.2-0.6 M

(NH4)2SO4, NaCl,

1.4-1.8 M

MPD, ethanol,
methanol
40-50%

or Na formate

------------------------------------------------------------------------

A trend which can be seen in all of these solutions is the


combination of either a high dielectric species (major precipitant)
with a species which lowers the dielectric (minor precipitant, salts
with PEG/MPD/organics), or a low dielectric species (major
precipitant) with species which raise the dielectric (minor
precipitant, PEG/MPD/organics with salts). Both types of
combinations may effectively maximize size exclusion and water
competition effects while preventing intermolecular charge
repulsion by increasing the solvent dielectric, thus improving both
the likelihood of crystallization and the quality of the crystals
which do appear. Additives may also act to salt in or salt out
macromolecules which are insoluble or especially soluble in the
major precipitant used. Biphasic solutions (e.g., such as those
created by mixing high concentrations of salts with high
concentrations of high molecular weight PEG's) may also be
effective in triggering nucleation and crystallization (Kuciel et al.,
1992; Ray and Bracker, 1986; Garavito et al., 1986)
Addition of metal ions (in particular, cadmium (Cd2+), cobalt
(Co2+), and manganese (Mn2+)) to crystallization solutions can be
used as a form of "intermolecular glue" to either trigger nucleation
or improve crystallization (McPherson, 1991). Cadmium ions tend
to form strongly covalent complexes between two adjacent
carboxylic acid groups in a square planar type arrangement with all
four carboxylic acid oxygens being involved in bonding. This type
of bond can act as a lattice contact in a crystal or as a nucleating
factor for protein crystal growth. Manganese has similar (though
weaker) interactions, while cobalt ions preferentially interact with
amines (lysines and histidines). Typically, 1 to 5 mM
concentrations of CdSO4, 1 to 50 mM MnSO4, or 1 to 50 mM
CoSO4 are used either alone or in conjunction with another
precipitant. Caution should be exercised when working with
transition metal ions due to their general toxicity. Cadmium is
especially dangerous as, with beryllium and thallium, it is one of
the three most toxic elements known.

Two other chemical factors which can be used to initiate or


enhance crystallization are viscosity altering compounds and antitwinning/solubilizing compounds (detergents and ethers). As noted
in the section on crystallization theory, the rate at which crystals
grow often affects the overall quality of the crystals due to the
inclusion of defective or misaligned molecules. By inclusion of
viscosity altering compounds (notably glycerin) in the
crystallization mother liquor the rate of crystal growth may be
retarded. In attempts to crystallize the 2 ADP complex of E. coli
adenylate kinase this did result in larger crystals, although they
were as poorly ordered as smaller crystals of the same
morphology. Generally, 1% glycerin is sufficient to greatly slow
the rate of crystallization and/or nucleation.
The use of detergents as additives in crystallization mother liquors
has been utilized mostly in attempts to crystallize membrane
proteins (Garavito et al., 1986; Gros et al., 1988). The hydrophobic
tail of the detergent molecule binds to the hydrophobic areas of the
protein which are usually embedded in the membrane, thus
solubilizing these areas through the exposed hydrophilic head
groups of the detergent molecules. At lower concentrations,
detergents, as well as simple ethers like dioxane, can be used in
conjunction with other precipitants either to increase the solubility
of poorly soluble molecules or to improve crystallization habit, and
particularly to reduce or eliminate twinning (Bergfors, 1993;
McPherson et al., 1986). In one case, McPherson and coworkers
reported an entirely new crystal form of a protein due simply to the
addition of 0.1-1% BOG (beta-octyl glucoside, a mild detergent) to
the crystallization conditions. Crystal size also appeared to be
increased, while nucleation decreased. The use of detergents is not
entirely harmless in all cases, as at too high a concentration they
quickly denature myoglobin, probably either by solubilizing the
oily heme pocket or by solubilizing the heme itself.

The most important factor in crystallizing proteins other than the


chemical composition of the mother liquor is its pH. For some
proteins (e.g. the closed conformation of E. coli adenylate kinase),
crystallization occurs over a very broad range of pH with little by
way of variation in crystal morphology. However, it is far more
typical for crystallization to occur over a fairly narrow range (< 1
pH unit). Crystal morphology, including various twinned and
polynucleated growth forms, often is directly related to pH.
Typically, there is a gradual improvement in crystal morphology as
the proper pH is approached, and a fall off of crystal quality on
either side of the optimal condition. In some cases, the pI of the
protein is the pH at which the best crystals grow, so this should, if
possible, be known in advance and tried as a condition.
Temperature is another factor which is of considerable import in
crystallization of protein. Crystallization of macromolecules has
been accomplished in a range of roughly 60 degrees C to < 0
degrees C, although the vast majority of molecules are crystallized
either at 4 degrees C or 22 degrees C (room temperature). Low
temperature tends to act as a preservative for sensitive proteins as
well as an inhibitor of bacterial growth. Solubility of proteins in
salt solutions tends to increase at low temperatures (4 degrees C),
while in PEG and MPD solutions, protein solubility generally
decreases with decreasing temperature. By increasing or
decreasing either precipitant or protein concentration,
crystallization should, at least in theory, be possible at either room
temperature or 4 degrees C, although the kinetics of crystallization
can be expected to vary in accord with temperature. Heating,
melting, and cooling of crystals or aggregates may also be tried in
order to enlarge crystals, although this is generally unsuccessful.
Overall, the mantra according to McPherson and others is to try
both 4 degrees C and 22 degrees C, protein permitting
(McPherson, 1992; Carter and Carter, 1979; Bergfors, 1993).

Chemical/biochemical modification of proteins is another factor


which may be used to change crystallization conditions.
Electrostatic surface characteristics play a large role in dictating
whether a protein crystallizes or not. Thus, modification of surface
charges by either chemical (derivitization) or biochemical
(mutagenesis) means can provide crystals where none were known
before (Rayment et al., 1993) or provide crystals in new space
groups, possibly with better resolution (McElroy et al., 1992).
Rayment reductively methylated lysines of chicken muscle myosin
subfragment-1 to improve crystallization and alter the space group
in which the protein crystallized.
Biochemical modification in the form of site directed mutagenesis
of surface residues has been utilized by McElroy and coworkers to
improve the crystallization characteristics of human thymidylate
synthase (McElroy et al., 1992). The protein initially crystallized in
such a way as to make it impossible to interpret the active site in
electron density maps due to disorder. Therefore, McElroy and
coworkers created single point mutations at non-conserved surface
residues at twelve different sites (one per mutant protein) which
were designed either to neutralize charges (mutation to
asparagine), reverse charges (arginine and lysine to glutamate or
aspartate, and vice-versa), or add charges (cysteine, proline,
leucine, and glutamine to aspartate, glutamate, or lysine). The
crystallization behavior for each point mutation was completely
unpredictable and multiple new crystal forms resulted. Whereas
the original crystal form was trigonal (P3121), single point mutants
crystallized in monoclinic (C2), orthorhombic (C222), and
tetragonal (P41212) space groups. The tetragonal space group was
confirmed as having a well defined binding pocket. What is
interesting in these results is that an example of each type of
mutation (neutralization, reversal, or addition of charge) produced
an alternate space group to the original trigonal form. As this
technique is far more controllable (although considerably more
expensive) than chemical modification, it suggests a possible

avenue for producing or improving crystals for proteins for which


recombinant expression systems exist.
------------------------------------------------------------------------

Physical methods of crystallization


Four methods are commonly employed to affect supersaturation in
the crystallization of macromolecules: vapor diffusion, free
interface diffusion, batch, and dialysis. Although each of these
techniques achieves supersaturation of the particular
macromolecule to be crystallized, the means by which
supersaturation is achieved in each case varies greatly. A
discussion of each of these techniques is presented below.
Vapor diffusion The vapor diffusion technique utilizes evaporation
and diffusion of water between solutions of different concentration
as a means of approaching and achieving supersaturation of
macromolecules. Typically, the solution containing the
macromolecule is mixed 1:1 with a solution containing the
precipitant at the final concentration which is to be achieved after
vapor equilibration. The drop containing the 1:1 mixture of protein
and precipitant (both of which have been diluted to 1/2 the original
concentration by mixing with the other) is then suspended and
sealed over the well solution, which contains the precipitant at the
target concentration, as either a hanging or sitting drop. [Glass
capillaries containing protein and precipitant at concentrations
below that required for crystallization can also be vapor
equilibrated against a well solution in a sealed test tube. DeMattei
and coworkers have shown capillary based vapor equilibration to
occur at rates up 102 times slower than drop based methods,
resulting in improved crystals (DeMattei et al., 1992).] The
difference in precipitant concentration between the drop and the
well solution is the driving force which causes water to evaporate
from the drop until the concentration of the precipitant in the drop

equals that of the well solution. Since the volume of the well
solution is much greater than that of the drop (1-3 ml as compared
1-20 ul) its dilution by the water vapor leaving the drop is
negligible.
Fowlis and coworkers have demonstrated that the rate of vapor
equilibration in normal (1 g) gravity is dependent strictly on the
rate of vapor diffusion of water in the space separating the drop
and the well (Fowlis et al., 1988). Due to convection effects
(caused by the increased concentration of the precipitant at the
edge of the drop as water evaporates), the rate of diffusion of a
water molecule in the suspended drop (in solution) is actually
greater than that of the vaporized water molecule. In microgravity
experiments (zero effective g), the rate of equilibration is based
solely on the rate of diffusion of water in the drop until crystal
nucleation, at which point in time Marangoni convection (due to
concentration gradients or surface tension around the crystal)
increases the rate of equilibration (Provost and Robert, 1991).
Provost and Robert report that the use of simple agarose gels can
offset convection currents under normal gravity producing
improved results with hanging drops.
Vapor diffusion is the optimal technique to use either when
screening a large number of conditions (by varying the
composition of each well solution) or when dearth of protein
prevents the use of other methods. Furthermore, this method can be
used to increase or decrease the concentration of protein in the
equilibrated state relative to its initial concentration. This is done
by varying the volume of protein mixed with the well solution
when the drop is initially setup. Since the drop equilibrates so that
the precipitant concentration matches that of the well solution, the
final volume of the drop will always equal that of the initial
amount of well solution mixed with the protein. Thus, if the
protein solution is mixed in a 2:1 ratio with the well solution, the

concentration of the protein at vapor equilibrium will be doubled


relative to its initial value.
One of the drawbacks to vapor equilibration is that it tends to form
smaller crystals than other methods. This may be due to small drop
volumes limiting the quantity of crystallizable solute present or
creating a higher level of impurities relative to other techniques
which utilize larger volumes. As discussed earlier in section 2.2, as
crystals grow, the concentration of defective molecules increases
relative to perfect molecules (which are being selected for by the
crystal). When this factor is combined with the higher probability
of impurities diffusing to the face of the crystal (due to the smaller
volumes), the likelihood of inclusion of defects into the growing
crystal increases. Thus, the production of X-ray quality crystals
may be better suited to the use of batch, free interface diffusion, or
dialysis techniques which utilize larger solution volumes at
equilibrium.
Free interface diffusion Layering of a low density solution onto
one of higher density, usually in the form of concentrated protein
onto concentrated salt, can be used as a means of growing large
crystals. Nucleation and crystal growth generally occurs at the
interface between the two layers, at which both the concentration
of salt and the concentration of protein are at their highest values.
The two solutions slowly intermix over time, and should be made
up so that at equilibrium, (at which point in time both solutions are
diluted to some fraction of their initial values), the concentration of
the precipitant is still high enough to promote crystal growth. Since
the solute to be crystallized must be concentrated, this method
tends to consume fairly large amounts of protein. This method was
used by Steve Althoff to grow large crystals of the E. coli
adenylate kinase/Ap5A complex (Althoff, 1987; Althoff et al.,
1988).

Batch In the batch method, concentrated protein is mixed with


concentrated precipitant to produce a final concentration which is
supersaturated in terms of the solute macromolecule and therefore
leads to crystallization. This can be done with up to ml amounts of
solution and typically results in larger crystals due to the larger
volumes of solute present and the lower chance of impurities
diffusing to the face of the crystal. Needless to say, this technique
is by far the most expensive in terms of consumption of the solute
macromolecule, and thus should not generally be used to screen
initial conditions for crystallization.
Dialysis Dialysis techniques utilize diffusion and equilibration of
small precipitant molecules through a semipermeable membrane as
a means of slowly approaching the concentration at which the
macromolecule solute crystallizes. Initially, the solute is contained
within the dialysis membrane which is than equilibrated against a
precipitant solution. Equilibration against the precipitant in the
surrounding solvent slowly achieves supersaturation for the solute
within the dialysis membrane, eventually resulting in
crystallization. Dialysis tubing can be used by itself, in the case of
large amounts of protein being available, or used to cover the
opening of a dialysis button, allowing diffusion of the surrounding
solvent in to the solute through the dialysis membrane. Dialysis
buttons themselves come in a variety of sizes from 7-200 ul. The
advantage of dialysis over other methods is in the ease with which
the precipitating solution can be varied, simply by moving the
entire dialysis button or sack from one condition to another.
Protein can thus be continuously recycled until the correct
conditions for crystallization are found (Carter et al., 1988). One
drawback of this method is that it does not work at all with
concentrated PEG solutions, as they tend to draw all the water out
of the button or sack faster than PEG dialyzes across the
membrane, thus resulting in precipitated protein.

------------------------------------------------------------------------

Strategies for screening crystallization conditions


Crystallization of uncharacterized or even well characterized
macromolecules is not nearly as straightforward as it might seem.
There are a vast number of possible conditions which must be
analyzed for their ability to produce crystals.
To exemplify this, consider an initial screen where only five
different precipitants are to be used. (This is a conservative number
to start with. Through combination of the major categories with
additives, there are probably hundreds of viable precipitants, each
with their own unique ability to influence or trigger
crystallization.) Each of these five precipitants are set up (with the
macromolecule) at four concentrations, and each of these groups is
set up at a different pH, from 4-9 at one pH unit intervals. At this
point 120 conditions are to be tried. If two different temperatures
(say, 4 degrees and 20 degrees C) and two macromolecule
concentrations are tried for all these conditions, the number of
initial trials approaches 500. Needless to say, the more variables
which are tested, the greater the expense in terms of labor and
macromolecule, all of which is for naught if the conditions are
incorrect for crystallization to occur.
To add to this, the kinetics of nucleation and crystal growth vary in
a seemingly random fashion. Thus, a condition which is viable for
crystal growth but unstable for nucleation may take days to years
to produce crystals, whereas another condition might produce
crystals in a matter of minutes. Worse yet, microscopic
crystallization might occur but go unnoticed due to its similarity in
appearance to precipitated solute. (The appearance of crystals in
initial screens is only rarely obvious. Most of the time
crystallization is completely inobvious.)

Various strategies have been set forth to initially screen conditions


for crystallization. These vary from somewhat rational approaches
(screening at the pI) to highly regimented approaches (successive
grid screening) to analytical approaches (incomplete factorials,
solubility assays, perturbation) to randomized approaches (sparse
matrices). Each of these strategies has their own particular use
based on whether the protein is unknown, well known, or
somewhat known (mutated). A short explanation of the various
approaches with references is provided below.
Screening at the pI The pI of a macromolecule (the pH at which it
has no net charge) is its point of lowest solubility. Therefore, it is
rational to assume this to be a point where crystallization may
occur. Screening at the pI may be done either by dialysis against
low concentrations of buffer (<20 mM) at the appropriate pH (see
section 2.3.2 on deionized water), or by the use of conventional
precipitants. In the case of adenylate kinases, however, each of the
four liganded states has a different pI, and usually only the pI of
the unliganded state is known, requiring the computational
determination of the pI for the unknown states.
Grid screening Grid screens are typically based on two
dimensional matrices with the two different factors to be tested as
the two axes. Screening takes place as an iterative process, starting
with a course grid over a wide range and ending with a very fine
grid over a narrow range. This type of procedure lends itself well
to robotic automation (Cox and Weber, 1988). Typically, the first
screen conducted in this method is precipitant concentration vs.
pH, as these are generally the two most important factors in the
crystallization of macromolecules. The result of the initial screen is
then graded on the lack or presence, as well as the type, of
precipitant, which can be flocculent (fluffy clouds), granular
(amorphousballs), dendritic (string), or crystalline (from micro

needles and needle balls to perfect crystals). A range of conditions


is then prepared which includes and brackets the best initial
condition, and this procedure is repeated until the two parameter
grid converges on a best condition. This condition then can be used
as an initial starting place to test additional factors (temperature,
additives). This method works well if crystals appear early in the
screening process, but otherwise becomes expensive in terms of
protein due to the need to exhaustively sample a wide range of
conditions.
Incomplete factorials Carter and Carter pioneered this approach as
a "rational exploration of cause and effect relationships governing
the crystallization of proteins" (Carter and Carter, 1979). The
incomplete factorial approach is to take an initial set of ~20
conditions, randomly assign combinations of these factors as
individual experiments, grade the success of the results of each
experiment based on an arbitrary objective scale, and statistically
evaluate the effects of each of the twenty factors on the
crystallization trials (Carter et al., 1988). Thus, the effects of
factors such as pH, temperature, precipitating agent, and cations
can be determined precisely with minimal subjective interpretation.
These experiments are carried out in dialysis buttons so that
protein can easily be recycled (by dialysis against buffer to remove
precipitant), and so that each experiment can be driven to
completion by changing the precipitant concentration until either
precipitation or crystallization occur. In the two sets of trials
reported by Carter and coworkers, final conditions yielding high
quality crystals, sometimes in multiple space groups, were
determined within 35 trials.
A similar approach is espoused by Stura and coworkers which
utilizes a solubility "footprint" to rationally determine the direction
of further trials (Stura et al., 1991). An initial series of six
precipitants at four different concentrations and three pH's yields

an initial direction for further incomplete factorial searches based


on the degree of solubility in each condition. This method can be
used to determine conditions at which the macromolecule
crystallizes even if the initial footprint does not yield crystalline
material.
Perturbation One or two dimensional crystals which cannot be
improved upon by further grid iteration have appeared several
times in the course of my attempts to crystallize mutant sperm
whale myoglobins and recombinant human hemoglobins (Whitaker
et al., 1995). In these cases,
------------------------------------------------------------------------

Table 1.3
Additives for perturbation screening
(In addition to the major precipitant, e.g. ammonium sulfate)
Increased
Other
Dielectric

Decreased

Detergents

Dielectric

-----------------------------------------------------------------------200 mM Na/K/LiCl
5-50 mM spermine

1-5% MeOH, EtOH,

0.25-1% BOG

isopropanol, or
tert-butanol
200 mM Na formate
1% 18-6 crown
ether

1% MPD

0.25-1% BNG

200 mM Na2/K2HPO4
5-50 mM CdSO4,
CoSO4, or NiSO4
20-50 mM urea,
1-5 mM DTT,
trichloroacetate,
cysteine or 1-20
guanidium HCl, or
mM AgNO3
KSCN

1-4% PEG 400,

0.15% 1,2,3

600, or 1000

heptanetriol

1-4% PEG MME 550,

0.1-1% dioxane

750, or 2000

-----------------------------------------------------------------------perturbation assays have been particularly effective in producing


usable three dimensional crystals. The perturbation approach
combines the condition which yields the one or two dimensional
crystals combined individually with a series of additives designed
to test the effects of altering the structure of bulk solvent and also
the solvent dielectric. Categorical list of additives are shown in
Table 1.3.
Sparse matrices Jancarik and Kim first described the use of a
sparse set of conditions for initial crystallization trials based on
conditions which were known to have crystallized proteins [the
Crystal Screen sold by Hampton is made up of solutions suggested
by Jancarik and Kim] (Jancarik and Kim, 1991; see also:
McPherson, 1992). The idea behind sparse matrix trials is to
provide a broad enough sampling of parameter space by random
(or near random) combination of conditions to yield initial crystals
which then may be improved upon. This is also a very cheap
method (if it works) in terms of consumption of the

macromolecule, and therefore is preferable under conditions where


the quantity of macromolecule is the limiting factor.
In attempts to crystallize recombinant human hemoglobins, I used
the program sparse which I wrote (sparse.f, in FORTRAN 77)
which reads in a series of pH's, precipitants, and additives and
outputs a series of randomized combinations. Several of the
conditions created by the program resulted in improved crystals.
The program utilizes an initial numerical seed based on the date
and time for random number generation, and therefore each time
the program is run it outputs a completely new series of conditions.
It is to be stressed that the best way to maximize the rate of initial
sparse matrix screening is to keep an inventory (e.g. solution
number, main precipitant, buffer, pH, additive, concentrations) of
all solutions which have been made and use these in initial trials.
Solutions should thus be made with 0.1% Na azide (WARNING:
TOXIC, don't eat it!) to prevent bacterial growth. It should be
noted that Na azide will bind to the iron of hemin.
------------------------------------------------------------------------

A cheap, fast, and effective initial screen


tray 1
----------------------------------------------------------------------------------------------------------------------------------------------PEG 8000
Ammonium Sulfate
pH 5.0,20%
pH 7.0,20%
7.0,2.0 M pH 8.8,2.0 M
pH 5.0,35%
pH 7.0,35%
7.0,2.5 M pH 8.8,2.5 M

pH 8.6,20%

pH 5.0,2.0 M

pH

pH 8.6,35%

pH 5.0,2.5 M

pH

MPD
Na Citrate
Na/K Phosphate
pH 5.8,30%
pH 7.6,30%
pH 5.8,1.3 M pH 7.5,1.3 M
pH 6.0,2.0 M pH 7.4,2.0 M
pH 5.8,50%
pH 7.6,50%
pH 5.8,1.5 M pH 7.5,1.5 M
pH 6.0,2.5 M pH 7.4,2.5 M

-----------------------------------------------------------------------------------------------------------------------------------------------

tray 2
----------------------------------------------------------------------------------------------------------------------------------------------PEG 2000 MME/0.2 M A.S.
for wells 31-48
pH 5.5,25% pH 7.0,25%
pH 8.5,25%
32
33
pH 5.5,40% pH 7.0,40%
pH 8.5,40%
35
36
41
47

37
43

38
42
44
48

Random
31
34

39

40

45

46

----------------------------------------------------------------------------------------------------------------------------------------------A cheap, fast, and effective initial screen The above screen is an
integration of the footprint screen of Stura with a grid approach,
plus some sparse conditions thrown in as well. It has been effective

numerous times in crystallizing unknowns. It requires two Linbro


trays and the use of hanging or sitting drops, although tray 2 can be
set up weeks after tray 1, to conserve protein. Wells 31-48 of tray 2
should be chosen randomly from a list of solutions or generated
using sparse methods, and should cover the range of precipitants,
additives, and pH's (emphasis 5.0-9.0).
------------------------------------------------------------------------

Seeding
Seeding is a method of insuring or triggering (in the case of
heterogeneous nucleation) nucleation and crystal growth (Stura
and Wilson, 1990, Thaller et al., 1981; McPherson, 1988).
Commonly a protein crystal or crystalline aggregate is placed in a
storage solution and then crushed to a fine powder (producing a
large number of microscopic crystals) and mixed to make a seed
solution. The seed solution is then added to the supersaturated
crystallization trial (e.g. added to the equilibrated hanging drop or
batch tube) thus ensuring the presence of nuclei for crystal growth.
If no crystalline material is available for a particular protein, seed
solutions from related proteins (e.g. the same protein, but from a
different species) may sometimes trigger crystallization of the
target protein.
Microanalytical seeding This technique is designed to analyze the
results of crystallization trials in which some or all of the
conditions failed to yield crystals. It is best done with hanging or
sitting drops but may be done with other methods as well. In short,
high concentrations of seeds are added to all trials and the results
are judged based on the morphology of the resultant crystals. In

general, nucleation will occur at the site of seed introduction and


will occur from a large number of nuclei, resulting in a multitude
of crystals. Seeds may be introduced either by adding 1-2 ul of the
seed solution directly to the equilibrated protein solution (and thus
diluting it) or by dipping a hair (from my dog) in the seed solution
and then streaking this across the surface of the drop (streak
seeding; see Stura and Wilson, 1990). I prefer the latter method
because it is faster and does not dilute the supersaturated solution
to any degree.
Microproduction seeding This is identical in all regards to
microanalytical seeding except that the seed solution is diluted
(with the crystal storage solution, not with water) to the point
where the addition of 1-2 ul or streak seeding results in the
addition of only a few nuclei, allowing for the growth of large
single crystals.
Macroseeding This is a technique whereby small crystals may be
enlarged to a size amenable for data collection (Thaller et al.,
1981). In short, small crystals are removed from the solution in
which they were grown, partially melted in a lower concentration
precipitant wash solution, and then introduced into equilibrated
protein/precipitant solutions in order to continue growth. There are
two tricks to this process. The first is to insure that the crystal is
melted enough in the wash solution to remove any potential nuclei
at its surface, without completely dissolving the crystal. The
second is to chose a precipitant and protein concentration which
allows for continued protein growth but not nucleation. The entire
process can be repeated multiple time to continue to increase the
size of the crystal.
------------------------------------------------------------------------

References
Althoff, S. M. The crystallization and preliminary X-ray analysis
of Escherichia coli adenylate kinase. Bachelors Thesis. University
of Illinois at Urbana. 1987.
Althoff, S., Zambrowicz, B., Liang, P., Glaser, M., and Phillips, G.
N. Jr. Crystallization and preliminary X-ray analysis of Escherichia
coli adenylate kinase. J. Mol. Biol. 199:665-666, 1988.
Arakawa, T. and Timasheff, S. N. Theory of protein solubility.
Methods in Enz. 114:49-77, 1985.
Bergfors, T. The crystallization lab manual. 1993.
Berry, M. B., Meador, B., Bilderback, T., Liang, P., Glaser, M.,
and Phillips, G. N., Jr. The closed conformation of a highly
flexible protein: the structure of E. coli adenylate kinase with
bound AMP and AMPPNP. Proteins: Structure, Function and
Genetics. 19:183-198, 1994.
Boistelle, R. and Astier, J. P. Crystallization mechanisms in
solution. J. of Cryst. Growth. 90:14-30, 1988.
Carter, C. W. Jr. and Carter, C. W. Protein crystallization using
incomplete factorial experiments. J. Biol. Chem. 254:1221912223, 1979.
Carter, C. W. Jr., Baldwin, E. T., and Frick, L. Statistical design of
experiments for protein crystal growth and the use of a
precrystallization assay. J. of Cryst. Growth. 90:60-73, 1988.
Conroy, M. J. and Lovrien, R. E. Matrix coprecipitating and
cocrystallizing ligands (MCC ligands) for bioseparations. J. of
Cryst. Growth. 122:213-222, 1992.

Cox, M. J. and Weber, P. C. An investigation of protein


crystallization parameters using successive automated grid
searches (SAGS). J. Cryst. Growth. 318-324, 1988.
DeLucas, L. J., Suddath, F. L., Snyder, R., Naumann, R., Broom,
M. B., Pusey, M., Yost, V., Herren, B., Carter, D., Nelson, B.,
Meehan, E. J., McPherson, A., and Bugg, C. E. Preliminary
investigations of protein crystal growth using the space shuttle. J.
Cryst. Growth. 76:681-693, 1986.
DeMattei, R. C., Feigelson, R. S., and Weber, P. C. Factors
affecting the morphology of isocitrate lyase crystals. J. Cryst.
Growth. 122:152-160, 1992.
Drenth, J. and Haas, C. Protein crystals and their stability. J. Cryst.
Growth. 122:107-109, 1992.
Feher, G. and Kam, Z. Nucleation and growth of protein crystals:
general principles and assays. Meth. Enz. 114:77-111, 1985.
Fowlis, W. W., DeLucas, L. J., Twigg, P. J., Howard, S. B.,
Meehan, E. J. Jr., and Baird, J. K. Experimental and theoretical
analysis of the rate of solvent equilibration in the hanging drop
method of protein crystal growth. J. Cryst. Growth. 90:117-129,
1988.
Garavito, R. M., Markovic-Housley, Z., and Jenkins, J. A. The
growth and characterization of membrane protein crystals. J. Cryst.
Growth. 76:701-709, 1986.
Gros, P., Groendijk, H., Drenth, J., and Hol, W. G. J. Experiments
in membrane protein crystallization. J . Cryst. Growth. 90:193-200,
1988.

Jancarik, J. and Kim, S-H.J. Sparse matrix sampling: a screening


method for crystallization of proteins. Appl. Cryst. 24:409-411,
1991.
Kuciel, R., Jakob, L., Lebioda, L., and Ostrowski, W. S.
Crystallization of human prostatic acid phosphatase using biphasic
systems. J. Cryst. Growth. 122:199-203, 1992.
Lin, S-X., Sailofsky, B., Lapointe, J., Zhou, M. Preparative fast
purification procedure of various proteins for crystallization. J.
Cryst. Growth. 122:242-245, 1992.
Littke, W. and John, C. Protein single crystal growth under
microgravity. J. Cryst. Growth. 76:663-672, 1986.
Lovrien, R. E., Conroy, M. J., and Richardson, T. I. Molecular
basis for protein separation. Protein-solvent Interactions:521-553.
McElroy, H. E., Sisson, G. W., Schoettlin, W. E., Aust, R. M., and
Villafranca, J. E. Studies on engineering crystallizability by
mutation of surface residues of human thymidylate synthase. J.
Cryst. Growth. 122:265-272, 1992.
McPherson, A., Koszelak, S., Axelrod, H., Day, J., Robinson, L.,
McGrath, M., Williams, R., and Cascio, D. The effects of neutral
detergents on the crystallization of soluble proteins. J. Cryst.
Growth. 76:547-553, 1986.
McPherson, A. and Schlichta, P. The use of heterogeneous and
epitaxial nucleants to promote the growth of protein crystals. J.
Cryst. Growth. 90:47-50, 1988.
McPherson, A. Current approaches to macromolecular
crystallization. Eur. J. Biochem. 189:1-23, 1990.

McPherson, A. A brief history of protein crystal growth. J. Cryst.


Growth. 110:1-10, 1991.
McPherson, A. Two approaches to the rapid screening of
crystallization conditions. J. Cryst. Growth. 122:161-167, 1992.
Polson, A., Potgieter, G. M., Largier, J. F., Mears, G. E. F.,
Joubert, F. J. The fractionation of protein mixtures by linear
polymers of high molecular weight. Biochim. Biophys. Acta.
82:463-475, 1964.
Provost, K. and Robert, M-C. Application of gel growth to hanging
drop technique. J. Cryst. Growth. 110:258-264, 1991.
Ray, W. J. Jr. and Bracker, C. E. Polyethylene glycol: catalytic
effect on the crystallization of phosphoglucomutase at high salt
concentration. J. Cryst. Growth. 76:562-576, 1986.
Ries-Kautt, M. M. and Ducruix, A. F. Crystallization of basic
proteins by ion pairing. J . Cryst. Growth. 110:20-25, 1991.
Robert, M. C. and Lefaucheux, F. Crystal growth in gels: principle
and application. J. Cryst. Growth. 90:358-367, 1988.
Sato, K., Fukuba, Y., Mitsuda, T., Hirai, K., and Moriya, K.
Observation of lattice defects in orthorhombic hen-egg white
lysozyme crystals with laser scattering tomography. J. Cryst.
Growth. 122:87-94, 1992.
Scopes, R. K. Protein purification. Springer-Verlag. 1994.
Sehnke, P. C., Harrington, M., Hosur, M. V., Li, Y., Usha, R.,
Tucker, R. C., Bomu, W., Stauffacher, C. V., and Johnson, J. E.
Crystallization of viruses and virus proteins. J. Cryst. Growth.
90:222-230, 1988.

Stura, E. A., Nemerow, G. R., and Wilson, I. A. Strategies in


protein crystallization. J. Cryst. Growth. 110:1-12, 1991.
Thaller, C., Weaver, L. H., Eichele, G., Wilson, E., Karlsson, R.,
and Jansonius, J. N. Repeated seeding technique for growing large
single crystals of proteins. J. Mol. Biol. 147:465-469, 1981.
Weber, P. C. Physical principles of protein crystallization. Adv.
Prot. Chem. 41:1-36, 1991.
Whitaker, T. L., Berry, M. B., Ho, E. L., Hargrove, M. S., Phillips,
G. N. Jr., and Olson, J. S. The D-helix in myoglobin and the
[[beta]] subunit of hemoglobin is required for the retention of
heme. Biochem. 34:8221-8226, 1995.

Das könnte Ihnen auch gefallen