Analytical Biochemistry 316 (2003) 223–231
`
`ANALYTICAL
`BIOCHEMISTRY
`
`www.elsevier.com/locate/yabio
`
`Detection and prevention of protein aggregation before, during,
`and after purification
`
`Sarah E. Bondos* and Alicia Bicknell1
`
`Department of Biochemistry and Cell Biology, Rice University, Houston, TX 77251-1892, USA
`
`Received 4 November 2002
`
`Abstract
`
`The use of proteins for in vitro studies or as therapeutic agents is frequently hampered by protein aggregation during expression,
`purification, storage, or transfer into requisite assay buffers. A large number of potential protein stabilizers are available, but de-
`termining which are appropriate can take days or weeks. We developed a solubility assay to determine the best cosolvent for a given
`protein that requires very little protein and only a few hours to complete. This technique separates native protein from soluble and
`insoluble aggregates by filtration and detects both forms of protein by SDS–PAGE or Western blotting. Multiple buffers can be
`simultaneously screened to determine conditions that enhance protein solubility. The behavior of a single protein in mixtures and
`crude lysates can be analyzed with this technique, allowing testing prior to and throughout protein purification. Aggregated proteins
`can also be assayed for conditions that will stabilize native protein, which can then be used to improve subsequent purifications. This
`solubility assay was tested using both prokaryotic and eukaryotic proteins that range in size from 17 to 150 kDa and include
`monomeric and multimeric proteins. From the results presented, this technique can be applied to a variety of proteins.
`Ó 2003 Elsevier Science (USA). All rights reserved.
`
`Keywords: Aggregation; Solubility; Inclusion bodies; Precipitation; Protein; Purification
`
`Recombinant proteins are required in biological re-
`search to investigate enzyme activity, ligand binding,
`protein interactions, or other functions in vitro. Many
`proteins are also potential pharmaceutical agents [1–5].
`One significant impediment to the study and utilization
`of proteins is their extreme sensitivity to solution con-
`ditions. Nonoptimal conditions during protein expres-
`sion, purification, storage, or handling can alter protein
`structure such that the protein irreversibly aggregates,
`with concomitant loss of activity [4–7]. Proteins fre-
`quently aggregate at the high concentrations required
`for structural studies, and small soluble aggregates can
`be incorporated into protein crystals as defects [8–10].
`Protein folding studies are often complicated by aggre-
`gation of intermediate and denatured states [9–13].
`Further, point mutations or deletion mutations required
`
`* Corresponding author. Fax: 1-713-348-6149.
`E-mail address: bondos@bioc.rice.edu (S.E. Bondos).
`1 Present address: UCSD Biology Student Affairs 038, 9500
`Gilman. Dr., La Jolla, CA 92093-0348, USA.
`
`for relevant protein studies may destabilize the native
`state and expedite aggregation [14–16]. The kinetics of
`aggregation may be an order of magnitude faster than
`folding kinetics, causing a significant fraction of the
`protein to be inactivated [10]. Competition between
`aggregation and folding can have biological ramifica-
`tions: protein aggregation in vivo is implicated in a va-
`including ParkinsonÕs disease,
`riety of disorders,
`AlzheimerÕs disease, and spongiform encephalopathies
`[15–18]. In vitro examination of the proteins involved in
`these diseases will require strategies to control aggre-
`gation [15,17]. Thus, protein aggregation is a problem
`common to biological systems, experimental research,
`and industrial and medical applications.
`To circumvent these problems, a wide variety of
`buffer cosolvents that can facilitate proper protein
`folding and solubility have been identified. Cosolvents
`exert their effects by either destabilizing aggregates or
`enhancing native protein stability [6,12,19–27]. Exam-
`ples of useful additives are listed in Table 1. Aggregate
`formation can be deterred by including cosolvents that
`
`0003-2697/03/$ - see front matter Ó 2003 Elsevier Science (USA). All rights reserved.
`doi:10.1016/S0003-2697(03)00059-9
`
`Page 1
`
`KASHIV EXHIBIT 1071
`IPR2019-00797
`
`

`

`224
`
`S.E. Bondos, A. Bicknell / Analytical Biochemistry 316 (2003) 223–231
`
`Table 1
`Agents that may promote protein solubility
`
`Kosmotropes
`
`Weak kosmotropes
`
`Chaotropes
`
`Amino acids
`
`Sugars and polyhydric alcohols
`
`Detergents
`
`Additive
`
`MgSO4
`ðNH4Þ2SO4
`Na2SO4
`Cs2SO4
`
`NaCl
`KCl
`
`CaCl2
`MgCl2
`LiCl
`RbCl
`NaSCN
`NaI
`NaClO4
`NaBr
`Urea
`
`Glycine
`LL-arginine
`
`Sucrose
`Glucose
`Lactose
`Ethylene glycol
`Xylitol
`Mannitol
`Inositol
`Sorbitol
`Glycerol
`
`Tween 80
`Tween 20
`Nonidet P-40
`
`Recommended concentration range
`
`Reference
`
`0–0.4 M
`0–0.3 M
`0–0.2 M
`0–0.2 M
`
`0–1 M
`0–1 M
`
`0–0.2 M
`0–0.2 M
`0–0.8 M
`0–0.8 M
`0–0.2 M
`0–0.4 M
`0–0.4 M
`0–0.4 M
`0–1.5 M
`
`0.5–2%
`0–5 M
`
`0–1 M
`0–2 M
`0.1–0.5 M
`0–60% v/v
`0–30% w/v
`0–15% w/v
`0–10% w/v
`0–40% w/v
`5–40% v/v
`
`0–0.2% w/v
`0–120 lM
`0–1%
`
`[19]
`[19]
`[19]
`[19]
`
`[7,19]
`[19]
`
`[19]
`[19,30]
`[19]
`[19]
`[19]
`[19]
`[19]
`[19,30]
`[6,12]
`
`[6]
`[6]
`
`[4,11]
`[21]
`[21]
`[38]
`[38]
`[38]
`[38]
`[29,30,38]
`[39]
`
`[4]
`[5]
`[42]
`
`destabilize protein–protein interactions. For example,
`low concentrations of charged cosolvents can screen
`protein electrostatic interactions that may facilitate ag-
`gregation [2,19,26]. Another strategy utilizes chaotropic
`species to interact with the peptide group, replacing or
`preventing the intermolecular interactions that lead to
`aggregation [6,12,25].
`An alternate approach is to stabilize native intramo-
`lecular protein interactions, thus out-competing the in-
`termolecular interactions that lead to aggregation. To
`this end, kosmotropes generally stabilize the native state
`of proteins [25–28]. Because kosmotropes increase the
`surface tension of the solvent, they are excluded from
`the protein–solvent surface. Therefore, the entropic cost
`of disturbing the distribution of small molecules to form
`the solvent cavity increases. Generally, the native state
`of globular proteins creates the smallest solvent cavity;
`therefore, kosmotropes raise the cost of hydrating in-
`termediate or denatured states relative to the native
`state. Sugars and polyhydric alcohols, in particular, in-
`teract with the protein more weakly than water [26].
`Optimization of the number of strong water–cosolvent
`interactions forces the cosolvent to be excluded from the
`
`thus stabilizing the state with the
`protein surface,
`smallest surface area [21,26,27,29,30]. The addition of
`such cosolvents not only stabilizes many proteins but
`also deters ice formation, thus inhibiting the deleterious
`effects of freezing on protein structure [4,5,26,27]. Fi-
`nally, small amino acids are also preferentially excluded
`from the protein surface. However, charged amino acid
`salts may also interact with the protein at certain pH
`and salt concentrations [2,26].
`Other types of buffer additives may also facilitate
`protein solubility. Dithiothreitol and b-mercaptoethanol
`are reducing agents that prevent aggregation of some
`proteins by inhibiting the formation of nonnative di-
`sulfide bonds. Importantly, reduced glutathione is not as
`effective; the reduced form often contains a small per-
`centage of oxidized glutathione which, ironically, may
`be sufficient to oxidize the protein [24]. Compounds such
`as trifluoroethanol or trichloroacetic acid prevent ag-
`gregation by stabilizing a-helical structure [23]. Ethanol
`has been used to stabilize a folding intermediate by
`weakening hydrophobic interactions that facilitate ag-
`gregation [13]. Membrane proteins may require deter-
`gents or micelles to form their native structure [20,22].
`
`Page 2
`
`

`

`S.E. Bondos, A. Bicknell / Analytical Biochemistry 316 (2003) 223–231
`
`225
`
`The list of potential cosolvents that influence protein
`solubility may appear staggering to one who must de-
`termine a starting point. The following strategy may be
`useful to rapidly identify optimal solvent conditions for
`a given protein. First, cosolvents or additive concen-
`trations that interfere with subsequent biochemical as-
`says should be eliminated. Previous experience with the
`protein or a member of the protein family may suggest
`cosolvents that are likely to succeed. For example, many
`nucleic acid binding proteins often require higher salt
`concentrations, as demonstrated in this paper. For novel
`proteins, a good choice is to select one cosolvent from
`each category in Table 1, using a concentration in the
`middle of the suggested range. Chemicals and concen-
`tration ranges within a promising category can then be
`optimized in a subsequent assay.
`When choosing a cosolvent, determination of the
`optimal concentration is critical. High concentrations of
`chaotropes will denature proteins, while high concen-
`trations of kosmotropes will salt proteins out of solu-
`tion. Consequently, removing ions from the buffer may
`enhance protein stability [17,31]. A list of many addi-
`tives and appropriate concentration ranges can be found
`in Table 1. Because any additive has the potential to
`alter protein conformation or activity, the effects of
`specific conditions on protein structure and function
`should be investigated by varying cosolvent concentra-
`tion or comparing with a second stabilizing cosolvent
`[26]. A more extensive list of protein stabilizing reagents
`and descriptions of their mechanisms of action can be
`found in several excellent reviews [25–27].
`The large number of potentially stabilizing cosolvents
`and the dependence of cosolvent effects on concentra-
`tion complicates determination of optimal buffer con-
`ditions for a given protein. Often, such efforts rely on
`trial and error during protein purification. Alternately,
`sample turbidity can be measured to assay for protein
`aggregation. However, turbidity measurements require
`high concentrations of protein and cannot detect low
`concentrations of aggregates or small, soluble aggre-
`gates [3,32]. Further, turbidity requires purified samples,
`prohibiting its use for proteins that aggregate during
`expression or purification.
`Here, we describe a facile method to identify buffers
`that maintain soluble, native protein. This technique
`can distinguish precipitates and small, soluble aggre-
`gates from native protein. Multiple buffers or proteins
`can be screened in just a few hours. Importantly, this
`solubility assay can be applied to protein mixtures and
`crude lysates, allowing assessment of protein stability
`throughout a protein purification protocol. This solu-
`bility assay was demonstrated for a variety of
`applications, including screening buffers to inhibit ag-
`gregation in functional assays, screening mutant pro-
`teins for aggregation, screening for solubility of a single
`protein in heterogeneous mixtures, assaying aggregated
`
`proteins for stabilizing conditions, and screening in-
`clusion bodies in crude cell lysates for stabilizing con-
`ditions. Both prokaryotic and eukaryotic proteins were
`tested,
`including monomeric and multimeric proteins
`ranging in size from 17 to 150 kDa. Therefore, this
`solubility assay can be utilized for a wide variety of
`proteins.
`
`Materials and methods
`
`Protein expression and purification
`
`The proteins used to test the aggregation assay are
`Bacilllus anthracis ATXA, Eschericia coli LacI, Dro-
`sophila melanogaster UDK-c, Drosophila melanogaster
`Ultrabithorax (Ubx), and Brachydanio rerio LMO4.
`ATXA protein was a gift of Dr. Edward Nikonowicz
`from Rice University. The LacI mutants L148F, S151P,
`G60+3-11, and Q60P-11 were provided by Hongli Zhan
`and Dr. Kathleen Matthews from Rice University.
`UDC-c was given by Daniel J. Catanese and Dr.
`Kathleen Matthews from Rice University. LMO4 crude
`lysates, purified protein, and GN5049 primary antibody
`were gifts of David Ji and Dr. Mary Ellen Lane from
`Rice University.
`Ubx contaminated with proteolysis products and full-
`length Ubx were purified for use in the solubility assay.
`The Ubx expression construct pET-Ubx-3c, a gift from
`Dr. Phillip A. Beachy (The Johns Hopkins University
`School of Medicine), was transformed into the E. coli
`strain BL21(DE3)pLysS. Twelve flasks containing 1 liter
`of Luria broth plus 50 lg/ml carbenicillin cultures were
`each inoculated with 10 ml of overnight culture and
`grown at 37 °C to mid-log. The cultures were cooled to
`room temperature prior to Ubx induction with 1 mM
`isopropyl b-LL-thiogalactoside. Cells were harvested 2 h
`after induction and frozen.
`To purify Ubx, a cell pellet was lysed in 20 ml of
`50 mM Tris–HCl, 4 mM DTT,2 800 mM NaCl, 20 lg/ml
`DNase, and 1 mM phenylmethyl sulfonyl fluoride. Lysis
`supernatant was treated with polyethyleneimine and
`centrifuged. The pH of the supernatant was adjusted to
`6.8 and centrifuged to remove precipitates. The final
`supernatant was loaded onto a phosphocellulose col-
`umn, washed with Buffer Z (10% glycerol, 0.5 mM DTT,
`0.1 mM EDTA, 25 mM NaH2PO4, 100 mM NaCl, pH
`6.8), and eluted with a 0 to 1 M NaCl gradient in Buffer
`Z. Ubx mixed with N-terminal proteolytic products was
`assayed for aggregation at this point in the purification.
`Fractions containing Ubx were dialyzed against 4 liters
`of 50 mM Tris–HCl, 100 mM NaCl, 1 mM DTT, 10%
`
`2 Abbreviations used: DTT, dithiothrectol; EMSA, electrophoretic
`mobility shift analysis.
`
`Page 3
`
`

`

`226
`
`S.E. Bondos, A. Bicknell / Analytical Biochemistry 316 (2003) 223–231
`
`Table 2
`Proteins used in the aggregation assay and relevant parameters
`
`Protein
`
`Source
`
`Filter used
`
`ATXA
`LacI:
`L148F
`S151P
`G60+3-11
`Q60+3-11
`UDKc-his
`UbxIb
`LMO4
`
`Bacillus anthracis
`Escherichia coli
`
`100 kDa MW cut-off Microcon
`Ultrafree—MC filter unit
`
`Drosophila melanogaster
`Drosophila melanogaster
`Brachydanio rerio
`
`100 kDa MW cut-off Microcon
`100 kDa MW cut-off Microcon
`100 kDa MW cut-off Microcon
`
`Initial
`concentration
`
`SDS–PAGE
`detection
`
`0.16 mM
`
`0.67 mM
`0.60 mM
`1.10 mM
`0.84 mM
`3.6 lM
`13 lM
`5 lM
`
`Silver stain
`Silver stain
`
`Western blotting
`Western blotting
`Western blotting
`
`Molecular weight
`and assembly
`
`50 kDa, monomer
`
`150 kDa, tetramer
`150 kDa, tetramer
`65 kDa, dimer
`65 kDa, dimer
`47 kDa, monomer
`40 kDa, monomer
`17 kDa, monomer
`
`glycerol, pH 8.0 for 1 h. Ni-NTA resin (Qiagen) was
`equilibrated in dialysis buffer, added to the dialysate,
`and mixed on a Nutator at 4 °C for 1 h. The resulting
`slurry was washed with 10 ml of buffer containing
`50 mM NaH2PO4, 300 mM NaCl, 5% glucose at pH 8.0.
`plus 10 mM imidazole followed by 25 ml of the same
`buffer with 20 mM imidazole. Ubx was eluted in 10 ml of
`the buffer containing 100 mM imidazole. Protein was
`stored at 4 °C after the addition of DTT to 5 mM and
`assayed the following day.
`
`Detection of aggregates is especially sensitive because
`nearly half of the recovered aggregated protein retained
`by the filter is loaded on the gel. The final protein con-
`centration required is dependent on the detection
`method used. Coomasie staining detects 0.05–0.5 lg of
`protein per band, silver staining detects 1–5 ng of pro-
`tein per band, and Western blotting detects less than
`1 pg of protein per band ([23], www5.amershambio-
`sciences.com).
`
`Aggregation assay
`
`Results
`
`The general strategy was to simultaneously test up to
`14 solution conditions on small samples of purified or
`unpurified protein. Soluble protein was then separated
`from aggregates and precipitates by filtration. The mo-
`lecular weight cut-off of the filter was chosen such that
`soluble protein was allowed to pass through the filter,
`while aggregate forms were retained.
`In more detail, protein was diluted or dialyzed into a
`series of buffer conditions such that the final volume was
`100–120 ll. Initial protein concentrations ranged from 3
`to 1 mM (see Table 2). The diluted protein was incu-
`bated at room temperature in the test buffer for 1 h.
`Soluble protein was then separated from aggregated
`protein using a Microcon concentrator (Millipore), with
`a molecular weight cut-off of 100 kDa. The Ultrafree–
`MC 0.1-lm filter unit (Millipore) was used for experi-
`ments on LacI mutants to accommodate the larger
`proteins. The Microcon or filter unit was spun in a
`desktop centrifuge at 16,000g for 15 min. Aggregated
`protein retained on the membrane was resuspended in
`30 ll of dH2O, pipeting repeatedly across the membrane
`to ensure that as much protein as possible is removed.
`Samples of 30 ll soluble protein and 30 ll aggregated
`protein were each mixed with 10 ll of 4 sample buffer
`(250 mM Tris–HCl, 40% glycerol, 140 mM SDS, 0.6 M
`b-mercaptoethanol, pH 6.8) and heated to 85–90 °C for
`10 min prior to SDS–PAGE. Because soluble protein
`and aggregated protein were assayed separately, small
`changes in either population can be readily observed.
`
`The solubility assay was developed for use in situa-
`tions spanning the lifetime of a protein from cell lysis,
`purification, and exchange into assay buffers. For sim-
`plicity, the straightforward application of assaying sol-
`uble protein for conditions that diminish aggregation is
`presented first. The subsequent experiments, including
`analyzing solubilization of aggregated protein and an-
`alyzing protein mixtures, increase in complexity. The
`section concludes with assaying inclusion bodies in
`crude cell lysates for conditions that will allow protein
`solubilization. This last application is the most useful,
`allowing optimization of cosolvents prior to purifica-
`tion.
`
`Example of screening buffers to enhance solubility of
`purified protein during functional assays
`
`Buffer conditions required for column binding and
`protein elution are often incompatible with functional
`and structural analysis. Initial purifications of the B.
`anthracis protein ATXA resulted in precipitation of
`some product. Further, gel retardation assays of DNA
`binding by the soluble protein fraction exhibited density
`in the wells, indicative of ATXA aggregation. Buffer
`conditions were screened for stabilization of native
`ATXA and prevention of aggregation in DNA binding
`assays (Fig. 1). Buffer additives were limited to salt and
`glycerol, which are normal components of buffers in
`DNA binding assays, and cosolvent concentrations
`
`Page 4
`
`

`

`S.E. Bondos, A. Bicknell / Analytical Biochemistry 316 (2003) 223–231
`
`227
`
`Fig. 1. Solubility assay of ATXA protein to identify an appropriate
`DNA binding buffer. To dilute the protein into assay buffers, 10 ll of
`0.16 mM protein was added to 100 ll of the five test buffers. All buffers
`contain 20 mM Tris–HCl, pH 7.5. ATXA was detected by silver
`staining. The soluble, native protein fraction is indicated by Sol, and
`the aggregated protein fraction is indicated by Agg. While ATXA
`aggregates in buffer containing 10% glycerol or 100 mM KCl, no ag-
`gregation was observed in buffer containing both 10% glycerol and 100
`or 200 mM KCl. The variation in total protein observed for each buffer
`condition is repeatable and thus most likely reflects adhesion to the
`filter used in the assay.
`
`were optimized within an acceptable range for DNA
`binding assays. Severe aggregation was detected in
`buffer containing only 20 mM Tris, pH 8.0. Aggregation
`was also visible when 100 mM KCl or 10% glycerol was
`added. However, inclusion of 200 mM KCl prevented
`aggregation. Thus, the protein is sensitive to low salt
`conditions. Interestingly, while neither 100 mM KCl nor
`10% glycerol alone can prevent aggregation, a combi-
`nation of both additives maintains soluble protein. The
`efficacy of combinations of cosolvents has been ob-
`served in other proteins also [6]. This assay provided
`two useful solution conditions for performing func-
`tional assays, which is important because DNA binding
`is profoundly affected by alterations in salt and glycerol
`concentrations [33–35]. Therefore, the solubility assay
`can successfully identify buffers that stabilize soluble
`protein.
`
`Example of screening mutant proteins for aggregation
`
`Mutations can dramatically alter the structure, sta-
`bility, or aggregation properties of a protein. Even
`though mutant proteins may purify like wild-type pro-
`tein, mutations may alter protein solubility. Proteins
`with mutations in the full-length tetrameric lactose re-
`pressor, LacI, and the dimeric deletion mutant, )11
`LacI [36,37], were assayed for aggregated contaminants
`after storage at )80 °C (Fig. 2). Ultrafree–MC 0.1-lm
`filter units were used to ensure sufficiently large pore size
`to accommodate the 65-kDa dimers and the 150 kDa
`monomers. The 0.1-lm pore size is approximately 15
`times the size of the )11 LacI dimer. Each mutant
`protein was diluted 1:100 into 20 mM Tris, pH 7.5,
`100 mM KCl, and 10% glycerol. As a positive control,
`the L148F LacI mutant was also diluted into water, thus
`forcing aggregation and precipitation. The high protein
`
`Fig. 2. (A) Solubility assay of LacI tetrameric and dimeric mutants.
`Each mutant was diluted 1:100 into buffer containing 20 mM Tris–
`HCl, pH 7.5, 100 mM KCl, and 10% glycerol. Initial protein concen-
`trations ranged from 0.6 to 1.1 mM. The positive control contains the
`L148F mutant diluted into Millipore water. LacI proteins were visu-
`alized by silver staining. The soluble protein is indicated by Sol, the
`filter unit wash by W, and the aggregated protein by Agg. (B) Tur-
`bidity of each sample, as measured by the optical density at 600 nm.
`The turbidity results correlate with the density in the corresponding
`Agg lane.
`
`concentrations used in this assay raised concerns that
`residual native protein left on the wet filter might be
`erroneously interpreted as aggregate. To prevent this
`problem, the membranes were washed by filtering with
`an additional 100 ll of buffer or water after separation
`of native and aggregated protein. Any native protein left
`on the filter after the wash should be below the detection
`limits of silver staining.
`As expected, the majority of the protein in the water
`control was retained by the filter, indicating that the
`filter was capable of separating native and aggregated
`protein. While the different LacI and )11 LacI mutants
`were primarily native, aggregation was detected in the
`tetrameric L148F and S151P samples. Both )11 mutants
`had little or no aggregates. Most of the positive control,
`LacI in water, was retained by the filter, demonstrating
`that a 0.1-lm filter can retain large amounts of aggre-
`gated LacI. Concerns that small concentrations of LacI
`aggregates might not be retained on filters with the very
`large pore size prompted comparison of the turbidity at
`600 nm prior to sample filtration (Fig. 2B). The turbidity
`measurements correlate very well with the density of the
`band in the retentates, verifying that dimeric mutants
`can be assayed using the 0.1-lm filter. Thus, the solu-
`bility assay works well using filters with larger pore sizes
`and can be applied to multimeric proteins.
`
`Page 5
`
`

`

`228
`
`S.E. Bondos, A. Bicknell / Analytical Biochemistry 316 (2003) 223–231
`
`Screening for aggregation of a partially purified protein
`
`Because solubility is monitored by SDS–PAGE, ag-
`gregation of a single component can be detected in a
`mixture of proteins. An interesting example is provided
`by the D. melanogaster protein Ultrabithorax Ib (Ubx).
`Ubx is not produced as inclusion bodies and does not
`precipitate during purification. However, a portion of
`the protein originally remained in the wells during
`electrophoretic mobility shift analysis (EMSA) to mea-
`sure DNA binding, an indication of protein aggregation.
`A partially purified sample with contaminating N-ter-
`minally degraded Ubx was used to assay buffers that
`potentially stabilize full-length, Ubx (Fig. 3A). Full-
`length Ubx aggregates were detected in buffers con-
`taining low concentrations of salt. However, none of the
`degradation products precipitated under low-salt con-
`ditions; therefore, either the salt-sensitive region of Ubx
`must reside in the N-terminal region or the structure
`must be altered upon removal of this region. Buffers for
`the remainder of the purification were adjusted to in-
`clude 100 mM NaCl and 5–10% glycerol or glucose.
`Full-length purified Ubx was also assayed (Fig. 3B). The
`results for purified, full-length Ubx match the protein
`mixture. These results indicate that interactions with
`proteolysis products do not influence the aggregation
`behavior of full-length Ubx and confirm that the N-
`terminus of Ubx contains a domain sensitive to salt
`concentration. The use of Ubx purified under the new
`
`buffer conditions for EMSA experiments abrogated the
`signal caused by protein trapped in the wells, thus
`confirming the solubility assay results (data not shown).
`Thus, aggregation of a single protein in a protein mix-
`ture can be reliably assessed using the solubility assay.
`
`Example of assaying aggregated protein for stabilizing
`conditions
`
`Many proteins precipitate upon purification to ho-
`mogeneity. Therefore, it is desireable to be able to use
`aggregated protein to assay buffers that will stabilize the
`native, soluble form. The D. melanogaster protein
`UDKc, also named UbxBP1, DIP1-c, and KLETT-c,
`originally precipitated after purification on a Ni-NTA
`column (Qiagen) via a histidine tag. The microfiltration
`solubility assay was used on purified, precipitated pro-
`tein to determine buffer conditions that would maintain
`native UDKc during subsequent purifications (Fig. 4A).
`Precipitation may be irreversible; therefore, the presence
`of any soluble protein indicates a condition that may
`prevent protein aggregation. UDKc-his, a histidine-
`
`Fig. 3. Solubility assay using a protein mixture. Ubx protein bands on
`a 10% SDS–PAGE gel were detected by Western blotting, using
`FP3.38 as the primary antibody [41]. Ubx, 10 ll, was diluted with
`100 ll of test buffers. All buffers contain 20 mM Tris, pH 8.0. (A)
`Aggregation assay of a mixture containing full-length Ubx and its N-
`terminal degradation products. The top band is full-length Ubx, and
`the bottom bands are N-terminal proteolysis products. Ubx aggregates
`in buffer with no additives, buffer with 50 mM KCl, and buffer with
`10% glycerol. However, Ubx did not aggregate in buffers with higher
`salt concentrations or with EDTA. (B) Aggregation assay of purified
`full-length Ubx. The behavior of purified Ubx matches the behavior of
`the mixture of full-length in a mixture with proteolyzed Ubx. Lanes are
`labeled as in Fig. 1.
`
`Fig. 4. Solubility assay of aggregated UDKc-his, a his-tagged RNA
`binding protein. Aggregated UDK-c was purified using Ni-NTA resin
`(Qiagen) and following the protocols therein. Protein bands on a 10%
`SDS–PAGE gel were detected by western blot using mouse anti tetra-
`his (Qiagen) as the primary antibody. (A) UDKc-his at 3.6 lM was
`dialyzed into four test buffers. Each buffer contains 50 mM sodium
`phosphate, 150 mM NaCl, 10 mM imidazole, pH 8.0, in addition to the
`cosolvents indicated. Lanes are labeled as in Fig. 1. Addition of both
`urea and 5% glucose increased the amount of soluble UDKc-his. (B)
`Aggregation assay of 100 ll of 5.1 lM UDKc-his similarly purified
`with all buffers containing 5% glucose. All of the protein is in the flow-
`through, demonstrating that inclusion of glucose in the purification
`buffers prevents aggregation of UDKc-his.
`
`Page 6
`
`

`

`S.E. Bondos, A. Bicknell / Analytical Biochemistry 316 (2003) 223–231
`
`229
`
`tagged construct that produces a 47-kDa protein, was
`dialyzed into different buffers to maintain the protein
`concentration at detectable levels. Buffer containing
`urea, and to a lesser extent 5% glucose, decreased ag-
`gregation of the protein. Because urea might affect
`subsequent biophysical characterization, 5% glucose
`was selected as a purification additive for subsequent
`purifcations. A second solubility assay using UDKc-his
`protein purified in buffers containing 5% glucose re-
`vealed that the protein did not aggregate under the new
`buffer conditions (Fig. 4B). In a single afternoon, ap-
`plication of the solubility assay to precipitated protein
`successfully predicted conditions that would stabilize the
`native state of the protein, allowing protein purification
`and long-term storage.
`
`Screening for solubilization of inclusion bodies
`
`Many proteins, including the B. rerio (zebrafish) zinc
`finger protein LMO4, are expressed as inclusion bodies
`in E. coli. Purification then has the added onus of re-
`storing the protein to its native, soluble state. Crude
`lysate containing 8 M urea was assayed for conditions
`that enhance LMO4 solubility during purification (Fig.
`5A). Due to the large number and high concentration of
`other proteins in the mixture, Western blotting was used
`to visualize LMO4 after SDS–PAGE. Because the pro-
`tein is produced as inclusion bodies, determination of
`buffer conditions that will maintain soluble, native
`LMO4 was expected to be difficult. Therefore 12 co-
`
`solvents that address a broad range of potential triggers
`for aggregation, such as exposure of hydrophobic
`groups, charge–charge interactions, and cysteine oxida-
`tion, were selected for examination using the solubility
`assay. All test buffers contained 50 mM Tris–HCl, pH
`8.0, 1 mM ZnSO4, and at least 100 mM KCl. The density
`of both aggregated and soluble LMO4 for several buffer
`conditions is very light. This light density is repeatable
`and therefore not an artifact of gel loading or aggregate
`resuspension. The most likely explanation is that LMO4
`adopts a conformation capable of irreversibly adhering
`to the membrane or the plastic casing in the filtration
`device under these buffer conditions. The only buffer to
`yield a substantial percentage of LMO4 in the soluble
`fraction contained 10 mM DTT, indicating that protein
`oxidation likely triggered LMO4 aggregation.
`To determine whether the proteins, lipids, and DNA
`in the crude lysate had influenced LMO4 solubility,
`precipitated LMO4 purified without DTT was assayed
`in the same buffers (Fig. 5B). While some LMO4 was
`observed in the flow-through of buffers containing urea,
`arginine, and trichloroacetic acid, all of the protein di-
`luted into buffer containing 10 mM DTT was in the fil-
`trate. Therefore, use of the solubility assay to analyze
`the solution behavior of proteins in mixtures and crude
`lysate can predict the behavior of purified proteins.
`Assaying crude lysates is an effective strategy for de-
`signing purification protocols for novel proteins or
`proteins with a history of aggregation or precipitation.
`Upon growth of a purification-scale batch of E. coli, a
`
`Fig. 5. Solubility assay using crude lysate. Ten microliters of both the crude lysate and the purified protein were diluted into 100 ll of test buffer. All
`buffers contain 50 mM Tris–HCl, pH 8.0, 1 mM ZnSO4, and at least 100 mM KCl. LMO4 protein was detected by Western blot using the rabbit anti-
`LMO4 antibody GN5049. (A) Analysis of LMO4 in crude lysate containing 8 M urea. Only the buffer containing 10 mM DTT contains a substantial
`percentage of LMO4 in the flow-through. (B) Analysis of LMO4 purified in buffer containing 8 M urea. All of the LMO4 protein is in the flow-
`through for the 10 mM DTT sample. Lanes are labeled as in Fig. 1.
`
`Page 7
`
`

`

`230
`
`S.E. Bondos, A. Bicknell / Analytical Biochemistry 316 (2003) 223–231
`
`small aliquot of protein-expressing cells should be fro-
`zen separately. This aliquot can be lysed and the protein
`solubility tested, allowing adjustment of the purification
`buffers to suit the needs of the protein prior to the initial
`protein purification. This strategy saves a substantial
`amount of time and supplies compared to an iterative
`trial and error approach to protein purification.
`
`Discussion
`
`Effective screening of many possible additives at
`various concentrations requires a rapid assay for protein
`solubility. However, efforts to identify optimal buffer
`conditions often rely on repurification or functional
`assays, a time- and protein-consuming trial and error
`approach. Alternately, the turbidity, or light scattered
`by precipitates at a nonabsorbing wavelength, can be
`used to rapidly detect insoluble protein aggregates.
`Turbidity measurements require large volumes of at
`least 10 lM final protein concentration [3]. The yields of
`many protein preparations are too low to allow
`screening by buffer dilution. Because turbidity is de-
`pendent on the molecular weight and the radius of gy-
`ration, the size or shape of the aggregates influence the
`outcome [32]. In addition, impure protein cannot be
`assayed. Finally, small soluble aggregates or low per-
`centages of aggregates can impact protein function or
`create point defects in crystal growth, but are not de-
`tectable with turbidity measurements [10].
`Here, we describe a sensitive method to simulta-
`neously screen a large number of conditions for soluble
`or insoluble aggregates in a few hours. Because the assay
`separates the species by size, small soluble aggregates
`can be separated from native protein and detected.
`Aggregation of a single protein in a mixture can be de-
`tected, allowing analysis of partially purified protein or
`unpurified lysates. The analysis of both partially purified
`Ubx- and LMO4-containing crude lysate matched the
`results from the corresponding purified protein. Thus,
`this solubil