A systematic approach to the large-scale production of protein crystals

Crystallization has recently emerged as a suitable process for the manufacture of biocatalysts in the form of cross-linked enzyme crystals (CLECs) or for the recovery of proteins from fermentation broths. In both instances it is essential to define conditions which control crystal size and habit, and that yield a reliable recovery of the active protein. Experiments to define the crystallization conditions usually depend on a factorial design (either incomplete or sparse matrix) or reverse screening techniques. In this work, we describe a simple procedure that allows the effect of three factors, for example protein concentration, precipitant concentration and pH, to be varied simultaneously and smoothly over a wide range. The results are mapped onto a simple triangular diagram where a 'window of crystallization' is immediately apparent, and that conveniently describes variations either in the crystal features, such as their yield, size, and habit, or in the recovery of biological activity. The approach is illustrated with two enzymes, yeast alcohol dehydrogenase (ADH I) and Candida rugosa lipase. For ADH the formation of two crystal habits (rod and hexagonal) could be controlled as a function of pH (6.5-10) and temperature (4-25 degrees C). At pH 7, in 10 to 16% w/v polyethylene glycol (PEG) 4000, only rod-shaped crystals formed whereas at pH 8, in 10 to 14% w/v PEG, only hexagonal crystals existed. For both enzymes, catalyst recovery was greatest at high crystallization agent concentrations and low protein concentration. For ADH, the greatest activity recovery was 87% whereas for the lipase crystals, by using 45% v/v 2-methyl-2,4-pentanediol (MPD) as the crystallization agent, a crystal recovery of 250 crystals per μl was obtained. For the lipase system, the use of crystal seeding was also shown to increase the crystal recovery by up to a factor of four. From the crystallization windows, the original conditions based on literature precedent (35% v/v MPD, 1 mM CaCl(2), 1.8 mg protein/ml) were altered (47.5% v/v MPD, 2 mM CaCl(2), 3 mg protein/ml). This led to an improved recovery of the lipase under conditions that scale reliably from 0.5 ml to 500 ml with no change in size, shape or recovery of the crystals themselves. Finally, these crystals were crosslinked with 5% v/v glutaraldehyde and mass and activity balances were calculated for the entire process of CLEC production. Up to 35% of the lipase activity present in the crude solid was finally recovered in the lipase CLECs after propan-2-ol fractionation, crystallization, and crosslinking.     

Stabilization and crystallization of Ca2+-ATPase in detergent-solubilized sarcoplasmic reticulum

Conditions were developed for the long-term stabilization of Ca2+-ATPase in detergent-solubilized sarcoplasmic reticulum, purified Ca2+-ATPase, and purified-delipidated Ca2+-ATPase preparations. The standard storage medium contains 0.1 M KCl, 10 mM K-3-(N-morpholino)propanesulfonate, pH 6.0, 3 mM MgCl2, 20 mM CaCl2, 20% glycerol, 3 mM NaN3, 5 mM dithiothreitol, 25 IU/ml Trasylol, 2 micrograms/ml 1,6-di-tert-butyl-p-cresol, 2 mg/ml protein, and 2-4 mg of detergent/mg of protein. Preparations stored under these conditions at 2 degrees C in a nitrogen atmosphere retain significant Ca2+-stimulated ATPase activity for periods of 5-6 months or longer when assayed in the presence of asolectin. The same conditions are also conducive for the formation of three-dimensional microcrystals of Ca2+-ATPase. Of the 49 detergents tested for solubilization, optimal crystallization of Ca2+-ATPase was obtained in sarcoplasmic reticulum solubilized with octaethylene glycol dodecyl ether at a detergent/protein weight ratio of 2, and with Brij 36T, Brij 56, and Brij 96 at a detergent/protein ratio of 4. Similar Ca2+-induced crystals of Ca2+-ATPase were obtained with purified or purified delipidated ATPase preparations at lower detergent/protein ratios. The stabilization of the ATPase activity in the presence of detergents is the combined effect of high Ca2+ (20 mM) and a relatively high glycerol concentration (20%). Ethylene glycol, glucose, sucrose, or myoinositol can substitute for glycerol with preservation of ATPase activity for several weeks in the presence of 20 mM Ca2+.Ca2+-induced association between ATPase molecules may be an essential requirement for preservation of enzymatic activity, both in intact sarcoplasmic reticulum and in solubilized preparations.     

Electron microscope observations on Ca2+-ATPase microcrystals in detergent-solubilized sarcoplasmic reticulum

Crystalline arrays of Ca2+-ATPase molecules develop in detergent-solubilized sarcoplasmic reticulum during incubation for several weeks at 2 degrees C under nitrogen in a medium of 0.1 M KCl, 10 mM K-3-(N-morpholino)propanesulfonate, pH 6.0, 3 mM MgCl2, 20 mM CaCl2, 20% glycerol, 3 mM NaN3, 5 mM dithiothreitol, 25 IU/ml Trasylol, 2 micrograms/ml 1,6-di-tert-butyl-p-cresol, 2 mg/ml protein, and 2-4 mg of detergent/mg of protein. Electron microscopy of sectioned, negatively stained, freeze-fractured, and frozen-hydrated Ca2+-ATPase crystals indicates that they consist of stacked lamellar arrays of Ca2+-ATPase molecules. Prominent periodicities of ATPase molecules within the lamellae arise from a centered rectangular lattice of dimensions 164 x 55.5 A. The association of lamellae into three-dimensional stacks is assumed to involve interactions between the exposed hydrophilic headgroups of ATPase molecules, that is promoted by glycerol and 20 mM Ca2+. Similar Ca2+-induced crystals were observed with purified or purified and delipidated Ca2+-ATPase preparations at lower detergent/protein ratios. Cross-linking of Ca2+-ATPase crystals with glutaraldehyde protects the structure against conditions such as low Ca2+, high pH, elevated temperature, SH group reagents, high concentration of detergents, and removal of phospholipids by extraction with organic solvents that disrupt unfixed preparations.     

Purification of erythrocyte band 4.1 and other cytoskeletal components using hydroxyapatite-Ultrogel

An improved method for purifying erythrocyte band 4.1, the protein which mediates the interaction between spectrin and actin, has been developed. The new procedure, using adsorption chromatography on hydroxylapatite crystals immobilized within a crosslinked agarose gel (HA-Ultrogel), is simple and reproducibly provides a high yield of band 4.1 which is essentially free of protein kinase. Other components eluted from the hydroxylapatite matrix include band 4.9, ankyrin, and a 35,000-Da polypeptide that appears to be glyceraldehyde-3-phosphate dehydrogenase that remains bound to the erythrocyte membrane in 150 mM NaCl.     

Influence of divalent cations in protein crystallization.

We have tested the effect of several cations in attempts to crystallize the ligand-bound forms of the leucine/isoleucine/valine-binding protein (LIVBP) (M(r) = 36,700) and leucine-specific binding protein (LBP) (M(r) = 37,000), which act as initial periplasmic receptors for the high-affinity osmotic-shock-sensitive active transport system in bacterial cells. Success was achieved with Cd2+ promoting the most dramatic improvement in crystal size, morphology, and diffraction quality. This comes about 15 years after the ligand-free proteins were crystallized. Nine other different divalent cations were tried as additives in the crystallization of LIVBP with polyethylene glycol 8000 as precipitant, and each showed different effects on the crystal quality and morphology. Cd2+ produced large hexagonal prism crystals of LIVBP, whereas a majority of the cations resulted in less desirable needle-shaped crystals. Zn2+ gave crystals that are long rods with hexagonal cross sections, a shape intermediate between the hexagonal prism and needle forms. The concentration of Cd2+ is critical. The best crystals of the LIVBP were obtained in the presence of 1 mM CdCl2, whereas those of LBP, with trigonal prism morphology, were obtained at a much higher concentration of 100 mM. Both crystals diffract to at least 1.7 A resolution using a conventional X-ray source.     

RecA protein rapidly crystallizes in the presence of spermidine: a valuable step in its purification and physical characterization

The RecA protein of Escherichia coli, whether pure or in a crude cell lysate, will rapidly form small crystals (microcrystals) in the presence of low concentrations of spermidine. We describe the conditions of time, pH, and polyamine concentration over which crystallization occurs. Microcrystal formation is inhibited by concentrations of chloride over 25 mM and concentrations of phosphate or sulfate ions as low as 2 mM. Crystallization is not inhibited by high concentrations of other proteins, and the RecA protein microcrystals are easily collected by brief centrifugation. This provides a powerful purification step with high yield. Using this novel property, we prepared over 200 mg of RecA protein at least 95% pure with a single-strand DNA-dependent ATPase activity of 98% from 65 g of cells in 2-3 days. Spermidine was easily removed from the RecA protein by dialysis.     

Growing calmodulin crystals for X-ray diffraction studies at room temperature in 2 days

Crystals of bovine brain calmodulin have been grown using a novel procedure which utilizes a mixture of alcohols as a precipitant. Crystals were grown by vapor diffusion at room temperature in the presence of 15% ethanol, 25% 2-methyl-2,4-pentanediol, 5.0 mM calcium chloride, and 10.0 mM sodium acetate buffer at pH 4.0. Crystals were visible within 4 h and grew to 0.75 X 0.3 X 0.125 mm in 2 days, without seeding. The crystals of calmodulin are isomorphous with those found by W. J. Cook, J. R. Dedman, A. R. Means, and C. E. Bugg (1980, J. Biol. Chem. 255, 8152-8153). The protein crystallizes in space group P1 with unit cell dimensions a = 29.92 A, b = 55.96 A, c = 24.75 A, alpha = 93.81 degrees, beta = 99.24 degrees, and gamma = 88.40 degrees.     

Some comparisons between solution and crystal properties of thiosulfate sulfurtransferase

The activity and crystal stability of the enzyme thiosulfate sulfurtransferase were studied as a function of ionic strength. At 2 $\text{M}$ ammonium sulfate, where the x-ray structural studies of this protein were done soluble enzyme has low activity (<16% of the activity of the enzyme at an ionic strength of 0.1) and crystals of the enzyme are stable when substrates are added. However, at 1.4 M ammonium sulfate, crystals of TST rapidly dissolve in 1 mM CN^? but are relatively stable in 1 mM $\text{S}{}_2\text{O}{}_3{}^{\mathrm{=}}$. These results are consistent with a conformational change on converting the sulfur substituted form of the enzyme (ES) to the sulfur-free form (E) and helps to explain why this change was not observed in the crystallographic studies.     

Partial purification and some biochemical properties of neonatal rat cutaneous glutathione S-transferases

1. Previous studies have demonstrated the presence of glutathione S-transferases in the skin of rodents and humans. This study represents the first attempt to purify cytosolic glutathione S-transferases from skin of 3-day-old rats. 2. A partial purification of the enzyme was achieved by a two-step procedure: affinity chromatography followed by HPLC. Two peaks, one major (P-1) and one minor (P-2), were resolved by HPLC containing about 82% and 10% of the recovered activity, respectively. 3. The major form exhibited an overall purification of about 2270-fold with a specific activity of about 73 mumoles/min/mg protein towards 1-chloro-2,4-dinitrobenzene. 4. The kinetic data for P-1 yielded mean Km values of 2.39 mM for 1-chloro-2,4-dinitrobenzene and 0.72 mM for reduced glutathione, while the respective average Vmax values were found to be 212 and 101 mumoles/min/mg protein. 5. Significantly inhibition of enzyme activity was noted in the presence of 0.2 mM HgCl2, 0.63 microM 1.2-naphthoquinone, 1.0 microM triphenyltin chloride, and 12.5 microM 17 beta-estradiol-3-sulfate.     

Crystallization of Ca2+-ATPase in detergent-solubilized sarcoplasmic reticulum

Microcrystalline arrays of Ca2+-transporting ATPase (EC 3.6.1.38) develop in detergent-solubilized sarcoplasmic reticulum upon exposure to 10-20 mM CaCl2 at pH 6.0 for several weeks at 2 degrees C, in a crystallization medium that preserves the ATPase activity for several months. Of 48 detergents tested, optimal crystallization was obtained with Brij 36T, Brij 56, and Brij 96 at a detergent:protein weight ratio of 4:1 and with octaethylene glycol dodecyl ether at a ratio of 2:1. Similar Ca2+-induced crystalline arrays were obtained with the purified or delipidated Ca2+-ATPase of sarcoplasmic reticulum but at lower detergent:protein ratios. The crystals are stabilized by fixation with glutaraldehyde and persist even after the removal of phospholipids by treatment with phospholipases A or C and by extraction with organic solvents. The crystals obtained so far can be used only for electron microscopy, but ongoing experiments suggest that under similar conditions large ordered arrays may develop that are suitable for x-ray diffraction analysis.     

Cloning, expression and characterization of an extracellular enolase from Leuconostoc mesenteroides

Enolase on the surface of streptococci putatively facilitates pathogenic invasion of the host organisms. The related Leuconostoc mesenteroides 512FMCM is nonpathogenic, but it too has an extracellular enolase. Purified isolates of extracellular dextransucrase from cultures of L. mesenteroides contain minute amounts of enolase, which separate as small crystals. Expression of L. mesenteroides enolase in Escherichia coli provides a protein (calculated subunit mass of 47 546 Da) catalyzing the conversion of 2-phsopho-D-glycerate to phosphoenolpyruvate. The pH optimum is 6.8, with Km and kcat values of 2.61 mM and 27.5 s(-1), respectively. At phosphate concentrations of 1 mM and below, fluoride is a noncompetitive inhibitor with respect to 2-phospho-D-glycerate, but in the presence of 20 mM phosphate, fluoride becomes a competitive inhibitor. Recombinant enolase significantly inhibits the activity of purified dextransucrase, and does not bind human plasminogen. Results here suggest that in some organisms enolase may participate in protein interactions that have no direct relevance to pathogenic invasion.     

Occurrence of glucocorticoid binding sites in solubilized microsomes from rat liver

Recent studies suggested the presence of specific glucocorticoid binding sites on rat liver microsomal membranes. We report here a new solubilization procedure which allows the physicochemical characterization of the microsomal glucocorticoid binding sites. Solubilization was achieved with 2 mM CHAPS in the presence of 5 mM benzamidine. Binding of [3H]cortisol showed a high affinity (Kd = 5.1 x 10(-9) M) and a limited capacity (0.72 pmol/mg of protein). The binding activity was abolished by elevated temperature and pronase. Competition experiments revealed that natural glucocorticoids and progesterone were highly potent competitors whereas dexamethasone and triamcinolone acetonide did not compete. Chromatography on DEAE Trisacryl and heparin Ultrogel confirmed that the solubilized protein is different from corticosteroid binding globulin and the cytosolic glucocorticoid receptor. Treatment of microsomal fractions with phosphatidyl inositol phospholipase C promoted the release of specific binding activity suggesting a putative glycosylphosphatidyl anchor for this protein. This finding may have interesting implications concerning the mechanism of glucocorticoid hormone action.     

Crystallization and preliminary x-ray characterization of thioredoxin reductase from Escherichia coli

Single crystals of thioredoxin reductase, suitable for x-ray diffraction studies, have been obtained at room temperature by vapor diffusion of 10-20 mg/ml protein solution against 35% polyethylene glycol containing 200 mM ammonium sulfate. Good quality crystals appear spontaneously only from a protein solution that had been stored for more than a year at 4 degrees C, although large single crystals are reproducibly obtained from fresh protein solutions by micro-seeding. The space group is P6(3)22 (a = b = 123.8 A, c = 81.6 A), with one monomer of the enzyme (34.5 kDa) in the crystallographic asymmetric unit. The crystals are well ordered and diffract to beyond 2 A resolution.     

The effect of some antineoplastic agents on glutathione S-transferase from human erythrocytes

Glutathione S-transferase was purified from human erythrocytes and effects of some antineoplastic agents were investigated on the enzyme activity. The purification procedure was composed of Glutathione-Agarose affinity chromatography after preparation of erythrocytes hemolysate. Using this procedure, the enzyme, having the specific activity of 16.00 EU/mg proteins, was purified 1143-fold with a yield of 80%. The purified enzyme showed a single band on the SDS-PAGE. The effects of paclitaxel, cyclophosphamide, and gemcitabine, are antineoplastic agents, were examined on the in vitro enzyme activity of glutathione S-transferase and were determined to be inhibitors for the enzyme. IC₅₀ values were 0.23 mM for paclitaxel, 5.57 mm for cyclophosphamide, and 6.35 mM for gemcitabine. These constants were 0.182 ± 0.028 mM and 0.162 ± 0.062 mM for paclitaxel, 6.97 ± 0.49 mM and 10.50 ± 5.43 mM for cyclophosphamide, and 6.71 mM and 7.93 mM for gemcitabine, with GSH and CDNB substrates, respectively. Inhibition types of all inhibitors were noncompetitive.     

Stimulation of ascites tumor RNA polymerase II by protein kinase.

The activity of purified RNA polymerase II from Novikoff ascites tumor cells is stimulated 5-7-fold by a purified protein factor. This protein factor, designated HLF2, has extensive protein kinase activity and catalyzed the incorporation of gamma-32G from ATP into protein under normal RNA polymerase assay conditions. Protein phosphorylation is totally dependent on the presence of HLF2 and is stimulated 2-3-fold by the presence of highly purified RNA polymerase II. The purification procedure developed for the isolation of the polymerase stimulatory factor resulted in a 4000-fold purification of a protein kinase. Chromatography on carboxymethylcellulose, phosphocellulose, and Sephadex G-100 did not resolve polymerase stimulatory activity from protein kinase activity. Adenylimidodiphosphate (AMP-PNP), an inhibitor of protein kinases, inhibited the stimulatory activity of purified factor by 80%. The heat denaturation profile of protein kinase was paralleled by the loss of polymerase stimulatory activity. Concentrations of (NH4)2SO4 which are known to inhibit polymerase stimulation (Lee and Dahmus, 1973) also inhibit protein kinase activity. The protein kinase activity associated with stimulatory factor catalyzes the phosphorylation of basic proteins such as protamine or histone. The protein kinase is not stimulated by cyclic 3', 5'-AMP or -GMP over a concentration range of 10(-6)-10(-4)M. Furthermore, protein kinase activity is not inhibited by either the regulatory subunit of rabbit muscle protein kinase or by the heat-stable inhibitor of cyclic 3', 5'-AMP-dependent protein kinases. Protein kinase activity is stimulated by KCl or NH4Cl and is inhibited by MnCl2. The apparent Km values, determined in the presence of 4 mM Mg2+, are 0.02 mM for ATP, and 4.1 mM for GTP.     

Protease activation of the entomocidal protoxin of Bacillus thuringiensis subsp. kurstaki.

Two isolates of Bacillus thuringiensis subsp. kurstaki were examined which produced different levels of intracellular proteases. Although the crystals from both strains had comparable toxicity, one of the strains, LB1, had a strong polypeptide band at 68,000 molecular weight in the protein from the crystal; in the other, HD251, no such band was evident. When the intracellular proteases in both strains were measured, strain HD251 produced less than 10% of the proteolytic activity found in LB1. These proteases were primarily neutral metalloproteases, although low levels of other proteases were detected. In LB1, the synthesis of protease increased as the cells began to sporulate; however, in HD251, protease activity appeared much later in the sporulation cycle. The protease activity in strain LB1 was very high when the cells were making crystal toxin, whereas in HD251 reduced proteolytic activity was present during crystal toxin synthesis. The insecticidal toxin (molecular weight, 68,000) from both strains could be prepared by cleaving the protoxin (molecular weight, 135,000) with trypsin, followed by ion-exchange chromatography. The procedure described gave quantitative recovery of toxic activity, and approximately half of the total protein was recovered. Calculations show that these results correspond to stoichiometric conversion of protoxin to insecticidal toxin. The toxicities of whole crystals, soluble crystal protein, and purified toxin from both strains were comparable.     

High-resolution hydroxyapatite chromatography of proteins

Hydroxyapatite chromatography has been used to separate all five isozymes of lactic dehydrogenase, six enzymatically active forms of bovine pancreatic DNase I, and a standard protein mixture. The proteins were eluted with a linear gradient of sodium phosphate. Enzyme activity recoveries were greater than 90%. Packing materials were obtained from commercial fine-particle-sized hydroxyapatite (DNA grade Bio-Gel HTP) by an elutriation procedure. Long columns packed with small crystals were run under low pressure at acceptable flow rates, and were used over prolonged periods.     

Protease activation of the entomocidal protoxin of Bacillus thuringiensis subsp. kurstaki

Two isolates of Bacillus thuringiensis subsp. kurstaki were examined which produced different levels of intracellular proteases. Although the crystals from both strains had comparable toxicity, one of the strains, LB1, had a strong polypeptide band at 68,000 molecular weight in the protein from the crystal; in the other, HD251, no such band was evident. When the intracellular proteases in both strains were measured, strain HD251 produced less than 10% of the proteolytic activity found in LB1. These proteases were primarily neutral metalloproteases, although low levels of other proteases were detected. In LB1, the synthesis of protease increased as the cells began to sporulate; however, in HD251, protease activity appeared much later in the sporulation cycle. The protease activity in strain LB1 was very high when the cells were making crystal toxin, whereas in HD251 reduced proteolytic activity was present during crystal toxin synthesis. The insecticidal toxin (molecular weight, 68,000) from both strains could be prepared by cleaving the protoxin (molecular weight, 135,000) with trypsin, followed by ion-exchange chromatography. The procedure described gave quantitative recovery of toxic activity, and approximately half of the total protein was recovered. Calculations show that these results correspond to stoichiometric conversion of protoxin to insecticidal toxin. The toxicities of whole crystals, soluble crystal protein, and purified toxin from both strains were comparable.     

Removal of salt from a salt-induced protein crystal without cross-linking. Preliminary examination of "desalted" crystals of phosphoglucomutase by X-ray crystallography at low temperature.

A model procedure for removing salt from relatively fragile salt-induced protein crystals is proposed. The procedure is based on physical principles and is validated by using millimeter-size crystals of rabbit muscle phosphoglucomutase grown from a 2.1 M solution of ammonium sulfate. Three types of operations are included in the procedure: initial transfer to salt solutions of reduced concentration; transfer to the organic-rich phase of an equilibrium biphasic mixture obtained with aqueous solutions of polyoxyethylene and the salt; and addition of various replacement cosolutes in aqueous solutions of polyoxyethylene to reduce osmotic stress on the crystal as the remaining salt is removed. A critical feature of the overall procedure is maintenance of near equilibrium throughout by using a large number of steps involving small changes in solute concentration. The conditions used in the actual transfer were adjusted to eliminate the fracturing of crystals by visually distinguishing between two opposing types of fracture patterns: those produced by osmotic crushing as opposed to osmotic expansion. Basic requirements for a successful procedure with other protein crystals are a high permeability toward small solutes and a relatively slow dissolution rate at salt concentrations for which biphasic mixtures can be obtained. Desalted crystals of phosphoglucomutase have no visible fractures, are stable in the final solution for at least a week, and exhibit no noticeable change in the resolution of their X-ray diffraction pattern. In fact, desalted crystals can be rapidly cooled to 160 K, whereas untreated crystals are almost completely disordered by the same cooling procedure. The component of the desalting mixture whose presence is crucial to the success of the cooling process is polyoxyethylene, which apparently impedes the formation of ice within the protein crystal. Diffraction data obtained with an area-detector diffractometer did not differ significantly, either in terms of quality or resolution range, between crystals in 2.3 M ammonium sulfate at room temperature and crystals at 160 K in which ammonium sulfate had been replaced by glycine. The successful use of the following replacement solutes, instead of glycine, also is documented: sucrose, glycerol, and a low molecular weight poly(ethylene glycol) (PEG-400).