Protein crystallization

Transcription

1 Protein crystallization Mirjam Leunissen

2 An essay on several aspects of protein crystallization research Lysozyme Mirjam Leunissen Department of solid state chemistry October 2001 Supervisor: Willem van Enckevort Front page: collage of protein crystals

3 Abstract This essay is about the crystallization of proteins. Since the first published observation of crystallizing proteins about 160 years ago, protein crystal growth has developed into an extensive research field with many applications, for instance in the pharmaceutical industry. In this report, an overview of the unique physical-chemical characteristics of protein molecules, their properties in solution and their properties in the crystalline state, together with a description of some important aspects of protein crystallization and research in this field, is given. Seven proteins that are the main models of crystallization studies are presented, just as the general (thermodynamic) principles applying to protein crystal growth and the main parameters influencing the process. The most commonly used protein crystallization techniques are described, together with their underlying principles and the approaches used to find suitable conditions for crystallization of a particular protein. Focusing on the fundamental studies of protein crystallization on a molecular level, the current knowledge of nucleation processes, crystal growth mechanisms and kinetics is summarized. Concluding this essay, an inventory is made of knowledge that is still missing, but that is crucial in order to develop a more directed crystallization approach. Suggestions for future research, aimed at obtaining this insight, are given.

4 Contents Abstract Chapter 1 Introduction 1.1 The importance of protein crystal growth A history of protein crystallization Aspects of protein crystallization research 4 Chapter 2 Features and properties of proteins 2.1 Protein structure Properties of proteins in solution Properties of protein crystals The complexity of protein crystallization Protein model systems 9 Chapter 3 Principles of protein crystallization 3.1 The thermodynamics of crystal growth Parameters influencing protein crystal growth 15 Chapter 4 Methods and approaches in protein crystallization 4.1 Protein crystallization methods Approaches to find crystallization conditions 25 Chapter 5 Nucleation and growth processes of protein crystals 5.1 Methods of investigation Formation of critical nuclei Growth mechanisms, the growth unit, transport processes and kinetics 30 Chapter 6 General discussion and conclusion 36 References

5 Chapter 1 Introduction This essay deals with the crystallization of the members of one particular, but very broad, class of biological macromolecules: the proteins. Protein crystallization forms a very extensive field of research, with many different aspects and applications. As you will learn in this essay, it is in some respects rather different from conventional crystallization of inorganic, small molecule compounds, for instance where the crystallization techniques are concerned, while in other respects they have many characteristics in common. An example of such similarity is formed by the crystal growth mechanisms. In order to get a bit acquainted with the subject, in this introductory chapter, I will first tell something about the importance and applications of protein crystallization and the development of the field of protein crystal growth from its starting point until now. Furthermore, focusing on the research component of the protein crystal growth field, I will indicate its many different aspects and which of them will be treated in this essay. 1.1 The importance of protein crystal growth So, question is: why would anybody want to spent (sometimes huge) efforts on the crystallization of proteins? While, in the past, purification of mixtures was the main goal in crystallizing experiments, at present the main answer to the question stated above is provided by the fact that studies of the atomic structure of biological macromolecules (i.e. proteins, DNA, RNA, etc.) have proved of great value in revealing structure/function relationships that are of importance in our understanding of how enzymes, nucleic acids and other macromolecules operate in biological systems. Recently, nuclear magnetic resonance (NMR) studies of protein solutions have yielded atomic structure information for some small proteins. However, for proteins with molecular weights in excess of 20,000 Da (Da=dalton=1.66x10-27 kg, this unit of mass is specifically applied to proteins) only X-ray and neutron diffraction techniques can provide structure information to atomic resolution. Such studies require single crystals of high structural perfection and with minimum dimension of several hundred micron. But, one may ask, is the amount of information obtained really worth the effort, time and money crystallization costs? Well, I would say yes, because crystallographic studies of biological macromolecules have become of considerable interest to pharmaceutical, biotechnological and chemical industries, as promising tools in protein engineering, drug design and other applications to biological systems (figure 1.1). Fig. 1.1 Use of protein crystallography. Diagram showing the path by which X-ray crystallography, in combination with computer graphics analysis and synthetic chemistry, may be rationally applied to create new chemical compounds such as medicinal drugs and pharmaceuticals, pesticides and herbicides and industrial fine chemicals. Source: [1]. 1

6 The elaborate information that can be obtained from the three-dimensional structure of a protein is useful in a variety of ways. From the basic biological view point, this information underlies our understanding of the mechanisms by which enzymes, receptors, hormones, etc. function in biological systems. Within the pharmaceutical industry, protein structure information can be helpful in the development of novel drugs. Since many pharmaceutical agents act by interacting with proteins, knowledge of the three-dimensional structure of a target protein can be used to design compounds that selectively bind to sites of this protein and thereby inhibit its activities. Another highly promising application of protein crystallography is in protein engineering. Using molecular biology techniques, investigators can specifically alter protein molecules by site-directed mutagenesis (i.e. by altering proteins in specific, selected regions by altering the DNA stretch encoding that part of the protein). In general, the most promising approaches to protein engineering depend upon detailed structural information about the proteins of interest. An additional use of protein crystallography is in the design of synthetic vaccines. Several recent studies have indicated that effective vaccines might be made from synthetic peptides (very small proteins) that are representative of protein segments found on the surfaces of target proteins. Protein crystallography provides one of the most effective techniques for locating those peptides. However, the crystallization of proteins is not only an excellent tool to obtain information about the three-dimensional protein structure, but also a very interesting subject for crystal growth studies in its own right. Compared to conventional, small molecule (often salts) crystal growth systems, proteins display relatively slow growth kinetics and have growth units with a large size (because individual protein molecules are very large). These facts allow one to more readily study crystal growth mechanisms, par example by means of in situ atomic force microscopy observations. Therefore, protein crystallization now serves as the best model system for general crystallization from solution. Concluding, one can say that there are two interests: a general interest in delineating processes that play a role in crystallization from solutions and an interest in the crystallization of certain proteins in particular, in order to obtain structural information about these proteins. These two interests do not stand on their own, however. As many proteins tend to resist their easy crystallization, more insight in the processes at work during crystal formation might enable a more directed approach with a higher chance of success. I will come back to this in the concluding chapter of this essay. 1.2 A history of protein crystallization This section is based on information in [1] and references therein. In table 1.1 (page 3) a chronology of protein crystal growth is given. The first protein crystals The history of (recorded) protein crystal growth started about 160 years ago. The first published observation of the crystallization of a protein appears to be by Hünefeld in The protein, hemoglobin from the earthworm, was obtained as flat plate-like crystals when the worm s blood was pressed between two slides of glass and allowed to dry very slowly. This observation clearly stated that protein crystals can be produced by the controlled evaporation of a concentrated protein solution, that is, protein crystals can be produced by slow dehydration. This is the basis for most of the techniques we use today and which are described in section 4.1. The first investigators that took up protein crystallization, focussed on hemoglobin (of many different animal species) for the next 15 years. Unfortunately, through 1850, all of the blood crystals reported, appeared to have grown more or less fortuitously and no investigator had suggested any general procedure for their directed growth. The first person to actually devise 2

7 successful and reproducible methods for the growth of hemoglobin crystals was Fünke, who published a series of articles on the purposeful growth of hemoglobin crystals. Following hemoglobin, the next class of proteins to be investigated, in the period from 1850 to about 1900, were the plant seed reserve proteins, principally the so called globulins. The methods that were developed to crystallize these proteins included: extraction of proteins into salt solutions followed by slow cooling, dialysis of a salt solution extract of the seeds exhaustively against distilled water and treatment of protein solutions with alcohol, acetone or ether. In these procedures we find for the first time the exploitation of several approaches now in common use: temperature variation under constant solution conditions, dialysis against low ionic strength solutions (to take advantage of the low solubility of many proteins at low ambient salt concentrations) and the use of organic solvents as precipitating or crystallizing agents. At almost the same time the work with plant seed proteins was carried out, similar efforts were done to crystallize the proteins hen egg albumin and horse serum albumin. The procedures for their crystallization used many of the suggestions of Hofmeister regarding the salting out of proteins by high concentrations of salt ions and the precipitation of proteins by careful regulation of ph (for an explanation, see section 3.2). The first enzyme, urease, was crystallized in 1925 by Sumner and at almost the same time the first hormone, insulin, was crystallized too. Crucial to the insulin crystallization was the addition of divalent zinc ions. This was one of the first examples of crystal growth promoted by the addition of metal ions. Table 1.1 A chronology of protein crystal growth. Source: [1]. 3

8 A change of attitude Until the 1930s (but also beyond that for many years), the rationale for crystallizing proteins, particularly enzymes, was to supply a technique for purifying a specific protein from a complex extract, or to demonstrate the purity of a preparation. In the late 1930s, however, certain X-ray diffractionists began to turn their attention to protein crystals as a source of structural information about biological macromolecules for reasons stated in the previous section. This interest carries on till present time. The interest of X-ray diffractionists was influential in promoting efforts to reproducibly grow high quality protein crystals, but also led to efforts to increase success rates and to automate the crystallization process. The background for the latter is formed by the fact that with the extraordinary advances in data collection and computing techniques and with the revolution in pharmacology and biotechnology, the ask for new macromolecular crystals very soon greatly surpassed their supply. At present the bottleneck in solving the problem of limited crystallization yields is mainly formed by a lack of insight. As you will see in this essay, most protein crystallization approaches are based on the trial-and-error principle, while insight in the fundamental processes at work is very restricted. As a consequence, many proteins still resist crystallization for unknown reasons. Clearly, at present, the challenge is to obtain enough knowledge about the processes at work and, based on this knowledge, to develop new, more directed methods in order to be able to readily crystallize any protein at will. 1.3 Aspects of protein crystallization research There are many areas of research related to protein crystallization, each focusing on different aspects. In this essay I will treat only a limited number of them. Chapter 2 provides an introduction to the subject common to all of the studies: protein molecules. The general physical-chemical characteristics of protein molecules, their properties in solution and their properties in the crystalline state are dealt with. From this information, the answer to the question why the crystallization of proteins is very complex will be extracted. Furthermore, in this chapter, the seven proteins that are the main models of crystallization studies are presented. In chapter 3, I will treat the general (thermodynamic) principles applying to (protein) crystallization and explain which parameters play an important role in the crystallization processes and how they exert their influences. Chapter 4 is an overview of modern protein crystallization techniques, their respective advantages and/or disadvantages and the underlying thermodynamic principles. Moreover, in this chapter, the approaches used at present to find suitable conditions for the crystallization of a particular protein are sketched. Then, in chapter 5, I will summarize the current knowledge of nucleation processes, crystal growth mechanisms and kinetics of proteins and how they can be studied experimentally. To conclude this essay, I will reflect on all previous chapters, in the general discussion and conclusion of chapter 6. This, in order to make an inventory of crucial knowledge that is still missing for a more directed approach. I will also sketch what, in my opinion, the direction of future research should be. Of course, there are many topics left which I will not treat, like: polymorphy, defects in protein crystals, sources of impurities, purification of protein samples, automation of crystallization trials and mutagenesis of proteins in order to transform it into a protein more suitable for crystallization. Also, the discussion presented here will not consider the possibility of conformational heterogeneity and flexibility of the protein molecular structure, which may be important for the perfection and utility of certain protein crystals. However, below, I will give references that can guide you to literature about some of these topics. A comment (short or more detailed) on virtually each of all aforementioned topics, together with lots of references to relevant literature can be found in a book written by McPherson [1]. Especially 4

9 if you are new to the field of protein crystallization I would highly recommend it. Other interesting references are: [2] and [3], if you are interested in protein molecular interactions in solutions and [4] for interactions in the crystalline state. For a general theoretical introduction to the problem of protein nucleation, see [4], [5], [6], [7], [8]. A comprehensive list of references on impurity effects and defect formation is provided in [9], but lots of information on these topics can also be found in [10] and [11]. A nice review of theoretical models that are used to study protein crystal formation and the major ideas used in their development is presented in [5]. Undoubtedly, the information in this essay and in the references given above will not cover the whole field of protein crystallization, but it will provide a good starting point for anybody interested in this extensive field. 5

10 Chapter 2 Features and properties of proteins 2.1 Protein structure Proteins are so called macromolecules, because the particle diameter is ~ Å as compared to ~3 Å for most inorganic particles. Proteins can be considered as polymers of amino acids, linked together in a chain-like arrangement. The number of amino acids constituting one protein molecule ranges approximately from 100 to 27,000 for the proteins known. The specific sequence of amino acids in a protein is called its primary structure. a R 1 + H 3 N C H + H 3 N COO - R 1 C H O C + N H + H 3 N R 2 R 2 C H COO - C COO - + H H 2 O Natural proteins are built up by multiple numbers of twenty different amino acids, sometimes referred to as residues. The general structure of an amino acid can be seen in figure 2.1a. In each amino acid, the amino and carboxyl groups make one bond each with the so called α-carbon atom, Cα. One of the other two bonds of Cα is occupied with hydrogen, while the other is occupied with a relatively small group, called the side chain, which is different for each of the twenty amino acids. These side chains consist of simple hydrocarbon groups, that may contain aromatic rings, nitrogen, oxygen and sulfur (figure 2.2). b R i O H ( NH CH C) n OH The so called peptide bonds between the amino and carboxyl groups of neighboring amino acids (figure 2.1a) form the backbone along the polypeptide chain. The side chains form its short branches (figure 2.1b). Fig. 2.1 Primary structure of proteins. a) Formation of a peptide bond between two amino acids that differ in their side chains (R1, R2). The carbon atom in the middle of each amino acid structure is called Cα. b) Overall formula for a polypeptide chain (backbone) with side chains Ri. Adapted from: [12]. The same amino acid sequence (that is, the same primary structure) can occur in different conformations through different azimuthal rotations of the side chains about peptide links. This can determine some secondary structure of the backbone, such as the formation of helical windings and sheet-like structures. In the latter case, the polypeptide chain is folded several times back and forth in a certain plane. For examples of these common types of secondary structure I would like to refer to figures 2.9a and 2.7 (page 11). Very important to the overall structure of a protein, is the fact that the amino acid side chains can both consist of polar functional groups or nonpolar functional groups (review Fig. 2.2 Natural amino acid side chains. Chemical structure of the side chains of the 20 amino acids that form proteins. Charged (polar) functional groups in bold. Source: [12]. 6

11 figure 2.2). In their natural aqueous environment, the polar groups prefer to associate with water. In contrast to these hydrophilic groups, the nonpolar hydrophobic groups prefer to associate with themselves. As a result of these preferences the molecule lowers its free energy by folding to a specific three-dimensional structure, the so called tertiary structure. The pictures in section 2.5 provide many examples of this overall three-dimensional structure of proteins. Molecules with comparable numbers of polar and nonpolar residues fold into globular shapes in which the hydrophilic residues tend to concentrate on the surface and the hydrophobic groups in the core. This three-dimensional structure is stabilized by various close-range interactions between atoms that are distant on the backbone. These include electrostatic interactions between charges and dipoles, van der Waals forces, hydrogen bonds, hydrophobic interactions and (covalent) disulfide bonds. The three-dimensional structure described above represents the final folding of a polypeptide chain, but some proteins display another level of organization, the so called quaternary structure. This refers to the fact that many functional proteins are complexes of smaller subunits. Each subunit is a polypeptide chain with its own specific tertiary, three-dimensional structure. To form the final functional protein a number of subunits associate to one big complex by means of noncovalent interactions. This arrangement of subunits is called the quaternary structure of a protein. A very nice example of the quaternary structure of a protein is displayed in figure 2.9 (page 11). 2.2 Properties of proteins in solution The properties of a protein are to a great extent determined by the amino acid side chains present on its surface. Within the group of polar residues there are two classes of charged amino acids: acidic and basic. At neutral ph, which centers the normal physiological range, the acidic amino acids are negatively charged and the basic amino acids carry a positive charge. Therefore, at ph ~6-7 and temperatures 0-40 C, typical of biological conditions, the molecular surface usually is charged. Like all polar side chains, the charged groups interact extensively with water and tend to solvate the protein. Moreover, exposed hydrophilic groups on macromolecules can bind, both transiently and in a stable manner, not only water molecules but also a variety of ions, both cations and anions. Thus, as a consequence of their polyionic character, the surfaces of macromolecules display a variegated pattern of positive and negative charge. As an example, the charge distribution on the surface of an arbitrary protein is visualized in figure 2.3. The actual net surface charge is dependent on the amino acid sequence and environmental factors, such as the ph of the solution. Lysozyme, for instance, has positive charges at ph 4.5 [13]. As a consequence, under these solution conditions repulsion dominates the interactions between these macro-ions. Fig. 2.3 Protein surface charge distribution. Positively and negatively charged amino acid side chains on a protein give rise to a characteristic electrostatic surface like that shown here for the protease from P. cyclopium. Blue represents positive field strength and red negative. The surface, however, can change as a function of the ph, which causes ionizable groups to gain or lose protons. Source: [1]. 7

12 At the iso-electric point, pi, (at ph=11.3 for lysozyme [14], between 5 and 7 for most other proteins) there are equal numbers of positively and negatively charged residues and as a consequence at this ph value the net surface charge vanishes. However, one must realize that this does not mean that individual surface patches do not carry a charge. The net surface charge of a protein molecule changes with the ph of the solution and consequently, as one can imagine, the solubility of the protein changes with the ph too. For a particular protein, with its own specific sequence of amino acid side chains, the solubility dependence on the solution ph is given by the following equation [15]: + + [ Pi] + [ Pj] = K ± [1 + γ ± ( Ki /[ H ] γi) + γ ± ([ H ]/ Kjγj)] exp( [ P] = [ P± ] + α 0s0 / kt ) i j i [P] = the total concentration of dissolved molecules α 0 = the surface free energy of a protein molecule in solution [P ± ] = the zwitterionic state of the protein molecule s 0 = the surface area of a protein molecule in solution [P i ] = negatively charged states of the protein molecule γ = activity coefficient [P j ] = positively charged states of the protein molecule K x = (de)protonation reaction constant side chain According to this formula, solubility reaches a minimum at proton concentrations corresponding to the isoelectric point and steeply increases at both lower and higher ph. The temperature dependence of solubility is determined by the temperature dependence of the protonation and deprotonation reaction constants of the side chains. For proteins, which exist predominantly in an aqueous environment, one free energy minimum is represented when they are fully solvated, but in extremely concentrated solutions where there is insufficient water to maintain hydration, the molecules may aggregate as an amorphous precipitate or they may crystallize. Let us therefore consider some properties of protein crystals. j 2.3 Properties of protein crystals Protein crystals differ from crystals of most other compounds in several respects [1], [16]. One of the major differences is the high solvent content of protein crystals, which is responsible for most of their unique features. On crystallization, proteins form a wide open network of macromolecules (figure 2.4). Depending on the specific protein, between 25 and 85 vol% of the crystal consists of channels. These are lined with protein-bound water, referred to as ordered or bound water [19], and possibly some precipitate ions. This bound water forms the ordered hydration shell of the protein and is a real component of the macromolecule s structure. Besides channels, also cavities are being formed and both are filled with so called disordered bulk water. Furthermore, water molecules of ill-defined character are present that provide continuity between the bulk and bound water. Much of the ordered water, localized in areas between adjacent protein molecules in the crystal lattice, plays a role in the formation of extended patterns of hydrogen bonds that supplement and strengthen the direct interactions which occur between the protein molecules [20]. Fig. 2.4 Structure of protein crystals. Schematic representation of protein crystal structure with wide open network of protein molecules lined by layers of bound water, and channels formed by the protein-bound water network filled with solution ( bulk water ). Source: [18]. Inset: view of monomer packing in the unit cell of a crystal of chymotrypsinogen, showing the large solvent channels penetrating the unit cell. Source: [17]. 8

13 Protein-protein contacts in crystals are complex, involving a delicate balance of specific and nonspecific, intrinsically flexible interactions [5], [20]. Hydrogen bonding and electrostatic interactions, involving the participation of flexible amino acid side-chains on the protein s surface, together with numerous solvent molecules or ions that are immobilized between molecules during crystal lattice formation, are examples of specific interactions. Nonspecific interactions are van der Waals interactions and hydrophobic interactions. It is not clear from the literature whether any particular type of interaction makes a dominant contribution to the energy of protein-protein contacts in crystals. It appears that interactions between proteins represent a delicate balance of many contributions whose relative importance varies from case to case. 2.4 The complexity of protein crystallization In general, the crystallization of proteins is a very complex process. Experiences of many investigators point out that most proteins are difficult to crystallize and even if a protein tends to crystallize relatively easily there are many parameters that must be taken into account. From the preceding sections one can identify multiple reasons for the difficulty of protein crystal growth. Apparently, protein molecules are very complex (large, flexible molecules often composed of several subunits), relatively chemically and physically instable (unfolding, hydration requirements, temperature sensitivity) and they have dynamic properties. If the solution changes, the molecule properties (e.g. conformation, charge and size) change too. Furthermore, every macromolecule is unique in its physical and chemical properties because every amino acid sequence produces a unique three-dimensional structure having distinctive surface characteristics. Thus, lessons learned by investigation of one protein only marginally apply to others. For a crystal to form, interactions between protein molecules must be suitable in their geometric arrangement, degree of specificity and strength. The size and complexity of proteins is reflected in their ability to form many different intermolecular contacts, yet only a few of these contacts must exist in a given crystal and consequently the correct selection must be made each time a molecule is incorporated [4]. Moreover, the ability to form the right interactions is not a property of the protein molecule alone, but a joint property of the protein and its solution environment. The diversity of chemical groups involved in the contacts implies that slight changes in ph, ionic strength, temperature or concentration of an auxiliary ion or molecule may strongly influence crystallization. It is not yet possible to predict the conditions required to crystallize a protein from its other physical properties. Changes in a single experimental parameter can simultaneously influence several aspects of a crystallization experiment. For example, temperature changes affect protein solubility, rates of nucleation and growth, and equilibration of the experimental apparatus. The interaction of parameters makes it difficult to design experiments to isolate individual effects and likewise complicates the interpretation of experimental results. Therefore, our insight in protein crystallization processes is still very limited and, consequently, protein crystal growth remains more of an art than a science. 2.5 Protein model systems As can be seen in the historical table of section 1.2 many different proteins have been crystallized. Nowadays, databases contain structures of more than 10,000 different proteins, determined by X- ray diffraction analysis of their crystals. However, for the study of the processes at work in protein crystal growth, only a couple of proteins are used as model systems. These proteins are relatively easy to crystallize and are thought to be representative of protein molecules in general. In the following, I will give a (short) description of the most commonly used model proteins. Each of these is accompanied by a list of references that contain further information. The space groups in which they crystallize are enlisted in table 2.1 (next page), together with the respective unit cell 9

14 dimensions and angles. Information on other proteins and their crystal structures can be found in the Protein Data Bank [21]. This is also the source of all pictures in this section. Table 2.1 Crystal structures of frequently used model proteins. Only angles 90 are denoted. Protein Space group a [Å] b [Å] c [Å] α [deg] β [deg] γ [deg] Lysozyme Canavalin Concanavalin A Thaumatin (Apo-)ferritin Catalase Insulin P P P P P P P R C I P P C P P C P F P P P I P P R P Fig. 2.5 Lysozyme. This ribbon model displays the folded structure of hen egg white lysozyme. Some secondary structure in the form of helices is clearly visible. Lysozyme The protein first studied by X-ray diffraction and still most widely used for crystal growth studies is hen egg white (HEW) lysozyme. This is an enzyme which hydrolyzes polysaccharides in bacterial cell walls. HEW lysozyme is composed of a single chain of 129 amino acids and has a molecular weight, M R, of 14,296 Da (Da=dalton= 1.66x10-27 kg). Figure 2.5 illustrates its three-dimensional, folded structure as a ribbon model of its peptide backbone. Lysozyme is particularly attractive for crystal growth research, because detailed information exists on its thermophysical properties (contrary to the situation for all other proteins). References: the crystallization of lysozyme was first published by Abraham and Robinson in [22]. Information on the solubility of the protein as a function of ambient salt concentration, ph and temperature can be found in [23], [24], [25], [26], [27], [28], [29], [30], [31]. 10

15 Canavalin Canavalin is the major reserve protein of the seeds of the jack bean. It is composed of three identical subunits, each with M R ~47,000 Da, arranged about a perfect threefold axis. Each of the subunits consists of two chains, called A and B (figure 2.6). a b References: the crystallization of canavalin was first published by Sumner and Howell [32]. More information about its crystallization can be found in [33] and [34]. Studies of the solubility of canavalin are [23] and [35]. Fig. 2.6 Canavalin. a) Ribbon model displaying the rather complex quaternary structure of canavalin. The protein is a complex of three subunits, densely packed together. b) Each of the subunits is composed of two chains, called A and B. Concanavalin A Concanavalin A is a derivative of the enzyme chitinase and is, just as canavalin, obtained from the jack bean. The protein consists of two chains (A and B) of 237 amino acids each ( and M R ~25550 Da). Figure 2.7 is a ribbon model of this protein. References: the crystallization of concanavalin A was first published by Sumner [36]. A later investigation of the protein s structure can be found in [37]. Solubility information was published by Mikol and Giegé [38]. Fig. 2.7 Concanavalin A. Ribbon model displaying the folded structure of the complex of two chains that form concanavalin A. Pronounced secondary structure in the form of sheets (indicated with arrows) is visible. Fig. 2.8 Thaumatin. Ribbon model. This relatively simple protein consists of a single chain only. Thaumatin Thaumatin (figure 2.8) is a 207 residues, 22,203 Da, sweet-tasting protein from the African serendipity berry and has commercial potential as a sweetener. It is problematic to crystallize under most conditions, except when tartrate is added to the solution, then it will readily crystallize. References: an X-ray analysis of thaumatin has been published in [39]. Phase diagrams of the protein can be found in [40]. a b Ferritin Ferritin is the iron storage protein of higher organisms. It binds and thereby stores iron intracellularly in the liver. Apoferritin is the same protein but with the iron ion (present in the structure of ferritin) removed. This protein is a very large oligomer composed of 24 identical, 174 residue subunits arranged in an octahedral complex, resulting in a total molecular weight of ~450,000 Da. An image of this impressive protein can be seen in figure 2.9. Ferritin is among the most readily crystallized of all proteins, because it crystallizes almost immediately when Cd 2+ ions are added to a solution of this protein. References: the crystallization of ferritin was first published by Laufberger [41]. Physical and chemical properties of the protein can be found in [42]. Fig. 2.9 Ferritin. a) Ribbon model of a single 174 residue subunit of the ferritin protein. Secondary structure almost entirely consists of helical windings. b) Ferritin, composed of 24 of the subunits displayed in (a). The helices of the 11 subunits are depicted as cylinders. Note the globular shape of the oligomer.

16 Insulin Insulin is a hormone from the pancreas. It consists of two chains (A and B) of respectively 21 and 30 amino acids (figure 2.10a). These chains are linked together by two (covalent) disulfide bonds. The protein s molecular weight is ~ 7,300 Da. Insulin molecules have a tendency to form dimers in solution due to hydrogen-bonding between the B chains. One such dimer is depicted in figure 2.10b. In the presence of zinc ions insulin may even form hexamers, as displayed in figure 2.10c. References: the crystallization of insulin was first published by Abel et al. in [43]. Further investigations on crystalline insulin are presented in [44] and [45]. a Fig Insulin. Ribbon model. a) The insulin monomer consists of two chains (A and B) covalently linked together via so called disulfide bonds. b) In aqueous environment insulin tends to form dimers due to hydrogen bonding between the B chains of the monomers. c) In the presence of divalent zinc ions hexamers are formed. The zinc ions are visible in the structure as grayish balls. b c Catalase Catalase is a detoxifying, heme-containing enzyme from the liver. It is responsible for the elimination of hydrogen peroxide without formation of free radicals. Catalase consists of 4 identical subunits of 484 residues and Da each (figure 2.11). References: the crystallization of catalase was first published by Sumner and Dounce [46]. a b Fig Catalase. Ribbon model. a) A single 484 residue subunit of the catalase protein. b) The complex quaternary structure of catalase. The protein consists of a packing of 4 of the subunits displayed in (a). 12

17 Chapter 3 Principles of protein crystallization In this chapter, I will introduce the general (thermodynamic) crystal growth principles and parameters influencing crystallization processes. The crystallization methods described in the next chapter are based on these principles and parameters. For a description of the specific nucleation processes, crystal growth mechanisms and kinetics of proteins I would like to refer to chapter The thermodynamics of crystal growth Crystallization is a complex process, involving multiple equilibria between different states of the crystallizing species [6], [47], [48]. The three stages of crystallization common to all molecules are nucleation, crystal growth and cessation of growth. During nucleation enough molecules associate in three dimensions to form a thermodynamically stable aggregate, the so called critical nucleus. These nuclei provide surfaces suitable for crystal growth, which can occur by a couple of different mechanisms. Crystal growth ceases when the solution is sufficiently depleted of protein molecules, deformation-induced strain destabilizes the lattice, or the growing crystal faces become poisoned by impurities. Both crystal nucleation and growth occur in supersaturated solutions where the concentration of the crystallizing species, in our case protein, exceeds its equilibrium solubility value. The region of solution parameter space suitable for crystallization is generally represented on the phase diagram by the solubility curve (figure 3.1). The supersaturation requirements for nucleation and crystal growth differ. This is shown on the phase diagram where the supersaturation region is further divided into regions of higher supersaturation (the labile region), where both growth and nucleation occur, and lower supersaturation (the metastable region), where only growth is supported. Solubility curve Supersaturation Supersolubility curve Labile region -stable nuclei spontaneously form and grow Metastable region -stable nuclei grow but do not initiate Fig. 3.1 Phase diagram applying to crystal growth. The solubility curve (solid) divides phase space into regions that support crystallization processes (supersaturated solutions) from those where crystals will dissolve (unsaturated solutions). The supersolubility curve (dashed) further divides the supersaturated region into higher supersaturation conditions where nucleation and growth compete (labile phase) and lower levels where only crystal growth will occur (metastable phase). Different kinds of solubility decreasing parameters will be treated in the next section. Adapted from: [47]. Unsaturation [Macromolecule] Now, the strategy employed to bring about crystallization is to guide the system toward a state of reduced solubility (i.e. supersaturation) by modifying the properties of the solvent or the character of the macromolecule (protein crystals are always grown from solutions, because this is the only environment in which protein molecules are stable, that is, remain folded in their native conformation). This can be done by changing one or more of many different parameters, using one of several methods possible. This will be explained in detail in sections 3.2 and

18 Fig. 3.2 Reversible molecular association reactions involved in the assembly of crystals. Monomers initially combine into small aggregates (here, called chains). The association of monomers into chains leads to the formation of prenuclear aggregates that continue to grow by further addition of monomers or chains. When sufficient molecules associate in three dimensions, a thermodynamically stable critical nucleus is formed. The addition of monomers and/or chains to critical nuclei eventually leads to the formation of macroscopic crystals. Source: [6]. Monomers Prenuclear aggregate Critical nucleus Crystal Associated chains The solubility must be decreased very slowly, otherwise amorphous precipitate will form instead of nicely ordered crystals (this is a major problem with macromolecular crystallization, while seldom encountered in the crystallization of conventional molecules). Moreover, care must be taken that the supersaturation level does not become too high. Indeed, a very high supersaturation level increases the chance of obtaining any crystals, but it has some possible deleterious consequences too. Too large numbers of nuclei may be produced, leading to many small crystals or the crystal growth process may become unfavorably rapid, leading to very low quality crystals. The ideal approach would be to induce nuclei at the lowest level of supersaturation which permits their formation, that is, just within the labile region. As these few nuclei begin to grow, the system will gradually enter the metastable region, due to depletion of solute from the solution. In the metastable region the few select crystals continue to develop, while no new nuclei or precipitate can occur. Indicative for the complexity of crystal growth processes is the fact that the formation of a critical nucleus may come about in many different ways. This is, because complex equilibria between many different states of a protein exist in a solution (figure 3.2). Inelastic light scattering investigations have shown that, during the course of crystallization, not only monomers, but also higher-molecular-weight aggregates and clusters of macromolecules with a broad distribution of sizes are present in solution. These may be discrete, ordered, prenuclear aggregates of subcritical size, they may be disordered clusters formed from a variety of nonspecific interactions between macromolecules, linear aggregates on a pathway toward precipitate, or only various dimers and trimers associating transiently through diverse interactions. With time, some of these aggregates may grow and ultimately give rise to crystal nuclei. Others may become ordered on the surfaces of existing crystals and contribute to their growth. Still others may form amorphous precipitates or other energetically favorable but for crystal formation competitive phases. Thus, the pathway from monomeric proteins to crystals is not necessarily a straight one, but generally involves many different processes and equilibria. 14

19 3.2 Parameters influencing protein crystal growth A wide variety of physical, chemical and biochemical parameters affect protein crystallization processes, as can be seen in table 3.1. Of course the physical-chemical characteristics of the crystallizing protein, described in chapter 2, are important, just as the purity of the sample, the properties of the solution used for crystallization (e.g. composition, temperature and ph) and the method of sample handling. In the following, I will shed some light on the role of many (mainly physical and chemical) factors, given a specific protein sample (with a particular purity, sequence, modifications etc.). Some factors will also be shortly commented on in section 4.1. An extensive review of the influence of all factors, including the biochemical ones, can be found in references [48] and [50]. Table 3.1 Crystallization parameters. Physical, chemical and biochemical factors affecting the outcome of a protein crystallization experiment. Adapted from: [48] and [49]. Physical factors Chemical factors Biochemical factors Temperature Precipitant type Sample purity Methodology Precipitant concentration Macromolecular impurities Time ph and buffer Aggregation Pressure Ionic strength Posttranslational modifications Gravity, convection and Reducing/oxidizing environment Sample source sedimentation Vibrations and sound Sample concentration Sample storage Magnetic fields Metal ions Proteolysis Electric fields Detergents Chemical modifications Dielectric properties Small molecule impurities Sequence modifications Viscosity Polyions Sample symmetry Equilibration rate Crosslinkers Sample pi Nucleants Heavy metals Sample history Volume Reagent source Ligands, co-factors, inhibitors Particulate/amorphous material Reagent purity Microbial contamination Surface of crystallization device Reagent formulation Purification method Sample handling ph and temperature As we learned in the previous section the objective in crystallizing a protein is to gradually force the macromolecule from solution by decreasing its solubility (that is, increasing its supersaturation). Many factors can influence protein solubility, which depends on the protein s surface charge (section 2.2). A protein s solubility is usually quite sensitive to ph and to temperature as well in many cases [51], [52]. For many years it was thought that the optimal ph target for crystallization of a protein should be its isoelectric point, pi. This was intuitively appealing because at its pi, a macromolecule carries an equal number of positive and negative charges and is therefore electrostatically neutral. This would seem to be the best situation for mutually attractive electrostatic interactions. In addition most macromolecules do have a pronounced solubility minimum at their pi values and some even precipitate at their pi. Eventually, however, an analysis was conducted based on an early version of the NIST/CARB BMC Data Base, which contains information on the crystallization conditions of many thousands of proteins [53]. The outcome revealed that no correlation between the pi of crystalline proteins and the ph at which they were crystallized existed. The pi even appears inclined toward formation of amorphous precipitate over the crystalline state. Rather, most proteins were successfully crystallized near their physiological ph. Therefore, at present, the seemingly best approach is to crystallize a protein from a solution near its physiological ph. This has as advantage that the risk of denaturation (i.e. the unfolding of the protein under conditions where it is no longer stable) is minimized. 15

20 In section 2.2, it was explained that the temperature dependence of protein solubility is determined by the temperature dependence of the protonation and deprotonation reaction constants of the amino acid side chains in the protein structure [15]. In spite of the fact that most proteins display a clear solubility dependence on temperature, this parameter is not very often used to control supersaturation, but is in the majority of the cases kept constant during the entire experiment. Crystallization has been reported to occur over the entire range from 0 to 40 C and in some cases even 60 C, although it is usually conducted at either 4 C or at room temperature, 25 C. Just as is the case with the ph, extremes in temperature tend to cause denaturation of proteins [1]. Precipitants Protein solubility can also be decreased by changing the composition of the solution, for instance by inclusion of additives such as alcohols, hydrophilic polymers and detergents. These solubilityinfluencing agents are commonly known as precipitants [1], [54]. Protein precipitants fall into four broad categories: salts, organic solvents, (long-chain) polymers and non-volatile organic compounds. In table 3.2 the most common precipitants in each of these categories (the last two have been taken together) are enlisted. Table 3.2 Precipitants used in macromolecular crystallization. Adapted from: [48] and [50]. Salts Organic solvents Polymers Ammonium or sodium sulfate Ethanol Poly(ethylene)glycol 1000, 3350, 6000, Lithium sulfate Isopropanol Jeffamine T Lithium chloride 1,3-propanediol Polyamine Sodium or ammonium citrate 2-methyl-2,4-pentanediol Sodium or potassium phosphate Dioxane Sodium or potassium or Acetone ammonium chloride Sodium or ammonium acetate Butanol Magnesium or calcium sulfate Acetonitrile Cetyltrimethyl ammonium salts Dimethyl sulfoxide Calcium chloride 2,5-hexanediol Ammonium nitrate Sodium formate Methanol 1,3-butyrolactone Poly(ethylene)glycol 400 Salts exert their influence as precipitant according to the salting out principle [55]. The solubility of a protein exponentially decreases as the ionic strength of the solution is increased. The rate of decrease is a function of the particular protein and ions involved. An example of the solubility of lysozyme as a function of varying concentrations of different salts is given in figure 3.3 (page 17). A simple explanation for the salting out effect is that water molecules, otherwise available for solvation of the protein, are monopolized to form bonds with the small ions, thus the salt dehydrates the protein. When the concentration of ions becomes sufficiently high, the proteins are driven to neutralize their surface charges by interacting with one another. This interaction may result in an ordered arrangement of the proteins in a crystalline lattice. The order of salting out effectiveness of (an)ions follows the Hofmeister series [56], SO 4 2- > HPO 4 2- > CH3 COO - > citrate 3- > tartrate 2- > HCO 3 - > CrCO3 - > Cl - > NO3 - > ClO 3 -, for negatively charged proteins, but is reversed for positively charged ones [52], [57]. Divalent and trivalent ions, such as sulfate and phosphate, are most commonly used. Cations are less effective in salting out. A theoretical discussion of the effect of ionic strength on the nucleation of protein crystals is presented in [58]. 16