Addressing the Protein Crystallization Bottleneck By Cocrystallization† Ashwini Warke and Cory Momany* Department of Pharmaceutical and Biomedical Sciences, College of Pharmacy, UniVersity of Georgia, Athens, Georgia 30602
CRYSTAL GROWTH & DESIGN 2007 VOL. 7, NO. 11 2219–2225
ReceiVed July 26, 2007; ReVised Manuscript ReceiVed August 3, 2007
ABSTRACT: The crystallization of macromolecules as an intermediate in the production of protein structures is often the ratelimiting factor. Achievements in the field of protein crystallization such as automation, availability of commercial crystallization kits, the application of developments in nanotechnology, and information mining of databases have significantly improved the throughput of crystallization screening. But this high throughput has not led to high output because of the crystallization bottleneck, and thus, there is a significant disparity between the pace of gene sequencing, protein production, and the determination of the gene product’s atomic structure. Development of complementary approaches that improve the crystallization potential of a protein would form a useful addition to the crystallographer’s toolbox. One such approach is to cocrystallize a target protein with another protein of known structure that specifically recognizes the target so that a more stable protein complex better suited for crystallization is formed. Such proteins are known as cocrystallization proteins (CCPs). In this review, the various mechanisms by which CCPs improve the odds of crystallizing a protein are explained. Additionally, we review some common CCPs such as Fabs (fragment antigen binding) and scFvs (single-chain fragment variable), advances, and current issues in the methodology of their generation and use, and discuss some recently developed novel CCPs such as ankyrin repeat proteins and affibodies. Introduction X-ray crystallography has played a fundamental role in connecting the dots between genomic data and biological function by providing accurate structural information to resolve several significant research problems. Solving protein structures by X-ray crystallography is contingent upon the availability of ordered, diffraction-quality crystals. However, according to the figures obtained from the Structural Genomics Information Portal maintained by the Research Collaboratory for Structural Bioinformatics (RCSB), the success rate of obtaining such highquality protein crystals relative to the total number of “cloned” target ORFs is a disappointing 4.4% (http://targetdb.pdb.org/ statistics/TargetStatistics.html#table1). Huge investments have been made in scaling the machinery of structure determinations (high-intensity X-ray synchrotrons, powerful computer software, and databases), and tremendous strides have been made in the fields of high-throughput cloning, protein production, purification, and automated crystallization screening techniques. Application of nanotechnology to the field has introduced microfluidic devices and biocompatible materials with new properties. Despite these strides, many proteins, in particular ones that have low solubility and easily aggregate, and even a large percent of soluble proteins, still fail to produce good diffraction-quality crystals. The membrane proteins and glycosylated proteins, which form a significant percent of the drug discovery targets, are disproportionately represented in the protein structure data banks. To address this bottleneck and improve the crystallization potential of the target protein, robust parallel approaches need to be developed. One useful approach to addressing the crystallization problem is to cocrystallize the target protein with another protein that specifically recognizes the target so that a stable complex is formed that is better suitable for crystallization.1,2 † Part of the special issue (vol 7, issue 11) on the 11th International Conference on the Crystallization of Biological Macromolecules, Quebec, Canada, August 16–21, 2006 (preconference August 13–16, 2006). * Corresponding author. Address: Room 372 Wilson Pharmacy Building, University of Georgia, Athens, GA 30602. E-mail:
[email protected]. Phone: (706) 542-5370. Fax: (706) 542-5358.
A cocrystallization protein (CCP) is a pre-existent protein molecule that can be modified to draw another protein target with it, leading to the assembly of the two macromolecules into a crystal lattice. Interaction with a CCP modifies the surfaces of the target in ways favorable for crystallization. This is especially true for proteins that contain limited hydrophilic domains, such as membrane proteins, because these polar domains are critical in the formation of important lattice contacts. Binding and subsequent cocrystallization of such a protein with a CCP extends the available crystallization surface so as to improve the probability of obtaining ordered crystals. For example, in the K+ channel-Fab (Fragment antigen binding) complex, the channel’s polar surface was extended by the Fab attached to its extracellular surface. Lattice-forming crystal contacts were formed by the adjacent Fab fragments and not by the four subunits of the K+ channel tetramer, allowing the channel to be suspended in the center in a natural biological conformation.3 Additionally, such CCP lattices provide enough space to accommodate large micelles of detergent solubilized proteins.1 In all of the membrane protein structures solved to date (Table 1), the antibodies have contributed major lattice contacts and filled in the interlattice spaces between the membrane proteins. Cocrystallization of a protein with its cognate CCP can also stabilize the protein by “locking down” flexible domains. For example, in the ternary CD4-gp120-Fab complex, the Fab stabilized the flexible regions of the CD4gp120 complex interface, which was found to be the chemokine receptor binding site that is absolutely essential for HIV-1 entry into cells.4 For several proteins, crystallization is inhibited because of uncontrolled aggregation or biologically significant oligomerization. Proteins such as the viral coat proteins like the HIV-1 capsid protein p24 have oligomerization domains through which they interact with each other to form the viral coat core. This natural feature of the viral capsid protein makes the proteins resistant to crystallization as they tend to aggregate in solution. The cocrystallization approach helps sterically block the protein interaction domains and maintains the protein in a relatively more monodisperse solution.2
10.1021/cg700702c CCC: $37.00 2007 American Chemical Society Published on Web 10/23/2007
2220 Crystal Growth & Design, Vol. 7, No. 11, 2007
Reviews
Table 1. Proteins Crystallized with the Help of a CCP (Cocrystallization Protein) or Where a Recombinant Selection System Was Used To Prepare the CCP protein and PDB ID aminoglycoside Phosphotransferase [2BKK]
CCP
resolution (Å) and comments
selection system
ref
ribosome display
2.15
30
Br3, B-cell surface receptor [2HFG] citrate transporter CitS
designed ankyrin repeat protein (DARPin) inhibitor AR_3A recombinant Fab CB3s various Recombinant Fabs
phage display phage display
52 37
citrate transporter CitS
various DARPins
ribosome display
ClC chloride channel [1OTS]
hybridoma
cytochrome bc1 complex [1EZV] cytochrome bc1 complex with cytochrome c [1KYO] cytochrome c oxidase [1AR1] Gla domain of human factor IX HIV type 1 Reverse Transcriptase [2HMI] HIV-1 capsid protein [1AFV] HIV-1 capsid protein [1E6J] Human Death Receptor 5 [2H9G; 1ZA3]
monoclonal Fab against S.typhimurium ClC monoclonal 18E11Fv monoclonal 18E11Fv monoclonal Fv7E2 monoclonal Fab10C12 monoclonal Fab28 monoclonal Fab25.3 monoclonal Fab13B5 recombinant Fab BDF1 Fab-YSd1
2.61 Fabs isolated, but no structure reported DARPins isolated, but no structure reported 2.51
influenza hemagglutinin (HA) [1QFU] KcsA K+ channel [1K4C; 1K4D]
monoclonal FabHC45 monoclonal Fab
hybridoma hybridoma
70 70 71 72 73 68 74 53 51 75 3
KvaP, a voltage-dependent cation channel [2A0L] KvAP, a voltage-dependent K+ channel [1ORQ; 1ORS]
monoclonal Fv33H1 monoclonal Fab6E1 monoclonal Fab33H1 DARPin off7 DARPin 1108_19 monoclonal Fab 184.1 recombinant protein A binding affibody ZSPA-1 monoclonal Fab2A2 recombinant scFv 1F9 B20–4 Fab G6 Fab
hybridoma hybridoma
2.3 2.97 2.7 successful crystallization 2.8 3.70 3.00 2.32 3.35 2.80 2.0 2.3 3.9 3.2 1.9 2.24 2.54 1.95 2.30 2.70 2.0 3.10 2.8
79 49 80
maltose binding protein (MBP) [1SVX] multidrug exporter, AcrB [2J8S] OspA, Lyme disease antigen [1OSP] protein Z [1LP1] S.aureus protein A domain D [1DEE] turkey egg white lysozyme (TEL) [1DZB] vascular endothelial growth factor [2FJH; 2FJG]
Apart from helping a protein’s crystallization process, a CCP crystal of known molecular structure can be used as a molecular replacement model3 or as a recipient of heavy atom labels to determine molecular phasing information for the target protein–CCP complex.1 Additional uses of CCPs include immunoaffinity purification of the target protein–CCP.5 Furthermore, there are instances in which the cocrystallization approach has not only rescued the structural studies of several interesting but noncrystallizing proteins but also helped in improving poorquality protein crystals and enhancing their X-ray diffraction patterns.6 Table 1 provides a list of proteins crystallized with the help of a CCP. Many more examples exist of macromolecular complexes that crystallized, but this table focuses on cases where the complex was directly used as the means of getting a structure of the target or where the structure represents successful use of applicable technologies to cocrystallization. To be an effective cocrystallization protein (CCP), the CCP should have reasonable solubility and should be easily overexpressed in large amounts. Additionally, the CCP should possess a surface topology capable of recognizing native conformational epitopes of target proteins such that binding does not interfere with the target’s biological activity. The ability of the CCP to form strong and specific noncovalent bonds with the target protein is desired so as to produce stable and dynamically restricted complexes. Figure 1 depicts the popularly used and novel CCP molecules. The space-filled regions demark the protein/antigen binding interfaces of the respective CCPs. Antibody Fragments as Cocrystallization Proteins Antibody structures and their interactions with antigens have been well-studied and reviewed.7,8 The basic IgG antibody
hybridoma hybridoma hybridoma hybridoma hybridoma hybridoma hybridoma phage display
ribosome display ribosome display hybridoma phage display hybridoma phage display phage display
37 69
76 77 63 65 78 33
consists of two light-chain monomers associated with two heavychain monomers through disulfide bonds. Structurally, they are divided into two major domains, a variable domain (VL and VH) that is linked to a constant domain (CL and CH) by a flexible linker or “elbow joint”.9 A potential of six loops, three from the heavy and three from the light chain, interact with antigens and confer the majority of ligand binding potential. These are the “complementarity determining regions” or CDR loops. Substantial diversity of sequences occurs in the CDR loops as a result of antibody gene rearrangements and somatic mutations. The framework segments, which lie between each CDR loop, are well-conserved structurally and define the β-barrel core of the antibody. Hinge region and elbow joints present a high degree of molecular motion within the different domains of an antibody. This high level of flexibility along with the bivalent binding mode of an intact antibody makes the crystallization of an intact antibody problematic. In the search to obtain smaller, rigid, and functional binding units, the antibody molecule is reduced by proteolysis to fragments of decreasing size. Most of the protein structures solved as cocrystallization complexes to date have used Fabs (Fragment antigen binding) generated proteolytically from monoclonal antibodies (Table 1). A Fab includes a single, full light-chain monomer consisting of a variable (VL) and constant domain (CL) and a single monomer of the first two domains of the heavy chain (VH and CH1). Antigen binding is associated with loops projecting from the VL–VH unit. Fv antibody fragments have only the variable regions of both chains. Proteolytically derived Fabs have the disadvantage of being a surprisingly heterogeneous population with multiple isoforms present that can often hinder crystallization attempts, if they are not resolved.2 Besides, there is no
Reviews
Crystal Growth & Design, Vol. 7, No. 11, 2007 2221
Figure 1. Molecules used for cocrystallizations and their binding surfaces. The various proteins commonly used for cocrystallizations are shown as ribbon representations with residues that interact (within 4 Å) with the binding partner shown in space-filling format. (A) Fab from a complex with HIV capsid protein p24 [PDB accession, 1AFV],68 (B) engineered ankyrin repeat from a complex with E. coli maltose binding protein (MBP) [1SVX],63 (C) scFv from a complex with turkey egg white lysozyme [PDB accession, 1DZB].49 and (D) affibody from a complex with Protein Z [PDB accession, 1LP1].33 The figure was prepared using PyMol (http://www.pymol.org).
scope to improve a Fab’s crystallization properties by genetic manipulation of the monoclonal cells. Another serious issue with Fabs derived directly from monoclonal antibodies is that the nature of the immune system does not inherently guarantee selection of antibodies to native conformations of the proteins. For B cells to be activated by T helper cells to produce specific antibodies, surface peptide presentation through the major histocompatibility complex is necessary. At this point, epitopes not associated with the native conformation are possible. Only through careful screening were Fabs derived from monoclonal antibodies found to the native conformation NhaA Na+/H+ antiporter from E. coli.10 Recombinant antibody technology offers a very convenient alternative to obtain versatile and high yields of homogeneous antibody fragments. The format used routinely for cocrystallization of proteins is the scFv (singlechain fragment variable) and that used less so is recombinant Fabs (Table 1). A scFv is a recombinant polypeptide consisting of a single chain composed of an antibody variable light chain (VL) tethered in-frame to a variable heavy chain (VH) by a designed peptide linker sequence.11 These antibody fragments lack a flexible elbow region that could be inhibitory to crystallization. However, use of scFvs may be complicated because of their sequence-dependent tendency to form highermolecular-weight dimers and trimers leading to a polydisperse solution and bivalent binding modes.12 Fabs, on the other hand, are more robust than scFvs or Fvs. Association of the two chains through the constant domains (CL and CH1) enhances the association of the two chains in the variable domains (VL and VH) where antigen binding occurs.13 As can be seen from Figure 1, Fabs have significantly larger surfaces than Fvs or scFvs. Fabs and scFv fragments are the most widely used antibody fragment formats for cocrystallization as compared to other formats such as single-domain antibodies,14,15 diabodies, and Fv fragments.16 An interesting permutation of these approaches is to use protein L as a lattice forming unit that can potentially carry antibody–target complexes into a predefined lattice.17,18 One recent variation of the Fab-based cocrystallization approach is to target surface loops by genetically introducing known sequences into the target protein for which monoclonal antibodies are readily available, such as a purification tag like
the FLAG polypeptide.19 Although the complex of the antiFLAG Fab and the detergent-solubilized K+ channel protein, KvPae, did not result in crystals of the target protein, the complex had improved solubility properties that suggested that the chances for crystallization of the membrane protein were improved. Nonimmunoglobulin Cocrystallization Proteins Since the early 1990s, antibody fragments selected by immune or nonimmune methods have been the first choice for obtaining high-affinity reagents for a wide variety of biomedical and biochemical applications. However, antibody fragments have some limitations to their usefulness. Yields of functional antibody fragments tend to be relatively low, they frequently form polydisperse solutions, and their folding pathways are dependent on accurate disulfide bond formation. These drawbacks have triggered the development of alternate natural and synthetic protein binders. Proteins that have diverse binding potential without an immunoglobulin fold have been identified that mediate immune responses via somatic mutation and selection.20 Thus, it is possible that proteins other than antibodies may be as suitable for molecular display techniques to generate high-affinity, specific binders.21 Ankyrin repeat proteins or ankyrin domains are widely distributed across several species and phyla in nature, mediate important protein–protein interactions, and have been isolated from various cellular environments such as intracellular, membrane-bound, and extracellular.22,23 Structurally, they are formed by repeated, 33 amino acid units composed of a β-turn followed by two antiparallel R-helices and a loop that connects to the turn of the next repeat. Four to six such repeat units stack onto each other, forming a continuous hydrophobic core and a large hydrophilic surface.24 The key to the widespread presence of these protein domains is the variability of amino acid residues that can occur throughout the proteins. Sequence analysis of several such proteins from different sources led to the determination that this variability was more clustered over the β-turn, R-helix, and loop region whereas the remainder formed the structural framework region. This gives these modular domains the ability to easily adapt
2222 Crystal Growth & Design, Vol. 7, No. 11, 2007
Reviews
their molecular surface to bind a large variety of target proteins.25,26 When compared to the CDR loops of antibody fragments, there is a much larger protein surface that can be randomized. In addition, ankyrin repeat proteins are highly soluble, stable, and monomeric and lack disulfide bonds. They can be overexpressed and produced from bacterial cytoplasms at levels in the range of 200 mg L-1. These positive biophysical properties have been incorporated into designed ankyrin repeat proteins (DARPins) composed of fixed framework amino acids to maintain the repeat fold with variation in the residues that form the interacting surface. Varying numbers of this repeat containing mutations at specific sites were assembled between N- and C-terminal capping repeats to yield combinatorial libraries that could be used in RNA display systems.27 Diversity of such libraries is not restricted by transformation efficiency, as is the case with recombinant antibody fragment libraries, and can be simply regulated by adjusting the number of repeat units incorporated into the library. High-affinity binders to several proteins have been isolated from these libraries,28,29 and several new protein crystal structures have resulted from cocrystallization with their cognate DARPins.30 Another nonimmunoglobulin, an in vitro designed class of molecules that has been used to cocrystallize cognate proteins for X-ray crystallography, is called an affibody.31 Out of the five homologous, three helix domains from the immunoglobulin (Fc region) binding region of Staphylococcal protein A (SPA), one domain called the Z-domain was isolated, designed, and randomized to create an affibody library.32 An affibody that binds to protein Z was selected from this library by phage display. Structural studies of this complex revealed that the complex binding interface had properties similar to those of the protein–antibody binding interface.33 Selection Techniques for Various CCP Reagents The production of Fabs for cocrystallization has traditionally used monoclonal hybridoma technology to produce monoclonal antibodies that are then treated with the proteases like papain and purified by anion exchange chromatography to remove the Fc region. Often, the resulting Fab crystallizes as a complex with its cognate protein. Recombinant antibody fragments for production in E. coli are usually made by isolating and cloning the genes from a monoclonal antibody obtained by conventional hybridoma technology.11,34 Generation of monoclonal antibodies by hybridoma technology, at a pace and scale necessary to match structural genomics initiatives, would involve a large investment of time, effort, and cost. Separate immunizations into mice would be required for each antigen, and the outcome (yield) would depend on the immune response of the animal to the target protein. Each antibody and the fragments generated from it would be different, necessitating sequencing and optimization for high recombinant yields and good crystallization properties. McCafferty et al. reported that antibodies could be displayed on the surface of phage and such phage libraries could be used to select highly specific antibodies to the target antigens.35 Unlike the conventional hybridoma technique, antibodies can be selected from a phage display library at a much faster pace independent of the immunized animal’s response. In this system, an antibody fragment library is expressed on the surface of a bacteriophage as fusions with a phage coat protein, which would be the proteins pIII or pVIII in the case of filamentous phage. The phage carries the genetic information encoding the antibody fragment genetically in-frame with its own surface protein. Thus, there is a coupling between the genetic information and the displayed proteins. Using a process termed “panning” illustrated
Figure 2. Selection systems used for obtaining cocrystallization proteins. (A) “Panning” cycle using phage display technology. Phage are shown as blue tubes expressing a CCPFab (yellow cylinders) as a fusion with the gene III product (green). The target protein, brown star, is immobilized on a solid medium. (B) Selection cycle using RNA display technology. The CCP selection protein, made on the ribosome, is shown as a yellow cylinder, whereas the target protein is a brown star. Pur represents a molecule of puromycin that is covalently attached to the 3′ end of the mRNA. When the ribosome encounters the puromycin, the puromycin-mRNA is transferred to the C-terminus of the peptide chain linking the mRNA to the protein. In ribosome display, no puromycin is used. Instead, conditions are controlled to ensure that the ternary complex does not dissociate before or during panning.
in Figure 2A, we incubated the library of phage bearing variants of antibody fragments with a target protein immobilized on a solid surface like a polystyrene plate or a particle like a magnetic bead. Following wash steps to remove phage that do not recognize the target, the bound phage are eluted off the immobilized protein and introduced to E. coli for amplification to produce fresh phage bearing a sublibrary of antibody
Reviews
fragments. These fresh phage are used to initiate a new panning cycle and after 3–5 such rounds, high-affinity antibody fragments to the target protein can be isolated and analyzed.36 Development of the antibody phage display technology to generate recombinant antibody fragments for macromolecular cocrystallization requires the optimization of a few criteria and resolution of some critical issues. First and foremost, the design of the phage display library should be such that all the variants in the library have a uniform template framework that has been optimized for stability, production, and purification.37 To this end, we have generated a nonimmune phagemid recombinant antibody fragment (rFab) library with a nominal diversity of 1.16 × 107 using synthetic approaches.38 Variation of the Fabs in this library is a result of mutations of the third CDR loops of both the light and heavy chains of a Fab engineered and optimized for stability, expression, and yield.39 As much as 15 mg of recombinant Fab can be purified to homogeneity from a liter of shake-flask E. coli culture with this system. Because the library is synthetic, there is no target bias and thus it can be used to select Fabs for membrane proteins or soluble proteins alike. This is a distinct advantage over both immune-derived libraries as well as hybridoma technology. An important issue in the application of this versatile technology for cocrystallization purposes is that of target protein presentation. To separate phage that display antibody fragments that recognize target proteins from phage that do not, it is necessary to immobilize the target protein on a solid support. Traditionally, target molecules have been immobilized onto polystyrene ELISA plates by passive adsorption. This coating technique is troublesome because bound proteins denature on polystyrene surfaces.40 Thus, the protein target is no longer presented in a native conformation. To resolve this issue for targets such as membrane proteins, selection approaches that maintain the target protein’s native structure, such as whole cell panning41 or panning on proteins reconstituted into proteoliposomes,42 have been used. However, these methods suffer from having only extracellular epitopes available and in the case of the whole cell display, have highly complex antigen surfaces. Additionally, the techniques are time-consuming, not easy to automate, and thus are not suitable for high throughput approaches. Several new materials for immobilizing proteins such as paramagnetic beads43 and Ni-NTA HisSorb plates40 have been used in biopanning to maintain protein native conformation for membrane as well as other proteins. Apart from maintaining the target proteins in their native conformation, paramagnetic beads such as Dynabeads from Invitrogen have other advantages such as improved mass-action properties, applicability to 96well formats and automation, a greater surface area for target binding, and availability in a wide variety of chemistries to be able to bind a wide variety of protein targets. The successful integration of the recombinant antibody fragment mediated cocrystallization approach into the structural genomics pipeline relies on the pace and specificity at which this approach can be used to generate antibody fragments and ultimately diffraction quality crystals of noncrystallizing proteins. The standard phage display methodology involves a series of complex steps such as overnight propagation (50–100 mL cultures per protein panned) and phage purification between adjacent rounds of selection that make this process tedious and time-consuming.36 Techniques that scale down this process to formats that can be automated so as to increase throughput are being developed.44,45 A good example is the URSA or ultra rapid selection of antibodies from a phage display library methodology.46 Bypassing the overnight large volume propaga-
Crystal Growth & Design, Vol. 7, No. 11, 2007 2223
tion and laborious phage purification steps, this technique recycles the phage produced by infected E. coli in the first few bacterial extrusions into the next selection round. Using a modified version of this method, we have been able to perform four rounds of selection in 2 days as compared to a previous 2 weeks. Thus, antibody phage display selection strategies can be designed to complement the pace and throughput of structural genomics programs. When coupled with advances in the production of recombinant antibody fragments, the technology is a powerful tool for the high-throughput generation of cocrystallization proteins (CCPs) that can be used to crystallize intransigent proteins. The key features of antibody phage display include simplicity of the technique, cyclic methodology amenable to high throughput, robustness of the phage particle, rapid bacterial growth rate, and the ability to efficiently display antibody fragments as fusions with several phage surface proteins (gene III and VIII products). Despite being such an attractive alternative to the use of hybridoma technology for the generation of antibodies for cocrystallization, only a limited number of antibody fragments selected by phage display have been cocrystallized with their antigen. These examples do not represent the unique application of the cocrystallization approach to get crystals of target proteins, though the method is actively being applied in structural genomics programs.47 Rather, the examples so far demonstrate only the feasibility of using phage-display-generated recombinant antibody fragments. In the case of the complex of turkey egg-white lysozyme with scFV 1F9, small fragile crystals diffracted to 2.2 Å and the structure showed that the correct monomeric species (rather than a undesirable homodimer) of the scFv was present.48 The scFv had a relatively low affinity (Kd ) 520 nM) for the native protein, though it was specific for the turkey lysozome with respect to the hen lysozyme. The phage-display-derived scFv also recognized different epitopes when compared to the crystal structures of other monoclonally produced Fabs complexed with lysozyme.49 The therapeutic drug discovery program at Genentech Inc. has led to several crystal structures of recombinant Fabs that were either identified using phage display selection or affinity matured by phage display. The affinity of humanized Fabs against vascular endothelial growth factor (VEGF) was increased 20-fold using phage display 48–50 The crystals of the improved recombinant Fab were isomorphous with the parent Fab-VEGF crystals, and the X-ray structure showed additional binding interactions relative to the parent Fab. The overall B value of 51.8 Å2 is relatively high, though lower than in the parent. The constant domain dimer (CL–CH1) is largely disordered in one subunit because of a lack of crystal contacts with this domain.50 Human death receptor 5 and BR3 B-cell activating receptor proteins were also crystallized with phage-display-selected Fabs.51–53 Intriguingly, one of these Fabs (YSD1) was selected from a highly restricted artificial Fab library, whereby all three of the CDR loops of the heavy chain were engineered with only the amino acids tyrosine and serine. The Fab-YSd1 had reasonably high affinity (Kd ) 60 nM) for it is target, though the diffraction limit of the resulting crystals was only 3.35 Å.51 Alternative selection systems exist to phage display. In phage display systems, the library encoding DNA has to be introduced to an in vivo bacterial environment. The assembly of the phage and expression of the protein library occurs inside the bacterial cell. Library size is thus restricted by transformation efficiency and poses a disadvantage for isolating high-affinity antibodies. Because the coupling of the phenotype and genotype occurs in vivo, negative selection pressures can affect the outcomes. If
2224 Crystal Growth & Design, Vol. 7, No. 11, 2007
the displayed protein interferes in some fashion with the phage assembly process at any step, that protein’s display might be less efficient and the production of clones that inhibit less would be enhanced. This can quickly lead to clones dominating the whole population after a few selection rounds. Displayed proteins that are toxic for the host or are prone to aggregation or proteolysis might cause slowed bacterial growth, lesser phage production and thus reduced library diversity.54 As an alternative, in vitro display and selection systems have been developed on the basis of immunoglobulin and nonimmunoglobulin oriented CCPs. The most widely used in vitro selection is based on a coupled transcription/translation system termed ribosome display55 or, with subtle modifications of the technique, RNA display (shown in Figure 2B).56 In ribosome display, an mRNA library is synthesized in vitro from a DNA template library encoding the library proteins using T7 RNA polymerase. A protein product is then synthesized by an in vitro translation system using the mRNA template. By removing termination signals in the mRNA and proper introduction of loop structures, the ribosome remains associated with the mRNA after the protein is synthesized. Thus, a ternary complex of mRNA, ribosome, and display protein is allowed to complex with a target molecule forming a quaternary intermediate complex. By previously linking the target to magnetic beads or other support, the complex can be selectively removed from the library and washed to remove nonspecific complexes. The captured mRNA is now reverse transcribed to DNA, which is used as a template to start the next cycle. RNA display differs in that coupling agents such as an RNA-binding protein57 or puromycin 58,59 are used to directly link the mRNA and displayed polypeptide. In practice, the reverse transcribed DNA–mRNA–puromycin– protein complex is the most desirable complex to be captured by the immobilized target, as it has the least competition of binding by the nonprotein components.60 Ribosome display has significant benefits over phage display as library size is limited only by the maximal number of ribosomes that can be provided in a selection round and thus large libraries of 1 × 1013 diversity are possible.61 Additionally, after each selection round, the isolated mRNA undergoes PCR amplification; this leads to introduction of mutations, variation, and sequence diversification at each step, thus making this system ideal for affinity maturation and molecular evolution.62 Ribosome display systems are very effective for display of nonimmunoglobulin CCPs such as ankyrin domains because of their ability to generate large, diverse libraries of designed proteins to mimic the evolutionary antibody diversity. The E. coli maltose binding protein was the first protein crystallized in a complex with a DARPin selected from a library by ribosomal display.63 Additional atomic structures of DARPin complexes quickly followed and include the inhibitor complex of aminoglycoside phosphotransferase30 and two membrane proteins: Na(+)-citrate symporter CitS64 and multidrug transport complex AcrB.65 Although crystal structures of AcrB were obtained without the use of ankyrin repeat units,66,67 the DARPin–AcrB complex crystals diffracted to higher resolution than the noncomplexed proteins and crystallized in a different space group. Critically, no atomic differences between the structures with and without DAPRin binders were observed. The stoichiometry of DARPin binders was unexpectedly two molecules per AcrB homotrimer. This asymmetry of the AcrB structure, emphasized and likely stabilized by the DARPin binders, was observed in the native protein and is biologically relevant.65 A unique comparison of ribosome display with phage display and the effective application of two CCPs, DARPins and Fabs, to the cocrystallization of the membrane protein CitS were presented
Reviews
by Huber et al..64 Both selection methods and both CCPs, as well as the application of appropriate experimental techniques like sandwich ELISAs and size-exclusion chromatography, were demonstrated to be effective for use with membrane proteins in the context of their associated detergents. Affinities of all the CCPs that bound to CitS were similar and in the low nanomolar range. Because the binding surfaces of Fabs and DARPins differ significantly from one another, with the Fabs being more able to bind unstructured epitopes, the two different CCPs will recognize completely different epitopes on the surfaces of their targets. The Fabs offer larger surfaces for potential crystal lattice interactions and project out from the target molecule. But DARPins can be produced in large quantities (>20 mg/L E. coli shake flask culture versus 3 mg/L for the Fab) and are stable in reducing agents. Huber et al.64 concluded that both selection methodologies (phage and ribosome display) and both CCPs (Fabs and DARPins) used together offer the most robust approach to fully explore cocrystallization space. One interesting challenge in the use of these selection systems for cocrystallization confronts industrial users. The various selection systems, whether phage, cell, or RNA display, are heavily patented by companies like Cambridge Antibody Technology (http://www.cambridgeantibody.com, Cambridge, U.K.) and Dyax (http://www.dyax.com, Cambridge, MA). For general research use, such as in academia, this is not a concern. But for pharmaceutical companies solving structures of therapeutic targets with high profit potential, the indirect path to a structure via a recombinant protein selected using patented technology is generally unacceptable, despite the ability to license the technology. Hybridoma technology for producing monoclonal antibodies is now well-established, and the critical patents have expired. Thus, industry will tend to exploit the older technology over the recombinant systems, except in the companies that have established licensing agreements. Conclusions Though the cocrystallization approach for crystallizing proteins that are otherwise difficult to crystallize has been around for quite some time, it has not been readily used because of the low-throughput methods of generating cognate cocrystallization protein reagents. During the last decade, much expertise has been gained in the fields of protein engineering and molecular display systems. Apart from the very popular antibody fragments, other protein scaffolds have been developed and show great promise in their use as CCP reagents. Selection methods for the isolation of high-affinity binders are not restricted to the use of hybridoma technology, and numerous powerful molecular display techniques have been designed for the purpose. Several groups, including ours, are trying to establish the use of molecular display techniques for the high-throughput generation of cocrystallization proteins.37,47,63 With this level of interest, investment, and effort, the recombinant cocrystallization protein technology is well on its way to becoming a powerful tool in a crystallographer’s repertoire.
References (1) Hunte, C.; Michel, H. Curr. Opin. Struct. Biol. 2002, 12, 503–508. (2) Kovari, L. C.; Momany, C.; Rossmann, M. G. Structure 1995, 3, 1291– 1293. (3) Zhou, Y.; Morais-Cabral, J. H.; Kaufman, A.; MacKinnon, R. Nature 2001, 414, 43–48. (4) Kwong, P. D.; Wyatt, R.; Robinson, J.; Sweet, R. W.; Sodroski, J.; Hendrickson, W. A. Nature 1998, 393, 648–659. (5) Kleymann, G.; Ostermeier, C.; Ludwig, B.; Skerra, A.; Michel, H. Biotechnology 1995, 13, 155–160.
Reviews (6) Laver, W. G. In Methods: A Companion to Methods of Enzymology; Birren, B., Lai, E., Eds.; Academic Press: San Diego, CA, 1990; Vol. 1, pp 70–74. (7) Wilson, I. A.; Stanfield, R. L. Curr. Opin. Struct. Biol. 1994, 4, 857– 867. (8) Harris, L. J.; Larson, S. B.; Skaletsky, E.; McPherson, A. Immunol. ReV. 1998, 163, 35–43. (9) Lesk, A. M.; Chothia, C. Nature 1988, 335, 188–190. (10) Padan, E.; Venturi, M.; Michel, H.; Hunte, C. FEBS Lett. 1998, 441, 53–58. (11) Bird, R. E.; Hardman, K. D.; Jacobson, J. W.; Johnson, S.; Kaufman, B. M.; Lee, S. M.; Lee, T.; Pope, S. H.; Riordan, G. S.; Whitlow, M. Science 1988, 242, 423–426. (12) Holliger, P.; Prospero, T.; Winter, G. Proc. Natl. Acad. Sci. U.S.A. 1993, 90, 6444–6448. (13) Röthlisberger, D.; Honegger, A.; Plückthun, A. J. Mol. Biol. 2005, 347, 773–789. (14) Ward, E. S.; Gussow, D.; Griffiths, A. D.; Jones, P. T.; Winter, G. Nature 1989, 341, 544–546. (15) Desmyter, A.; Transue, T. R.; Ghahroudi, M. A.; Thi, M. H.; Poortmans, F.; Hamers, R.; Muyldermans, S.; Wyns, L. Nat. Struct. Biol. 1996, 3, 803–811. (16) Ostermeier, C.; Iwata, S.; Ludwig, B.; Michel, H. Nat. Struct. Biol. 1995, 2, 842–846. (17) Stura, E. A.; Graille, M.; Housden, N. G.; Gore, M. G. Acta Crystallogr., Sect. D 2002, 58, 1744–1748. (18) Stura, E. A.; Taussig, M. J.; Sutton, B. J.; Duquerroy, S.; Bressanelli, S.; Minson, A. C.; Rey, F. A. Acta Crystallogr., Sect. D 2002, 58, 1715–1721. (19) Roosild, T. P.; Castronovo, S.; Choe, S. Acta Crystallogr., Sect. F 2006, 62, 835–839. (20) Pancer, Z.; Amemiya, C. T.; Ehrhardt, G. R.; Ceitlin, J.; Gartland, G. L.; Cooper, M. D. Nature 2004, 430, 174–180. (21) Hosse, R. J.; Rothe, A.; Power, B. E. Protein Sci. 2006, 15, 14–27. (22) Bork, P. Proteins 1993, 17, 363–74. (23) Kobe, B.; Kajava, A. V. Curr. Opin. Struct. Biol. 2001, 11, 725–732. (24) Sedgwick, S. G.; Smerdon, S. J. Trends Biochem. Sci. 1999, 24, 311– 316. (25) Marcotte, E. M.; Pellegrini, M.; Yeates, T. O.; Eisenberg, D. J. Mol. Biol. 1999, 293, 151–160. (26) Kohl, A.; Binz, H. K.; Forrer, P.; Stumpp, M. T.; Plückthun, A.; Grütter, M. G. Proc. Natl. Acad. Sci. U.S.A. 2003, 100, 1700–1705. (27) Binz, H. K.; Stumpp, M. T.; Forrer, P.; Amstutz, P.; Plückthun, A. J. Mol. Biol. 2003, 332, 489–503. (28) Amstutz, P.; Binz, H. K.; Parizek, P.; Stumpp, M. T.; Kohl, A.; Grütter, M. G.; Forrer, P.; Plückthun, A. J. Biol. Chem. 2005, 280, 24715– 24722. (29) Amstutz, P.; Koch, H.; Binz, H. K.; Deuber, S. A.; Plückthun, A. Protein Eng. Des. Sel. 2006, 19, 219–29. (30) Kohl, A.; Amstutz, P.; Parizek, P.; Binz, H. K.; Briand, C.; Capitani, G.; Forrer, P.; Plückthun, A.; Grütter, M. G. Structure 2005, 13, 1131– 1141. (31) Nilsson, B.; Moks, T.; Jansson, B.; Abrahmsen, L.; Elmblad, A.; Holmgren, E.; Henrichson, C.; Jones, T. A.; Uhlen, M. Protein Eng. 1987, 1, 107–113. (32) Nord, K.; Nilsson, J.; Nilsson, B.; Uhlen, M.; Nygren, P. A. Protein Eng. 1995, 8, 601–608. (33) Högbom, M.; Eklund, M.; Nygren, P.-A.; Nordlund, P. Proc. Natl. Acad. Sci. U.S.A. 2003, 100, 3191–3196. (34) Huston, J. S.; Levinson, D.; Mudgett-Hunter, M.; Tai, M.-S.; Novotny, J.; Margolies, M. N.; Ridge, R. J.; Bruccoleri, R. E.; Haber, E.; Crea, R.; Oppermann, H. Proc. Natl. Acad. Sci. U.S.A. 1988, 85, 5879–5883. (35) McCafferty, J.; Griffiths, A. D.; Winter, G.; Chiswell, D. J. Nature 1990, 348, 552–554. (36) Phage Display: A Laboratory Manual; Barbas, C. F., Ed.; Cold Spring Harbor Laboratory Press: Cold Spring Harbor, NY, 2001. (37) Röthlisberger, D.; Pos, K. M.; Plückthun, A. FEBS Lett. 2004, 564, 340–348. (38) Kelley, L. L.; Momany, C. Biotechniques 2003, 35, 750–756. (39) Nadkarni, A.; Kelley, L. L.; Momany, C. Protein Expression Purif. 2007, 52, 219–229. (40) Padan, E.; Venturi, M.; Michel, H.; Hunte, C. Qiagen News 1999, 17–20. (41) Shadidi, M.; Sioud, M. Biochem. Biophys. Res. Commun. 2001, 280, 548–552. (42) Mirzabekov, T.; Kontos, H.; Farzan, M.; Marasco, W.; Sodroski, J. Nat. Biotechnol. 2000, 18, 649–654. (43) McConnell, S. J.; Dinh, T.; Le, M. H.; Spinella, D. G. Biotechniques 1999, 26, 208210, 214.
Crystal Growth & Design, Vol. 7, No. 11, 2007 2225 (44) Vanhercke, T.; Ampe, C.; Tirry, L.; Denolf, P. J. Biomol. Screening 2005, 10, 108–117. (45) Walter, G.; Konthur, Z.; Lehrach, H. Comb. Chem. High Throughput Screening 2001, 4, 193–205. (46) Hogan, S.; Rookey, K.; Ladner, R. Biotechniques 2005, 38, 536–538. (47) Shea, C.; Bloedorn, L.; Sullivan, M. A. J. Struct. Funct. Genomics 2005, 6, 171–175. (48) Küttner, G.; Keitel, T.; Gieimann, E.; Wessner, H.; Scholz, C.; Höhne, W. Mol. Immunol. 1998, 35, 189–194. (49) Ay, J.; Keitel, T.; Küttner, G.; Wessner, H.; Scholz, C.; Hahn, M.; Höhne, W. J. Mol. Biol. 2000, 301, 239–246. (50) Chen, Y.; Wiesmann, C.; Fuh, G.; Li, B.; Christinger, H. W.; McKay, P.; de Vos, A. M.; Lowman, H. B. J. Mol. Biol. 1999, 293, 865–881. (51) Fellouse, F. A.; Li, B.; Compaan, D. M.; Peden, A. A.; Hymowitz, S. G.; Sidhu, S. S. J. Mol. Biol. 2005, 348, 1153–1162. (52) Lee, C. V.; Hymowitz, S. G.; Wallweber, H. J.; Gordon, N. C.; Billeci, K. L.; Tsai, S. P.; Compaan, D. M.; Yin, J.; Gong, Q.; Kelley, R. F.; DeForge, L. E.; Martin, F.; Starovasnik, M. A.; Fuh, G. Blood 2006, 108, 3103–3111. (53) Li, B.; Russell, S. J.; Compaan, D. M.; Totpal, K.; Marsters, S. A.; Ashkenazi, A.; Cochran, A. G.; Hymowitz, S. G.; Sidhu, S. S. J. Mol. Biol. 2006, 361, 522–536. (54) Ling, M. M. Comb. Chem. High Throughput Screening 2003, 6, 421– 432. (55) Hanes, J.; Plückthun, A. Proc. Natl. Acad. Sci. U. S. A. 1997, 94, 4937–4942. (56) Wilson, D. S.; Keefe, A. D.; Szostak, J. W. Proc. Natl. Acad. Sci. U.S.A. 2001, 98, 3750–3755. (57) Sawata, S. Y.; Suyama, E.; Taira, K. Protein Eng. Des. Sel. 2004, 17, 501–508. (58) Roberts, R. W. Curr. Opin. Chem. Biol. 1999, 3, 268–273. (59) Kurz, M.; Gu, K.; Lohse, P. A. Nucleic Acids Res. 2000, 28, E83. (60) Gold, L. Proc. Natl. Acad. Sci. U.S.A. 2001, 98, 48254826.. (61) Dufner, P.; Jermutus, L.; Minter, R. R. Trends Biotechnol. 2006. (62) Hanes, J.; Jermutus, L.; Weber-Bornhauser, S.; Bosshard, H. R.; Plückthun, A. Proc. Natl. Acad. Sci. U.S.A. 1998, 95, 141305.. (63) Binz, H. K.; Amstutz, P.; Kohl, A.; Stumpp, M. T.; Briand, C.; Forrer, P.; Grütter, M. G.; Plückthun, A. Nat. Biotechnol. 2004, 22, 575–82. (64) Huber, T.; Steiner, D.; Röthlisberger, D.; Plückthun, A. J. Struct. Biol. 2007, 159, 206–21. (65) Sennhauser, G.; Amstutz, P.; Briand, C.; Storchenegger, O.; Grütter, M. G. PLoS Biol. 2006, 5, e7. (66) Murakami, S.; Nakashima, R.; Yamashita, E.; Matsumoto, T.; Yamaguchi, A. Nature 2006, 443, 173–9. (67) Seeger, M. A.; Schiefner, A.; Eicher, T.; Verrey, F.; Diederichs, K.; Pos, K. M. Science 2006, 313, 1295–8. (68) Momany, C.; Kovari, L. C.; Prongay, A. J.; Keller, W.; Gitti, R. K.; Lee, B. M.; Gorbalenya, A. E.; Tong, L.; McClure, J.; Ehrlich, L. S.; Summers, M. F.; Carter, C.; Rossmann, M. G. Nat. Struct. Biol. 1996, 3, 763–70. (69) Dutzler, R.; Campbell, E. B.; MacKinnon, R. Science 2003, 300, 108– 12. (70) Hunte, C.; Koepke, J.; Lange, C.; Rossmanith, T.; Michel, H. Structure 2000, 8, 669–84. (71) Ostermeier, C.; Harrenga, A.; Ermler, U.; Michel, H. Proc. Natl. Acad. Sci. U.S.A. 1997, 94, 1054710553.. (72) Xiaoli, S.; Yongdong, L.; Chuanbing, B.; Gengxiang, Z.; Mingdong, H. Acta Crystallogr., Sect. D 2005, 61, 701–703. (73) Ding, J.; Das, K.; Hsiou, Y.; Sarafianos, S. G.; Clark, J. A. D.; JacoboMolina, A.; Tantillo, C.; Hughes, S. H.; Arnold, E. J. Mol. Biol. 1998, 284, 1095–1111. (74) Berthet-Colominas, C.; Monaco, S.; Novelli, A.; Sibai, G.; Mallet, F.; Cusack, S. EMBO J. 1999, 18, 1124–36. (75) Fleury, D.; Barrere, B.; Bizebard, T.; Daniels, R. S.; Skehel, J. J.; Knossow, M. Nat. Struct. Biol. 1999, 6, 530–4. (76) Lee, S.-Y.; Lee, A.; Chen, J.; MacKinnon, R. Proc. Natl. Acad. Sci. U.S.A. 2005, 102, 1544115446.. (77) Jiang, Y.; Lee, A.; Chen, J.; Ruta, V.; Cadene, M.; Chait, B. T.; MacKinnon, R. Nature 2003, 423, 33–41. (78) Li, H.; Dunn, J. J.; Luft, B. J.; Lawson, C. L. Proc. Natl. Acad. Sci U.S.A. 1997, 94, 35843589.. (79) Graille, M.; Stura, E. A.; Corper, A. L.; Sutton, B. J.; Taussig, M. J.; Charbonnier, J.-B.; Silverman, G. J. Proc. Natl. Acad. Sci. U.S.A. 2000, 97, 53995404.. (80) Fuh, G.; Wu, P.; Liang, W.-C.; Ultsch, M.; Lee, C. V.; Moffat, B.; Wiesmann, C. J. Biol. Chem. 2006, 281, 6625–6631.
CG700702C
