Antibody-structure-based design of pharmacological agents

372 reviews Antibody-structure-based design of pharmacological agents William C. Dougall, Norman C. Peterson and Mark I. Greene The well-studied antigen-combining sites of antibody molecules hold considerable promise as a model system for the design of bioactive peptides. These small, immunoglobulin-derived peptides can be used in the development of alternative treatments for disease and in diagnostic strategies. The general principles derived from the design of small pharmacological agents based on the structural features of antibodies may also be extended to the design of other bioactive peptides. The fine specificity and affinity of antigen recognition and binding by antibody molecules is understood to be conferred by amino acid residues within the hypervariable regions of the immunoglobulin (Ig) protein (see Glossary). Significant advances have recently been made in the elucidation of the antigen-combining sites of antibodies through the determination of the threedimensional structure of antibody-antigen complexes and other experimental methods, including mutational analysis of the antibody-combining site. An understanding of these structures is necessary for the rational design of diagnostic reagents and drugs based on the activities of antibodies. The advent of monoclonal antibody (mAb) technology I showed that Igs have a wide range of biological activities and specificities, and indicated the tremendous potential of these molecules in the treatment of disease and/or in diagnosis. The use of the large Ig protein to obtain desired binding or biological effects in the clinic has proved difficult, largely as a result of technical problems that include a host antiantibody immune response. To overcome this problem, 'humanized' antibodies have been synthesized by inserting the rodent hypervariable regions into human framework regions (see Glossary) (Ref. 2). Mternatively, other researchers have generated mAbs by selecting antibodies from human recombinatorial phage-display libraries 3. Despite these modifications, the use of the Ig protein as a clinical tool has remained problematic, largely because of its extremely large size (-150 kDa). In addition, antibody molecules show minimal penetration of poorly vascularized tissues, including tumors 4-6. Immunoglobulin domains, particularly the constant region, outside the antigen-binding sites have detrimental effects on the specificity of antibody binding by mediating binding to membrane Fc receptors on monocytes, macrophages and natural killer T cells present at inflammatory sites7. These unfavorable properties of whole Ig molecules have limited their applications in the development of drugs and their use as diagnostics. The application of a reductionist approach to protein design and structural biology has enabled the minimal functional domains of various proteins to be determined. This strategy should be particularly useful for the design of antibody-derived small molecules, with the goal of generating stable pharmacological agents that lack the disadvantages of intact Igs, but yet retain the critical binding characteristics and biological activities of the original antibody. The synthesis and use of smaller functional subunits derived from the structure of antibodies would also enable simple modifications and structural analysis to be carried out, which would accelerate drug development. In this article, we shall not endeavor to review comprehensively the literature on antibody-antigen interactions, as they have been the subject of two recent reviews8,9, but rather we will discuss this structural information and other relevant data that relate to the question: what makes up the smallest functional unit of an antibody? For the purposes of the discussion, we will define antibody function as that directly related to antigen binding and the resultant biological effects, without the contribution of immunological effector functions, including complement- or cell-mediated cytotoxic mechanisms. G e n e r a l structure o f t h e a n t i g e n - b i n d i n g sites W. C. DougalI, N. C. Petersonand M. L Greeneare at the Division of Immunology and Centerfor ReceptorBiology, Department of Pathology and LaboratoryMedicine, Universityof Pennsylvania School of Medicine, 252John Morgan Building, 36th and Hamilton Walk, Philade~hia, PA 19104, USA. TIBTECHSEPTEMBER1994 (VOL 12) Variable region Primary-sequence comparisons and computer modeling have been used in conjunction with crystallographic data to elucidate the structure of the immunoglobulin domains. Sequence analysis of the Ig © 1994, Elsevier Science Ltd 373 reviews Glossary Antigen-combining site - The portion of an antibody (composed of light and heavy chain hypervariable determinants) that binds to antigen. region (CDR} - The portion of the antibody variable domain that binds to antigen. It is usually composed of hypervariable sequences, three from each immunoglobin (Ig) chain (synonymous with antigen-combining site), Complementarity-determining Constant domains - The conserved domains of the Ig heavy chain (CH1, CH2, CH3), and the light chain CL1; typically this region is located at the carboxyl-ends of the two protein chains. Fab fragment (fragment antigen binding) - A proteolytic fragment of the Ig molecule resulting from digestion with the enzyme papain. The Fab fragment comprises the VH1-CH1 domain of the heavy chain and the VLI-CL1 domain of the light chain, joined by a single interchain disulfide bond. The fragment is monovalent with respect to antigen binding (see Fig. 2). F v fragment - Immunoglobulin fragment comprising only the VL1 and VH1 domains (see Fig. 2). Framework region - The portion of the antibody variable region that is less divergent in amino acid sequence. Frame- work region sequences usually comprise IB-pleated sheets, which help to stabilize the overall structure of the variable region. I-lypervariable region - The portion of the antibody-variable domain that is most divergent in terms of amino acid sequence. Immunoglobulin (Ig} - The immunoglobulin molecule, composed of two light and two heavy protein chains, comprising the antibody. The light chain is - 25 kDa, and the heavy chain is -50 kDa. The term Ig refers to the immunity-conferring portion of the gamma-globulin fraction. Mimetics - Small organic compounds that resemble the structure of an antibody molecule, and which 'mimic' the bind- ing and biological activities of that antibody; also called 'CDR mimetics'. Monoclonal antibody {mAb) - Antibodies produced by hybridoma cell lines derived from the fusion of an immortal- ized myeloma cell and a normal antibody-producing B cell. The hybridoma secretes a single antibody population. Single chain Fv {scFv} - Immunoglobulin fragment comprising only the VL1 and VH1 domains connected by a short peptide linker (see Fig. 2). Variable heavy domain (VH} - The portion of the heavy chain that comprises part of the variable (or divergent in amino acid sequence) region. This domain constitutes one half of the variable region. Variable light domain (VL} - The portion of the variable domain encoded by. the light chain Ig. polypeptide indicates that the amino acid sequence is very diverse at the amino-termini of both the heavy (H) and the light (L) chains, and is highly conserved within the carboxyl-portion of both chains. The amino acid sequences of the amino-terminal portions of the molecule are referred to as the 'variable' (V H and VL) regions (see Glossary) in order to distinguish them from the conserved (or 'constant') regions found at the carboxyl-ends. Within the variable region, the V H and Vt chains are composed of alternating stretches of moderately conserved and highly variable residues, referred to as the framework regions (FRs) and the complementarity determining regions (CDRs), respectively (see Glossary) 1°. These two domains comprise two distinct functionalities within the variable regions; the CDRs provide the important determinants necessary for antigen binding, and the laR residues supply structural constraints for the entire variable region (Fig. 1). The V c and V H domains have homologous tertiary structures. Each dOmain is composed of two [3-pleated sheets formed by four FRs that are interconnected by three CDRs. The CDRs loop out from the FRs to form highly exposed surfaces, which enable them to interact with epitopes. The C D R loops assume struc- tures from a limited set of conformations, referred to as 'canonical structures '11. The canonical structure of many of the CDRs of an Ig molecule can be predicted from specific conserved residues located within the CDRs and, occasionally, including adjacent F R residues 11,12. In general, the canonical CDRs, with the exception of the heavy chain C D R H 3 , have reverseturn conformations, which can sometimes have the features of [3-turns. As a result of the complex genetic mechanisms that control the repertoire of long- or medium-length loops for the third C D R (CDRH3), the patterns of antigen interaction mediated by this C D R are equally diverse. The ability to produce smaller functional antibody fragments was first realized in the late 1950s when Porter isolated Fab and Fc fragments (see Glossary) from proteolytically cleaved rabbit gamma globulins 13. The Fab fragments comprise heterodimers of V H- C H1 and Vc-Ccl domains, covalently linked by disulfide residues in the constant domains (Fig. 2). If the loss in binding affinity attributed to bivalency is disregarded, Fabs still bind antigen with reasonable affinities. Fv fragments (see Glossary) are produced by further proteolytic digestion of the Ig molecule and, more recently, by recombinant techniques. The Fv is -50% TIBTECHSEPTEMBER1994 (VOL 12) 374 reviews Figure 1 Computer-generated predicted model of the Fv variable region of an anti-p185/ c-erbB-2 antibody molecule (7.16.4) (provided by Pedro Alzari). The heavy chain is at the top (brown) and the light chain is at the bottom (magenta). The complementarity determining regions (CDRs)of both the heavy and light chain are also shown (white, green and blue). The amino acid sequences of the CDRs are hypervariable and the CDRs themselves project from the surface of the antibody in order to interact with antigen. The antigen-combiningsite can be composed of multipleCDR determinants from both heavyand light chains. The framework regions of each chain are more highly conserved within the entire Fv variable region and often form 13-pleated sheets, which are important for the overall structure of the Fv region. smaller than the Fab and is composed ofnoncovalently associated V H and V L domains (Fig. 2). As a result of their smaller size, antigen-Fab (or Fv) complexes can be crystallized and thus have contributed much to our understanding of antibody structure and antigenantibody interactions. Specific antibody structures important for antigen binding Recent high-resolution crystal structures ofFv fragments'.and Fab-antigen complexes have revealed general features of the complex that will be useful in the design of antibody-derived small molecules 8,9. The size, shape and chemical composition of the interactive surfaces between antibody and antigen give some indication as to the minimal structure required for antibody binding. Topographically, these antibody interfaces (paratopes) are generally flat or concave for larger protein antigens, and comprise more of a defined 'binding cleft' for smaller antigens and haptens. Upon closer inspection, these flat surfaces are not uniform, with protuberances and depressions formed TIBTECHSEPTEMBER1994 (VOL 12) by C D R residues defining the chemical- and shapecomplementarity with the antigen. The antibody surface buried upon complex formation ranges in size from 150 A 2 to over 900 A 2, with 700 A 2considered to be average 14. However, relative to the protein-protein interactions within the core of various proteins, the antibody-antigen interfaces may not be as tightly packed as was initially thought. High-resolution (1.8 A) analysis of free and lysozyme-bound antibody Fv D1.3 indicates that the interface uses a network of water molecules to complete the antigen-antibody complementarity 15. Water in the interface forms hydrogen-bonded networks, which play a critical role in increasing the packing density and decreasing the conformational entropy of antibody residues. Water molecules have also been identified at the interface of the NC41-neuraminidase complex 16, although the contribution of water to chemical complementarity was not defined. The relatively large size of the interactive surface suggests that multiple noncovalent interactions are required for 'adequate' binding energies and, thus, the interaction of antibody-derived small molecules with antigen would be thermodynamically unfavorable. However, each C D R is not equally important in any given antibody; analyses of the relative contributions of the C D R loops have indicated a predominant role for heavy-chain determinants (in particular, CDRH3) in the number of specific contacts with antigen 17. Wilson and Stanfield 9 compiled a database of known antibody-antigen complexes and have reported that the minimum number of hypervariable loops used in terms of atomic contacts in antigen binding is four. However, closer examination of the three-dimensional structures of antibody-antigen complexes illustrates that there are specific examples in which a hmited number of C D R determinants predominate in chemical bonding (or the loss of exposed surface area) on binding to antigen TM. These examples include C D R H 3 in Fab D1.3-HEL (Ref. 19) and C D R H 3 of Fab 26-10-digoxin 2°. In order to address the individual contributions of antibody residues and C D R determinants for antigen binding, the specific thermodynamics of antigen recognition conferred by the structure of the antibody-combining site must be taken into account. The specificity and affinity of antigen-antibody interactions exclude covalent interactions and primarily involve 'local' forces such as hydrogen-bonding, van der Waals forces and charge-charge interactions. It is extremely diflScult to dissect the energies contributed by these different interactions because of the complex nature of the interface, as described above. Titration calorimetry can be used to determine the overall enthalpic and entropic changes; however, these experimental methods only give the sum of energetic changes conferred by antibody-antigen binding. Estimates of the Gibbs free-energy changes in the formation of antigen-antibody complexes have been made using systematic calculations based on known atomic coordinates of antibody-antigen complexes 21. a v /f CH1 375 reviews b CH1 ~ * cN2ii CH 3 Immunoglobulin I CDR peptides I Peptide mimetic I Figure2 Antibody-derived small molecules. (a) Conventional diagram of an immunoglobulin molecule. The arrows show the breakdown of the molecule into its smaller derivatives, as discussed in the text. (b) Fab molecule produced by papain digestion of an immunoglobulin. (c) Fv fragment comprising noncovalently associated variable domains of the heavy and light chains of an immunoglobulin. (d) scFv: single Chain Fv molecule, with the VHand VL regions joined by a peptide linker. (e) Complementarity-determining region (CDR) peptides, which are synthetic peptides of antibody-binding determinants. (f) Peptide mimetic. This is a conformationally restricted organic ring that mimics the structure of a CDR loop and preserves the antigen-interactive side-chains. These theoretical calculations have indicated that the heavy chain and, specifically, only a few discrete residues within each antibody-combining site contribute significantly to binding energies. Favorable free-energy changes involved a subset of the total paratope for the antibodies D1.3, McPc 603 and Hy Hel-5. One single amino acid (H52) within McPc 603 and a discrete, contiguous three amino acid segment (H99-H101) of D1.3 were thought to provide the majority of the binding energies for their respective antibodies. In the case of D1.3, this observation has been confirmed by mutational analysis of residues H99-H101 (Ref. 22). The other CDtL residues contributed less significantly to binding energies, and probably play a more indirect role by forming a complementarity shape between the molecular surfaces of the antibody and antigen21, or by forming electrostatic fields23. These observations suggest that individual subdomains of antibodies are predominantly inVolved in antigen binding and also raise the possibility that antibody-derived small molecules can retain functionality. In order to apply this information to the rational design of such molecules, the individual contribution of CDR.s to the binding energy must be viewed quantitatively on a case-by-case basis. Another important criterion in identifying candidate antibody determinants for the development of antibody-derived small molecules is the ability to assess the surface accessibility of the combining-site determinants. In the absence of the three-dimensional structure, computer modeling can be used to predict surface-accessible hypervariable determinants. Clearly, isolated antibody fragments that are composed of moieties that are normally accessible will be more likely to retain binding activity. However, small molecules removed from the structural context of the whole antibody will also present pharmacological information that may have been buried by other hypervariable structures, potentially resulting in unexpected or altered activities. This is pertinent as small CDR-derived peptides require additional FR residues to adopt the correct conformation for binding (see below). Recombinant manipulations of Ig sequence to produce smaller antibody molecules must not perturb the canonical structure, if any semblance to the parent structure or function is to be maintained. Generic FR residues, which are TIBTECHSEPTEMBER1994(VOL12) 376 reviews themselves chemically inert, can be used to help restrict the mobility of the pharmacologically active C D R side-chains in smaller peptides. The synthesis of small molecules based on antibody stucture has indicated that the orientation of the side-chain functional groups is more important than backbone geometry for biological specificity24, so there is probably some flexibility in the choice of FR residues for small molecules. Recombinant antibody fragments Recent progress in Ig-gene cloning strategies and the adaptation of bacterial expression systems25,26 has provided additional information on the structure of Ig, and has presented new opportunities to test the function of smaller antibody domains. Information obtained from studies of antigen-antibody interactions, crystal structures and computer modeling provides guidelines for the construction of recombinant antibody fragments. The design of Ig-genespecific polymerase chain reaction (PCR) primers has been instrumental in the cloning of specific V region genes and the production of Vii and V L libraries3, 27. Although Fvs are significantly smaller than Fab fragments, their potential clinical application has been limited by a strong tendency for the V H and Vt chains to dissociate because they are noncovalently associated. This drawback has been largely overcome by the design of a flexible peptide linker placed between the carboxyl terminus of the V H region and the amino terminus of the V L to produce a 'single chain Fv' (scFv) (Ref. 4) (Fig. 2). Antigen-binding affinities within an order magnitude of their related Fabs have been achieved with several scFvs (Refs 28-32). In addition, in vivo studies with xenografted athymic mice have demonstrated that a variety of antitumor scFvs have faster blood-clearance rates, improved tumor-to-normal tissue ratios and better tumor penetrance when compared with Fabs (Refs 5, 6, 32). The VII and V L cDNAs, which serve as templates for the production of scFvs, have generally been derived from specific antibody-producing hybridomas. However, this technology can be used to produce expression libraries from which clones that produce antibody fragments with a desired specific affinity can be selected. Marks et al. 33 demonstrated that Fv fragments with high binding affinities for lysozyme and phenyloxazol (107 M-1 and 2 × 106 M-1, respectively) could be isolated from a combinatorial Fv library derived from nonimmunized human peripheral blood c.ells. In addition, Ward et al. 34 screened a VII expression library and characterized two different antilysozyme V H domains, which have reported affinities in the 20 nM range. Indeed, naturally occurringfunctional heaw-chain Igs (devoid of light chains) have been discovered in camels35, indicating that smaller components of an Ig can function independently. These results indicate that screening combinatorial and single-domain libraries for specific antigen binding may be a potentially effective means of generating antigen-binding proteins of 70 kDa or smaller. IBTECHSEPTEMBER1994 (VOL12) Pessi et al. 36 relied heavily on structural modeling to produce a 'minibody' that bound nickel with an affinity of ~106M-1 Their construct consisted of three [3-strands, CDI~H1 and C D R H 2 of an antibody V H domain. The stabilizing residues of the minibody were conserved, and hydrophobic residues that normally interact with the V L chain were replaced. In order to bind nickel, three exposed residues, which were in the proper orientation and proximity for optimal binding, were replaced with histidine residues. The success of this approach demonstrates the potential for the design of other minibodies with binding activities directed against more-complex epitopes. Development of antibody-derived small molecules CDR peptides As previously discussed, several well-studied antigen-antibody associations appear to involve a limited number of interactions between the residues in the Ig CDRs and the epitope. This implies that smaller antigen-binding peptides can be designed by isolating these residues in a context that conserves their orientation with respect to the binding site. Information acquired from structural analysis of antigen-antibody complexes and computer modeling is helpful in determining potential sites of interaction. One approach is to design constrained peptide analogs based on antibody structure, and to screen these for similar bioactive effects to those seen with antireceptor mAbs (Ref. 37) (Fig. 2). Once the sequence and structure of antireceptor antibodies have been determined, the CDRs can be identified and peptide analogs can be generated. The synthesis of overlapping peptides deduced from the C D R sequence enables the residues that are important for activity to be determined, with the first level ofscreemng typically involving the inhibition of the parental mAb binding (Fig. 3). This type of strategy has been used to develop small peptides that are capable of antigen binding. Welling and colleagues designed a 13-residue peptide analog of the C D R H 2 (Re£ 38) and a 10-residue peptide representing the CDRL3 (Re£ 39) of an anti-lysozyme mAb. W h e n these peptides were immobilized on a sepharose column, antigen binding was demonstrated by the successful purification of lysozyme by immunoafffinity chromatography. Levi et al. 4° synthesized and analyzed six peptides, which included each of the CDRs of an anti-HIV-1 antibody, to determine activity against the viral envelope protein. They found that C D R H 3 was most effective in inhibiting binding of the parental mAb, and in preventing syncitia formation in infected cells. In addition, when peptides derived from the other CDRs were individually combined with the C D R H 3 peptide, a synergistic inhibition of antibody binding was observed. These results indicate that the CDRs of antibodies whose binding activity results in a physiological effect may serve as potential models for the design ofpeptides or other pharmaceuticals that retain similar effects. 377 reviews Alternatively, CDR-peptide sequences can be compared with protein sequence databases to identify potential motifs that are commonly involved in receptor-ligand interactions 41. This concept of hgand mimicry was initially used by Williams and colleagues24 to develop biologically active C D R peptides. The prototype model system has been the mAb 87.92.6, which was developed to bind the type 3 reovirus receptor on cells, and competitively inhibits the binding of this virus to these cells42. Monoclonal antibody 87.92.6 also mimics the biological effects of the reovirus (i.e. inhibition of D N A synthesis in fibroblasts, neuronal cells and lymphocytes), suggesting that a structure in the mAb is similar to the natural ligand for reovirus type 3 receptors. The deduced amino acid sequence of a combined determinant of C D R L 2 and C D R H 2 of mAb 87.92.6 showed similarity with the viral hemagluttinin antigen43, and synthetic peptides derived from the C D R L 2 (17 residues) were able to inhibit the binding of both mAb 87.92.6 and reovirus type 3 to the cellular receptors 44. Further examples of biological activities exhibited by peptides derived from antibody structure have been reported. A 21-residue C D R peptide (which included a sequence of three amino acids within the C D R H 3 domain of an anti-platelet receptor antibody, which was similar to the fibrinogen-binding site of the platelet receptor) effectively inhibited binding of the parental mAb and fibrinogen to the platelet receptor 4s. In another study41, a CDRL1 and a CDR.H2 of two different anti-idiotypic antibodies to thyroid stimulating hormone (TSH) were discovered to contain peptide sequences that were similar to the ot and [3 chains of TSH. These peptides were constructed and shown to inhibit binding of the parental antibody, and also to mimic TSH by increasing thymocyte cAMP concentrations. One general feature of these peptide studies is the known or predicted reverse-turn conformation of the active antibody (or ligand) structure. As a peptide in aqueous solution is likely to adopt many different conformations, a constrained structure similar to that in the intact Ig (such as a reverse turn) will, potentially, have enhanced activity. Peptides can be cyclized by incorporation of cysteine residues and oxidation (or incorporation of organic linkers) in order to mimic more accurately the secondary structure. In the reovirus type 3-mAb 87.92.6 system, cyclic peptides were found to have significantly higher (40-fold) binding affinities than their linear counterparts 24. Further substitution of individual amino acids, coupled with computer-aided modeling, defined the residues that are critical for assuming the necessary conformation for receptor binding and resulted in a peptide derived from mAb 87.92.6 that not only inhibited mAb-receptor interactions, but also retained the biological activity of the mAb (i.e. inhibition of conA-induced lymphocyte mitogenesis). Taken together, these studies indicate that it is possible to develop receptor-binding peptides based on the structure of antibody CDP,.s. Amino acid sequence of antibody Structural data of antigen-antibody interaction / 1 Amino acid sequence motif analysis Computer modelling / CDR peptide synthesis Peptide cyclization Inhibition of parental antibody binding assays (Ability to mimic parental antibody) Organic cyclic mimetic Figure 3 Flow diagram depicting the overall design strategy for antibody-derived peptides and organic mimetics. The development of this design strategy necessitates a knowledge of the antibody primary sequence (obtained by cDNA sequencing) and structural analysis or modeling of the antibody complementarity-determining regions (CDRs). The efficacy of small molecule (peptide or organic cyclic mimetic) binding to antigen is determined by competitive inhibition and biological assays. Organic macrocyclic CDR mimetics The peptides described above retain their antigenbinding properties. However, because of their peptidic nature, linear and cyclic peptides are still subject to proteolysis, may possess some immunogenicity, and are often insoluble. The synthesis of soluble organic compounds that retain the overall structure that is important for CDP,_ binding and activity, but yet are formed without natural amino acid side-chains and peptide bonds, offers distinct advantages. A simple strategy has been developed to synthesize organic CDR_ structures on a macrocyclic ring stucture that can be altered to mimic the predicted and biologically active conformation of antibody hypervariable-loop structures. Structural information provided by the biological screening of constrained peptides and computer-aided molecular modeling will facilitate the design of these nonpeptidic CDRs. The design strategy for CDP,. peptides and mimetics is outlined in Fig. 3. Each successive modification of peptide or small molecule (mimetic) structure is analyzed repeatedly for binding activity using a competitive-inhibition assay of the parental antibody binding to antigen. This provides a rapid and simple screen for alterations in binding affinities. Such C D R mimetics have been shown to be resistant to peptidases, and are water-soluble and nontoxic 46. This general design strategy has been used to synthesize an organic mimetic of the C D R L 2 of mAb 87.92.6. A small (624 g mo1-1) macrocyclic compound (87.1 mimetic) competitively (and specifically) inhibited binding of the parental mAb to antigen, and TIBTECH SEPTEMBER 1994 (VOL 12) 378 reviews inhibited lymphocyte proliferation in a similar manner to mAb 87.92.6 (Ref. 46). The construction of C D R mimetics is achieved by a retrosynthetic modular-component synthesis strategy, which has been described in detaiD7,46. This strategy introduces natural and synthetic amino acid side-chains onto a ring structure. Stereochemical modifications at the base of the constrained [3-loop structure in the 87.1 mimetic have improved the structure of the mimetic such that its binding affinities are in the low nM range (M. I. Greene, unpublished). A synthetic-turn mimetic derived from a CD4 CDR-like structure inhibited HIV gp120 binding to CD4 + cells 47, and demonstrates that this technology has wide application in the design of antibody C D R mimetics and mimetics of other members of the Ig gene superfamily. Future prospects In order for antibody-derived small molecules to be used in a clinical context, they must maintain reasonable affinities and specificities in antigen binding. To improve upon the loss of binding often seen with the recombinant, monovalent scFv fragments, at least two groups have proposed strategies for creating bivalent and bifunctional molecules26, 48. In a similar fashion, C D R peptides or mimetics can be oligomerized using nonflexible linkers of various dimensions to link two domains to give a multivalent small molecule 24. Conceivably, this linkage approach should also enable the appropriate spatial reorientation of different C D R structures that react with separate regions of the epitope to create a molecule with enhanced binding relative to a single C D R unit. An analogous design has been used successfully with different C D R peptides that react to HIV-1 (Ref. 40). In addition to the advantages described above, small molecules derived from antibody structure may also have improved metabolic stability, better oral absorption, biodistribution and other pharmacological features. The efficacy of these reagents excludes a requirement for the Fc portion of the antibody (i.e. complement activation and antibody-dependent cellmediated cytotoxicity). If it proves necessary to enhance the biological effects of these molecules, toxins, drugs (or prodrugs), or radionuclides can be linked to the antigen-binding domain to give targeted immuno-therapies or diagnostics. The design of bioactive peptides and mimetics based, on antibody structure offers the advantage of yielding a substantial database of critical protein structural motifs that are important for antigen binding and biological activities. In this way, pharmacological improvements in peptide design can be based upon previously determined structural information from other antibody-antigen interactions. This type of rational drug design can be used to complement additional strategies to develop mAbs with a particular binding or biological activity, such as phagedisplay technologies2-4. These studies have shown the feasibility of this design strategy in constructing TIBTECH SEPTEMBER1994 (VOL 12) smaller subdomains of antibodies, thus eliminating the problems inherent in the use of the whole Ig molecule, while retaining the antigen binding, specificity and biological activities of the parental antibody. References 1 Kohler, G. and Milstein, C. (1975) Nature 256, 52-55 2 Reichmann, L., Clark, M., Waldmann, G. and Winter, G. (1988) Nature 332, 332-327 3 Huse, W. and Lerner, R. (1989) Science 246, 1534-1536 4 Pluckthun, A. (1992) Methods 2, 88-96 5 Yokata, T., Milenic, D. E., Whitlow, M. and Schlom, J. (1992) Cancer Res. 52, 3402-3408 6 Colcher, D. et al. (1990)J. Natl Cancer Inst. 82, 1191-1197 7 Bird, R. E. et al. (1988) Science 242, 423-426 8 Mariuzza, R. A. and Poljak, R.J. (1993) Curr. Opin. Immunol. 5, 50-55 9 Wilson, I. A. and Stanfield, R. (1993) Curr. Opin. Struct. Biol. 3, 113-118 10 Wu, T. and Kabat, E. (1970)J. Exp. Med. 132, 211 11 Chothia, C. and Lesk, A. M. (1987)J. Mol. Biol. 196, 901-917 12 Chothia, C. et al. (1989) Nature 342, 877-883 13 Porter, R. R. (1959) Biochem.J. 73, 119 14 Mariuzza, R. A., Phillips, S. E. V., and Poljak, R. J. (1987) Annu. Rev. Biophys. Biophys. Chem. 16, 139-159 15 Bhat, T. N. etal. (1994) Proc. NatlAcad. Sci. USA 91, 1089-1093 16 Tulip, W. R., Varghese, J. N., Laver, W. G., Webster, R. G. and Colman, P. M. (1992)J. Mol. Biol. 227, 122-148 17 Kabat, E. A. and Wu, T. T. (1991)J. Immunol. 147, t709-1719 18 Davies, D. R., Padlan, E. A. and Sheriff, S. (1990) Annu. Rev. Biochem. 59, 439-473 19 Amit, A. G., Mariuzza, R. A., Phillips, S. E. V. and Poljak, R. J. (1986) Science 233, 747-755 20 Jeffrey, P. D. etal. (1993) Pro& NatlAcad. Sci. USA 90, 10310-10314 21 Novotny, J., Bruccoleri, R. E. and Saul, F. A. (1989)Biochemistry 28, 4735-4749 22 Hawkins, R. E., Russell, S. J., Baier, M. and Winter, G. (1993) J. Mol. Biol. 234, 958-964 23 Novotny, J. and Sharp, K. (1992) Prog. Biophys. Mol. Biol. 58, 203-224 24 Williams, W. V., Kieber, E. T., VonFeldt, J., Greene, M. I. and Weiner, D. B. (1991)J. Biol. Chem. 266, 5182-5190 25 Huston, J. et al. (1991) Methods Enzymol. 203, 46-87 26 Pluckthun, A. (1992) Immunol. Rev. 130, 151-188 27 Orlandi, R., Gussow, D. H.,Jones, P. T. and Winter, G. (1989)Proc. Natl Acad. Sci. USA 86, 3833-3837 28 Tai, M. S. et al. (1990) Biochemistry 29, 8024-8030 29 Milenic, D. E. et al. (1991) CancerRes. 51, 6363-6371 30 Laroche, Y. et al. (1991)J. Biol. Chem. 266, 16343-16349 31 Huston, J. et al. (1988) Proc. Natl Acad. Sci. USA 85, 5879-5883 32 Wels, W., Harwerth, I. M., Mueller, M., Groner, B. and Hynes, N. E. (1992) CancerRes. 52, 6310-6317 33 Marks, J. D., Hoogenboom H. R. and Bonnet, T. P. (1991)J. Mol. Biol. 222, 581-597 34 Ward, E. S., Gussow, D., Griffiths, A. D., Jones, P. T. and Winter, G. (1989) Nature 341,544-546 35 Hamers-Casterman, C. et al. (1993) Nature 363, 446-448 36 Pessi, A., Bianchi, E., Crameri, A., Venturini, S., Tramontano, A. and Sollazzo, M. (1993) Nature 362, 367-369 37 Saragovi, H. U., Greene, M. I., Chrusciel, R. A. and Kahn, M. (1992) Bio/Technology 10, 773-778 38 Welling, G. W. et al. (1990)J. Chromatogr. 512, 337-343 39 Welling, G. W., Van Gorkum, J., Damhof, R. A. and Drijfllout, J. W. (1991)J. Chromotogr. 548, 235-242 40 Levi, M. et al. (1993) Proc. Natl Acad. Sci. USA 90, 4374-4378 41 Taub, R. and Greene, M. I. (1992) Biochemistry 31, 7431-7435 42 Williams, W. V., Weiner, D. B. and Greene, M. I. (1989) Methods Enzymol. 178, 321-341 43 Bruck, C. et al. (1986)J. Biol. Chem. 83, 6578-6582 379 reviews 44 Williams, W. V., Kieber, E. T., Weiner, D. B., P,ubin, D. H. and Greene, M. I. (1991)J. Biol. Chem. 266, 9241-9250 45 Taub, P,.. et al. (1989)J. Biol. Chem. 264, 259-265 46 Saragovi, H. U., Fitzpatrick, D., P,.aktabutr, A., Nakanishi, H., Kahn, M. and Greene, M. I. (1991) &ience 253, 792-795 47 Chen, S. et al. (1992) Proc. Natl Acad. Sci. USA 89, 5872-5876 48 Holliger, P., Prospero, T. and Winter, G. (1993) Proc. NatlAcad. Sci. USA 90, 6444-6448 The cellulosome - a treasuretrove for biotechnology Edward A. Bayer, Ely Morag and Raphael Lamed The cellulases of many cellulotytic bacteria are organized into discrete multienzyme complexes, called cellulosomes. The multiple subunits of cellulosomes are composed of numerous functional domains, which interact with each other and with the cellulosic substrate. One of these subunits comprises a distinctive new class of noncatalytic scaffolding polypeptide, which selectively integrates the various cellulase and xylanase subunits into the cohesive complex. Intelligent application of cellulosome hybrids and chimeric constructs of cellulosomal domains should enable better use of cellulosic biomass and may offer a wide range of novel applications in research, medicine and industry. Cellulose is the most abundant renewable source of carbon and energy on the Earth. The chemical simplicity of cellulose belies its structural complexity and stability - properties that account for its contribution to modern society in the form of paper and, as cellulosic waste, as a major pollutant worldwide 1. As cellulose is a very stable polymer, effective hydrolysis of it requires the cellulolytic enzymes of several different microorganisms to act synergistically2. For the greater part of this century, cellulolytic microbes and their cellulase systems have been considered for use in the industrial conversion of cellulosic biomass. However, eventually, it became apparent that natural systems are not necessarily compatible with industry. Advanced engineering techniques alone are not sufficient to enable viable processes for solubilizing cellulosics to be designed, and it was realized that more had to be learnt about the enzymes and microbes that mediate cellulolysis. In recent years, multienzyme complexes, known as cellulosomes (see Glossary), have been identified in many cellulolytic microorganisms3,4. These complexes are dedicated to the efficient degradation of cellulose E. A. Bayer and E. Morag are at the Department of Biophysics, The Weizmann Institute of Science, Rehovot, Israel. R. Lamed is at the Department of Molecular Microbiology and Biotechnology, University of Tel Aviv, Ramat Aviv, Israel. © 1994, Elsevier Science Ltd and hemicellulose. This article describes some of the structural features of cellulosomes that have been elucidated from combined biochemical and molecular biology studies and discusses the impact of an improved understanding of cellulosome structure and function on the controlled hydrolysis of lignocellulose. Other potential applications ofcellulosome technology are also discussed. The cellulosic ecosystem Cellulolytic microorganisms do not occupy an ecological niche on their own; they exist in concert with many other cellulolytic and noncellulolytic strains of bacteria and fungi 5 (Fig. 1). Nevertheless, the cellulolytic strain(s) plays a vital role in the cellulosic ecosystem as the predominant polymer-degrading species. In the plant, cellulose is usually coated by other polymers, predominantly xylan and lignin, which also hinder cellulolysis. Each of these protective polymers is characterized by a different intrinsic chemical and structural arrangement and, consequently, different groups of microorganisms and enzymes are required to degrade the different types of polymer. Xylan is degraded quite readily by xylanases6; however, hgnin poses more of a problem 7, in that the degradation process requires molecular oxygen, and the degradation products are often toxic to the cellulolytic microorganisms and inhibitory to their enzymes: TIBTECHSEPTEMBER1994 (VOL 12)