* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download Genetically Engineered Antibodies
Survey
Document related concepts
Innate immune system wikipedia , lookup
Immunoprecipitation wikipedia , lookup
Duffy antigen system wikipedia , lookup
Adaptive immune system wikipedia , lookup
Anti-nuclear antibody wikipedia , lookup
Adoptive cell transfer wikipedia , lookup
Immunocontraception wikipedia , lookup
Molecular mimicry wikipedia , lookup
DNA vaccination wikipedia , lookup
Polyclonal B cell response wikipedia , lookup
Cancer immunotherapy wikipedia , lookup
Transcript
CLIN.CHEM.35/9, 1849-1853 (1989) Genetically Engineered Antibodies GordonP. Moore The technologyneeded to geneticallyengineer antibodiesis 6-8), the possibility is now enhanced that the undoubted of immunotherapy can someday be realized. evolving rapidly and the potential utility of these novel reagents is being explored with vigor. The process includes promise cloningof the antibodygenes, their in vitromanipulationand mutagenesis, expression in a suitable host/vectorsystem, and, for commercial production,scale-up, purification,and productevaluation.At each step, significantadvances have been achieved recently.Forexample:at first,antibodygenes were cloned from genomic librariesby using adjacent DNA probes;techniquesfor rapidsequencingby pnmerextension of total mRNA allowed more specificscreeningwith synthesized oligomers;finally,antibodygenes can now be created de novo by chemical synthesis. Moreover, such synthesis allows total control over the antibody sequence so that molecules of any configuration can be produced. New reagents created in this way includemunne antibodieswhose constantregionsand variable-regionframeworkshave been replacedwith human sequence to enhance immunocompat- Genetic EngineerIng of AntIbodies ibility with patients, to switch immunoglobulin class, or both. Additional Keyphrases: monoclonal antibodies sis expression vector immunotherapy bodies gene synthechimeric anti- Monoclonal antibodies have been available for more than 15 years (1). Despite their great utility in the laboratory and in diagnostic testing, and notwithstanding their high promise and the enormous research effort expended on them, they have yet to make a significant impact as clinical tools. Monoclonals are not used routinely to treat or image tumors, to provide passive immunization against infectious diseases, or for immunosuppression after organ transplant-all uses for which they have been intended. Among the problems encountered in the effort to translate the diversity and specificity of monoclonals into clinically useful reagents are the following: #{149} The human immune response to foreign antibodies. #{149} Difficulty in making physical contact between the antibody and the target antigen. . #{149} Change in antigen structure during tumor development or through mutation, as well as antigen shedding. #{149} Low affinity, inappropriate isotype, or nonoptimal systemic half-life of immunotherapeutic antibodies. #{149} Difficulty in producing sufficient quantities of antibody for therapy. In 1983, Oi et al. reported (2) that lymphoid cells can express cloned, transfected immunoglobulin genes. This demonstration opened the way to a new technology that may provide solutions to some of the problems listed above, i.e., the cloning, engineering, and expression of genes that code for antibodies. Because these methods have been shown capable of creating new kinds of antibody molecules (e.g., 3-5), and some of these are clinically relevant (e.g., Integrated Genetics, Inc., One Mountain Road, Framingham, MA 01701. Received April 13, 1989; accepted June 9, 1989. The process by which genetically engineered antibody variants can be created has been reviewed extensively elsewhere (e.g., 9, 10). The major goals of such engineering include: #{149} Creating immunocompatible reagents by joining rodent variable regions to human constant regions. #{149} Altering systemic half-life by truncation or amino acid addition. #{149} Isotype switching. #{149} Increasing affinity or avidity. #{149} Using gene fusion to generate bispecific antibodies. #{149} Using gene fusion to generate molecules that bind antigen but also possess additional activities. #{149} Altering effector functions such as F receptor binding and complement activation. #{149} Reducing anti-idiotypic response to rodent antibodies by replacing rodent variable-region framework sequence with human sequence. For each antibody of interest, engineering begins by cloning the variable regions, which dictates antigen specificity. DNA is isolated from hybridoma cells and cloned, usually in Aphage vectors. The phage library is screened by hybridization with a radioactively labeled probe. Because the genomic rearrangement that results in production of functional immunoglobulin genes leaves common sequences between the variable (V) and constant (C) regions, any functional heavy-chain gene can be isolated with a single (“universal”) probe.’ The same is true for the light chain. These clones can be verified with synthesized oligomere as described below. Once the desired V regions have been isolated, they must be cloned into a vector appropriate for expression. If immunocompatibility is an objective, it is desirable that the expression vector contain a human C region; given that most available hybridomas are murine, the V region is usually from mouse. Moreover, panels of expression vectors now exist, each containing a different human C region. Thus, the isotype of the antibody can be chosen at will. Expression vectors have been designed for use in bacteria, yeast, or mammalian cells (2, 11, 12). The most common hosts are murine myeloma cells, which, through mutation, have lost the ability to make endogenous antibody. Vectors for myeloma cells usually contain heavy(H)and light(L)-chain genes individually, and these are cointroduced into the host (6, 7). In addition to the murine V and human C regions, the vectors contain a bacterial origin of replication and drug-selectable markers to identify which cells have taken up and expressed the vectors. The immunoglobulin genes are usually controlled by their own promoters and enhancers. These constructions may contain ‘Nonstandard abbreviations: V, variable; C, constant; H, heavy; L, light; and CDR, complementarity-determining region. CLINICAL CHEMISTRY, Vol. 35, No.9, 1989 1849 other features as described below. When antibody genes have been engineered as desired and placed in expression vectors, they are transfected into myeloma cell hosts. Transfection is by electroporation (13) or protoplast fusion (14). Although both H and L chain vectors are individually selectable, it is interesting that co-transfection is so frequent that selection with only one drug is necessary. Transfection frequency is about one in 1000. Drug-resistant colonies are grown, tested for secretion of antibody by enzyme-linked immunosorbent assay, and high-producing lines are identified. A B C 200- 92 566 2450- Fig.2. Purification of chimericB6.2 immunoglobulin by (A) nonresodiumdodecyl sulfate-polyacrylamide gelelectrophoresis; (C) isoelectric focusing ducingpolyacrylamidegel electrophoresis;( Purificationof Murine V/Human C ChimericAntibodies Once a cell line secreting the engineered antibody has been identified, it must be scaled-up for production. This can be done in tissue culture or by growing the cells in mice and collecting antibody from ascites fluid. Various purification schemes have been reported (e.g., 15), some of which utilize the binding activity of the antibody. My colleagues and I have used methods that rely on properties of the human C regions that distinguish chimeric from contaminating antibodies. In the case of tissue culture media, the major contaminant is bovine antibody found in serum. As shown in Figure 1 (C. Shearman and D. Lawrie, Integrated Genetics, unpublished), elution from Protein A-Sepharose can clearly separate a murine/human chimeric antibody from contaminating bovine immunoglobulins. In ascites fluid, murine antibodies constitute a major contaminant. These are removed effectively (Figure 2) by binding to Protein A-Sepharose, elution, then passage over Sepharose complexed with anti-murine IgG (16). Both these methods yield chimeric antibody of sufficient purity for subsequent functional analysis. Moreover, these methods can be applied to any murine V/human C chimeric antibody without regard to its antigen specificity. A and B: lane 1, molecular mass standards; lane 2, unpurIfied ascites fluid from mice injected with cB6.2-producing transfectoma cells; lane 3, ascites fluid after binding and elution of Igs from Protein A-agarose; lane 4, cB6.2 after purification from ascites fluid by binding to Protein A, followed by chromatography on Sepharose-bound sheep anti-munnelgG; lane 5, B6.2 purified by “Fast Protein Liquid Chromatography’ (LKB, Bromma, Sweden). C: lane 1, unpurified ascites from mice injected withB6.2-producing transfectoma cells; lane 2, the same material after binding and elution from Protein A; lane 3, the same material after passage over anti-murine lgG-Sepharose; lane 4, purIfied B6.2; upper arrow, purified B6.2; lower arrow, positionof purified cB6.2. From reference 16; used with permission are laborious and costly, it is critical that novel variants be screened by various in vitro assays or in animal studies. The nature of such analysis is dependent upon the particular antibody in question. To illustrate, I refer to a murine V/human C chimera of the murine antibody B6.2 (6), which has specificity for human breast, colon, and lung carcinomas (17). Comparisons of chimeric and murine B6.2 with Characterization of Chimeric Antibodies As described above, a virtually limitless array of new antibody variants can be produced by genetic engineering. The utility of each reagent for immunotherapy must ultimately be determined in clinical trials. Because such trials I 3 5 7 s II 3 IS I? FUC3IOI II 21 23 23 25 27 25 21 33 33 55Usd Fig. 1. Purificationof a murine V/human C chimeric antibody by chromatography on a column of ProteinA-Sepharose About 400 mL of tissue culture supemate containingabout 5 mg of chimenc antibodyper milliliter was passed over a 2.0-mL column. Bound antibodywas eluted witha linearpH gradientof 0.1 mot/Lcitrate bufferfrom pH 6.5 to 3.0. Fractions (1.0mL) were assayed forbovine arid chimeric antibody by ELISA. The symbols designate (11) total protein, (+) bovine, and () chimeric antibody (Chip Shoarman and Dawson Lawrie, unpublished) 1850 CLINICAL CHEMISTRY, Vol. 35, No. 9, 1989 Fig. 3. Indirectimmunofluorescentlabeling of human tumor cells boundto murineor chimericB6.2 Top of each pair of photographs (upper-case lettering): cells bound with cB6.2, then stained with rtiodamine-conjugated goat anti-human K. Bottom photographs (lower.case lettering): cells of the same line bound with 86.2, then stained with fluorescein isothiocyanate.conjugated goat anti-munne lgG. The cell lines are A549 (A,a),9812 (B,b), LS174T (C,c), SW900(E,e), and human fibroblasts (F,!). InD,d, A549 cells were bound with total human (A) or murine (a) lgG, then stained as above. From reference 16; used with permission Fig. 5. Scintigraphicimages of mice bearing human tumors,24 h after injectionof ‘251-IabeledcB6.2 In antibody-positiveLS174 tumor (A) the radionuclideis localized in the xenograft.In antibody-negative HCT-15tumor(B) the radlonuclideis foundin the blood pool. In A, a small amountoffree 1251 hasbeenIncorporatedintothe thyroid. Tumors were implanted in the right hind limb. From reference 16;used with premission rig Unbeled Competitor general, their properties are similar to the parent murine Their utility in immunotherapy is being tested clinical trials, the results of which should be available over the next few years. It is hoped that, because of their presumed greater immunocompatability in patients, they will outperform the murine antibodies from which they are derived. Meanwhile, new kinds of antibody variants are being produced and tested (see below). monoclonal. in various Increasing the Expression Level of Engineered Antibodies o.a 1.0 Unbdsd Competito #{231}sg) Fig.4. Competitive bindingof murine and chimencB6.2 to human tumorantigen (A) 66.2 was purified, labeled with I, and incubated with antigen coated on microtiterwells in the presence of increasing amounts of unlabeled 66.2 (0), cB6.2 (to), or anti-horseradish peroxidase (anti-HRP), an unrelated antibody (0). (B) cB6.2 was purified, labeled with 1251, and Incubated with increasing amounts of 66.2 (0), cB6.2 (A), or anti-HRP, an unrelated antibody (0). The membrane was prepared from LS174T human colon carcinoma cells; 60-80% of the input counts were bound.(C) cB6.2 wasboundto intact LS174Tcellsin the presence of increasing amounts of 66.2 (0), or OKT9 (0), and the cells were labeled by incubation with fluorescein isothiocyanate(FITC)-conjugated goat anti-human lgG. 86.2 was bound to LS1741 cells in the presence of increasing amounts of 86.2 (I), and the cells were labeled by incubation with FITC-conjugated goat anti-munne lgG. Fluorescein-labeled OKT9 was bound toLS174Tcellsin the presence of increasing amountsof 66.2. (#{149}). Labeling was quantifiedby flow cytometry; 70-85% of the cells were labeled. From reference 16; used with permission respect to a variety of biochemical, biological, and immunologic characteristics (6, 16) showed that, for example, the murine and chimeric antibodies bound to the same panel of cell types (Figure 3). Most important, competition studies carried out in the cell sorter or with in-vivo-labeled antibody demonstrated that the chimeric and murine molecules competed for binding to antigen to an approximately equal extent (Figure 4). This important result implied that exchange of the human and murine C regions had little or no effect on the ability of the antibody to bind antigen. In a further test, chimeric B6.2 was used successfully to image a human tumor implanted in a mouse (Figure 5). Murine V/human C chimeric antibodies of many specificities now have been produced in several laboratories. In A major issue in the widespread clinical application of genetically engineered antibodies will be the ability to produce them in large amounts. This is particularly important because many immunotherapeutic regimens involve repeated administration of relatively large doses of antibodies. Unfortunately, engineered antibodies expressed by transfection into eukaryotic hosts, e.g., myeloma cells, often are not produced efficiently. Most “transfectomas” produce 2-10 pg/1O6 cells per day; this is only 1-10% of the production level of the best murine hybridomas. The probable reason is that the transfected genes integrate into the host chromosomes at random and, therefore, at loci that are not necessarily optimal for expression. Although antibody genes have been expressed in yeast and bacteria, there are problems in chain association, secretion, glycation, and biological activity. We have attempted to increase the expression of transfected antibody genes by linking them to genes that can be amplified by a slow increase in the concentrations of various drugs (18). As the number of copies of the amplifiable gene increases in response to the drug, the linked antibody gene(s) is also amplified in copy number. This amplification results in an increase in gene expression and thus increased secretion of antibody. A commonly utilized gene-drug combination is the gene coding for dihydrofolate reductase, which is amplified by the drug methotrexate. We have adapted this combination for use in myeloma cells; a representative vector is shown in Figure 6. Using this vector, we have achieved a large increase in the expression of transfected immunoglobulin genes (18). Moreover, similar ampliflable vectors are now being used to express nonimmunoglobulin genes in myeloma cells (19, and M. Hendricks, Integrated Genetics, unpublished). CLINICAL CHEMISTRY, Vol. 35, No. 9, 1989 1851 B. SV40 SVIO SOIl Fig. 6. Structuresof plasmidscontainingchimericIg genes (derived from munne hybridoma B6.2) and the methotrexate (MTX)-resistant dihydrofolatereductase (DHFR) gene Chimencheavy chain (A) or chimeric light chain (B)wereclonedadjacentto mDHFR and designated DHFR/H and DHFR/L, respectively. From reference 18;used with permission The Use of DNA Synthesis In AntIbody EngIneering The chemical synthesis of DNA of desired nucleotide sequence is an important new technology in molecular biology. The process has now been automated, and fragments about 100 nucleotides long can be made without great difficulty. We have been applying this technique to antibody engineering in a variety of ways. The process of cloning desired V regions is laborious because genomic libraries must be constructed and the tiny minority (perhaps one in 106) of target clones must be purified from the background. Moreover, a serious problem is that aberrantly rearranged, and hence nonfunctional, V regions are often obtained from the screen (20). These problems can be circumvented by synthesizing the desired V regions de novo rather than cloning them. Obviously, the DNA sequence of the region must be determined before synthesis. Because mRNA that specifies immunoglobulin is so prevalent in hybridoma cells, this can be accomplished directly (21). Poly A mRNA is isolated from the hybridoma and hybridized with a short “universal” DNA primer synthesized with the knowledge of the sequence of the common C region. The duplex is then extended by using reverse transcriptase, and the product is sequenced. This sequence can be used to make specific probes to screen libraries. The more innovative approach is to synthesize the entire V region. Although the sequence is too long (400-500 nucleotides) to synthesize in one piece, several alternative methods are available to assemble the synthesized fragments. One technique is to synthesize pairs of opposite strands with restriction sites at their ends and short overlaps. The overlaps are annealed, then the structure is filled in with DNA polymerase. The resulting double-stranded fragments are then ligated through the restriction sites at their ends to form the finished structure. Depending on the length, intermediate cloning may be required to generate sufficient material for the next step. Care must be taken that the engineered restriction sites do not interfere with either the desired amino acid sequence or with physiologic codon usage. The number of known restriction sites is so large, however, that this is rarely a problem. The advantage of polymerase is that it reduces the amount of DNA synthesis required. The disadvantage is that the error rate is relatively high. Because even a one-base error can render the construct nonfunctional, each 1852 CLINICAL CHEMISTRY, Vol. 35, No. 9, 1989 piece must be sequenced after synthesis and after each cloning step. A second method involves synthesis of both complementary strands in their entirety. A 10- to 15-nucleotide “overhang” allows ligation of serial double-stranded fragments. Surprisingly, one simply mixes a relatively large number of single-stranded, overlapping fragments. This complex reaction results simultaneously in opposite-strand hybridization and overlapping pair ligation to form the completed structure in a single step. The synthesized V region is then cloned, sequenced (for verification), and inserted into an appropriate expression vector. As described above, one advantage of total V region synthesis is ease and speed. A second advantage is avoidance of cloning aberrant V regions. More importantly, however, this method gives absolute control over the structure of the resulting gene. Thus, any variant can be created easily. Further, mixing of nucleotides at nucleotide addition steps during the gene synthesis generates multiple amino acids at particular positions. This new form of in vitro mutagenesis is extremely powerful. Replacement of Compiementarity-Determinlng Regions Any antibody variant can be made by using the technique of total gene synthesis. Technical limitations include possible problems with message or protein stability, chain association, glycation, or secretion. The major limitations, however, are the creative imagination, the knowledge of antibody structure/function needed to design improved antibodies, and the resources necessary to test them. A major goal of antibody engineering is to make immunocompatible therapeutic agents. I have described chimeric antibodies containing human C regions. These reagents, however, are still capable of eliciting host immune response against portions of the rodent V regions (22). As an example of the use of total gene synthesis, I now describe a new antibody variant designed to reduce further, or eliminate, host response to therapeutic antibodies. It has been known for some time that most antigen specificity resides in defined portions of the V regions called “hypervariable” or “complementarity-determining” regions (CDRs) (23). Spatially, the CDRs are looped away from the V region framework, which consists of a pleated sheet (24). These CDR loops form the antigen-combining sites. Comparison of V region sequence among immunoglobulins from a wide variety of sources shows that sequence variation is far greater in the CDRs than elsewhere in the V region (25). Consequently, it should be possible to transplant the CDRS from a rodent antibody into a human framework (3,10) and to maintain the antigen specificity of the rodent parent antibody while eliminating host response to the V region framework. Further, because the CDRs are so variable, even within a species, they are unlikely to elicit an immune response. Thus, host immune response should be reduced greatly, or eliminated, by CDR replacement. The three H-chain and three L-chain CDRS are interspersed with framework sequence. Thus, total gene synthesis represents the most practical way to make CDR-replaced variants. In our hands, the process is carried out as shown in Figure 7. Poly A mRNA is isolated from the target rodent hybridoma and sequenced as described previously. A computer search is made to identify a closely related pair of human H and L chains. Alternatively, one can utilize the most related H and L chains and assume that they will later be capable of association. The next step Steps __ Leader 1. in CDR Replacement 14 I O)IU Isolate poly __ 4 I rF’ A4’ mRNA CDR2 f FR3 CDR3 FR4 from the target hybridoma and sequence by primary extension. the H and L chain V regions 2. Computer search to related human H and L chain identify V 3. Design civilized imarine CDR into restriction sites at physiologic codon 4. Synthesize S. Ligate civilized V cxons transfect and select. 6. Test the civilized engineered CP, Hellstrom I. Chimeric mouse-human IgGi antibody that can mediate lysis of cancer cells. Proc Nat! Acad Sci USA 1987;84:3429-43. 9. Morrison SL, Wims LA, Wallick S, Tan L, Oi VT. Genetically engineered antibody molecules and their application [Review]. Ann NY Acad Sci 1988;507:187-98. 10. Riechmann L, Clark M, Waldman H, Winter G. Reshaping human antibodies for therapy. Nature (London) 1988;332:323-7. 11. Boss MA, Kenton JH, Wood CR, Emtage JS. Assembly of H and I. chain V exonswhich incorporate a human framework. Includeunique the CDR/FR usage. V exons boarders and maintain and confmn into expression by sequencing. rectors. functional antibodies from immunoglobulin heavy and light chains synthesized mE. coli. Nucleic Acids Res 1984;12:3791-806. 12. Wood CR, Boss MA, Kenton JH, Calvert JE, Roberts NA, Emtage JS. The synthesis and in vivo assembly of functional then product. Fig. 7. Procedure to generate a CDR-replaced antibody; arrows indicatethe positionsof engineered,uniquerestrictionsites sites rather than by synthesis and assembly antibodies in yeast. Nature (London) 1985;314:44&-9. 13. Potter H, Weir L, Leder P. Enhancer dependent expression of human K immunoglobulin genes introduced into mouse pre-B lymphocytes by electroporation. Proc Natl Acad Sci USA is to design the CDR-replaced variant by incorporating the rodent CDRS into the human frameworks. Unfortunately, the exact borders of the CDRs are not clear and probably vary among different antibodies. Thus, it is not always obvious exactly which amino acids to change to human and which to leave as rodent. The goal, of course, is maximal antigen binding with minimal host immune response. Much of our research is directed towards learning rules that will help dictate these choices. When suitable CDRreplaced sequences for H and L chains have been designed, they are synthesized, cloned, sequenced, expressed, and tested. A further feature of the scheme shown in Figure 7 is that unique restriction sites have been engineered at the CDR/framework borders. In the future, new antibody specificities will be generated by insertion of synthesized CDRS at these possessing novel effector functions. Nature (London) 1984;312:604-8. 4. Jones PT, Dear PH, Foote J, Neuberger MS, Winter G. Replacing the complementarity-determining regions in a human antibody with those from a mouse. Nature (London) 1986;321:522-5. 5. Morrison SL, Johnson MJ, Herzenberg LA, Oi VT. Chimeric human antibody molecules: mouse antigen-binding domains with human constant region domains. Proc Nat! Acad Sci USA 1984;81:6851-5. 6. Sahagan BG, Dorai H, Saltzgaber-Muller J, et al. A genetically engineered murine/human chimeric antibody retains specificity for human tumor-associated antigen. J Imunol 1986;137:1066-74. 7. Beidler CB, Ludwig TR, Cardenas HJ, et al. Cloning and high level expression of a chimeric antibody with specificity for human carcinoembryonic antigen. J Immunol 1988;141:4053-60. 8. Liu AY, Robinson RR, Helistrom KE, Murray Jr ED, Chang of the entire V regions. This will result in a significant saving of effort. One CDR-replaced antibody has been described by Riechmann et al. (10), and others are being evaluated in our laboratory and elsewhere. Many other engineered antibody variants are also being produced and studied. At the least, this work will result in a greater understanding of immunoglobulin structure and function. New molecules will be produced that can be applied successfully for immunotherapy. References 1. Kohler G, Milstein C. Continuous cultures of fused cells secreting antibody of predefined specificity. Nature (London) 1975:256:495-7. 2. Oi VT, Morrison SL, Herzenberg LA, Berg P. Immunoglobulin gene expression in transformed lymphoid cells. Proc Natl Acad Sci USA 1983;80:825-9. 3. Neuberger MS, Williams GT, Fox RO. Recombinant antibodies 1984;81:7161-5. 14. Sandri-Goldin RM, Goldin AL, Levine M, Glorioso JC. Highfrequency transfer of cloned herpes simplex virus type 1 sequences to mammalian cells by protoplast fusion. Mo! Cell Biol 1981;1:743- 52. 15. Morrison SL, Canfleld S, Porter S, Lee TK, Tao M, Wims LA. Production and characterization of genetically engineered anti- body molecules. Clin Chem 1988;34:1668-75. 16. Brown BA, Davis GL, Saltagaber-Muller J, et a!. Tumorspecific genetically engineered murine/human chimeric monoclonal antibody. Cancer Res 1987;47:3577-83. 17. Kufe DS, Nadler L, Sargent L, et al. Biological behavior of hmnan breast carcinoma-associated antigens expressed during cellular proliferation. Cancer Res 1983;43:851-9. 18. Dorai H, Moore GP. The effect of dihydrofolate reductasemediated gene amplification on the expression of transfected immunoglobulin genes. J bumunol 1987;139:4232-41. 19. Hendricks MB, Banker MJ, McLaughlin A. A high efficiency vector for expression 1988;64:43-51. of foreign genes in myeloma cells. Gene 20. Yancopoulos GD. Alt FW. Regulation of the assembly and expression of variable-region genes [Review]. Annu Rev Immunol 1986;3:339-68. 21. Hanlyn PH, Brownlee GO, Cheng CC, Gait MJ, Milstein C. Complete sequence of constant and 3’ noncoding regions of an immunoglobulin mRNA using the dideoxynucleotide method of RNA sequencing. Cell 1978;15:1067-75. 22. Jerne NK. Towards a network theory of the immune system. Ann Imniunol (Paris) 1974;125c:373-89. 23. Kabat EA, Wu T1. Attempts to locate complementaritydetermining residues in the variable positions of light and heavy chains. Ann NY Acad Sci 1971;190:382-93. 24. Davies DR, Padlan EA. Correlations between antigen-binding specificity and the three dimensional structure of the antibody combining site. In: Haber E, Krause RM, eds. Antibodies in human diagnosis and therapy. New York: Raven Press, 1977:119-28. 25. Wu ‘VF, Kabat EA. An analysis of the sequences of the variable regions of Bence Jones proteins and myeloma light chains and their implication for antibody complementarity. J Exp Med 1978;132:211-50. CLINICAL CHEMISTRY, Vol. 35, No. 9, 1989 1853