THE JOURNAL OF BIOLOGICAL CHEMISTRY Vol. 279, No. 8, Issue of February 20, pp. 6213–6216, 2004 © 2004 by The American Society for Biochemistry and Molecular Biology, Inc. Printed in U.S.A.
Accelerated Publication Engineered Human IgG Antibodies with Longer Serum Half-lives in Primates* Received for publication, October 27, 2003, and in revised form, December 19, 2003 Published, JBC Papers in Press, December 29, 2003, DOI 10.1074/jbc.C300470200 Paul R. Hinton‡, Mary G. Johlfs, Joanna M. Xiong, Kelly Hanestad, Kelly C. Ong§, Chuck Bullock, Stephen Keller, Meina Tao Tang, J. Yun Tso, Max Va´squez, and Naoya Tsurushita From Protein Design Labs, Inc., Fremont, California 94555
The neonatal Fc receptor (FcRn) plays an important role in regulating the serum half-lives of IgG antibodies. A correlation has been established between the pH-dependent binding affinity of IgG antibodies to FcRn and their serum half-lives in mice. In this study, molecular modeling was used to identify Fc positions near the FcRn binding site in a human IgG antibody that, when mutated, might alter the binding affinity of IgG to FcRn. Following mutagenesis, several IgG2 mutants with increased binding affinity to human FcRn at pH 6.0 were identified at Fc positions 250 and 428. These mutants do not bind to human FcRn at pH 7.5. A pharmacokinetics study of two mutant IgG2 antibodies with increased FcRn binding affinity indicated that they had serum half-lives in rhesus monkeys ⬃2-fold longer than the wild-type antibody.
Antibody therapy is coming of age, with 15 monoclonal antibodies approved for therapeutic use in the United States and many others currently undergoing clinical trials (1). The advent of antibody engineering over the past two decades has contributed to the recent clinical success of therapeutic antibodies. The development of chimeric (2) and humanized (3) antibodies not only reduced the potent immunogenicity of rodent antibodies in humans but also improved the serum halflives and efficacy of such therapeutics compared with rodent antibodies. Phage display (4) and other display technologies have led to the ability to increase the affinity of antibodies for their target antigens. More recently, antibody engineering has been used to modify the effector functions of antibodies by altering their binding to C1q (5) and various Fc␥ receptors (6). The neonatal Fc receptor (FcRn)1 is a heterodimer that comprises a transmembrane ␣ chain with structural homology to * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. ‡ To whom correspondence should be addressed. Tel.: 510-574-1688; Fax: 510-574-1500; E-mail:
[email protected]. § Present address: Genentech, Inc., 1 DNA Way, South San Francisco, CA 94080. 1 The abbreviations used are: FcRn, neonatal Fc receptor; 2m, 2microglobulin; IgG2M3, M3 variant of human IgG2; HBV, hepatitis B virus; FACS, fluorescence-activated cell sorter; FBB, FACS binding buffer; PBS, phosphate-buffered saline; R-PE, R-phycoerythrin; PK, pharmacokinetics; CL, clearance; AUC, area under the curve. This paper is available on line at http://www.jbc.org
the extracellular domains of the ␣ chain of major histocompatibility complex class I molecules, and a soluble light chain consisting of 2-microglubulin (2m) (7). FcRn mediates both transcytosis of maternal IgG to the fetus or neonate and IgG homeostasis in adults (8). Evidence for the latter role initially came from studies indicating an unusually short serum halflife for IgG antibodies in 2m-deficient mice (9 –11). This observation led to the generation of mutant mouse hinge-Fc fragments with enhanced binding to FcRn and increased serum persistence in mice (12). Recently, several studies have identified human IgG1 mutants with enhanced FcRn binding (6, 13), although no improvement in the serum half-lives of these mutants was observed in mice (13) or reported in primates. The binding of IgG to FcRn is sharply pH-dependent; IgG binds to FcRn under mildly acidic conditions and is released under slightly basic conditions (14). It has been hypothesized that pinocytosed IgG antibodies are captured by FcRn in acidified endosomes, rescued from degradation in lysosomes, recycled back to the cell surface, and returned to the circulation (8). Mutagenesis studies have identified both the mouse (15, 16) and human (17) Fc residues believed to be important in mediating pH-dependent binding. The results of the mutagenesis studies are consistent with the interpretation of a crystallographic study of the Fc䡠FcRn interaction (18). In the current study, molecular modeling was used to identify residues in the human IgG Fc near the FcRn binding site that, when mutated, might alter binding to FcRn without affecting the pH dependence of this interaction. Following exhaustive mutagenesis at these positions, several IgG2 mutants were identified with improved binding to FcRn at pH 6.0 that retained the property of pH-dependent release. A pharmacokinetics study in rhesus monkeys showed that two mutant IgG2 antibodies with increased FcRn binding affinity had considerably longer serum half-lives than the wild-type antibody. EXPERIMENTAL PROCEDURES
Molecular Modeling—Molecular models of the human Fc䡠FcRn complex were generated based on the crystal structures (18, 19) and a model (20) of the rat Fc䡠FcRn complex. First, the rat 2m and FcRn ␣ chains of the complex were replaced, respectively, with the human 2m (21) and FcRn ␣ chains (22). Next, the rat Fc residues of the complex were replaced with the corresponding human IgG1 Fc residues (23), and then energy minimization calculations were done using SEGMOD and ENCAD (24, 25) to produce a model of the human IgG1 Fc䡠FcRn complex. The process was repeated to produce a model of the complex of human FcRn and the M3 variant of human IgG2 (IgG2M3) (26). IgG Mutagenesis, Expression, and Purification—The light and heavy chain cDNAs from a trioma cell line expressing the human anti-hepatitis B virus (HBV) antibody OST577 (27) were cloned by PCR. The light and heavy chain V-genes were converted by PCR into mini-exons and subcloned, respectively, into pV2, a derivative of pVk (28) containing the human 2 constant region, and the M3 variant of pVg2.D.Tt (26). Overlap extension PCR (29) was used to generate random amino acid substitutions at positions 250, 314, or 428 (numbered according to the EU index (23)) in the heavy chain of OST577-IgG2M3, until each possible amino acid at these positions was obtained. The resulting PCR fragments were subcloned into pVAg2M3-OST577, a derivative of the M3 variant of pVg2.D.Tt containing the pUC18 replication origin (30). Human kidney cell line 293-H (Invitrogen) was transiently cotransfected with the antibody expression plasmids using the LipofectAMINE 2000 reagent (Invitrogen). Culture supernatants were concentrated and buffer exchanged into PBS, pH 6.0, with Vivaspin centrifugal concentrators (Vivascience, Hannover, Germany). Mouse myeloma cell line Sp2/0 (American Type Culture Collection, Manassas, VA) was stably cotransfected by electroporation. OST577-IgG2M3 antibodies were pu-
6213
6214
Engineered Antibodies with Longer Serum Half-lives
FIG. 1. Model of human IgG2M3 Fc䡠FcRn complex. The Fc chains are colored gray (proximal to FcRn) and yellow (distal to FcRn). For the receptor, 2m is dark blue, and the FcRn ␣ chain is light blue. Thr-250 is depicted in red and Met-428 in green.
rified from culture supernatants by protein A column chromatography. For in vivo studies, the protein A eluate was further purified by size exclusion chromatography. FcRn Cloning and Surface Expression—Human 2m and FcRn ␣ chain cDNAs were cloned by PCR from human peripheral blood mononuclear cells and subcloned into pDL172, a derivative of pVk containing a glycosylphosphatidylinositol linkage signal from human decay-accelerating factor (31) fused to the FcRn ␣ chain, resulting in pDL208. Mouse myeloma cell line NS0 (European Collection of Animal Cell Cultures, Salisbury, Wiltshire, UK) was stably transfected with pDL208 by electroporation, yielding cell line NS0-HuFcRn. Competitive Binding Assays—Transiently expressed IgG2M3 wildtype and mutant antibodies were tested for binding to human FcRn in a single-point competitive binding assay. Briefly, 2 ⫻ 105 NS0-HuFcRn cells/test were washed with FACS binding buffer (FBB) (PBS containing 0.5% bovine serum albumin, 0.1% NaN3), pH 8.0, and then with FBB, pH 6.0, and resuspended in 120 l of biotinylated OST577IgG2M3 (8.3 g/ml) and concentrated supernatant (containing 8.3 g/ml of competitor antibody) in FBB, pH 6.0. After 1 h on ice, the cells were washed with FBB, pH 6.0, and resuspended in 25 l of 2.5 g/ml streptavidin-conjugated R-PE (BioSource, Camarillo, CA) in FBB, pH 6.0. After 30 min on ice, the cells were washed with FBB, pH 6.0, resuspended in 1% formaldehyde in PBS, and analyzed by flow cytometry using a FACSCalibur flow cytometer (BD Biosciences). Purified IgG2M3 wild-type and mutant antibodies were further tested in a competitive binding assay using purified competitor antibody (2-fold serial dilutions from 208 to 0.102 g/ml) as described above. IC50 values were calculated using GraphPad Prism, version 3.02 (GraphPad Software, San Diego, CA). Rhesus Pharmacokinetics Study—A non-GLP (“good laboratory practice”) pharmacokinetics (PK) study was approved by the Institutional Animal Care and Use Committee and conducted at the California National Primate Research Center (University of California, Davis). Twelve male rhesus macaques were randomized by weight and assigned to one of three study groups. Each animal received a single intravenous dose of wild-type or one of two mutants of OST577-IgG2M3 at 1 mg/kg by infusion over a span of 15 min. Blood samples were drawn prior to dosing on day 0, at 1 and 4 h after dosing, and at 1, 7, 14, 21, 28, 42, and 56 days. The concentrations of the OST577-IgG2M3 wildtype and mutant antibodies in rhesus serum samples were determined using a qualified enzyme-linked immunosorbent assay. Appropriately diluted serum samples and calibrators were captured with a mouse anti-OST577-IgG1 idiotype monoclonal antibody (OST577-␥1 anti-id; Protein Design Labs, Inc.) and detected with horseradish peroxidaseconjugated goat anti-human light chain antibody (Southern Biotechnology Associates, Birmingham, AL). The serum antibody concentration data were fitted with a two-compartment model using WinNonlin Enterprise Edition, version 3.2 (Pharsight, Mountain View, CA). RESULTS AND DISCUSSION
Identification of IgG Mutants with Altered Binding to FcRn—Molecular models of the human Fc䡠FcRn complex (Fig. 1) guided the selection of positions 250, 314, and 428 of the human IgG heavy chain for mutagenesis. Although the wildtype amino acids at these positions are located near the
Fc䡠FcRn interface, it does not appear likely that they directly contribute to the pH-dependent interaction between Fc and FcRn. Inspection of the molecular models suggested that amino acid substitutions at these positions might increase or decrease the affinity of Fc for FcRn, without disrupting pH-dependent binding and release, by affecting the conformation of Fc amino acids that do interact with FcRn. Since the amino acids at and around these positions are conserved among all four human IgG subtypes (23), it is reasonable to expect that the FcRn binding phenotype resulting from an amino acid substitution in one IgG subtype could be transferred to the other three IgG subtypes. In this study, an IgG2M3 form (26) of the human anti-HBV monoclonal antibody OST577 (27) was chosen for mutagenesis because it would not be expected to bind either to antigen or Fc␥ receptors in HBV-free primates. PCR mutagenesis was used to generate all 19 single amino acid substitutions at each position. Transiently expressed OST577-IgG2M3 mutants were screened for binding to NS0HuFcRn cells in a single-point competitive binding assay (Fig. 2). Several of the mutants at positions 250 (e.g. Glu and Gln) (Fig. 2A) and 428 (e.g. Phe and Leu) (Fig. 2C) appeared to be stronger competitors in this assay than the wild-type antibody, indicating that these mutants have increased binding to FcRn at pH 6.0. At position 250, it appears that the presence of a hydrogen bond acceptor in the appropriate geometry (e.g. Glu or Gln) is important for better FcRn binding, since smaller but chemically related residues (e.g. Asp or Asn) reduced binding to FcRn. At position 428, a large hydrophobic amino acid confers better FcRn binding. None of the mutations at position 314 (Fig. 2B) resulted in increased binding to FcRn. In a previous study (13), human IgG1 Fc position 428 was randomly mutated, and mutants were screened for binding to mouse FcRn by phage display; however, no mutants at this position with increased binding to mouse FcRn were obtained. This could be because of biases in the construction of the library, the loss of certain sequences during propagation in Escherichia coli, or incomplete sampling of the library during screening. Another explanation is that mouse rather than human FcRn was used to pan the human IgG1 Fc mutant library in the previous study (13). Because the amino acids at and around position 428 are not conserved between the mouse and human Fc regions, it follows that an Fc mutant at position 428 may interact differently with mouse and human FcRn. The best competitors among the mutants at positions 250 and 428 were stably expressed either alone (T250Q, M428L) or in combination (T250Q/M428L), and purified antibodies were compared in a competitive binding assay to human FcRn (Fig.
Engineered Antibodies with Longer Serum Half-lives
6215
FIG. 3. Competitive binding assay of OST577-IgG2M3 wild-type and mutant antibodies to human FcRn. The binding of biotinylated OST577-IgG2M3 antibody to NS0-HuFcRn cells in the presence of increasing concentrations of purified OST577-IgG2M3 wild-type (●), T250Q (ƒ), M428L (‚), or T250Q/M428L (䡺) competitor antibodies in FBB, pH 6.0, was detected with streptavidin-conjugated R-PE and analyzed by flow cytometry. The mean channel fluorescence was plotted against the competitor concentration (g/ml) for each antibody. The data are representative of three independent experiments.
FIG. 2. Single-point competitive binding assay of OST577IgG2M3 wild-type and mutant antibodies to human FcRn. The binding of biotinylated OST577-IgG2M3 antibody to NS0-HuFcRn cells in the presence of wild-type or mutant OST577-IgG2M3 competitor antibodies in FBB, pH 6.0, was detected with streptavidin-conjugated R-PE and analyzed by flow cytometry. The mean channel fluorescence and standard deviation of replicate samples are shown for wild-type OST577-IgG2M3 (far left bar in each panel) compared with all 19 mutants at positions 250 (A), 314 (B), or 428 (C). The data in each panel are representative of two independent experiments.
3). Comparison of the IC50 values indicated that the single mutants T250Q and M428L showed an increase in binding to human FcRn at pH 6.0 of ⬃3- and 7-fold, respectively, whereas the double mutant T250Q/M428L showed an increase in binding of ⬃28-fold. To confirm that binding was pH-dependent, the antibodies were allowed to bind to NS0-HuFcRn cells at pH 6.0 and were removed by washing the cells at pH 6.0, 6.5, 7.0, 7.5, or 8.0. As the pH value of the washes in successive samples was raised from pH 6.0 to 8.0, the binding of the wild-type and mutant antibodies to human FcRn was comparably diminished, with essentially no binding observed at pH 7.5 or above (data not shown). Similar results were obtained when these mutant antibodies were tested for binding to rhesus FcRn. The binding of the T250Q, M428L, and T250Q/M428L IgG2M3 mutants to rhesus FcRn at pH 6.0 was ⬃4-, 8-, and 27-fold better than the wild-type antibody, respectively, and binding to rhesus FcRn was pH-dependent (data not shown). Pharmacokinetics of IgG Wild-type and Mutant Antibodies in Rhesus Monkeys—The PK behavior of the OST577-IgG2M3
FIG. 4. Pharmacokinetics of OST577-IgG2M3 wild-type, M428L, and T250Q/M428L antibodies in rhesus monkeys. The modeled data (simulated based on each group’s geometric mean of the primary pharmacokinetic parameters) as well as the observed mean serum antibody concentration (g/ml) and the standard deviation for each group of four animals were plotted as a function of time (days after infusion) for the OST577-IgG2M3 wild-type (●), M428L (‚), or T250Q/ M428L (䡺) antibodies.
wild-type, M428L, and T250Q/M428L antibodies was examined in rhesus monkeys. The PK profiles of the two mutants are clearly distinct from that of the wild-type (Fig. 4). The mean serum antibody concentrations of the M428L and T250Q/ M428L mutants were maintained at higher levels than wildtype OST577-IgG2M3 at all time points. Because the mean maximum serum antibody concentration (Cmax, Table I) was very similar among the three test groups, indicating that the administered antibodies were distributed to the circulation in a similar manner, the higher concentrations of the mutant IgG2M3 antibodies thereafter are attributable to their increased persistence in the serum. Analysis of the mean clearance (CL) indicated that this was the case. The mean CL, the volume of serum antibody cleared per unit of time, was ⬃1.8fold lower for the M428L mutant (0.0811 ⫾ 0.0384 ml/h/kg; p ⫽ 0.057), and ⬃2.8-fold lower for the T250Q/M428L mutant (0.0514 ⫾ 0.0075 ml/h/kg; p ⫽ 0.029) compared with wild-type OST577-IgG2M3 (0.144 ⫾ 0.047 ml/h/kg) (Table I), indicating a significant decrease in the clearance of the OST577-IgG2M3 M428L and T250Q/M428L mutants. The PK profiles of the OST577-IgG2M3 wild-type and mu-
6216
Engineered Antibodies with Longer Serum Half-lives
TABLE I Summary of pharmacokinetic parameters Pharmacokinetic parameters were calculated using WinNonlin (Pharsight). The group mean ⫾ S.D. is shown for each parameter. Mann-Whitney tests were done using GraphPad Prism (GraphPad Software) to compare the statistical significance of differences in the pharmacokinetic parameters between the wild-type (WT) group and each mutant group. Cmax, maximum serum antibody concentration; t1⁄2, elimination (-phase) half-life. *, indicates a significant difference (p ⬍ 0.060). OST577-IgG2M3
Cmax
CL
g/ml
ml/h/kg
h䡠g/ml
h
WT M428L T250Q/M428L
36.7 ⫾ 12.8 36.5 ⫾ 20.1 39.9 ⫾ 6.8
0.144 ⫾ 0.047 0.0811 ⫾ 0.0384* 0.0514 ⫾ 0.0075*
7710 ⫾ 3110 15200 ⫾ 8700* 19800 ⫾ 2900*
351 ⫾ 121 642 ⫾ 205 652 ⫾ 28*
tant antibodies were further analyzed by calculating other parameters. Since the area under the curve (AUC) is inversely proportional to CL, it follows that the mean AUC, the area under the concentration-time curve extrapolated from time zero to infinity, was ⬃2.0-fold higher for the M428L mutant (15,200 ⫾ 8,700 h䡠g/ml; p ⫽ 0.057) and ⬃2.6-fold higher for the T250Q/M428L mutant (19,800 ⫾ 2,900 h䡠g/ml; p ⫽ 0.029) compared with wild-type OST577-IgG2M3 (7,710 ⫾ 3,110 h䡠g/ ml) (Table I), indicating a significant increase in the total exposure of the OST577-IgG2M3 M428L and T250Q/M428L mutants. Finally, the mean elimination (-phase) half-life (t1⁄2) was ⬃1.8-fold longer for the M428L mutant (642 ⫾ 205 h) and ⬃1.9-fold longer for the T250Q/M428L mutant (652 ⫾ 28 h; p ⫽ 0.029) compared with wild-type OST577-IgG2M3 (351 ⫾ 121 h) (Table I). The elimination half-life for wild-type OST577IgG2M3 in this study is similar to that for OST577-IgG1 (324 ⫾ 85 h) in a previous PK study in rhesus monkeys (27). Although there is an indication from the CL and AUC parameters that the T250Q/M428L mutant may have increased serum persistence compared with the M428L mutant, this difference may not be significant. Since the elimination half-lives of the two mutants appear similar, it is possible that a maximal increase in serum persistence has been achieved in rhesus monkeys with these mutants. The T250Q mutant, which showed a modest increase in binding to rhesus and human FcRn, may be expected to show an intermediate increase in serum half-life in primates. While the M428L and T250Q/M428L amino acid substitutions are presumed to account for the observed increase in the elimination half-lives of the IgG2M3 antibodies, it is possible that post-translational modifications (e.g. glycosylation, Met oxidation) might affect their PK properties. However, it is unlikely that glycosylation differences would account for the observed half-life differences because previous studies have indicated that glycosylated and aglycosylated IgG antibodies have similar half-lives in mice (32) and chimpanzees (33). Moreover, carbohydrate analysis of the OST577-IgG2M3 wildtype and mutant antibodies revealed only minor differences in their glycosylation patterns (data not shown). Engineered antibodies with increased serum half-lives might prove valuable in antibody therapy. For example, it may be possible to reduce the frequency of administration of such antibodies. This will be a great benefit to patients undergoing long-term antibody therapy. Based on the results described in this study, it is reasonable to expect that human IgG1, IgG3, and IgG4 antibodies with longer serum half-lives also may be engineered by transferring the M428L or T250Q/M428L mutations into these IgG subtypes. In addition, it should now also be possible to alter the serum half-lives of other IgG-related therapeutics such as IgG Fc fusion proteins using this approach. Provided that the mutations described in this study do not substantially increase the immunogenicity of these therapeutics in humans, IgG antibodies and Fc fusion proteins with
AUC
t1⁄2
longer serum half-lives should represent a potent new class of human therapeutics. Acknowledgments—We thank Tina Balsara, Michael Cole, Sharyn Farnsworth, Brett Jorgensen, Sheri Kostelny, Lili Liu, David Maciejewski, Julie Mikkelsen, Paul Motchnik, Tom Robinson, Matt Sweeney, Mark Wesson, Don Young, and Wenge Zhang for valuable assistance and discussion. We also thank Rich Murray and Cary Queen for comments on the manuscript. REFERENCES 1. Waldmann, T. A. (2003) Nat. Med. 9, 269 –277 2. Morrison, S. L., Johnson, M. J., Herzenberg, L. A., and Oi, V. T. (1984) Proc. Natl. Acad. Sci. U. S. A. 81, 6851– 6855 3. Jones, P. T., Dear, P. H., Foote, J., Neuberger, M. S., and Winter, G. (1986) Nature 321, 522–525 4. Hoogenboom, H. R., and Chames, P. (2000) Immunol. Today 21, 371–378 5. Idusogie, E. E., Wong, P. Y., Presta, L. G., Gazzano-Santoro, H., Totpal, K., Ultsch, M., and Mulkerrin, M. G. (2001) J. Immunol. 166, 2571–2575 6. Shields, R. L., Namenuk, A. K., Hong, K., Meng, Y. G., Rae, J., Briggs, J., Xie, D., Lai, J., Stadlen, A., Li, B., Fox, J. A., and Presta, L. G. (2001) J. Biol. Chem. 276, 6591– 6604 7. Simister, N. E., and Mostov, K. E. (1989) Nature 337, 184 –187 8. Ghetie, V., and Ward, E. S. (2000) Annu. Rev. Immunol. 18, 739 –766 9. Ghetie, V., Hubbard, J. G., Kim, J. K., Tsen, M. F., Lee, Y., and Ward, E. S. (1996) Eur. J. Immunol. 26, 690 – 696 10. Junghans, R. P., and Anderson, C. L. (1996) Proc. Natl. Acad. Sci. U. S. A. 93, 5512–5516 11. Israel, E. J., Wilsker, D. F., Hayes, K. C., Schoenfeld, D., and Simister, N. E. (1996) Immunology 89, 573–578 12. Ghetie, V., Popov, S., Borvak, J., Radu, C., Matesoi, D., Medesan, C., Ober, R. J., and Ward, E. S. (1997) Nat. Biotechnol. 15, 637– 640 13. Dall’Acqua, W. F., Woods, R. M., Ward, E. S., Palaszynski, S. R., Patel, N. K., Brewah, Y. A., Wu, H., Kiener, P. A., and Langermann, S. (2002) J. Immunol. 169, 5171–5180 14. Raghavan, M., Bonagura, V. R., Morrison, S. L., and Bjorkman, P. J. (1995) Biochemistry 34, 14649 –14657 15. Kim, J. K., Tsen, M. F., Ghetie, V., and Ward, E. S. (1994) Eur. J. Immunol. 24, 542–548 16. Medesan, C., Matesoi, D., Radu, C., Ghetie, V., and Ward, E. S. (1997) J. Immunol. 158, 2211–2217 17. Kim, J. K., Firan, M., Radu, C. G., Kim, C. H., Ghetie, V., and Ward, E. S. (1999) Eur. J. Immunol. 29, 2819 –2825 18. Martin, W. L., West, A. P., Jr., Gan, L., and Bjorkman, P. J. (2001) Mol. Cell 7, 867– 877 19. Burmeister, W. P., Huber, A. H., and Bjorkman, P. J. (1994) Nature 372, 379 –383 20. Weng, Z., Gulukota, K., Vaughn, D. E., Bjorkman, P. J., and DeLisi, C. (1998) J. Mol. Biol. 282, 217–225 21. Saper, M. A., Bjorkman, P. J., and Wiley, D. C. (1991) J. Mol. Biol. 219, 277–319 22. West, A. P., Jr., and Bjorkman, P. J. (2000) Biochemistry 39, 9698 –9708 23. Kabat, E. A., Wu, T. T., Perry, H. M., Gottesman, K. S., and Foeller, C. (1991) Sequences of Proteins of Immunological Interest, 5th Ed., National Institutes of Health, Bethesda, MD 24. Levitt, M. (1992) J. Mol. Biol. 226, 507–533 25. Levitt, M. (1983) J. Mol. Biol. 168, 595– 620 26. Cole, M. S., Anasetti, C., and Tso, J. Y. (1997) J. Immunol. 159, 3613–3621 27. Ehrlich, P. H., Moustafa, Z. A., Justice, J. C., Harfeldt, K. E., Kelley, R. L., and Ostberg, L. (1992) Hum. Antibodies Hybridomas 3, 2–7 28. Co, M. S., Avdalovic, N. M., Caron, P. C., Avdalovic, M. V., Scheinberg, D. A., and Queen, C. (1992) J. Immunol. 148, 1149 –1154 29. Higuchi, R. (1989) in PCR Technology: Principles and Applications for DNA Amplification (Erlich, H. A., ed) pp. 61–70, Stockton Press, New York 30. Yanisch-Perron, C., Vieira, J., and Messing, J. (1985) Gene 33, 103–119 31. Caras, I. W., Davitz, M. A., Rhee, L., Weddell, G., Martin, D. W., Jr., and Nussenzweig, V. (1987) Nature 325, 545–549 32. Tao, M. H., and Morrison, S. L. (1989) J. Immunol. 143, 2595–2601 33. Simmons, L. C., Reilly, D., Klimowski, L., Raju, T. S., Meng, G., Sims, P., Hong, K., Shields, R. L., Damico, L. A., Rancatore, P., and Yansura, D. G. (2002) J. Immunol. Methods 263, 133–147
