Abstract
Conventional methods for humanizing animal-derived antibodies involve grafting their complementarity-determining regions onto homologous human framework regions. However, this process can substantially lower antibody stability and antigen-binding affinity, and requires iterative mutational fine-tuning to recover the original antibody properties. Here we report a computational method for the systematic grafting of animal complementarity-determining regions onto thousands of human frameworks. The method, which we named CUMAb (for computational human antibody design; available at http://CUMAb.weizmann.ac.il), starts from an experimental or model antibody structure and uses Rosetta atomistic simulations to select designs by energy and structural integrity. CUMAb-designed humanized versions of five antibodies exhibited similar affinities to those of the parental animal antibodies, with some designs showing marked improvement in stability. We also show that (1) non-homologous frameworks are often preferred to highest-homology frameworks, and (2) several CUMAb designs that differ by dozens of mutations and that use different human frameworks are functionally equivalent.
This is a preview of subscription content, access via your institution
Access options
Access Nature and 54 other Nature Portfolio journals
Get Nature+, our best-value online-access subscription
$29.99 / 30 days
cancel any time
Subscribe to this journal
Receive 12 digital issues and online access to articles
$99.00 per year
only $8.25 per issue
Buy this article
- Purchase on Springer Link
- Instant access to full article PDF
Prices may be subject to local taxes which are calculated during checkout
Similar content being viewed by others
Data availability
All data generated and analysed during the study are available within the paper and its Supplementary Information, except for human antibody germline sequences, which were taken from the IMGT reference database (https://www.imgt.org/vquest/refseqh.html), and sequences for FDA-approved humanized antibodies, which were taken from the TheraSAbDab database (http://opig.stats.ox.ac.uk/webapps/newsabdab/therasabdab/search). The crystal structure of hαQSOX1.4 is available through the Protein Data Bank (PDB; https://www.rcsb.org), with accession ID 8AON. The crystal structures of mαQSOX1 and mαHEWL are available through the PDB, with accession IDs 4IJ3 and 1MLC, respectively. Source data are provided with this paper.
Code availability
RosettaScripts59 xml files for running CUMAb and code for generating humanized designs can be found at https://github.com/Fleishman-Lab/CUMAb, with detailed explanations.
References
Raybould, M. I. J. et al. Thera-SAbDab: the therapeutic structural antibody database. Nucleic Acids Res. 48, D383–D388 (2020).
Schroff, R. W., Foon, K. A., Beatty, S. M., Oldham, R. K. & Morgan, A. C. Jr Human anti-murine immunoglobulin responses in patients receiving monoclonal antibody therapy. Cancer Res. 45, 879–885 (1985).
Shawler, D. L., Bartholomew, R. M., Smith, L. M. & Dillman, R. O. Human immune response to multiple injections of murine monoclonal IgG. J. Immunol. 135, 1530–1535 (1985).
Hwang, W. Y. K. & Foote, J. Immunogenicity of engineered antibodies. Methods 36, 3–10 (2005).
Kuramochi, T., Igawa, T., Tsunoda, H. & Hattori, K. in Human Monoclonal Antibodies: Methods and Protocols (ed. Steinitz, M.) 213–230 (Springer, 2019).
Saldanha, J. W., Martin, A. C. & Léger, O. J. A single backmutation in the human kIV framework of a previously unsuccessfully humanized antibody restores the binding activity and increases the secretion in cos cells. Mol. Immunol. 36, 709–719 (1999).
Jones, P. T., Dear, P. H., Foote, J., Neuberger, M. S. & Winter, G. Replacing the complementarity-determining regions in a human antibody with those from a mouse. Nature 321, 522–525 (1986).
Baca, M., Presta, L. G., O’Connor, S. J. & Wells, J. A. Antibody humanization using monovalent phage display. J. Biol. Chem. 272, 10678–10684 (1997).
Lazar, G. A., Desjarlais, J. R., Jacinto, J., Karki, S. & Hammond, P. W. A molecular immunology approach to antibody humanization and functional optimization. Mol. Immunol. 44, 1986–1998 (2007).
Padlan, E. A. A possible procedure for reducing the immunogenicity of antibody variable domains while preserving their ligand-binding properties. Mol. Immunol. 28, 489–498 (1991).
Choi, Y., Hua, C., Sentman, C. L., Ackerman, M. E. & Bailey-Kellogg, C. Antibody humanization by structure-based computational protein design. MAbs 7, 1045–1057 (2015).
Kipriyanov, S. M., Moldenhauer, G., Martin, A. C., Kupriyanova, O. A. & Little, M. Two amino acid mutations in an anti-human CD3 single chain Fv antibody fragment that affect the yield on bacterial secretion but not the affinity. Protein Eng. 10, 445–453 (1997).
Safdari, Y., Farajnia, S., Asgharzadeh, M. & Khalili, M. Antibody humanization methods—a review and update. Biotechnol. Genet. Eng. Rev. 29, 175–186 (2013).
Warszawski, S. et al. Optimizing antibody affinity and stability by the automated design of the variable light-heavy chain interfaces. PLoS Comput. Biol. 15, e1007207 (2019).
Shire, S. J., Shahrokh, Z. & Liu, J. Challenges in the development of high protein concentration formulations. J. Pharm. Sci. 93, 1390–1402 (2004).
Foote, J. & Winter, G. Antibody framework residues affecting the conformation of the hypervariable loops. J. Mol. Biol. 224, 487–499 (1992).
Makabe, K. et al. Thermodynamic consequences of mutations in vernier zone residues of a humanized anti-human epidermal growth factor receptor murine antibody, 528. J. Biol. Chem. 283, 1156–1166 (2008).
Janeway, C. A., Travers, P., Walport, M. & Shlomchik, M. J. Immunobiology: The Immune System in Health and Disease (Garland Science, 2005).
Giudicelli, V., Chaume, D. & Lefranc, M.-P. IMGT/GENE-DB: a comprehensive database for human and mouse immunoglobulin and T cell receptor genes. Nucleic Acids Res. 33, D256–D261 (2005).
MacCallum, R. M., Martin, A. C. & Thornton, J. M. Antibody-antigen interactions: contact analysis and binding site topography. J. Mol. Biol. 262, 732–745 (1996).
Dondelinger, M. et al. Understanding the significance and implications of antibody numbering and antigen-binding surface/residue definition. Front. Immunol. 9, 2278 (2018).
Leaver-Fay, A. et al. ROSETTA3: an object-oriented software suite for the simulation and design of macromolecules. Methods Enzymol. 487, 545–574 (2011).
O’Meara, M. J. et al. A combined covalent-electrostatic model of hydrogen bonding improves structure prediction with Rosetta. J. Chem. Theory Comput. 11, 609–622 (2015).
Grossman, I., Alon, A., Ilani, T. & Fass, D. An inhibitory antibody blocks the first step in the dithiol/disulfide relay mechanism of the enzyme QSOX1. J. Mol. Biol. 425, 4366–4378 (2013).
Feldman, T. et al. Inhibition of fibroblast secreted QSOX1 perturbs extracellular matrix in the tumor microenvironment and decreases tumor growth and metastasis in murine cancer models. Oncotarget 11, 386–398 (2020).
Marks, C., Hummer, A. M., Chin, M. & Deane, C. M. Humanization of antibodies using a machine learning approach on large-scale repertoire data. Bioinformatics 37, 4041–4047 (2021).
Prihoda, D. et al. BioPhi: a platform for antibody design, humanization, and humanness evaluation based on natural antibody repertoires and deep learning. MAbs 14, 2020203 (2022).
Starr, C. G. et al. Ultradilute measurements of self-association for the identification of antibodies with favorable high-concentration solution properties. Mol. Pharm. 18, 2744–2753 (2021).
Makowski, E. K., Wu, L., Desai, A. A. & Tessier, P. M. Highly sensitive detection of antibody nonspecific interactions using flow cytometry. MAbs 13, 1951426 (2021).
Xu, Y. et al. Addressing polyspecificity of antibodies selected from an in vitro yeast presentation system: a FACS-based, high-throughput selection and analytical tool. Protein Eng. Des. Sel. 26, 663–670 (2013).
Kelly, R. L. et al. Chaperone proteins as single component reagents to assess antibody nonspecificity. MAbs 9, 1036–1040 (2017).
Kelly, R. L. et al. High throughput cross-interaction measures for human IgG1 antibodies correlate with clearance rates in mice. MAbs 7, 770–777 (2015).
Kingsbury, J. S. et al. A single molecular descriptor to predict solution behavior of therapeutic antibodies. Sci. Adv. 6, eabb0372 (2020).
Almagro, J. C. et al. Second antibody modeling assessment (AMA-II). Proteins Struct. Funct. Bioinf. 82, 1553–1562 (2014).
Sircar, A., Kim, E. T. & Gray, J. J. RosettaAntibody: antibody variable region homology modeling server. Nucleic Acids Res. 37, W474–W479 (2009).
Dunbar, J. et al. SAbPred: a structure-based antibody prediction server. Nucleic Acids Res. 44, W474–W478 (2016).
Leem, J., Dunbar, J., Georges, G., Shi, J. & Deane, C. M. ABodyBuilder: automated antibody structure prediction with data-driven accuracy estimation. MAbs 8, 1259–1268 (2016).
Norn, C. H., Lapidoth, G. & Fleishman, S. J. High-accuracy modeling of antibody structures by a search for minimum-energy recombination of backbone fragments. Proteins 85, 30–38 (2017).
Kodali, P., Schoeder, C. T., Schmitz, S., Crowe, J. E. Jr & Meiler, J. RosettaCM for antibodies with very long HCDR3s and low template availability. Proteins https://doi.org/10.1002/prot.26166 (2021).
Mirdita, M. et al. ColabFold: making protein folding accessible to all. Nat. Methods 19, 679–682 (2022).
Evans, R. et al. Protein complex prediction with AlphaFold-Multimer. Preprint at bioRxiv https://doi.org/10.1101/2021.10.04.463034 (2022).
O’Brien, T. J. et al. The CA 125 gene: an extracellular superstructure dominated by repeat sequences. Tumour Biol. 22, 348–366 (2001).
Duffy, M. J. et al. CA125 in ovarian cancer: European Group on Tumor Markers guidelines for clinical use. Int. J. Gynecol. Cancer 15, 679–691 (2005).
Lloyd, K. O. & Yin, B. W. Synthesis and secretion of the ovarian cancer antigen CA 125 by the human cancer cell line NIH:OVCAR-3. Tumour Biol. 22, 77–82 (2001).
Kashmiri, S. V. S., De Pascalis, R., Gonzales, N. R. & Schlom, J. SDR grafting—a new approach to antibody humanization. Methods 36, 25–34 (2005).
De Pascalis, R. et al. Grafting of ‘abbreviated’ complementarity-determining regions containing specificity-determining residues essential for ligand contact to engineer a less immunogenic humanized monoclonal antibody. J. Immunol. 169, 3076–3084 (2002).
Braden, B. C. et al. Three-dimensional structures of the free and the antigen-complexed Fab from monoclonal anti-lysozyme antibody D44.1. J. Mol. Biol. 243, 767–781 (1994).
Chao, G. et al. Isolating and engineering human antibodies using yeast surface display. Nat. Protoc. 1, 755–768 (2006).
Knappik, A. et al. Fully synthetic human combinatorial antibody libraries (HuCAL) based on modular consensus frameworks and CDRs randomized with trinucleotides. J. Mol. Biol. 296, 57–86 (2000).
Apgar, J. R. et al. Beyond CDR-grafting: structure-guided humanization of framework and CDR regions of an anti-myostatin antibody. MAbs 8, 1302–1318 (2016).
Lowe, D. et al. Aggregation, stability, and formulation of human antibody therapeutics. Adv. Protein Chem. Struct. Biol. 84, 41–61 (2011).
Dudgeon, K. et al. General strategy for the generation of human antibody variable domains with increased aggregation resistance. Proc. Natl Acad. Sci. USA 109, 10879–10884 (2012).
Perchiacca, J. M., Bhattacharya, M. & Tessier, P. M. Mutational analysis of domain antibodies reveals aggregation hotspots within and near the complementarity determining regions. Proteins 79, 2637–2647 (2011).
Wu, T. T., Johnson, G. & Kabat, E. A. Length distribution of CDRH3 in antibodies. Proteins 16, 1–7 (1993).
Raybould, M. I. J. et al. Five computational developability guidelines for therapeutic antibody profiling. Proc. Natl Acad. Sci. USA 116, 4025–4030 (2019).
Baran, D. et al. Principles for computational design of binding antibodies. Proc. Natl Acad. Sci. USA 114, 10900–10905 (2017).
Borenstein-Katz, A. et al. Biomolecular recognition of the glycan neoantigen CA19-9 by distinct antibodies. J. Mol. Biol. 433, 167099 (2021).
Eddy, S. R. Profile hidden Markov models. Bioinformatics 14, 755–763 (1998).
Fleishman, S. J. et al. RosettaScripts: a scripting language interface to the Rosetta macromolecular modeling suite. PLoS ONE 6, e20161 (2011).
Ye, J., Ma, N., Madden, T. L. & Ostell, J. M. IgBLAST: an immunoglobulin variable domain sequence analysis tool. Nucleic Acids Res. 41, W34–W40 (2013).
Baker, N. A., Sept, D., Joseph, S., Holst, M. J. & McCammon, J. A. Electrostatics of nanosystems: application to microtubules and the ribosome. Proc. Natl Acad. Sci. USA 98, 10037–10041 (2001).
Madeira, F. et al. Search and sequence analysis tools services from EMBL-EBI in 2022. Nucleic Acids Res. 50, W276–W279 (2022).
Waterhouse, A. M., Procter, J. B., Martin, D. M. A., Clamp, M. & Barton, G. J. Jalview Version 2—a multiple sequence alignment editor and analysis workbench. Bioinformatics 25, 1189–1191 (2009).
Alon, A. et al. The dynamic disulphide relay of quiescin sulphydryl oxidase. Nature 488, 414–418 (2012).
Javitt, G., Kinzel, A., Reznik, N. & Fass, D. Conformational switches and redox properties of the colon cancer‐associated human lectin ZG16. FEBS J. 288, 6465–6475 (2021).
Adams, P. D. et al. PHENIX: a comprehensive Python-based system for macromolecular structure solution. Acta Crystallogr. D 66, 213–221 (2010).
Emsley, P., Lohkamp, B., Scott, W. G. & Cowtan, K. Features and development of Coot. Acta Crystallogr. D 66, 486–501 (2010).
Noronha, A. et al. AXL and error-prone DNA replication confer drug resistance and offer strategies to treat EGFR-mutant lung cancer. Cancer Discov. 12, 2666–2683 (2022).
Chen, Y. et al. Armed antibodies targeting the mucin repeats of the ovarian cancer antigen, MUC16, are highly efficacious in animal tumor models. Cancer Res. 67, 4924–4932 (2007).
Liang, W.-C. et al. Cross-species vascular endothelial growth factor (VEGF)-blocking antibodies completely inhibit the growth of human tumor xenografts and measure the contribution of stromal VEGF. J. Biol. Chem. 281, 951–961 (2006).
Acknowledgements
We thank A. Mechaly (Israel Institute for Biological Research, Department of Infectious Diseases) for a critical reading of the paper, and R. Diskin (Weizmann Institute of Science, Department of Chemical and Structural Biology) and O. Khersonsky (Weizmann Institute of Science, Department of Biomolecular Sciences) for advice. Research in the Fleishman lab was supported by the European Research Council through a Consolidator Award (815379), the Dr Barry Sherman Institute for Medicinal Chemistry, and a donation in memory of Sam Switzer. Research in the Fass lab was supported by the European Research Council through a Proof-of-Concept grant (825076). Research in the Tessier lab was supported by the National Institutes of Health (RF1AG059723 and R35GM136300) and the National Science Foundation (1804313). We acknowledge the European Synchrotron Radiation Facility for the provision of beam time on ID30B, and A. McCarthy for assistance. The collaboration between the Yarden and Fleishman labs was supported by the Weizmann Institute’s BINA framework. Funding for this research was provided by Teva Pharmaceutical Industries Ltd as part of the Israeli National Forum for BioInnovators (NFBI). This work was supported in part by a grant from the Manya Igel Centre for Biomedical Engineering and Signal Processing and the Moross Integrated Cancer Center.
Author information
Authors and Affiliations
Contributions
A.T. and S.J.F. wrote the paper, with input from all authors. A.T. developed the humanization algorithm and designed all humanized sequences. L.K. performed experimental work related to anti-QSOX1 antibodies except for the developability experiments. L.K., N.Y. and D.F. performed crystallography-data collection, and L.K. and D.F. analysed the data. E.K.M. performed developability experiments for anti-QSOX1 antibodies, and E.K.M. and P.M.T. analysed the data. A.N. performed the IP assay for anti-AXL antibody. M.L. and I.Z. expressed and purified anti-AXL antibodies. R.K. performed experimental work related to anti-MUC16 antibodies, and R.K. and J.A. analysed the data. A.T. and L.K. expressed and purified anti-HEWL antibodies. A.T., Y.F.S. and Y.G.-W. performed nano-DSF and SPR characterization of anti-AXL antibodies and analysed the data together with S.J.F., M.L. and Y.Y. A.T., Y.F.S. and Y.G.-W. performed nano-DSF and SPR characterization of anti-HEWL antibodies and analysed the data together with S.J.F.
Corresponding author
Ethics declarations
Competing interests
A.T. and S.J.F. are named inventors in a patent filing on the CUMAb method (WO2023012807A1). A.T., S.J.F., L.K. and D.F. are named inventors in a patent filing on humanized anti-QSOX1 designs. The other authors declare no competing interests.
Peer review
Peer review information
Nature Biomedical Engineering thanks Charlotte Deane and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Extended data
Extended Data Fig. 1 Additional biophysical and structural characterization of humanized anti-QSOX1 antibodies.
a, Dot-blot analysis of 15 CUMAb designs. 12 designs exhibit comparable expression levels to the AbLIFT18 design of the anti-QSOX1 antibody14. CUMAb designs are labelled according to heavy chain V-gene subgroup and then by light chain subgroup according to IMGT. If there are two chains from the same subgroup, they are labelled as ‘a’ or ‘b’ arbitrarily. ‘Mock’ refers to a control transfection where no DNA was added. ‘Prev’ refers to a construct from a prior version of CUMAb. b, Functional screen of 15 CUMAb designs, labelled as in A. Presence of an upper band, corresponding to MBP-ZG16 modified with PEG-mal 5000, indicates remaining unoxidized MBP-ZG16 as a result of inhibited QSOX1 activity. As evident from the ‘no QSOX1’ lane, the MBP-ZG16 used in the experiment was not fully reduced, but the reduced population was sufficient to distinguish between the mock-transfected sample, which lacks antibody, and the antibody-containing samples, which all show some PEG-mal-modified species indicative of QSOX1 inhibition. Designs marked with * were further analyzed. c, Titration of relative QSOX1 activity at different antibody concentrations for mαQSOX1 and four CUMAb designs (aggregate data shown in main Fig. 2B). Points represent the average of two technical repeats and errors bars represent the standard deviation. d, Thermal denaturation of mαQSOX1 and four CUMAb designs using nano differential scanning fluorimetry. Shown is the average of two technical repeats. e, Crystal structure of oxidoreductase fragment of human QSOX1 in complex with parental mouse antibody (QSOX1 in dark gray, parental mouse antibody Fv in light gray, PDB entry 4IJ3) aligned with crystal structure of oxidoreductase fragment of human QSOX1 in complex with hαQSOX1.4 (QSOX1 in purple, hαQSOX1.4 Fv in magenta, PDB entry 8AON. Structures are nearly identical (0.7 Å) despite 51 Fv mutations.
Extended Data Fig. 2 Additional controls for the FACS analyses of anti-MUC16 antibodies.
a, Non-transfected HEK293 cells were stained with a commercial anti-tag antibody, cαMUC16_Ab1, hαMUC16_Ab1, cαMUC16_Ab2, hαMUC16_Ab2.1, and hαMUC16_Ab2.2, demonstrating that none of the five antibodies showed significant binding to cells not expressing the MUC16 construct. b, HEK293 cells expressing the MUC16 construct were stained either without a primary antibody or with a commercial anti-tag antibody, demonstrating that the secondary antibody alone does not label the cells and that the cells do display the intended construct. c, Acetone-fixed OVCAR-3 cells were stained either without a primary antibody or with a control anti-MUC16 antibody69, demonstrating that the secondary antibody alone does not label the cells and that the cells do display MUC16.
Extended Data Fig. 3 Yeast surface-display expression and binding analyses of anti-HEWL antibodies.
a, Expression levels of negative control (the G6 anti-VEGF70 antibody), mouse anti-HEWL antibody (mαHEWL), and 6 CUMAb designs, all formatted as scFv. b, Binding levels to HEWL of the eight antibodies.
Extended Data Fig. 4 Sequence alignments of V genes to the human germline V genes.
Sequence alignments of V genes in successful (a) anti-QSOX1 and (b) anti-HEWL designs to the human germline V genes with the highest sequence identity.
Supplementary information
Supplementary Tables
Supplementary Tables 1–9.
Source data
Source Data Figs. 2–5 and Extended Data Fig. 1
Source data and unprocessed gels.
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Tennenhouse, A., Khmelnitsky, L., Khalaila, R. et al. Computational optimization of antibody humanness and stability by systematic energy-based ranking. Nat. Biomed. Eng 8, 30–44 (2024). https://doi.org/10.1038/s41551-023-01079-1
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1038/s41551-023-01079-1
This article is cited by
-
Opportunities and challenges in design and optimization of protein function
Nature Reviews Molecular Cell Biology (2024)
-
Designing stable humanized antibodies
Nature Biomedical Engineering (2023)