科技科学

Computational design of metalloproteases

Nature正文已收录本站

Data availability

All data generated or analysed in this study are available in the paper, its Supplementary Information and at Zenodo (https://doi.org/10.5281/zenodo.22654831)62. Crystal structures have been deposited in the PDB (https://www.rcsb.org/) under accession codes 11CE, 11CM, 11CP, 11CR, 11CU, 11CY and 11DU.

Code availability

The code used in this study is available at Zenodo (https://doi.org/10.5281/zenodo.22654831)62.

References

  1. Studer, S. et al. Evolution of a highly active and enantiospecific metalloenzyme from short peptides. Science 362, 1285–1288 (2018).

    Article  ADS  CAS  PubMed  Google Scholar 

  2. Kim, D. et al. Computational design of metallohydrolases. Nature 649, 246–253 (2026).

    Article  ADS  CAS  PubMed  Google Scholar 

  3. Bauer, M. S. et al. design of phospho-tyrosine peptide binders. Preprint at bioRxiv https://doi.org/10.1101/2025.09.29.678898 (2025).

  4. Butcher, J. et al. Design of all-atom biomolecular interactions with RFdiffusion3. Preprint at bioRxiv https://doi.org/10.1101/2025.09.18.676967 (2025).

  5. Radzicka, A. & Richard, W. Rates of uncatalyzed peptide bond hydrolysis in neutral solution and the transition state affinities of proteases. J. Am. Chem. Soc. 118, 6105–6109 (1996).

    Article  ADS  CAS  Google Scholar 

  6. Hooper, N. M. Families of zinc metalloproteases. FEBS Lett. 354, 1–6 (1994).

    Article  CAS  PubMed  Google Scholar 

  7. Lipscomb, W. N. & Sträter, N. Recent advances in zinc enzymology. Chem. Rev. 96, 2375–2434 (1996).

    Article  CAS  PubMed  Google Scholar 

  8. de Almeida, L. G. N. et al. Matrix metalloproteinases: from molecular mechanisms to physiology, pathophysiology, and pharmacology. Pharmacol. Rev. 74, 714–770 (2022).

    Article  Google Scholar 

  9. Page-McCaw, A., Ewald, A. J. & Werb, Z. Matrix metalloproteinases and the regulation of tissue remodelling. Nat. Rev. Mol. Cell Biol. 8, 221–233 (2007).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  10. Edwards, D. R., Handsley, M. M. & Pennington, C. J. The ADAM metalloproteinases. Mol. Aspects Med. 29, 258–289 (2008).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  11. Wang, Z., Li, W., Chen, S. & Tang, X. X. Role of ADAM and ADAMTS proteases in pathological tissue remodeling. Cell Death Discov. 9, 447 (2023).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  12. Iwai, M. & Horiuchi, M. Devil and angel in the renin-angiotensin system: ACE-angiotensin II-AT1 receptor axis vs. ACE2-angiotensin-(1-7)-Mas receptor axis. Hypertens. Res. 32, 533–536 (2009).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  13. Wang, J., Xu, A., Wan, Y. & Li, Q. Purification and characterization of a new metallo-neutral protease for beer brewing from Bacillus amyloliquefaciens SYB-001. Appl. Biochem. Biotechnol. 170, 2021–2033 (2013).

    Article  CAS  PubMed  Google Scholar 

  14. Jabeen, A., Kandadai, R. M., Kannikannan, M. A. & Borgohain, R. Guidelines for the use of botulinum toxin in movement disorders and spasticity. Ann. Indian Acad. Neurol. 14, S31–S34 (2011).

    Article  PubMed  PubMed Central  Google Scholar 

  15. Leippold, T., Reitz, A. & Schurch, B. Botulinum toxin as a new therapy option for voiding disorders: current state of the art. Eur. Urol. 44, 165–174 (2003).

    Article  CAS  PubMed  Google Scholar 

  16. Satriyasa, B. K. Botulinum toxin (Botox) A for reducing the appearance of facial wrinkles: a literature review of clinical use and pharmacological aspect. Clin. Cosmet. Investig. Dermatol. 12, 223–228 (2019).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  17. Zerner, B., Bond, R. P. M. & Bender, M. L. Kinetic evidence for the formation of acyl-enzyme intermediates in the α-chymotrypsin-catalyzed hydrolyses of specific substrates. J. Am. Chem. Soc. 86, 3674–3679 (1964).

    Article  ADS  CAS  Google Scholar 

  18. Bordwell, F. G. Equilibrium acidities in dimethyl sulfoxide solution. Acc. Chem. Res. 21, 456–463 (1988).

    Article  ADS  CAS  Google Scholar 

  19. Nagy, N. et al. 4-methylumbelliferone treatment and hyaluronan inhibition as a therapeutic strategy in inflammation, autoimmunity, and cancer. Front. Immunol. 6, 123 (2015).

    Article  PubMed  PubMed Central  Google Scholar 

  20. Bailey-Darland, S., Krueger, T. D. & Fang, C. Ultrafast spectroscopies of nitrophenols and nitrophenolates in solution: from electronic dynamics and vibrational structures to photochemical and environmental implications. Molecules 28, 601 (2023).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  21. Mahesh, S., Tang, K.-C. & Raj, M. Amide bond activation of biological molecules. Molecules 23, E2615 (2018).

    Article  Google Scholar 

  22. Watson, J. L. et al. De novo design of protein structure and function with RFdiffusion. Nature 620, 1089–1100 (2023).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

  23. Glögl, M. et al. Target-conditioned diffusion generates potent TNFR superfamily antagonists and agonists. Science 386, 1154–1161 (2024).

    Article  ADS  PubMed  PubMed Central  Google Scholar 

  24. Vázquez Torres, S. et al. De novo designed proteins neutralize lethal snake venom toxins. Nature 639, 225–231 (2025).

    Article  ADS  PubMed  PubMed Central  Google Scholar 

  25. Bennett, N. R. et al. Atomically accurate de novo design of antibodies with RFdiffusion. Nature 649, 183–193 (2026).

    Article  ADS  CAS  PubMed  Google Scholar 

  26. Vassiliou, S. et al. Structure-guided, single-point modifications in the phosphinic dipeptide structure yield highly potent and selective inhibitors of neutral aminopeptidases. J. Med. Chem. 57, 8140–8151 (2014).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  27. Abramson, J. et al. Accurate structure prediction of biomolecular interactions with AlphaFold 3. Nature 630, 493–500 (2024).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

  28. Grams, F. et al. Structure of astacin with a transition-state analogue inhibitor. Nat. Struct. Biol. 3, 671–675 (1996).

    Article  CAS  PubMed  Google Scholar 

  29. Xue, F. et al. Improving protein sequence design through Designability Preference Optimization. Preprint at https://doi.org/10.48550/arXiv.2506.00297 (2025).

  30. Sahtoe, D. D. et al. Reconfigurable asymmetric protein assemblies through implicit negative design. Science 375, eabj7662 (2022).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  31. Tronrud, D. E., Monzingo, A. F. & Matthews, B. W. Crystallographic structural analysis of phosphoramidates as inhibitors and transition-state analogs of thermolysin. Eur. J. Biochem. 157, 261–268 (1986).

    Article  CAS  PubMed  Google Scholar 

  32. Kim, H. & Lipscomb, W. N. Comparison of the structures of three carboxypeptidase A-phosphonate complexes determined by X-ray crystallography. Biochemistry 30, 8171–8180 (1991).

    Article  CAS  PubMed  Google Scholar 

  33. Bartlett, P. A. & Marlowe, C. K. Phosphonamidates as transition-state analogue inhibitors of thermolysin. Biochemistry 22, 4618–4624 (1983).

    Article  CAS  PubMed  Google Scholar 

  34. Hoffnagle, A. M. & Tezcan, F. A. Atomically accurate design of metalloproteins with predefined coordination geometries. J. Am. Chem. Soc. 145, 14208–14214 (2023).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

  35. Der, B. S. et al. Metal-mediated affinity and orientation specificity in a computationally designed protein homodimer. J. Am. Chem. Soc. 134, 375–385 (2012).

    Article  ADS  CAS  PubMed  Google Scholar 

  36. Schullek, J. R. & Wilson, I. B. The binding of zinc to angiotensin-converting enzyme. Arch. Biochem. Biophys. 265, 346–350 (1988).

    Article  CAS  PubMed  Google Scholar 

  37. van Kempen, M. et al. Fast and accurate protein structure search with Foldseek. Nat. Biotechnol. 42, 243–246 (2024).

    Article  ADS  PubMed  Google Scholar 

  38. Wolz, R. L. A kinetic comparison of the homologous proteases astacin and meprin A. Arch. Biochem. Biophys. 310, 144–151 (1994).

    Article  CAS  PubMed  Google Scholar 

  39. Ryan, J. W. et al. A radiochemical assay for aminopeptidase N. Anal. Biochem. 210, 27–33 (1993).

    Article  CAS  PubMed  Google Scholar 

  40. Batatia, I. et al. MACE-POLAR-1: a polarisable electrostatic foundation model for molecular chemistry. Preprint at https://doi.org/10.48550/arXiv.2602.19411 (2026).

  41. Chen, S.-L., Li, Z.-S. & Fang, W.-H. Theoretical investigation of astacin proteolysis. J. Inorg. Biochem. 111, 70–79 (2012).

    Article  CAS  PubMed  Google Scholar 

  42. Kapust, R. B., Tözsér, J., Copeland, T. D. & Waugh, D. S. The P1’ specificity of tobacco etch virus protease. Biochem. Biophys. Res. Commun. 294, 949–955 (2002).

    Article  ADS  CAS  PubMed  Google Scholar 

  43. Mei, M. et al. A combinatorial strategy for HRV 3C protease engineering to achieve the N-terminal free cleavage. Int. J. Biol. Macromol. 265, 131066 (2024).

    Article  CAS  PubMed  Google Scholar 

  44. Fang, B., Boross, P. I., Tozser, J. & Weber, I. T. Structural and kinetic analysis of caspase-3 reveals role for s5 binding site in substrate recognition. J. Mol. Biol. 360, 654–666 (2006).

    Article  CAS  PubMed  Google Scholar 

  45. Radisky, E. S. & Koshland, D. E. Jr. A clogged gutter mechanism for protease inhibitors. Proc. Natl Acad. Sci. USA 99, 10316–10321 (2002).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

  46. Yiallouros, I., Grosse Berkhoff, E. & Stöcker, W. The roles of Glu93 and Tyr149 in astacin-like zinc peptidases. FEBS Lett. 484, 224–228 (2000).

    Article  CAS  PubMed  Google Scholar 

  47. Beaumont, A. et al. The role of histidine 231 in thermolysin-like enzymes. A site-directed mutagenesis study. J. Biol. Chem. 270, 16803–16808 (1995).

    Article  CAS  PubMed  Google Scholar 

  48. Phillips, M. A., Fletterick, R. & Rutter, W. J. Arginine 127 stabilizes the transition state in carboxypeptidase. J. Biol. Chem. 265, 20692–20698 (1990).

    Article  CAS  PubMed  Google Scholar 

  49. Kunugi, S., Hirohara, H. & Ise, N. pH and temperature dependences of thermolysin catalysis. Catalytic role of zinc-coordinated water. Eur. J. Biochem. 124, 157–163 (1982).

    Article  CAS  PubMed  Google Scholar 

  50. Johnson, L. L. et al. A rationalization of the acidic pH dependence for stromelysin-1 (Matrix metalloproteinase-3) catalysis and inhibition. J. Biol. Chem. 275, 11026–11033 (2000).

    Article  CAS  PubMed  Google Scholar 

  51. Arai, T. et al. TDP-43 is a component of ubiquitin-positive tau-negative inclusions in frontotemporal lobar degeneration and amyotrophic lateral sclerosis. Biochem. Biophys. Res. Commun. 351, 602–611 (2006).

    Article  ADS  CAS  PubMed  Google Scholar 

  52. Westermark, G. T., Fändrich, M. & Westermark, P. AA amyloidosis: pathogenesis and targeted therapy. Annu. Rev. Pathol. 10, 321–344 (2015).

    Article  CAS  PubMed  Google Scholar 

  53. Levin, O. S. & Vasenina, E. E. 25 years of the amyloid hypothesis of the origin of Alzheimer’s disease: advances, failures, and new perspectives. Neurosci. Behav. Physiol. 47, 1065–1070 (2017).

    Article  CAS  Google Scholar 

  54. Guenther, E. L. et al. Atomic structures of TDP-43 LCD segments and insights into reversible or pathogenic aggregation. Nat. Struct. Mol. Biol. 25, 463–471 (2018).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  55. Babinchak, W. M. et al. The role of liquid–liquid phase separation in aggregation of the TDP-43 low-complexity domain. J. Biol. Chem. 294, 6306–6317 (2019).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  56. Sharma, S. K. & Bagshawe, K. D. Antibody directed enzyme prodrug therapy (ADEPT): trials and tribulations. Adv. Drug Deliv. Rev. 118, 2–7 (2017).

    Article  CAS  PubMed  Google Scholar 

  57. Silva, D.-A. et al. De novo design of potent and selective mimics of IL-2 and IL-15. Nature 565, 186–191 (2019).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

  58. Shin, H. et al. De novo design of protease-activatable cytokine prodrugs. Preprint at bioRxiv https://doi.org/10.64898/2026.09.19.752834.

  59. Mansour, A. M. et al. A new approach for the treatment of malignant melanoma: enhanced antitumor efficacy of an albumin-binding doxorubicin prodrug that is cleaved by matrix metalloproteinase 2. Cancer Res. 63, 4062–4066 (2003).

    CAS  PubMed  Google Scholar 

  60. Escobar-Rosales, M. et al. De novo design of peptide masks enables rapid generation of conditionally-active miniprotein binders. J. Am. Chem. Soc. 147, 45495–45505 (2025).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

  61. Ratnikov, B. I. et al. Basis for substrate recognition and distinction by matrix metalloproteinases. Proc. Natl Acad. Sci. USA 111, E4148–E4155 (2014).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  62. Chen, A. Computational design of metalloproteases. Zenodo https://doi.org/10.5281/zenodo.22654831 (2026).

Download references

Acknowledgements

We thank L. Goldschmidt for maintaining the computational resources, and K. Van Wormer, R. Ticzon, H. Nunez-Ortegaand and E. Weston for maintaining the wet laboratory resources at the Institute for Protein Design. We thank I. Kalvet, H. Yu, B. Belmont, A. Broerman, B. Liu, L.T. Yu, C. Liu, W. Ahern, F. Xue and C. Wang for helpful discussions during the development of the project. We thank D. Whittington and R. Lundeen for helping with mass spectrometry. We thank I. Kalvet and Y. Liu for providing reagents. We thank M. A. Kennedy for reading and editing drafts of the article. We thank L. Stewart for managing the Institute for Protein Design and securing funding resources. This work is based on research conducted at the Northeastern Collaborative Access Team beamlines, which are funded by the National Institute of General Medical Sciences from the National Institutes of Health (grant no. P30 GM124165). The Eiger 16M detector on the 24-ID-E beam line is funded by a NIH-ORIP HEI grant (no. S10OD021527). This research used beamtime awards (https://doi.org/10.46936/APS-190993/60014759) from the Advanced Photon Source, a US Department of Energy (DOE) Office of Science User Facility operated for the DOE Office of Science by Argonne National Laboratory under Contract No. DE-AC02-06CH11357. This research used resources 17-ID-1 (AMX)/17-ID-2 (FMX) of the National Synchrotron Light Source II, a US Department of Energy (DOE) Office of Science User Facility operated for the DOE Office of Science by Brookhaven National Laboratory under Contract No. DE-SC0012704. The Center for BioMolecular Structure (CBMS) is primarily supported by the National Institutes of Health, National Institute of General Medical Sciences (NIGMS) through a Center Core P30 grant (no. P30GM133893) and by the DOE Office of Biological and Environmental Research (grant no. KP1605010).

Funding

This work was funded by the following: Bill and Melinda Gates Foundation grant no. INV-043758; Defense Threat Reduction Agency grant no. HDTRA1-19-1-0003; Helmsley Charitable Trust Type 1 Diabetes (T1D) Program grant no. 2019PG-T1D026; Howard Hughes Medical Institute. The Audacious Project at the Institute for Protein Design; The Open Philanthropy Project Improving Protein Design Fund; The National Institutes of Health’s National Institute of Allergy and Infectious Disease, grant no. R0AI160052; The National Institutes of Health’s National Institute on Aging, grant no. R01AG063845; the Matchmakers team were supported by the Cancer Grand Challenges partnership funded by Cancer Research UK (grant no. CGCATF-2023/100008) and the National Cancer Institute 1 (grant no. OT2CA297288-01), and the Mark Foundation for Cancer Research; The Advanced Research Projects Agency for Health APECx Program Award grant no. 1AY1AX000036; The Gates Foundation grant no. INV-043758; The National Institute for Allergy and Infectious Diseases, National Institutes for Health grant no. 1U19AI181881-01; Wu Tsai Translational Investigator Fund; Grantham Foundation for the Protection of the Environment; Schmidt Family Foundation; MICIU/AEI/10.13039/501100011033 and FEDER, EU (grant no. PID2023-151988OB-I00); ‘la Caixa’ Foundation (grant no. LCF/BQ/PR25/12110008); European Research Council (ERC) under the European Union’s Horizon Europe research and innovation programme (grant agreement 101077370); The National Natural Science Foundation of China (NSFC) grant no. 22377107; Wolfson College, Cambridge through a Junior Research Fellowship and ARIA, DSIT and Pillar VC under the Encode AI for Science Fellowship.

Author information

Author notes

  1. These authors contributed equally: Anqi Chen, Kejia Wu

Authors and Affiliations

  1. Department of Biochemistry, University of Washington, Seattle, WA, USA

    Anqi Chen, Kejia Wu, Hojae Choi, Preetham Venkatesh, Samuel J. Pellock, Nikita Hanikel, Hojeong shin, Declan Evans, Brian Coventry, Donghyo Kim, Seth M. Woodbury, Pengfei Ji, Shingo Honda, Asim K. Bera, Hannah Nguyen, Alex Kang, Yanqing Wang, Xinting Li, Stacey Gerben, Lemuel Chang, Anthony A. Hyman & David Baker

  2. Institute for Protein Design, University of Washington, Seattle, WA, USA

    Anqi Chen, Kejia Wu, Hojae Choi, Preetham Venkatesh, Samuel J. Pellock, Nikita Hanikel, Hojeong shin, Declan Evans, Kieran Didi, Brian Coventry, Donghyo Kim, Seth M. Woodbury, Pengfei Ji, Shingo Honda, Asim K. Bera, Hannah Nguyen, Alex Kang, Yanqing Wang, Xinting Li, Stacey Gerben, Lemuel Chang & David Baker

  3. Graduate Program in Biological Physics, Structure and Design, University of Washington, Seattle, WA, USA

    Hojae Choi, Preetham Venkatesh & Seth M. Woodbury

  4. Department of Computer Science, University of Oxford, Oxford, UK

    Kieran Didi

  5. NVIDIA Corp, Santa Clara, CA, USA

    Kieran Didi

  6. Cavendish Laboratory, Department of Physics, University of Cambridge, Cambridge, UK

    Lars L. Schaaf

  7. Department of Materials, Imperial College London, London, UK

    Lars L. Schaaf

  8. Institut Químic de Sarrià, Universitat Ramon Llull, Barcelona, Spain

    Cristina Díaz-Perlas & Benjamí Oller-Salvia

  9. Howard Hughes Medical Institute, University of Washington, Seattle, WA, USA

    Brian Coventry & David Baker

  10. Department of Chemistry, Zhejiang University, Hangzhou, China

    Pengfei Ji

  11. Max Planck Institute of Molecular Cell Biology and Genetics, Dresden, Germany

    Xiao Yan

  12. Laboratory of Organic Chemistry, ETH Zurich, Zurich, Switzerland

    Donald Hilvert

Authors

  1. Anqi Chen
  2. Kejia Wu
  3. Hojae Choi
  4. Preetham Venkatesh
  5. Samuel J. Pellock
  6. Nikita Hanikel
  7. Hojeong shin
  8. Declan Evans
  9. Kieran Didi
  10. Lars L. Schaaf
  11. Cristina Díaz-Perlas
  12. Brian Coventry
  13. Donghyo Kim
  14. Seth M. Woodbury
  15. Pengfei Ji
  16. Shingo Honda
  17. Asim K. Bera
  18. Hannah Nguyen
  19. Alex Kang
  20. Yanqing Wang
  21. Xinting Li
  22. Stacey Gerben
  23. Lemuel Chang
  24. Xiao Yan
  25. Benjamí Oller-Salvia
  26. Anthony A. Hyman
  27. Donald Hilvert
  28. David Baker

Contributions

A.C., K.W., P.V., D.H. and D.B. conceptualized the project. A.C., K.W., H.C., P.V. and S.J.P. developed the computational design pipeline. N.H. performed the Zn(II)-dissociation constant experiments. H.S. designed and performed the conditional cytokine mimic activation experiments. B.O.-S. and C.D.-P. designed and performed the masked conditional antagonist experiments. B.C. trained RFD2-MI. D.E. performed the molecular dynamics simulations. K.D. and L.L.S. performed the MLFF calculations. A.K.B., H.N., A.K. and Y.W. performed protein crystallography. P.J., D.K., S.M.W. and S.H. contributed ideas and code. A.C., K.W. and L.C. produced and characterized the designs. A.C. and K.W. analysed the experimental results. X.L. performed intact protein mass spectrometry. S.G. provided training on equipment and advice on experimental protocols. X.Y. and A.A.H. produced and characterized full-length human TDP-43. K.W., D.H. and D.B. supervised the project. A.C., K.W., D.H. and D.B. wrote the paper. All authors reviewed, commented on and edited the paper.

Corresponding authors

Correspondence to Kejia Wu, Donald Hilvert or David Baker.

Ethics declarations

Competing interests

Provisional patent applications 64/047,504 and 63/921,483 have been filed related to aspects of the methods and designs described in this work, listing A.C., K.W., H.C., P.V., S.J.P., D.H., A.K.B. and D.B. as inventors or contributors. The other authors declare no competing interests.

Peer review

Peer review information

Nature thanks James Fraser 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 figures and tables

Extended Data Fig. 1 Detailed computational pipeline for the de novo design of metalloproteases.

The generation process involves one round of initial motif-scaffolded backbone generation, followed by sequence design and AF3 filtering, two rounds of partial diffusion refinement, followed by sequence design and AF3 filtering and one round of iteration between sequence design and AF3 filtering. See Supplementary Methods for all details of the computational pipeline.

Extended Data Fig. 2 Gene maps of the open reading frames (ORF) of four constructs used in the experimental sections.

a, The cis construct used for the screen of functional active sites. b, The substrate chain of the trans constructs used for the screen of active proteases. c, The protease chain of the trans constructs used for the screen of active proteases. d, Substrate construct with an N terminal sfGFP and a C terminal mScarlet used in the validation and characterization of active designs.

Extended Data Fig. 3 Computational metrics predictive of cross-reactivity.

a, The heatmap pattern for experimental specificity shown in Fig. 4a, where the diagonal elements and the elements in the last row are highlighted, is observed with multiple computational metrics from AF3 predictions of the [Zn+2].O ES complex, plotted in the same heatmap format as Fig. 4a. b, Correlation between experimentally measured normalized digested fraction and computational metrics. High chain C (Zn(II) ion) pLDDT, high complex pLDDT, high iptm, low mean interface PAE and low minimum interface PAE all are predictive of high cross reactivity.

Extended Data Fig. 4 Additional crystal structures for metalloprotease designs.

a, Apo crystal structure of Zn5 (red) superimposed with the design model (grey) and zoomed in view of the active site. b, Crystal structure of Zn5 E75Q-Zn(II) complex (red) superimposed with the design model (grey) and zoomed in view of the active site. c, Crystal structure of apo TDPr3 E45Q (green) superimposed with the design model (grey) and zoomed in view of the active site. d, Crystal structure of Zn48 E75Q-Zn(II) complex (red) superimposed with the design model (grey) and zoomed in view of the active site.

Extended Data Fig. 5 Structural similarity of metalloprotease designs with best matches found in pdb.

Searches were performed with: a, the Zn5-Zn(II) crystal structure; b, the TDPr3 E45Q-Zn(II) crystal structure; c, the ZnO7 E32Q-Zn(II)-substrate crystal structure; d, the Zn48 E117Q-Zn(II)-substrate crystal structure.

Extended Data Fig. 6 Screening of the active site single site saturation mutagenesis library.

a, Design model of Zn45 and the substrate of S_ZnO36 showing 21 residue positions chosen for single site saturation in sticks in blue. The five catalytic residues are shown in wheat. b, Reaction progress curves for the 378 single variants with WT (blue) and buffer (black) controls. All variants were purified by strep-tag affinity chromatography and concentrations were normalized. Variants in red were further purified using HPLC, tested again to confirm improvements before mutations were recombined (Supplementary Methods 3.19). c, Activity heatmap of all single variants. d, Michaelis-Menten kinetics measurements for six improved single variants, each being the most active variant at the corresponding residue position in the activity screen. v0: reaction rate in the initial linear region (nM/s). [E0]: enzyme concentration, 500 nM. Measurements were made in triplicate. Black dots and error bars represent mean ± std. e, Single mutations selected for recombination. f, MPNN probabilities for residues where beneficial mutations are discovered. Blue: wild-type AA. Orange: the sum of probabilities for all beneficial substitutions. Individual measurements are represented as black dots and bars and error bars represent mean ± std; n = 5. g, Correlation between activity fold improvement and AF3 metrics. Many deleterious mutations can be predicted by metrics such as complex pLDDT, ipae min and pae interface.

Extended Data Fig. 7 Screening of active site combinatorial library.

a, Design model of Zn45 and the substrate of S_ZnO36 showing 6 residue positions chosen for combinatorial saturation in sticks in red. The five catalytic residues are shown in wheat. b, Reaction progress curves for the 148 combinatorial variants with WT and base KO controls. c, The relationship between activity and number of mutations. Box plots show the median (centre line), 25th and 75th percentiles (box limits), and whiskers extending to the most extreme values within 1.5× the interquartile range. Points represent individual observations. The number of variants with mutation numbers 1-6: 11, 38, 52, 34, 10, 1. d, Total turnover number (TTN) for Zn45_v2. A series of substrate concentrations of 160 µM, 80 µM, 40 µM, 20 µM, 10 µM, 5 µM, were incubated with or without 20 nM Zn45_v2 at 37 °C for 5 h and the TTN was determined by densitometry. This experiment was performed three times with similar results. Full gel image in Supplementary Fig. S13. e, Reaction progress curves of Zn45_v2 and the individual catalytic residue knockout variants. Measurements were made in triplicate. Solid lines and shaded regions represent the mean ± std. f, pH dependence of the activity of Zn45_v2 in the low substrate concentration regime (3 µM). Measurements were made in triplicate. Black dots and error bars represent mean ± std. g, Michaelis Menten kinetics of three other combinatorial variants. Measurements were made in triplicate. Black dots and error bars represent mean ± std. h, Correlation between activity fold improvement and AF3 metrics. Overall the correlation between activity improvement and AF3 metrics is weak.

Extended Data Fig. 8 Biological relevance of cleaving human TDP-43 and characterization of human full length TDP-43 in oligomeric and monomeric formats.

a, TDP-43 contains structured domains (NTD, RRM1, RRM2) essential for nuclear function and a long disordered C-terminal domain (CTD) that governs liquid–liquid phase separation (LLPS) and oligomerization/aggregation. Full-length degradation risks loss of essential functions, whereas cleavage in the CTD can modulate LLPS and/or oligomerization without ablating function. b, Redesigning existing scaffold to target TDP43. c, Substrate specificity of the TDP43 redesign, TDPr3. Experiments were performed in triplicates. The solid line and the shaded area represent the mean ± std. d, Dynamic light scattering (DLS) determined the average size of oligomeric TDP-43 (Peak 1) as dozens of molecules solubly clustered together, while monomeric TDP-43 (Peak 2) remains monomeric; measurements were made in triplicate. Dots and error bars represent mean ± std. e, TDP-43 form aggregates after incubation at 37C 6h, 12h, 24 and monomeric TDP-43 at 37C 6h, 12, 24h in buffer 50 mM HEPES, 100 mM NaCl, 50 uM ZnSO4 samples in under microscope, monomers started to aggregate while oligomers remained unchanged. f, SDS-PAGE gel analysis after incubating human TDP-43 with TDPn3 or the base knockout (KO). In the WT, the band intensities of the C and N fragments (CO, NO for oligomer, CM, NM for monomer) gradually increase and the full-length (FLO for oligomer, FLN for monomer) is depleted in time. In base KO, no products are detected. The loss of full-length intensity likely is a result of precipitation as shown by the high molecular weight smears. This experiment was performed three times independently with similar results.

Extended Data Fig. 9 De novo metalloproteases targeting Serum amyloid A and amyloid-β peptide.

a,d, Design model of a de novo zinc protease, SAAc8, against serum amyloid A (SSRSFFSFLG, a) and a de novo zinc protease, AbetaF3, against amyloid-β peptide (KGAIIGLMVG, d). b,e, Reaction progress curves of SAAc8 and AbetaF3 on their cognate substrates, and on other substrates presented in this work. Measurements were made in triplicate. Solid lines and shaded regions represent the mean ± std. c, Kinetics of SAAc8 and AbetaF3 each on their cognate substrates. v0: reaction rate in the initial linear region (nM/s). [E0]: enzyme concentration, 500 nM. Measurements were made in triplicate. Black dots and error bars represent mean ± std.

Supplementary information

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Chen, A., Wu, K., Choi, H. et al. Computational design of metalloproteases. Nature (2026). https://doi.org/10.1038/s41586-026-11167-2

Download citation

  • Received:

  • Accepted:

  • Published:

  • Version of record:

  • DOI: https://doi.org/10.1038/s41586-026-11167-2