Search
Search articles by title, abstract,author or doi
Advanced search

Designing substrate-specific metalloproteases

Haiyan Liu

Vita ››

PDF (437KB)
Vita > Cutting Edge > DOI: 10.15302/vita.2026.08.0063
Vita Open AccessPublished:

Designing substrate-specific metalloproteases

Author information +
History +
Vita () Cite this article
PDF (437KB)

In a recent study published in Vita, Qu et al. leverage ProteusFlow, a flow-based backbone generation method, to design substrate-encapsulating metalloproteases targeting three distinct cleavage sites in the amyloid-β peptide, achieving programmable substrate recognition and high catalytic efficiency in a subset of designs.

登录浏览全文

4963

注册一个新账户 忘记密码

De novo enzyme design seeks to build proteins that catalyze a reaction of interest ab initio rather than by mutagenesis engineering of existing enzymes. Early efforts rely on transplanting functional groups arranged in catalytically competent geometric configurations into structurally compatible natural protein scaffolds1. This strategy is inherently limited by the finite diversity and functional constraints of naturally occurring folds. Now, deep learning–driven methods support the direct generation of protein architectures that fold around and stabilize a user-specified active site, through two complementary computational approaches2-4. The first, known as “hallucination”, inverts protein structure prediction models to jointly sample protein structures and sequences2. The second leverages generative diffusion or flow-matching models to directly produce backbone coordinates3,4, followed by sequence design via inverse folding.
A longstanding challenge in the field is whether enzymes with catalytic efficiency and specificity on par with natural counterparts can be designed entirely computationally — that is, without costly experimental optimization via directed evolution or high-throughput screening. Recent advances suggest growing feasibility2-6. For instance, a fragment-based, TIM-barrel-guided workflow successfully generated Kemp eliminases5; a “family-wide hallucination” strategy produced compact, hyperstable artificial luciferases2; and Riff-Diff — a method integrating diffusion-based backbone generation with fragment and rotamer sampling — has been applied to design retro-aldol and Morita–Baylis–Hillman catalysts3. In these studies, artificial enzymes with catalytic efficiencies approaching natural counterparts have been obtained as direct results of computational design. Notably, RFdiffusion2 introduced an active-site–focused backbone generation model that eliminates the need to predefine both sequence locations and backbone coordinates for catalytic residues, enabling more flexible, function-driven design4. This approach has been successfully applied to generate zinc metallohydrolases exhibiting substantially higher activity than prior de novo metalloenzymes6.
In stark contrast to catalytic efficiency, substrate selectivity remains profoundly underexplored in de novo enzyme design. The vast majority of published enzyme designs are benchmarked against a single model substrate; when selectivity is reported, it typically arises as a by-product of pocket complementarity rather than as an explicitly programmable objective. Achieving rationally controllable substrate choice, epitomized by proteases that cleave peptide bonds at specific substrate sites, largely remains an open frontier.
The substrate specificity or selectivity of natural proteases is essential for their roles as regulators of sophisticated cellular functions. The ability to design and control the specificity of proteases toward particular peptide sequences has far-reaching implications in both biotechnology and medicine. In biotechnology, proteases serve as indispensable biocatalysts — from detergent formulations and food processing to precision protein engineering in biopharmaceutical manufacturing. Yet their utility is frequently constrained by prior substrate preferences: either overly promiscuous or excessively restrictive. In medicine, rationally designed proteases with tunable, site-specific activity could enable next-generation therapeutics such as targeted protein degradation, conditional activation of prodrugs, or selective ablation of pathogenic proteins. In these applications, high on-target cleavage efficiency and minimal off-target proteolysis are critical requirements for both safety and therapeutic efficacy.
To overcome the inherent substrate constraints of natural proteases, researchers have widely adopted rational mutagenesis and directed evolution to rewire their specificity. Although prior studies have achieved notable successes7, such approaches typically yield only minor modifications to the enzyme’s native substrate specificity — most commonly replacing one or two key residues. Reprogramming an existing protease to recognize and cleave new substrate sequences with user-specified profiles remains exceptionally challenging.
In a recent study, Qu and colleagues sought to design de novo metalloproteases targeting three distinct cleavage sites in amyloid-β (Aβ) peptides8 (Fig. 1). To this end, they employed ProteusFlow — a flow-based backbone generation program — to co-generate the structures of both the protease and its substrate. Their designs leveraged the active site configurations from two existing metalloproteases: PPEP-1 and Fungalysin. For the PPEP-1-based designs, the authors regenerated the peptide-binding pockets (termed “PP designs”) while retaining the core structure. For the Fungalysin-based designs, they retained only the minimal catalytic motif and regenerated the entire surrounding protein scaffold (termed “DP designs”).
The authors reasoned that high-fidelity substrate recognition demands tight encapsulation of the substrate at the enzyme–substrate interface, minimizing steric clashes and maximizing packing density. To achieve this, they implemented a “two-step encapsulation” strategy: first, generating the protein backbone around the catalytic glutamate; second, building the scaffold around the oxyanion hole. This approach significantly enhanced substrate recognition.
After experimentally testing on average 26 designed variants per strategy–site pair, the DP strategy yielded active enzymes targeting all three Aβ cleavage sites, whereas the PP strategy produced active enzymes for one site. Kinetic analysis revealed substantial differences in catalytic efficiency among the active designs: DP622-S2 exhibited the highest activity, with a kcat/Km of 325.26 M−1s−1. Moreover, PP507-S1 displayed strict substrate selectivity, cleaving exclusively its intended Aβ site. Other designs showed varying degrees of cross-reactivity. Notably, the DP221-S3 H191A variant achieved remarkably enhanced selectivity, effectively eliminating cleavage at the two unintended substrate sites. The activity of the designed enzymes was further assessed on full-length synthetic Aβ42 peptide. A cocktail comprising selected variants was shown to simultaneously process Aβ42 without mutual interference.
The catalytic mechanisms of the designed proteases were further validated by site-directed mutagenesis. Additionally, cryo-EM structures were determined for three representative designs (PP507-S1, DP622-S2, and DP221-S3, each targeting a distinct cleavage site) in complex with full-length Aβ42 peptide. The resolved substrate densities confirmed that each protease specifically recognizes its intended Aβ segment.
Guided by structural insights, the authors performed a two-stage refinement of DP622-S2 to enhance catalytic efficiency. This iterative optimization yielded variants with improved substrate selectivity and catalytic activity. The most optimized variant, OP669-S2, achieved a kcat/Km of 3045.14 M−1s−1 — a nearly tenfold increase over DP622-S2 — demonstrating the power of structure-guided, iterative design to substantially improve catalytic performance.
In summary, the study by Qu et al. establishes a promising framework for designing programmable proteases for targeted peptide hydrolysis. Their structural and biochemical analyses not only validate the design strategy but also underscore the persistent challenge of engineering catalytic and substrate-binding microenvironments that simultaneously support high catalytic efficiency and strict substrate selectivity.

[1]

Kiss, G., Çelebi-Ölçüm, N., Moretti, R., Baker, D. & Houk, K.N. Angew. Chem. Int. Ed. Engl. 52, 5700–5725 (2013).

[2]

Yeh, A.H.W. et al. Nature 614, 774–780 (2023).

[3]

Braun, M. et al. Nature 649, 237–245 (2026).

[4]

Ahern, W. et al. Nat. Methods 23, 96–105 (2026).

[5]

Listov, D. et al. Nature 643, 1421–1427 (2025).

[6]

Kim, D. et al. Nature 649, 246–253 (2026).

[7]

Blum, T.R. et al. Science 371, 803–810 (2021).

[8]

Qu, Y.N., Wang, C.T., Zhu, H.L., Wang, Y.J. & Cao, L.X. Vita https://doi.org/10.15302/vita.2026.07.0055 (2026).

RIGHTS & PERMISSIONS

The Author(s) 2026. Published by Higher Education Press. This is an Open Access article distributed under the terms of the CC BY license (https://creativecommons.org/licenses/by/4.0/).

Cite this article

Download citation ▾
Liu, H.  Designing substrate-specific metalloproteases  Vita https://doi.org/10.15302/vita.2026.08.0063 ()
AI Summary AI Mindmap
PDF (437KB)
Sections
Figures
References

209

Accesses

0

Citation

Detail

Recommended

AI思维导图

/

〈 〉