Proteases catalyze one of the most fundamental biochemical reactions: the hydrolysis of peptide bonds in peptides and proteins. Ubiquitous in nature – comprising approximately 2 % of an organism's proteome - proteases play critical roles in nearly all cellular processes. Their diverse physiological and pathological functions make them key targets in numerous therapeutic strategies (Bleuez et al., 2022; Cheng et al., 2022; Craven et al., 2018; Drag and Salvesen, 2010; Faucher et al., 2023; Liu et al., 2022; Ofori et al., 2015; Rahbar Saadat et al., 2021; Rudzińska et al., 2021; Vizovisek et al., 2021; Widen et al., 2021). Similarly, harnessing the diverse specificity of proteases has been instrumental in advancing various applications. In medicine, recombinant protease replacement therapies have resulted in 12 FDA-approved drugs to treat hematological and muscular malignancies (Craik et al., 2011). In chemical biology and biotechnology, proteases are useful reagents for recombinant protein production (Frey and Görlich, 2014; Jakob et al., 2013; Rosano et al., 2019). In chemical and synthetic biology, proteases are essential signal processors in protein circuits in living cells (Chen and Elowitz, 2021; Gao et al., 2018).

Proteases have become integral to nearly every domain of molecular biotechnology, driving a surge of interest and innovation in protease engineering. Early efforts in this field primarily targeted enhancements in enzyme stability and recombinant expression, leading to the development of bacterial proteases widely used in laundry detergents (Maurer, 2004) and in protease replacement therapies for hematological, muscular, and digestive disorders (Craik et al., 2011). Despite these advances, progress in reprogramming core protease attributes—such as substrate specificity, activation mechanisms, and catalytic efficiency—remained limited for many years. However, recent advances made over the last fifteen years in protein engineering and directed evolution, structural biology, bioinformatics, and computational and machine learning approaches have begun to overcome these challenges, ushering in a new era of sophisticated protease design and application.

Traditionally, protease substrate specificity was viewed through a narrow lens: most proteases are believed to recognize and cleave short amino acid motifs—typically between one and eight residues—within their target proteins. These residues are conventionally labeled as PX–P1/P1’–PX’, with P1/P1’ denoting the scissile bond (Schechter and Berger, 1967). Depending on the protease, X can be 4–2 (Fig. 1). Conversely, each substrate residue occupies a corresponding enzyme subsite labeled SX-S1/S1’-SX’. This simplified version of substrate recognition hints that one could rationally switch protease substrate specificity, for example, by introducing mutations in the subsites that interact with the mutated substrate. However, early attempts at engineering proteases - converting trypsin into Chymotrypsin (Hedstrom, 2002), MMP16 into MMP17 (Ratnikov et al., 2014) - through rational subsite redesign yielded limited success. These challenges reinforced the findings that (1) active site mutations alone cannot impart proteases with novel activities and specificities, (2) subsite residue crosstalk influences substrate recognition, and (3) mutations far away from active sites (for example, exosites) play a role in substrate binding and catalytic efficiency. A similar interwoven engineering solution is faced when engineering proteases for conditional activation.

Successful protease specificity and activity reprogramming depend on methods that effectively explore a protease's fitness landscape and select enzyme variants with desired properties. This review discusses the tools and principles for engineering proteases with novel activities and specificities and highlights their applications in biotechnology and biomedicine. First, we describe and evaluate the high-throughput screening and selection strategies implemented in E. coli, yeast, and cell-free protein expression systems. Second, we discuss computation-based strategies to engineer split proteases. Lastly, we highlight emerging approaches that leverage protein-protein interactions to modify protease activity and specificity. Although protease substrate specificity engineering and profiling go hand in hand, this review does not provide an in-depth description of genetic and proteomic tools to profile protease substrate specificity. Overall, we aim to illustrate the importance and necessity of the evolution of proteases with novel activities to expand the field's understanding of how proteases can drive biotechnology and therapeutic advances.