Enzymes constantly undergo structural transformations in response to external perturbations such as temperature, pressure, and solution (Gao et al., 2018; Karjiban et al., 2010; Nucci et al., 2014; Tang et al., 2017; Tang et al., 2020). Mutations can induce structural changes in enzymes, often leading to functional alterations affecting their stability, activity, specificity, and enantioselectivity. However, most structural changes are detrimental, eventually producing unfolded and non-functional enzymes, thereby resulting in negative experimental outcomes (Tracewell and Arnold, 2009). For instance, enzymes with 200 amino acid residues can undergo numerous mutations, including 3800 possible single-, 14,367,800 double-, and 5,405,166,600 triple-point mutations. Approximately 95% of the mutants have either neutral or negative impacts (Goldsmith and Tawfik, 2013). Traditional error-prone PCR and DNA shuffling typically generate large libraries of relatively inefficient mutants, resulting in time-consuming, labor-intensive, and costly screening processes in the absence of high-throughput screening methods (Yang et al., 2019). Recent advancements in biocomputing and advanced algorithms have enabled rational and semi rational design methodologies guided by insights from sequences, structures, and functions. This paradigm has demonstrated remarkable potential to build “smart-but-small libraries,” thus facilitating the evolution of enzyme properties (Reetz et al., 2010). With the reliable prediction of target loci, these smart libraries improve the efficiency of enzyme evolution by increasing the ratio of beneficial to detrimental mutations. These libraries yield results comparable to those of traditional directed evolution achieved through multiple rounds of experiments (Yu et al., 2022). Thus, the identification of suitable hotspot regions can facilitate more efficient acquisition of enzymes with the desired properties.

Atomic packing defects with different local compressibility properties in proteins play an irreplaceable role in the execution of functions and the maintenance of structural stability. This unique behavior depends on a range of interactions with varying strengths and frequencies, including disulfide bonds, hydrogen bonds, hydrophobic interactions, charge-charge interactions, and van der Waals interactions. The combined effects of these forces facilitate effective packaging of enzymes for various biological functions. X-ray diffraction analysis reveals that enzymes exhibit tight packing, resulting in an average packing density (0.74) comparable to that of organic molecular crystals (0.70 to 0.78) (Liang and Dill, 2001). This solid-like characteristic has been confirmed by compressibility and high-pressure crystallography experiments (Collins et al., 2007). However, enzymes also possess fluid-like properties, which can be attributed to the presence of various cavities on both the surface and interior of the rigid structure. These cavities, sometimes called voids, pockets, or pores, drive the movement of the enzyme molecules.

Cavities in enzymes generally arise from insufficient packing density of amino acids in their native state (Bhattacharyya and Varadarajan, 2013). The presence of these cavities is often ascribed to variations in the hydrophilicity of the residual side chains. Enzymatic cavities are classified into two types based on their spatial position: internal cavities located within domains and interfacial cavities situated between protein domains or subunits (Hubbard and Argos, 1994; Hubbard et al., 1994). Internal cavities in proteins are generally small, typically tens of Å3, and their geometric properties strongly influence protein stability. Conversely, interfacial cavities tend to be larger, often exceeding 100 Å3, and are closely associated with enzyme catalytic activity, protein-protein interactions, and ligand binding. Although classical models of induced-fit and selected-fit describe the catalytic activity of enzymes with active sites located at the protein interface (Chakraborty and Di Cera, 2017; Johnson, 2008), a substantial majority of known enzymes (>60%) house their catalytic active sites within internal cavities connected to the surrounding solvent via pathways (Marques et al., 2016; Pravda et al., 2014). Consequently, cavity engineering not only influences enzyme stability, but also plays a crucial role in its catalytic function.

Cavities can exhibit varying degrees of closure, including complete enclosure, as observed in most icosahedral viral capsids (Chwastyk et al., 2016). However, incomplete closure, resulting in openings that connect with the external solvent, is common. An illustrative example is the pathogenesis-related class 10 protein (PR-10) (Chwastyk et al., 2014). Similarly, ribosomes feature an opening in the form of an exit channel for polypeptide chains. The internal cavity housing the active sites is typically connected to the external bulk solvent through one or more pathways, represented by tunnels and channels (Kokkonen et al., 2019; Wu et al., 2023). The cavity and pathways may be distinct but not exclusive, resulting in potential overlap. In some cases, the internal cavity may be integrated within the pathway as a cohesive assembly, thereby maintaining its integrity (Stank et al., 2016). Alternatively, tunnels can connect the internal cavity with the external solvent. Both cases highlight the pivotal role of cavities in the catalytic cycle of the enzyme.

Prokop et al. (2012) proposed a “Schottkhole-lock-key” model that offers insights into enzymatic catalysis when the active sites are buried within internal cavities connected to tunnels. The importance of tunnels in enzyme functions, including enhancing substrate specificity, regulating the entry of co-substrates, and protecting against transition metal-induced damage, has been discussed in a recent review (Kokkonen et al., 2019). However, it neglected the conformational changes that occur within the cavity throughout the catalytic cycle. The opening of the internal cavity connected to the substrate tunnel evolves concurrently with the tunnel dynamics and vice versa. The functions of tunnels are frequently intertwined with those of cavities harboring buried active sites, and their synergistic action contributes to the evolution of enzymatic properties. Consequently, it is challenging to precisely demarcate and define the tunnel and internal cavity functions (Pravda et al., 2014; Stank et al., 2016).

To illustrate the catalytic effect of active sites within buried cavities on enzymes, we introduce the “cavity fit model.” Fig. 1 shows a simplified and generalized depiction of the enzymatic cycle within the enzyme cavity. The catalytic cycle is initiated with the substrate entering the activated state of the cavity (Fig. 1B), followed by cavity reshaping to encapsulate the substrate through catalytic site-substrate interactions (Fig. 1C). Next, the substrate binds to the active site leading to the generation of the product (Fig. 1D) (Kokkonen et al., 2019). Subsequently, the active center of solvation causes the product to unbind and release from the cavity (Fig. 1E). Finally, the cavity returns to its natural state and is prepared for the next catalytic cycle (Fig. 1F). Notably, specific processes across various enzyme-catalyzed systems can vary because of the unique characteristics of the cavities in which the active site is buried. And not all catalytic systems encompass the steps depicted in Fig. 1.

Cavities have garnered the attention of enzyme engineers as essential mediators for fine-tuning the enzyme's stability and functionality. Cavity engineering is being widely used across various enzymes (Guo et al., 2020; Montemiglio et al., 2021; Roche et al., 2012; Xue et al., 2019). Modifying the size, spatial placement, and geometry of cavities harboring active sites can alter enzyme function. Concurrently, the interactions, spatial distribution, and spatial epistasis within higher-dimensional multicavities also play distinct and critical roles. However, current reviews on the cavity modification of enzyme properties are limited. This review endeavors to furnish an exhaustive overview of cavity engineering approaches aimed at optimizing enzyme properties, including 1) engineering of cavity-creating from large to small; 2) engineering of cavity filling from small to large; 3) potential future trajectories in cavity engineering, underscoring systematic critical region screening, rigorous statistical analysis, and intelligent library design; and 4) an overview of feasible tools for cavity calculation and design. We envisage that cavity engineering strategies will enable the development of new specialty enzymes for industrial applications.