Nature News · Feb 11, 2026 · Collected from RSS
MainEnzymatic inhibitors are indispensable tools for dissecting biological pathways and developing therapeutic interventions1. They are broadly categorized by their binding sites and mechanisms of action. Among these, orthosteric inhibitors, which bind to the catalytic site and directly compete with substrates, have been extensively explored due to their predictable structure–activity relationships. However, such inhibitors are typically substrate-agnostic, as their mechanism relies solely on blocking the active site. By contrast, substrate-dependent inhibitors, which achieve selectivity by engaging allosteric sites or exosites, can modulate enzyme activity in a substrate-specific manner2. Yet their design remains challenging due to the complex structural and dynamic determinants governing these interactions. Ideally, combining the tractability of orthosteric inhibitors with the precision of substrate-dependent modulation would offer a powerful strategy—but whether such a hybrid approach is feasible has remained unclear.The CSN is a multi-subunit protein complex evolutionarily conserved across all eukaryotic species3,4. It plays a crucial role in regulating almost every aspect of cellular functions by modulating the cullin-RING ubiquitin ligases (CRLs) and is required for targeted protein degradation induced by emerging protein degraders5,6,7,8,9,10,11,12. As the largest family of E3s, the function of CRLs hinges on the dynamic modification of the cullin scaffolds by a ubiquitin-like protein, NEDD8 (N8), also known as neddylation13. Although cullin neddylation enhances the E3 activity of CRLs, CSN-mediated cullin deneddylation is thought to protect the CRL substrate receptors from auto-ubiquitination and promote their exchange on the cullin scaffolds through an adaptive CRL assembly cycle14,15.COP9 signalosome comprises eight core subunits (CSN1–CSN8), including the catalytic subunit CSN5 (Fig. 1a)16. The crystal structure of the CSN holoenzyme revealed that CSN5 and CSN6 form a heterodimer, which is affixed to the other six subunits17. The active site of CSN5 is occluded by its insertion-1 (Ins-1) loop in the CSN holoenzyme structure, suggesting that the iso-peptidase complex might be in an inactive state. Multiple subsequent cryogenic electron microscopy (cryo-EM) studies have captured the overall architecture of CSN in complexes with four different N8~CRLs18,19,20,21,22. These structures revealed a catalytic hemisphere of CSN formed among CSN2, CSN4, CSN5 and CSN6 that undergoes major conformational changes following substrate engagement (Fig. 1b). All of these structures, however, suffer from a limited resolution, especially within the catalytic hemisphere. How the N8~cullin iso-peptide linkage is recognized by the catalytic site of CSN in its pre-catalytic state remains to be elucidated.Fig. 1: The pre-catalytic state of CSN in complex with N8~CRL1.a, Domain composition of CSN subunits, N8 and the CUL1–RBX1 E3 scaffold. b, Schematic drawings of CSN in its apo and N8~cullin-RING-bound forms with the CSN subunits in the catalytic hemisphere and the N8~CRL substrate coloured. Major conformational changes in CSN following substrate engagement are indicated by lock and clamp icons. c, The chemical structure of CSN5i-3. d, Cryo-EM map of the CSNDM–N8~CRL1 complex. e, The three-molecular interface among CSN5DM (surface in dark blue), N8 (cartoon in orange) and the CUL1–WHB domain (carton in green). The iso-peptide bond formed between the N8 C-terminal carboxyl group and the lysine residue of CUL1 (Lys120) is indicated. f, Recognition of the iso-peptide bond formed between N8 and Lys720 of the CUL1–WHB domain by CSN5DM with its reshaped Ins-1 loop. g, A close-up view of the catalytic site in isolated CSN5 occupied by its Ins-1 loop. PDB ID: 4D10. h, A close-up view of the interface between CSN5DM and the CUL1–WHB domain with the substrate’s iso-peptide bond housed at the catalytic site. i, A close-up view of the WHB–RBX1 interface. j, Two close-up views of the exosite interface between N8 (orange) and CSN5DM (slate).Full size imageCSN5i-3 has recently been developed as a small molecule CSN5 inhibitor that can potently block cullin deneddylation by CSN (Fig. 1c)23. Although the compound directly targets the CSN5 active site, it has recently been reported to inhibit the deneddylase complex via a surprisingly uncompetitive mechanism24. Unlike a non-competitive inhibitor, an uncompetitive (also known as anti-competitive) inhibitor is expected to bind only to the enzyme–substrate complex, not to the free enzyme25. Here we discover that CSN5i-3 unexpectedly acts as a molecular glue, which not only stabilizes the CSN–N8~CRL complex, but also gains its high potency in a substrate-dependent manner. This unusual mechanism of action leads us to establish the concept of orthosteric molecular glue (OMG) inhibitors.Pre-catalytic state of CSNDM–N8~CRL1As a metalloprotease, CSN5 features a catalytic Zn2+ ion at the end of a hydrophobic cleft, which has been predicted to recognize the C-terminal tail of N8 conjugated to cullins18. A crystal structure of isolated CSN5 bound to CSN5i-3 revealed that the compound directly coordinates the Zn2+ ion and occupies the enzyme active site as an orthosteric inhibitor23. Before reconciling its orthosteric nature and its reported uncompetitive mechanism, we first aimed to experimentally determine how the N8~CRL1 iso-peptide linkage is physically engaged with the CSN5 catalytic cleft in the absence of the compound. The catalytic zinc ion in CSN5 is coordinated by an aspartate (Asp151) and two histidine residues (His138 and His140), which are joined by an upstream water-activating glutamate residue (Glu76)16,17,26. Using a catalytically impaired CSN mutant harbouring a CSN5 H138A mutation, past studies have not been able to resolve the binding mode of the N8~cullin conjugates at the CSN5 catalytic site. In an enzymatic assay with N8~CRL1 as a substrate, we found that the CSNCSN5-H138A mutant, as well as several other documented CSN mutants, still retain a detectable iso-peptidase activity (Extended Data Fig. 1a). By contrast, a CSN5 E76A/D151N double mutation completely abolishes the catalytic activity of CSN towards N8~CRL1. Leveraging this CSN mutant (hereafter referred to as CSNDM), we determined the cryo-EM structure of a CSNDM–N8~CRL1 complex at 3.2 Å resolution (Fig. 1d, Extended Data Fig. 1b and Extended Data Table 1).Distinct from all previously reported CSN–N8~CRL structures, the catalytic hemisphere of the CSNDM–N8~CRL1 complex is clearly resolved in the three-dimensional reconstruction map with the substrate iso-peptide linkage stably trapped at the CSN5 catalytic site (Fig. 1e). The resulting structural model confirms the previously described global architectural changes in CSN following the binding of N8~CRLs; such changes are highlighted by the pivotal movement of the CSN5–CSN6 dimer, which is induced by the association of CUL1–CTD–RBX1 with CSN2 and CSN4. The complex structure also reveals a cascade of protein–protein interactions around the catalytic centre, involving N8, CSN5, CUL1–WHB and the RBX1 RING domain. Importantly, the resolution of the structure is high enough to resolve the complete N8 C-terminal tail conjugated to the side chain of the CUL1–WHB Lys720 residue, which together span the entire CSN5 catalytic cleft. In comparison to the apo form of CSN, the central region of the CSN5 Ins-1 loop is dislodged from the catalytic cleft and adopts a β-hairpin structure, forming a three-stranded anti-parallel β-sheet with N8 C-terminal tail (Fig. 1f,g). Such a structural arrangement positions the N8~CRL1 iso-peptide bond right above the catalytic zinc ion binding site, representing the pre-catalytic state of the enzyme–substrate complex.Pre-catalytic state protein interactionsSimilar to all previously reported CSN–N8~CRL structures, the polypeptide sequence connecting the CUL1 WHB domain to the CTD has no visible density and is presumably flexible in structure. The CUL1 WHB domain, nevertheless, is well resolved in the cryo-EM map, forming an extensive interface with CSN5 centred around the neddylation site, CUL1–Lys720 (Fig. 1h and Extended Data Fig. 1c). Part of this interface is made between CUL1–WHB and the CSN5 MPN core domain, where the aliphatic side chain of CUL1–Lys720 is buttressed by two aromatic residues of CSN5, Tyr143 and Trp146. These two CSN5 residues, at the same time, interact with Met721 and Lys723 of CUL1–WHB through hydrophobic and π–cation interactions, respectively. Two further CSN5 residues at the tip of the Ins-1 loop β-hairpin, Glu104 and Thr105, further strengthen the interface by interacting with Arg717 of CUL1–WHB via a salt bridge and van der Waal packing (Fig. 1h). Following engaging N8~CRL1, the rotation movement of the CSN5–CSN6 dimer pivots the CSN5 Ins-2 loop away from the helical bundle. Adopting a new conformation, this second CSN5 insertion loop joins Trp146 to wrap around the aliphatic side chain of CUL1–Lys723 (Extended Data Fig. 1c). Two of its hydrophobic residues—Ile211 and Val216—further augment the interface via hydrophobic packing.To test the functional importance of this CSN5–CUL1–WHB interface, we measured the enzymatic activity of CSN with the CSN5 Tyr143 and Trp146 residues individually mutated to alanine. Both mutants were catalytically compromised, albeit to different degrees (Extended Data Fig. 1d). Similarly, an internal truncation of the CSN5 Ins-2 loop (Δ205–217), markedly attenuated the activity of the iso-peptidase complex. In fact, converting Val216, a hydrophobic residue in the Ins-2 loop, to a charged amino acid was sufficient to achieve the same effect. These results suggest that the CSN5–CUL1–WHB interaction is functionally coupled with the engagement of the iso-peptide linkage at the CSN5 catalytic cleft.By securing the CUL1 WHB domain against CSN5, the active site-engaged iso-peptide linkage also stabilizes the