An allosteric inhibitor of substrate recognition by the SCF^Cdc4 ubiquitin ligase

The ubiquitin-proteasome system (UPS) mediates the intracellular degradation of many proteins through a cascade of enzyme activities, termed E1, E2 and E3, which serially activate and then transfer ubiquitin to substrate proteins3. E3 enzymes, also referred to as ubiquitin ligases, specifically recognize discrete sequence motifs in substrates termed degrons. The human genome encodes at least 600 E3 enzymes, each of which has the potential to recognize multiple substrates4 V体育平台登录. The largest class of E3 enzymes, the cullin-RING ligases (CRLs), were discovered through identification of the multi-subunit Skp1–Cdc53/Cullin–F-box protein (SCF) complexes1, 2. A large family of F-box proteins recruit substrates to the core SCF complex via protein interaction domains, typically leucine rich repeats (LRRs) or WD40 repeats, often in a phosphorylation dependent manner1, 2, 5-7. The SCF enzymes likely target hundreds of different substrates4, 8-10 and thus hold untapped potential for drug discovery4.

The WD40 repeat is an ancient conserved motif that functions in many different cellular processes11, 12. Tandem arrays of five to eight WD40 repeats form a circularly permuted β-propeller domain structure13. In yeast, recognition of the cyclin-dependent kinase (CDK) inhibitor Sic1 by the WD40 domain of the F-box protein Cdc4 depends on phosphorylation of multiple Cdc4 phospho-degron (CPD) motifs in Sic16, 14. SCFCdc4 also targets other substrates including Far1, Cdc6 and Gcn41. Human Cdc4, also known as Fbw7, recruits a number of important regulatory factors for ubiquitination including cyclin E, Myc, Jun, Notch, SREBP and presenilin9 VSports注册入口. Cdc4 is a haploinsufficient tumor suppressor that is mutated in many cancer types9, 15, and also likely influences stem cell renewal by virtue of its effects on Myc and other factors16. Given the central role of Cdc4/Fbw7 in growth and division, we sought to identify small molecules that inhibit substrate recognition by Cdc4.

We adapted a previously established fluorescence polarization (FP) assay to monitor the displacement of a fluorescein-labeled CPD peptide (Kd ≈ 0 V体育官网入口. 2 μM) from yeast Cdc4 (Supplementary Fig. 1a)14. The FP assay achieved a Z-factor of 0. 8, based on negative (DMSO solvent only) and positive (unlabelled CPD peptide) controls. A screen against a 50,000 compound library enriched for drug-like molecules17 yielded 44 hits that inhibited the CPD-Cdc4 interaction by at least 50% (Fig. 1a). Two of these compounds, denoted SCF-I2 and SCF-I6, strongly inhibited the interaction of full length phospho-Sic1 with Cdc4 and prevented Sic1 ubiquitination by SCFCdc4 (Fig. 1b). We pursued only SCF-I2 because SCF-I6 appeared to cause non-specific loss of Skp1-Cdc4 complex from the capture resin (Fig 1b). SCF-I2 corresponds to 1-(2-carboxynaphth-1yl)-2-naphthoic acid, which is a derivative of 1,1′-binapthyl-2,2′diol, also known as BINOL, a bi-planar axially chiral atropisomer that is widely used as a scaffold in chiral synthesis18. The two hydroxyl groups of BINOL are substituted by carboxylic acid groups in SCF-I2 (Fig. 1c). The form of 1-(2-carboxynaphth-1-yl)-2-naphthoic acid) used in our all of our assays was an undefined racemic mixture of the R- and S- enantiomers, which are non-interconvertable at even high temperature18. SCF-I2 was 10-fold less potent than unlabeled CPD peptide in the FP assay, with an IC50 = 6. 2 μM versus 0. 5 μM, respectively (Fig. 1c). SCF-I2 inhibited binding and/or ubiquitination of both full length Sic1 and Far1 with an IC50 of ~60 μM (Supplementary Fig. 1b,c); the weaker apparent affinity of SCF-I2 in these assays may reflect differences in the interaction of peptides and full length substrates with Cdc4. SCF-I2 did not affect the activity of the closely related E3 enzyme SCFMet30, which recruits its substrate Met4 via the WD40 domain of the F-box protein Met30 (Supplementary Fig 1d)19.

We determined the crystal structure of SCF-I2 bound to a Skp1-Cdc4 complex20 to 2. 6 Å resolution (see Supplementary Table 1 for data collection and refinement statistics). Unbiased difference electron density maps revealed that SCF-I2 binds to the WD40 repeat domain of Cdc4 at a site that is 25 Å distant from the CPD binding pocket (Fig. 2a). The eight WD40 repeat motifs of Cdc4 form a canonical propeller structure in which each propeller blade consists of four anti-parallel β-strands and intervening loop regions (Supplementary Fig. 2)20. SCF-I2 embeds in a deep pocket on the lateral surface of the β-propeller between blades 5 and 6 (Fig. 2a,b; Supplementary Fig. 2). Cdc4 engages only one of two enantiomers of SCF-I2, the (R)-(+) equivalent of BINOL. The top napthalene ring system of SCF-I2 inserts deeply between blades 5 and 6 forming extensive hydrophobic contacts with Leu628, Ile594, Leu634, Trp657, and Ala649 (Fig. 2b) V体育2025版. In addition, the carboxyl group of the top ring system hydrogen bonds to the NH group of the Trp657 side chain and forms a salt bridge with the side chain of Arg664. The bottom napthalene ring system is more solvent exposed and forms a stabilizing co-planar stacking interaction with the side chain of Arg664 and van der Waals interactions with the side chains of Ser667 and Arg655. The carboxyl group of the bottom ring system also forms ionic interactions with the side chains of Arg655 and His631.

In the apo–Skp1-Cdc4 structure, there is no obvious pre-existing pocket that might anticipate the binding mode of SCF-I2 (Fig. 2c) VSports app下载. Rather, the SCF-I2 binding pocket is induced by separation of blades 5 and 6 and a drastic shift of the β21-β22 linker that connects the two blades (Fig. 2d). The reorientation of the β21-β22 linker entails a 5 Å shift of the main chain and a massive 13 Å shift of the side chain of His631 from a buried to solvent-exposed position (Fig. 2d,e). These large conformational alterations create an inter-blade void that is filled by the rearrangement of residues proximal to the CPD binding pocket (Fig. 2d,e). The void is filled in part by a swap of side chain positions between Val635, which is normally buried and adjacent to His631, and the normally solvent-exposed Leu634 side chain; as a consequence of this rearrangement, the side chains of Val635 and Leu634 traverse 6 and 8 Å, respectively. The position vacated by Leu634 in turn is filled by rotation of the side chain of Tyr574. Critically, both Tyr574 and Leu634 comprise part of the highly conserved CPD binding infrastructure. In the CPD peptide–Skp1-Cdc4 complex20, Tyr574 and Leu634 line the hydrophobic P-2 binding pocket within the central pore (Fig. 2e) and thereby dictate the preference for hydrophobic residues at the P-2 position of the CPD consensus motif14, 20. The P-2 pocket is thus severely distorted by the reoriented side chains of Tyr574 and Leu634 in the SCF-I2 bound structure. In addition, the hydroxyl group of Tyr574 participates in stabilizing H-bond interactions with the side chain of Arg572, one of the four invariant essential Arg residues found in all Cdc4 orthologs20. Arg572 stabilizes the orientation of Tyr548, which in turn directly hydrogen bonds to the CPD phosphate group in the P0 position. Thus, SCF-I2 critically compromises the main binding pockets for the P-2 and P0 positions of the CPD consensus sequence14, 20. As predicted by this structural model, the effects of SCF-I2 are mimicked by a Tyr574Ala mutation, which results in a 20-fold reduction in affinity of Cdc4 for the CPD peptide (Fig. 2f).

We explored the determinants of the SCF-I2–Cdc4 interface. The two carboxylic acid groups of SCF-I2 exhibit marked charge complementarity with the guanidinium side chains of Arg655 and Arg664. Mutation of each Arg residue individually to Ala attenuated the inhibition of Cdc4 by SCF-I2 by at least 50-fold (Fig. 2g). Alleles bearing either mutation fully complemented Cdc4 function in vivo, indicating that this region of Cdc4 does not normally play a critical role in substrate recognition or SCF catalytic activity (Supplementary Fig. 3a). To investigate the structural features of SCF-I2 required for Cdc4 inhibition, we tested a panel of available BINOL analogs for activity in the FP assay (Supplementary Fig. 3b,c). This series demonstrated the importance of the napthalene ring systems that participate in numerous hydrophobic interactions and the carboxylate groups that form electrostatic interactions with the two Arg residues on Cdc4. These mutational and structure-activity results validate the binding mode for SCF-I2 observed in the crystal structure.

We next assessed the activity SCF-I2 towards human Fbw7. The key Cdc4 residues Arg655 and Arg664 are replaced in Fbw7 by Lys and Cys respectively, suggesting that Fbw7 might be resistant to inhibition by SCF-I2. This proved to be the case as SCF-I2 inhibited the CPD-Fbw7 interaction only at high concentrations (Fig. 3a). The residual inhibitory activity of SCF-I2 towards Fbw7 might be due to the conservative Arg-to-Lys substitution and the conservation of most other residues that form the induced SCF-I2 binding pocket (Fig 3b; Supplementary Fig. 2). Alignment of all human WD40 domains revealed that, aside from the two surface Arg residues, the pattern of SCF-I2 contact residues is often conserved (Supplementary Fig. 4). We are currently exploring whether the BINOL scaffold can be modified to more potently interact with Fbw7 and other human WD40 domain proteins.

The β subunits of heterotrimeric G proteins, which transduce signals from a host of G protein coupled receptors (GPCRs), are the most thoroughly studied WD40 domain proteins²¹. Notably, the interaction of the regulatory protein phosducin with the G_Tβ subunit of the heterotrimeric G-protein transducin also causes significant structural rearrangements between adjacent WD propeller blades²². These rearrangements induce a binding pocket for the C-terminal farnesyl moiety of the partner G_Tγ subunit, which may serve to regulate membrane association of the G_Tβγ complex²². Comparison of our SCF-I2–Cdc4 structure and the phosducin-Gβγ structure reveals three highly similar features. First, the ligand-bound forms of both structures exhibit an analogous buried-to-exposed transition of the conserved His residue at the apex of the connector between the affected blades (Fig. 3c). Second, the Cdc4 and Gβ structures show a remarkably close juxtaposition of induced binding pockets for the SCF-I2 and farnesyl ligands, respectively (Fig. 3c). Third, these rearrangements occur between blades 5 and 6 for both WD40 structures. That two functionally unrelated and evolutionarily distant proteins undergo similar induced conformational changes hints that allosteric responsiveness may be an intrinsic and conserved feature of the WD40 domain.

In contrast to conventional protein interaction inhibitors that directly block the substrate binding site, such as the p53-Mdm2 inhibitor nutlin²³, SCF-I2 elicits its effect by an allosteric mechanism. A structural feature of WD40 domains, and other β-propeller structures such as the Kelch domain, is the variability in blade number, which in known WD40 structures ranges from five to eight blades per domain¹³. The circular β-propeller structure can exhibit inter-blade separation²⁴ and structural tolerance to artificial insertion of an additional repeat²⁵. WD40 domains may thus be inherently susceptible to disruption by insertion of appropriately configured small molecules between adjacent blades. Although it remains to be determined whether all WD40 domains exhibit allosteric responsiveness, in other protein families ultra-conserved residues can transmit long range allosteric effects²⁶.

To our knowledge, SCF-I2 represents the first example of a WD40 domain inhibitor. As our data with Cdc4 and Fbw7 shows, allosteric inhibition can discriminate between even highly related domains that recognize identical substrate motifs; it may be thus feasible to design other inhibitors that are selective for particular WD40 domain proteins. Moreover, allosteric inhibitors may be combined with conventional binding pocket inhibitors to increase potency²⁷. The yeast genome encodes at least 113 proteins with WD40 or WD40-like domains that function in signaling, transcription, chromatin remodeling, mRNA splicing, DNA replication and repair, protein synthesis, the ubiquitin system, autophagy, vesicle trafficking, the cytoskeleton and organelle biogenesis (Supplementary Table 2). In humans, WD40 domains occur in at least 256 different proteins and perform similarly diverse functions (Supplementary Table 3). Biomedically important WD40 domain proteins include the F-box proteins Fbw7 and β-TrCP⁸, target of rapamycin (TOR) kinase complex subunits²⁸ and Gβ-subunits of heterotrimeric G proteins^{21, 27}. Our findings suggest that the WD40 domain may be generally accessible to allosteric modulation by small molecules.