The study of fast and intricate enzyme reactions requires methods that have the speed and sophistication to match. Such an approach reveals the way in which proteins are tagged with ubiquitin for destruction.
The ubiquitin–proteasome system is the primary pathway through which the stability of cellular regulatory proteins is controlled. Proteins marked with a chain of ubiquitin molecules are recognized by a cellular machine called the proteasome, and are rapidly degraded. The ubiquitins are added by an E1–E2–E3 enzyme cascade. This simple model has dominated the field for several years, but the actual mechanisms that control the rates and types of ubiquitin-chain extensions have remained poorly understood1. In particular, a minimum of four ubiquitins must be added to the substrate destined for destruction before it can be delivered to the proteasome2. As such, four ubiquitins must be added during a single encounter with the E3, lest the substrate dissociate prematurely and its ubiquitins be rapidly removed by cellular deubiquitylating enzymes.
There are two competing models for this chain extension — stepwise addition of ubiquitin monomers, and en-bloc transfer, wherein the ubiquitin chain is first built on the enzymatic machinery before being transferred to the substrate in a single step1,3,4. Differentiating between these models has been difficult owing to the speed of ubiquitin-transfer reactions. On page 615 of this issue5 and in a companion paper in Cell6, Raymond Deshaies and colleagues describe how they have dissected the ubiquitin-transfer reaction on a millisecond timescale, and provide compelling evidence in favour of the stepwise model.
Ubiquitylation of proteins occurs via transfer of ubiquitin from a ubiquitin-activating enzyme (E1) to a ubiquitin-conjugating enzyme (E2), which, in collaboration with a substrate-selective ubiquitin ligase (E3), covalently attaches ubiquitin to a lysine amino-acid residue on the substrate. The primary E2–E3 system employed by Deshaies and colleagues is the Cdc34–SCFCdc4 complex, whose substrates include cyclin E (CycE). SCFCdc4 is the prototypical form of a family of enzymes known as cullin-RING ubiquitin ligases (CRLs)7,8, and has been used as a model for some 300 CRLs in mammals9.
Using an assay that allowed visualization of the number of ubiquitin molecules attached to a model CycE peptide substrate, Pierce et al.5 observed that three or more ubiquitins are linked within 30 seconds of mixing the CycE– SCFCdc4 complex with ubiquitin-charged Cdc34 (Cdc34∼Ub, where ∼ represents a thioester bond between Cdc34 and ubiquitin). However, the precise manner in which Cdc34 transfers ubiquitin to CycE — stepwise, en bloc or combinations thereof — was not discernible. So the authors developed a theoretical model to test potential pathways of chain elongation, on the basis of the observed number of ubiquitins conjugated to CycE during a single encounter with the SCF. From a consideration of the distribution of ubiquitin molecules on a pre-assembled chain, and the number of transfer events from Cdc34∼Ub to substrate, it was apparent that only stepwise or en-bloc transfer was occurring.
With stepwise transfer, one would expect to observe Cdc34 molecules charged with a single ubiquitin, whereas the accumulation of charged Cdc34 molecules containing multiple ubiquitins would favour the en-bloc model (Fig. 1). Using mass spectrometry, Pierce et al.5 found that Cdc34 is linked with a single ubiquitin in vitro, consistent with stepwise transfer. These results are in agreement with studies10 indicating that human Cdc34 is charged with a single ubiquitin in vivo.
To distinguish further between stepwise and en-bloc transfer, Pierce et al.5 performed mixing reactions on a millisecond timescale in a quench-flow device, and subsequently visualized the reaction products. In this setting, the CycE substrate is modified with a single ubiquitin within 10 milliseconds, and subsequent conjugates appear sequentially. These data, along with the fact that Cdc34 is charged with a single ubiquitin, argue that Cdc34 transfers ubiquitin to the substrate in a stepwise manner (Fig. 2). In extending the approach to SCFβ-TrCP and its substrate β-catenin, Pierce et al. show that the stepwise mechanism is generalizable to other CRLs.
Several observations emerge from this study that potentially explain how CRLs modulate substrate degradation half-lives in vivo. First, most SCF–substrate encounters may be unproductive because the substrate dissociation rate (koff) is faster than the rate of transfer of the first ubiquitin (kUb1). Second, the rate of ubiquitin transfer is not constant. Once a substrate is monoubiquitylated, the rate of chain elongation increases dramatically before eventually declining (Fig. 2), thereby allowing multiple ubiquitin transfers before substrate dissociation5,11. This decline in rate probably reflects topological constraints conferred by the growing ubiquitin chain. Third, individual kUb1 values differ for SCFCdc4 and SCFβ-TrCP by an order of magnitude, whereas koff values for their individual peptide substrates are similar regardless of ubiquitin-chain length5. A substrate with a larger kUb1 and a smaller koff is more likely to productively encounter the SCF. Given the diversity of substrates and mechanisms of recognition by CRLs, it seems likely that the kUb1 and koff values, and hence rates of polyubiquitylation and turnover, will vary significantly with full-length substrates in vivo.
The term 'processivity' describes the number of ubiquitins transferred before the substrate dissociates, and the process is key to the production of chains that are long enough for proteasomal targeting. Because there is a single RING domain in the subunit (Rbx1) of the CRL used for Cdc34∼Ub recruitment, discharged Cdc34 must dissociate from the CRL after ubiquitin transfer to allow for recruitment of the next Cdc34∼Ub. Processivity does not depend on the rate of Cdc34 recharging as such, because most Cdc34 in the cell is pre-charged with ubiquitin under steady-state conditions10. However, Cdc34∼Ub binds tightly to the SCF, raising the question of how Cdc34 balances affinity with processivity. Insight into this conundrum comes from an analysis of Cdc34–SCF interaction kinetics by Kleiger and colleagues6.
In addition to the well-known interaction between Cdc34 and Rbx1, Kleiger et al.6 identify a new interaction between an acidic carboxy-terminal tail on Cdc34 and a basic 'canyon' in the cullin subunit (Fig. 2). Remarkably, association of Cdc34 with SCF is about two orders of magnitude faster than the diffusion-controlled limit. Coupled with a fast dissociation rate and the abundance of Cdc34∼Ub, this allows for many encounters with substrate-bound SCF before substrate dissociation. Mutations in either the basic canyon or the acidic tail substantially impede processivity6, indicating that electrostatic complementarity has a key role in increasing the frequency of productive encounters. Thus, engagement of Cdc34∼Ub by the SCF can be viewed as involving two major features: initial rapid recognition, perhaps depending mainly on the tail–canyon interaction, and a secondary positioning interaction via Rbx1 that orients the ubiquitin thioester for transfer to the substrate. The finding that the canyon exists across the cullin protein family makes it likely that this kinetic mechanism will be used by all CRLs.
CRL enzymes surmount seemingly impossible odds to polyubiquitylate substrates, thereby establishing the dynamics of much of the proteome. By exploiting small regions of charge conservation, CRLs have evolved to maintain high-affinity interactions as well as rapid on/off kinetics to sequentially ubiquitylate substrates. The acidic tail of Cdc34 is unique among the cellular E2 enzymes examined, so further studies will be required to determine the extent to which analogous mechanisms are at work for other E2–E3 pairs. The new papers5,6 provide a framework for examining this question, while delivering an unprecedented view of how Cdc34–SCF generates a stairway to the proteasome.
About this article
Dynamics of ubiquitin-mediated signalling: insights from mathematical modelling and experimental studies
Briefings in Bioinformatics (2016)