Research Article

Specificity of the anaphase-promoting complex: A single-molecule study

See allHide authors and affiliations

Science  10 Apr 2015:
Vol. 348, Issue 6231, 1248737
DOI: 10.1126/science.1248737

Single-molecule assay of ubiquitylation

Many biological processes in cells are regulated by ubiquitin peptides that are attached to proteins. Measurement of single fluorescent molecules in cell extracts can be used to trace the kinetics of such reactions. Lu et al. refined assay conditions to follow ubiquitination by an E3 ubiquitin ligase (see the Perspective by Komander). They visualized the activity of the anaphase-promoting complex (APC), a ubiquitin ligase critical for control of the cell division cycle. The processive initial reaction catalyzed by the APC was replaced by slower reactions. The results show how small, commonly occurring recognition motifs can guide specific and highly controlled enzymatic events. In a companion paper, Lu et al. explored how the number and arrangement of added ubiquitin chains affected the interaction of ubiquitylated proteins with the proteasome (a protein complex that recognizes ubiquitylated proteins and degrades them). The extent of ubiquitylation determined the strength of interaction of a substrate protein with the proteasome, and the arrangement of the ubiquitin chains determined the movement of the protein into the proteasome and thus the rate of degradation.

Science, this issue 10.1126/science.1248737, 10.1126/science.1250834; see also p. 183

Structured Abstract


Regulation of biological processes is generally thought to be conveyed by structurally complementary interactions between molecules, resulting in a bound state that is the most energetically stable. However, the limited size of most recognition motifs poses a very general question as to how specificity is actually achieved. In the cell cycle, timing of events from exit of mitosis through passage into G1 is regulated by the anaphase-promoting complex (APC), an enzyme that catalyzes ubiquitin conjugation to multiple lysines on substrate molecules, for proteasome-mediated degradation. The difference in how APC ubiquitylates its substrates leads to the ordering of events. APC-substrate interactions are also mediated through very short and redundant sequences, such as the well-studied D and KEN boxes found in 69% of human proteins. It is unclear how the APC can distinguish a relatively small number of substrates from nonsubstrates having the same interaction motifs. At a total concentration of approximately 1 mM, many of these nonsubstrates should compete for APC binding, thus reducing the rate of ubiquitylation and degradation of the real targets.


Simple binding based on structural complementarity seems inadequate to explain the specificity in APC-mediated ubiquitylation. We therefore looked beyond a single binding step to the extensive network of reactions that take place at multiple sites on a substrate molecule and in a cellular environment with competing reactions. To describe these, we developed single-molecule fluorescence assays, capable of identifying multiple reaction intermediates, and applied these methods to the study of the kinetics of APC-mediated ubiquitylation in both cytosolic extracts and in purified reaction systems. Our goal was to understand how the APC selects the right substrates among competing molecules and generates effective ubiquitin configurations that can be recognized for degradation.


In this assay, the rates of ubiquitylation and APC-substrate interaction can be measured simultaneously and traced to individual substrate molecules. We find that the APC-mediated ubiquitylation initially involves a highly processive reaction conjugating several ubiquitins on a substrate molecule, followed by multiple encounters and reactions at a slower rate. The initial ubiquitylation greatly enhances substrate-binding affinity with the APC in subsequent reactions, by both increasing the on-rate and decreasing the off-rate. Results of kinetic studies in bulk assays further substantiate this positive feedback mechanism.


The process of converting the reactivity of ubiquitin on a substrate to binding affinity can complement weak recognition by the D/KEN-box motifs. Cycles of positive feedback achieve high specificity for substrates with short recognition motifs in a cellular environment of competing molecules. This process, called processive affinity amplification (PAA), iteratively increases conjugated ubiquitins on substrates with both the recognition motifs and receptor lysine residues. Reaction schemes similar to PAA may be responsible for maximizing specificity, while maintaining efficiency in other biological processes.

Positive feedback enhances specificity and efficiency in APC-mediated ubiquitylation.

The intrinsic affinity between APC and D-box–containing decoy substrates is low, which prevents APC from inhibition by a high concentration of those substrates in the cytosol (A). The initial ubiquitylation on real substrates is highly processive, likely due to their preponderance in the availability and reactivity of receptor lysines. Conjugated ubiquitins enhance substrate’s affinity for the APC and promote further ubiquitylation, enabling proteasomal degradation (B).


Biological processes require specific enzymatic reactions, paradoxically involving short recognition sequences. As an example, cell-cycle timing depends on a sequence of ubiquitylation events mediated by the anaphase-promoting complex (APC) based on short redundant motifs. To understand the origin of specificity, we designed single-molecule fluorescence assays that capture transient ubiquitylation reactions. We find that the APC-mediated ubiquitylation involves a highly processive initial reaction on the substrate, followed by multiple encounters and reactions at a slower rate. The initial ubiquitylation greatly enhances the substrate’s binding affinity in subsequent reactions, by both increasing the on-rate and decreasing the off-rate. We postulate that these cycles of positive feedback enable high specificity for substrates with short recognition motifs in a complex cellular environment.

Regulation of biological processes requires that proteins distinguish the proper binding partners or substrates from nonsubstrates. Emil Fischer famously viewed the substrate as the key and the enzyme as the lock. There are many proven examples of specificity arising from such complementary binding interactions. However, specificity cannot be attributed completely to differential binding affinity alone, as has been demonstrated by the discrepancy between the error rate of protein synthesis and the measured binding affinities of the correct and incorrect aminoacyl–tRNAs on mRNA. In a conceptual theory for resolving such discrepancies, called “kinetic proofreading,” the energy from adenosine 5′-triphosphate or guanosine 5′-triphosphate hydrolysis is used to substantially reduce error rates (1, 2). The kinetic proofreading mechanism has been invoked to explain the high degree of specificity of various biological processes governed by relatively weak binding interactions (3, 4).

The importance of recognition and specificity in protein degradation is increasingly appreciated. During mitosis, cyclin A, cyclin B, securin, polo kinase, and UbcH10 are ubiquitylated sequentially by the anaphase-promoting complex (APC) and degraded by the proteasome, contributing to an ordered execution of cell-cycle events (5, 6). The precise timing of protein degradation at cell-cycle transitions demands high specificity in APC-substrate interaction. Paradoxically, these interactions are mediated by exceedingly short recognition sequences, such as the D [RXXLXXXX(N)] and KEN (KEN) boxes (where X is any amino acid) (79). These recognition sequences are highly redundant, appearing in 69% proteins in the human genome, whereas only 50 to 100 real substrates for the APC are expected from genome-wide studies (10). This preponderance of “decoy substrates” poses an additional challenge. At a total concentration of ~1 mM, they should compete for APC binding and reduce the rate of ubiquitylation and degradation of real substrates. Yet, this does not appear to happen. APC substrates, such as securin and cyclin B, can be efficiently degraded in a few minutes in yeast and mammalian cells. Although these small recognition sequences are required, they are not sufficient to confer susceptibility when transferred to other proteins (7). Thus, real APC targets must provide other specificity features, even though no additional sequence features are discernible. Therefore, the simple “key‐lock” mechanism appears to be inadequate to explain the specificity in APC-mediated ubiquitylation. To explain the required specificity, we looked beyond single binding interactions to the network of reactions that take place in the cellular environment, where there are competing substrates and competing reactions.

The APC interacts with substrates transiently and recurrently, conjugating complex and heterogeneous ubiquitin configurations on substrates molecules (11). These features make it difficult to understand the reaction kinetics and mechanisms with conventional biochemical methods. In this study, we have developed single‐molecule (SM) fluorescence assays capable of identifying multiple reaction intermediates. Using these assays, we could examine reactions at individual steps to study APC-mediated ubiquitylation in both purified systems and in crude cell extracts. We distinguished different modes of substrate binding, reflected in different kinetics for initial ubiquitin conjugation and subsequent elongation steps. We identified a reaction pathway based on the evolving ubiquitylation status of the substrate that can explain how substrate specificity and reaction efficiency can be achieved in a network of low-specificity interactions.


A single-molecule ubiquitylation assay

To observe reactions of single molecules, we used ubiquitin and substrates chemically labeled with different chemical fluorescent dyes (Fig. 1, A and B). The labeled proteins were incubated in concentrated extracts of human HeLa cells in the G1 phase of the cell cycle. Endogenous APC was labeled with a fluorescently tagged antibody; the substrates were C‐terminally labeled with biotin and tethered with streptavidin to polyethylene glycol and albumin passivated glass slides. Because each ubiquitin molecule was conjugated with a single fluorophore at its N terminus, the number of conjugated ubiquitins on a substrate molecule could be measured as the total fluorescence intensity. A major problem in developing the SM methods for ubiquitylation in Fig. 1B was the background fluctuation from fluorescently labeled ubiquitin in solution and the nonspecific binding of fluorescent species to the slide surface. We used combinatorial strategies to increase the signal-to-noise ratios and specificity of the detection (fig. S1); these allowed us to distinguish the addition and removal of single ubiquitins at a 95% confidence level in the presence of 3 µM fluorescent ubiquitin (materials and methods), near its physiological range of 5 to 10 µM. The chemical fluorophores with their high flux of photons, and a relatively long exposure time of 3 s afforded by the natural slow reaction rates, along with strong laser excitation, allowed us to reduce Poisson noise. Adding 10 mM imidazole greatly reduced nonspecific binding of all fluorescent species and lowered the background fluctuation without affecting kinetics of ubiquitylation (fig. S2). Intensity values of single molecules were calculated by a method of local background subtraction (figs. S3 and S4 and materials and methods).

Fig. 1 The single-molecule ubiquitylation assay.

(A) Major reaction pathways in the experiments. S: substrates; Dubs: deubiquitylating enzymes. (B) Schematics of the experimental setup. Substrates (S), ubiquitin, and the APC were fluorescently labeled. Substrates were also C-terminally biotinylated and immobilized on slides. Substrate-coated slides were incubated with functional cell extracts or purified APC reaction components. The kinetics of ubiquitylation and APC-substrate interaction were monitored on a TIRF microscope. (C) A representative trace of photobleaching a fluorescent polyubiquitin chain synthesized in an E2-25K reaction (more traces are in figs. S29 and S30). Right: Pairwise intensity difference distribution of 168 photobleaching traces. The position of the first peak has been normalized to “1.” The photobleaching experiment was performed in the absence of free ubiquitin. (D) Biotinylated, purified APC molecules on the slide were incubated with fluorescent antibody against APC4 to analyze the stability of antibody-APC interaction. Left: a representative trace; right: histogram of apparent interaction durations.

To distinguish substrate-specific ubiquitylation signals from nonspecific binding of ubiquitylated proteins from cell extracts, we limited our analysis to ubiquitylation events that were coincident with binding of the fluorescently labeled APC to a fluorescently labeled substrate. Because of the sparseness of nonspecific binding events, the chance of background binding of ubiquitin or ubiquitylated proteins coinciding with the labeled APC was less than 10−5 (materials and methods). The accuracy and linearity of the ubiquitylation measurement were assessed and confirmed by photobleaching of preformed ubiquitin chains, a process that randomly inactivates single fluorophores (Fig. 1C and figs. S5 and S29 to S30). The APC antibody labeling was sufficiently stable to allow observation of relatively transient substrate‐APC interactions (Fig. 1D). So that each event of ubiquitin conjugation was recorded, we completely depleted endogenous free ubiquitin from extracts by adsorption to beads containing the E2‐UbcH10, without affecting ubiquitylated proteins (fig. S6); the extracts were then supplemented with 3 µM Alexa Fluor 467 (alexa647)–labeled ubiquitin (fig. S7).

Ubiquitylation mediated by the APC at a single-molecule level

Geminin is an APC substrate that functions in control of DNA replication. When geminin was immobilized on a slide as the substrate, we observed a greatly enhanced ubiquitin fluorescence signal, compared to background, that was coincident with APC binding events (fig. S8 and movies S1 and S2). Overall ubiquitylation level was strongly reduced in the presence of the APC inhibitor Emi1, suggesting that APC is the primary E3 for geminin ubiquitylation (fig. S9). We recorded time traces of APC binding and ubiquitylation on single geminin molecules. In cell extracts, the ubiquitylation signal was dynamic, showing features that can be attributed to enzymatic deubiquitylation in the extract, proteasomal degradation of the substrate, and a slow rate of photobleaching (half-life = 8.6 min; fig. S10). The SM method fully reported on the heterogeneity in the reactions: After the binding of an APC molecule, ubiquitylation on individual geminin molecules occurred with variable rates and reached different numbers of conjugated ubiquitins at the moment of APC disassociation. We present a representative raw trace of these features in Fig. 2, A and B, and 20 more raw traces for each substrate [geminin, the cyclin B–N-terminal domain (cyclin B–NT), and cyclin A in cell extracts; cyclin B–NT, securin, and K64–cyclinB–NT in a purified system] in the supplementary materials (figs. S31 to S36). The SM kinetics of ubiquitylation on geminin and cyclin B–NT were very similar. Each is known to be a processive substrate, where processivity is defined as the average number of ubiquitins conjugated to a substrate molecule, regardless of its degree or configuration of ubiquitylation, in each round of APC binding (4).

Several observations indicate that the ubiquitin signal on immobilized substrates represents bona fide ubiquitylation by the APC, rather than the binding of ubiquitylated APC or ubiquitylation by other E3s. We did not observe a sharp decrease of ubiquitin signal when the APC dissociated from the substrate, suggesting that the APC was not ubiquitylated. Either the APC inhibitor Emi1 or methylation of lysines on substrates, which prevents ubiquitylation, strongly reduced the ubiquitin signal (figs. S9 and S11). In about 22% cases, we observed increased ubiquitin fluorescence in the absence of detectable APC binding. This may have been caused by either APC conjugated to an unmodified (and hence nonfluorescent antibody) or free APC unconjugated to antibody. However, we cannot exclude a low level of APC‐independent ubiquitylation. Ubiquitylated substrates on the slide were competent to interact with the proteasome (labeled with a fluorescent antibody) in the cell extract, suggesting that ubiquitin configurations that are competent for proteasomal degradation had been formed during the reaction (fig. S12).

In the SM traces, there were always two phases of ubiquitylation. The first was highly processive, soon after the initial binding of the APC, resulting in multiple ubiquityl residues transferred to a substrate molecule within 5 s (the time resolution of the experiment) (Fig. 2A and figs. S31 and S32). The second phase was much less processive, as the rate of ubiquitin transfer was lower and was interrupted by APC disassociations. Rebinding of APC was also frequently unproductive (“u” in Fig. 2, A and B), generating no detectable ubiquitin conjugations. Processive ubiquitylation could reoccur after the substrate had been partially or fully deubiquitylated. To illustrate the biphasic ubiquitylation statistically, we aligned 315 traces of cyclin B molecules ubiquitylated in cell extracts by their initiation points and plotted the median number of conjugated ubiquitins as a function of time. The processive initial reaction transferred an average of five ubiquityl moieties, followed by a slower transfer rate of ~1 ubiquitin per minute (Fig. 2C and fig. S13). The actual transfer rate may be slightly higher due to signal loss by photobleaching (fig. S10).

Fig. 2 Kinetics of APC-mediated ubiquitylation in cell extracts and in purified reaction systems.

(A) Representative traces of ubiquitylation and APC binding on geminin, cyclin B–NT, and cyclin A molecules in HeLa cell G1 extracts. Ubiquitylation signal (black) has been converted to the number of ubiquitin molecules as on the y axis. The APC traces (red) are plotted in arbitrary units. More examples are shown in the supplementary materials. p: productive binding; u: unproductive binding. (B), as in (A), but using wt-securin and single-lysine K64–cyclin B–NT in purified APC ubiquitylation reactions. (C) 315 traces of ubiquitylation on individual cyclin B–NT molecules in cell extracts with either wt-Ub (black) or zero-lysine-Ub (Ub0K; red), or cyclin A with wt-Ub (green) were aligned by the ubiquitylation initiation point (arbitrarily set at t = 2) of each substrate molecule. The median number of ubiquitins on a substrate molecule is plotted versus time. Gray traces show the boundary of the standard deviation of the mean. (D) As in (C), but using single-lysine K64–cyclin B–NT with or without 100 nM Ube2S in purified APC reactions. (E) Histogram of the maximum number of ubiquitins on each substrate molecule during a 15-min experiment in the cell extract. Upper, on cyclin B with either wt-Ub or zero-lysine-Ub; Lower, on either cyclin B or cyclin A with wt-Ub.

To distinguish the kinetics of monoubiquitylation from that of chain elongation, we replaced the endogenous free ubiquitin in cell extract with zero-lysine ubiquitin (Ub0K) to suppress chain formation. In this case, three rather than five ubiquityl moieties were transferred in the initial reaction, suggesting that some short ubiquitin chains were formed during the initial ubiquitylation when wild-type ubiquitin (wtUb) was used (Fig. 2C). In the second (or slower) phase, the rate of ubiquitin conjugation using wtUb was three times higher than that with Ub0K, suggesting that chain elongation occurred mainly after the initial processive reaction (fig. S13). It is unlikely that Ub chains were transferred en bloc (12, 13), because the vast majority of the E2 UbcH10 was charged with a single ubiquitin (fig. S14).

To generalize this result, we compared histograms of the maximum number of ubiquitins conjugated to individual cyclin B molecules during the experiment using wtUb or Ub0K. For cyclin B, wtUb provided twice the level of ubiquitylation as Ub0K (Fig. 2E), consistent with earlier results by mass spectrometry (11). Kinetics of binding of the APC was substrate‐dependent and strongly affected by substrate ubiquitylation status (see below). On average, 4.1 APC‐binding cycles were recorded on a cyclin B molecule in a 15‐min experiment, and 3.7 for geminin (Fig. 3A, fig. S15, and movie S3). To understand the contribution of each successive APC-binding cycle to substrate ubiquitylation, we analyzed the median number of ubiquitins, before, within, and after, the binding of APC on cyclin B–NT molecules in the cell extract. The first APC-binding cycle conjugated on average seven ubiquitins, including both the initial processive and the secondary distributive phases (fig. S16). Subsequent APC-binding cycles conjugated fewer ubiquitins but were important for chain elongation and for maintaining cyclin B in various ubiquitylated states in the presence of active deubiquitylation. To reduce the complexity due to deubiquitylating and proteasomal activities in the cell extract, we performed the same experiments in a purified system, composed of purified APC and other recombinant components. As in the cell extract, we found that for the processive substrates, securin and cyclin B–NT, ubiquitylation was biphasic and required multiple encounters with the APC (Fig. 2B, figs. S34 to S36, and movie S4).

Fig. 3 Kinetics of APC-substrate interaction.

(A) Upper: Histogram of the duration of APC-substrate interaction in the cell extract, plotted on a log-log scale. Only productive binding events have been considered. Lower: Histogram of the number of APC binding events on cyclin B–NT molecules in the cell extract during a 15-min experiment. (B) Time traces of the fraction of cyclin B–NT molecules bound by the APC in HeLa cell G1 extracts under the indicated conditions. About 2 × 105 cyclin B–NT molecules were immobilized as the substrate. The dashed line represents a curve fitted using a simple exponential mode described in the text. (C) As in (B), but performed in a purified APC ubiquitylation system with wt-alexa647-Ub. Results of cyclin B–NT, geminin, and securin are shown. The same analysis but using K64–cyclin B–NT as the substrate in purified APC reactions, with or without 100 nM E2 Ube2S, is shown in (D).

Differences in the processivity of ubiquitylation have been used to explain the order of degradation for APC substrates during the cell cycle (4). Multiple steps in the reaction pathway could contribute to processivity, and these can be identified by SM methods. We compared a well-characterized processive substrate cyclin B and a distributive substrate cyclin A for their kinetics of ubiquitylation and APC binding using SM assays. Cyclin A–APC interactions were transient, lasting for on average 7 s compared to 30 s for cyclin B–NT in cell extracts (Fig. 3A). Accordingly, less ubiquitin was transferred to cyclin A in the initial phase or accumulated during the experiment (Fig. 2, C and E). Nonetheless, ubiquitin transfer was still highly efficient on cyclin A molecules once APC bound (Fig. 2A and fig. S33). Therefore, the processivity of cyclin A seems to be limited by relatively short APC interaction and also perhaps by cyclin A having fewer ubiquitylatable lysine residues.

The elongation of ubiquitin chains on APC substrates is facilitated by another E2, Ube2S. It has been debated whether UbcH10‐mediated ubiquitin nucleation or Ube2S‐mediated chain elongation is rate‐limiting for ubiquitylation (1416). Using single‐lysine substrate K64–cyclin B–NT that can form only a single ubiquitin chain, we analyzed the two steps individually. In the absence of Ube2S, K64cyclinB‐NT essentially received no more than three ubiquityl moieties in an APC reaction (fig. S17). Ube2S promoted chain elongation in a concentration-dependent manner (fig. S17). In the SM assay with Ube2S at its physiological concentration of 100 nM, we found that chain elongation beyond three ubiquitins was usually slow and stepwise, compared to the highly processive initial ubiquitylation mediated by UbcH10 (Fig. 2D). We conclude that UbcH10 initiates monoubiquitylation and short chains with high processivity, but fails to form long ubiquitin chains. Further chain elongation by Ube2S is less processive than the initial ubiquitylation by UbcH10, when these E2s are compared at their physiological concentrations.

Positive feedback in APC-mediated ubiquitylation

APC-substrate interactions are highly dynamic. The kinetics of enzyme-substrate complex formation is commonly used in conventional assays to obtain the reaction rate constants, such as the off-rate (koff), by fitting a simple binding model between molecule A and B, described as [AB](t) ∝ 1 – exp(–koff *t) given Kd >> [A],[B], where [A], [B], and [AB] refer to the concentrations of A, B, and AB, respectively; Kd refers to the dissociation constant. Under this condition, the time scale of the binding curve is dominated by the off-rate, as detailed in materials and methods. Rate parameters obtained in this way should be consistent with the values calculated from analyzing SM traces; a discrepancy may indicate a time-dependent modulation of the interaction in the population.

In cell extracts, the number of APC molecules interacting with a fixed number of substrate molecules increased gradually during the experiment and eventually reached a steady state. This was a common feature for all the processive substrates tested in this study (movies S2, S4, S5, and S7). The parameter needed to fit the APC–cyclin B binding curve (Fig. 3B, blue, and movie S5), as measured by the number of APC molecules on the slide as a function of time, is 0.12 min−1. However, this is only 6% of the value for the off-rate obtained from analyzing SM traces of APC–cyclin B interaction in the presence of ubiquitylation (Fig. 3A). Therefore, the rate parameter of the APC–cyclin B binding curve does not represent the actual off-rate, but rather represents the time scale of an evolving factor that modulates the APC–cyclin B interaction, such as the varying state of ubiquitylation of the substrate, which was suggested by the overall parallel of ubiquitylation with the APC binding kinetics (fig. S18). Furthermore, low concentrations of free ubiquitin (the cyan curve in Fig. 3B), where no additional ubiquitin was added, led to reduced amounts of APC–cyclin B complex formation, suggesting that substrate ubiquitylation enhances its affinity with the APC (Fig. 3B and movie S6). This behavior is not an artifact due to the use of fluorescent ubiquitin, because unlabeled ubiquitin added to a Ub‐depleted cell extract restored the APC-binding kinetics measured with fluorescent ubiquitin (Fig. 3B and movie S7). This affinity enhancement was also observed in experiments with purified APC, and with other substrates (Fig. 3C), excluding a major role for unidentified cofactors in the cell extract. It is unlikely that E2s charged with ubiquitin facilitate APC binding to substrates, because we observed a reduced amount of substrate‐bound APC when we used a cell extract substituted with Ub0K in which the E2s were still charged with ubiquitin (Figs. 3B and 2C and movie S8). Because Ub0K cannot form chains, the result shows that ubiquitin chains on the substrate may facilitate, but are not absolutely required for, enhancement of APC binding. When the elongation factor E2 Ube2S, was added to an in vitro ubiquitylation reaction, it increased the number of conjugated ubiquitins on single‐lysine K64–cyclin B–NT and resulted in an enhanced binding of the APC (Fig. 3D). These results suggest that the number of conjugated ubiquitins, rather than the presence of specific ubiquitin configurations, determines the binding enhancement of the APC. In conclusion, the apparent discrepancy between SM and ensemble measurements is resolved if we consider the effect of ubiquitylation on increasing the substrate’s affinity with the APC; such a binding enhancement occurs with various ubiquitin configurations, including those without chains.

Because of constitutive deubiquitylation and degradation, ubiquitylated substrates usually represent only a small fraction of the total substrate concentration in the cell extract (fig. S19). Because Kd = [APC] × [Substratetotal]/[SubstrateAPC bound], a small increase in bound APC molecules may reflect a large decrease in Kd for ubiquitylated substrates with the APC. We estimated Kd values for the binding of ubiquitylated and unmodified substrate by the SM assay (materials and methods). Ubiquitylation reduced the Kd of the binding of cyclin B–NT by a factor of 60 and reduced the Kd for geminin by a factor of 40 (Fig. 4A). Consistently shown by SM traces, the interaction of APC with unmodified cyclin B was transient compared to that with ubiquitylated cyclin B and usually characterized by long intervals between successive binding cycles (fig. S20). To separate the contribution of on-rate and off-rate to the Kd reduction, we measured these kinetic constants by SM methods on cyclin B and related them to the ubiquitylation status of the same substrate. We used a purified system to prevent interference from the proteasome and deubiquitylating enzymes. Ubiquitylation simultaneously decreased koff and increased kon; together they accounted for the reduction in Kd (Fig. 4B). A similar result was obtained with securin as the substrate (fig. S22).

Fig. 4 Ubiquitylation on a substrate enhances its affinity with the APC.

(A) The average Kd values for cyclin B–NT, geminin, and their ubiquitylated forms with the APC in cell extracts. Error bars represent the SD of results from four experiments. (B) APC binding rate constants as a function of ubiquitylation on cyclin B–NT in a SM experiment using purified APC components. About 2300 cyclin B–NT molecules were analyzed. Inset: A schematic diagram explaining the calculation of binding rate constants. koff was calculated from APC binding time tb distribution, and kon was calculated from binding gap time tg. The distribution of tb and tg is shown in fig. S21.

To address potential concerns about the SM experimental design affecting enzyme behaviors at solid-liquid interface, we compared the SM results to conventional assays in bulk solution. Excess APC substrates compete for ubiquitylation by the APC and consequently inhibit proteasomal degradation. Our SM measurements indicate that a protein substrate that is nonubiquitylatable should have a lower affinity for the APC than do ubiquitylated substrates. Lysine methylation completely blocks ubiquitin conjugation on substrates, while preserving structural stability and the D-boxes on the protein. We used cyclin B–NT, geminin, and their methylated forms as competitive inhibitors in degradation assays in cell extracts or as inhibitors of ubiquitylation reactions by purified APC. Indeed, methylated cyclin B and geminin exhibited weaker inhibitory effects in both assays with a low concentration of radioactively labeled cyclin B–NT or geminin as the reporter (Fig. 5, A and B). A consistent but smaller effect was observed with securin as the substrate (fig. S23). We also tested the inhibition of APC by preubiquitylated cyclin B–NT using radioactive cyclin B–NT as the reporter. Preubiquitylated cyclin B–NT was more effective in reducing APC activity in in vitro ubiquitylation assays than unmodified cyclin B–NT, with the difference diminishing at lower concentrations (fig. S24).

Fig. 5 Substrate ubiquitylation affects APC binding kinetics in bulk assays.

(A) The rate of radioactive geminin degradation in HeLa cell G1 extracts, with indicated concentrations of unlabeled cyclin B–NT, geminin, or their methylated forms as a competitive inhibitor. The first set of data is an experimental replicate with no inhibitor. (B) Radioactive cyclin B–NT was ubiquitylated by purified APC in the presence of unlabeled cyclin B–NT or methylated cyclin B–NT as a competitive inhibitor. (C) Assay for the effect of substrate ubiquitylation on binding with the APC. Cyclin B–NT (2 µM) was incubated with purified APC reaction components (CycB-Ub) or without E2 (CycB) and was allowed to bind to beads. The amount of APC on beads was quantified by Western blot, shown below. Error bars represent the SD of results from three blots. SN: supernatant, B: beads-bound.

To assess directly the effect of ubiquitylation in modulating the interaction with APC, we performed a coprecipitation assay. Ubiquitylated cyclin B–NT interacted more strongly with APC than unmodified cyclin B–NT (Fig. 5C and S25). APC interacted very weakly with substrates, even after substrate ubiquitylation (Fig. 4A). This makes direct binding measurements difficult by conventional methods, which are prone to interference from nonspecific binding (fig. S25). Therefore, we would not expect the binding and competition experiments to quantitatively reproduce the results of SM measurements; nonetheless, both lines of evidence support a stronger interaction of ubiquitylated substrates with the APC.

If ubiquitins conjugated to substrates increase affinity for the APC, increased affinity should facilitate further ubiquitylation on the same substrate molecule during subsequent rounds of binding with the APC. To test this expectation, we fused a single ubiquitin to the N terminus of K48-securin to obtain a homogenous population of a monoubiquitylated substrate. Ub–K48-securin, but not K48-securin alone, was ubiquitylated efficiently in the presence of a high concentration of competitive inhibitors (Fig. 6A). We also tested the effect of conjugated ubiquitins on subsequent ubiquitylation in a reaction without competitive inhibitors by performing a chase experiment on partially ubiquitylated cyclin B–NT. In the middle of an in vitro ubiquitylation reaction on cyclin B–NT, we added excess hemagglutin (HA)–tagged ubiquitin and measured the distribution of HA‐ubiquitin incorporated onto the substrate by Western blot. HA‐ubiquitin was first incorporated into high–molecular weight conjugates of the total cyclin B–Ub population. In contrast, it was distributed more uniformly if HA‐ubiquitin was added before the reaction (Fig. 6B). It is unlikely that incorporation into the high–molecular weight conjugates is due to partially ubiquitylated cyclin B–APC complex at the point of HA‐ubiquitin addition, because APC turnover on cyclin B is much faster than the sampling rate (Fig. 3A). Rather, these results suggest that partially ubiquitylated substrates are the preferred targets for subsequent ubiquitylation by the APC. This conclusion, in conjunction with the effect of conjugated ubiquitins to promote binding with the APC, provides evidence of a positive feedback in the APC reaction.

Fig. 6 APC preferentially reacts with ubiquitylated substrates.

(A) The effect of N-terminal ubiquitin on the selectivity of ubiquitylation. Radioactive K48-securin or K48-securin with an N-terminal ubiquitin fusion (Ub-K48Sec) was used as the substrate in an APC ubiquitylation reaction in the presence (negative) or absence (positive) of 50 µM unlabeled wt-securin as a competitive inhibitor. (B) A chase experiment to determine the preference of addition of ubiquitin to substrates. Cyclin B–NT (200 nM) was subjected to ubiquitylation by purified APC. After 5 min when ~40% cyclin B–NT had been ubiquitylated, excess HA-ubiquitin was added to the reaction (Chase). In Control, HA-ubiquitin was added at the beginning. Samples were taken at indicated time points and analyzed by Western blot with anti-HA. Line profiles of corresponding lanes at each time point are compared in the right panels. The asterisk (*) indicates a nonspecific band. (C) The degree of ubiquitin conjugation in single- and multiple-encounter assays. Radioactive single-lysine K48-securin was used as the substrate in an APC ubiquitylation reaction including 100 nM E2 Ube2S. In the single-encounter assay (left), 50 µM unlabeled wt-securin was added to prevent rebinding of K48-securin to the APC, and sampled after 1 and 2 min. Negative: The competitive inhibitor was added together with labeled substrate. Sec-KMDBM: Using 50 µM wt-securin whose D box and KEN box had been mutated as a competitive inhibitor. Positive: No competitive inhibitor was added. The multi-turnover assay (right) is a standard APC ubiquitylation reaction with no competitive inhibitor. The average number of conjugated ubiquitins on each substrate molecule was quantified using a PhosphorImager.


Specificity in biology is often thought to be conveyed by complementary interactions between chemical species, with the degree of specificity simply reflecting the free-energy difference of binding. However, in certain biological systems, recognition may require more than complementarity. This may be true of substrate specificity in the ubiquitin system, which is largely determined by E3 enzymes. For one E3 enzyme, APC, substrates are characterized by the requirement for short recognition sequences, which are found throughout the genome on many nonsubstrates. Studies of the reactions in this system have been stymied by heterogeneities of the products and challenges in measuring the dynamics. To confront these difficulties, we designed single‐molecule fluorescence assays for detecting ubiquitin conjugation and E3-substrate interaction. We used multiple strategies, including minimizing fluorescent nonspecific binding and long exposure times to stabilize the local background fluctuation in total internal reflection fluorescence (TIRF) microscopy due to our use of near-physiological concentrations of free ubiquitin. We could obtain adequate sensitivity and signal-to-noise ratio to observe directly how the APC promoted polyubiquitylation through multiple encounters with a substrate molecule. Using this approach, we examined the mechanism of APC‐mediated ubiquitylation reactions in cell extracts and in purified systems and revealed unusual kinetic features; these features provide a basis for understanding how the APC can function efficiently in a complex cytosolic environment with numerous D/KEN-box–containing “decoy” substrates and deubiquitylating enzymes. As discussed below, these experiments suggest a mechanism for how APC can choose substrates among a very large number of proteins with identical recognition sequences.

Substrate molecules are likely to encounter the APC multiple times before degradation. Although we have not accurately measured the rate of proteasomal degradation in the SM assay, the half‐life of cyclin B and geminin is about 30 min in bulk assays under our experimental conditions, a time sufficient for multiple encounters to occur. The duration of substrate‐APC interactions is very heterogeneous even for molecules with the same number of conjugated ubiquitins, as expected for stochastic processes. Different ubiquitin configurations on the substrate may also contribute to the heterogeneity (11). In both extracts and purified systems, we found an initial highly processive ubiquitylation process, involving both multi‐monoubiquitylation and chain elongation. We did not resolve the processive phase into single ubiquitin steps. This does not represent a limitation of the SM detection method per se because we could observe stepwise addition of single ubiquitins in certain cases, presumably because of heterogeneity in the reaction rates (fig. S26); rather, it reflected insufficient time resolution due to limitations in the speed of detection.

Our study on a mutant of cyclin B with a single lysine showed a slower ubiquitin transfer rate by the elongation factor Ube2S than the initial ubiquitylation by UbcH10. We expect that Ube2S contributes less to the overall stoichiometry of ubiquitylation in a physiological context where the length of ubiquitin chains is restricted by the action of deubiquitylating enzymes (11). However, this small incremental contribution by Ube2S may be critical in selecting the substrate for degradation, because it solely promotes chain elongation.

Single-molecule kinetic studies demonstrate a highly processive initial ubiquitylation for APC-substrates. Furthermore, we find that conjugated ubiquitins bias further ubiquitylation by increasing kon and decreasing koff with the APC. In this way, the chemical reactivity of lysine residues (which themselves do not seem to contribute appreciably to substrate binding) is converted into enhanced affinity with the APC. This would recursively promote the binding and reaction of the substrate with the APC. The decrease in koff may result from ubiquitin interacting directly with the APC. The APC likely contains weak ubiquitin‐interacting regions whose identities have not been fully examined; however, in recent studies, ubiquitin has been found to interact with both RING domain and HECT domain of E3 ligases in the E2‐ubiquitin‐E3 crystal structures (17, 18). Specifically, the RING domain of APC has been shown to contain separate regions interacting with both donor and receptor ubiquitins (19). Kinetic studies also show that APC-Ube2S complex can interact and track emerging ubiquitin chains on substrates during elongation (20). Modifications of kon have not been typically encountered in biochemistry. In ubiquitylation reactions, an enhanced kon may be a consequence of an increased interaction cross-section between ubiquitylated substrates and the APC, specifically if the latter contains ubiquitin receptors.

The results of our experiments and mathematical modeling suggest a mechanism that would explain the specificity in APC-mediated ubiquitylation. At minimum, specificity should reflect sufficient interaction between enzyme and substrate. The SM measurements indicate that the intrinsic affinity of APC for unmodified substrates is very low, which may reflect the limited size of APC-binding motifs. A low substrate-binding affinity with corresponding high off-rates may be required to ensure the functions of APC. Otherwise, the APC would be completely inhibited by the many D- or KEN-box–containing decoy substrate proteins, the concentration of which reaches about 1 mM in the cell. However, low affinity should result in a low reaction rate and specificity for real substrates, which would result in very low reaction specificity. The SM experiments show that when the APC encounters a real target that contains sufficient ubiquitylatable lysine residues (see below), a wave of processive ubiquitylation occurs, and this leads to higher on-rates and lower off-rates for ubiquitylated proteins with the APC, which makes the ubiquitylated protein a “preferred” target in subsequent binding and reaction cycles (Fig. 7A). Such a positive feedback would amplify the initial weak interaction between the APC and substrate. We termed such process “processive affinity amplification” or “PAA” (Fig. 7A). The pronounced effect of PAA on increasing substrate specificity can be demonstrated with a simple ordinary‐differential‐equation model simulating substrate‐E3 binding, ubiquitylation, deubiquitylation, and degradation steps using physiologically appropriate parameters (Fig. 7B).

Fig. 7 Processive affinity amplification enhances the specificity of APC.

(A) A schematic diagram showing how the PAA mechanism could enhance the specificity of APC in cytosolic environments. See the text for details. (B) Left: The ratio of substrate specificity of the APC in the presence and absence of PAA as a function of the amplification factor, calculated from an ODE model (Methods). Sp: specificity, defined as the ratio of concentration of APC interacting with the real substrates and APC interacting with decoy substrates. The amplification factor represents the maximum fold of Kd reduction due to PAA. Right: The maximum rate of substrate degradation as a function of total substrate concentration, calculated from the ODE model with or without the PAA mechanism. (C) The average number of lysine residues within a certain distance (x axis) next to the D box is plotted for 74 published APC substrates from multiple species whose D box has been identified, and for all D-box–containing proteins in human proteome as a comparison.

PAA should also increase the efficiency of the APC‐ubiquitylation system, when it is challenged by deubiquitylating enzymes in the cell. Without PAA, these enzymes would substantially reduce the ubiquitylation and degradation efficiency, especially at high concentrations of substrates (including decoy substrates) (21). APC substrate molecules require multiple encounters with the E3 before they accumulate enough ubiquitins for degradation (Fig. 3A and figs. S15 and S16). In the presence of a high concentration of substrates and decoy substrates, there will be increased delays between successive binding cycles, long enough for deubiquitylating enzymes to partially or completely remove conjugated ubiquitin; this would drastically increase the time it takes to accumulate sufficient ubiquitins for degradation. Such an inhibitory effect can be demonstrated in the mathematical model as above, varying the concentration of total substrates (Fig. 7B). At the concentration of total D- or KEN box–containing proteins (about 1 mM) in the cell, APC efficiency for any real substrate drops to less than 1% of its maximum. The PAA mechanism, where partially ubiquitylated substrates bind more quickly and with higher affinity than nonubiquitylated substrates, would increase APC efficiency by focusing the ubiquitylation reactions on a subset of molecules, increasing their complement of conjugated ubiquitin until they are recognized by the proteasome. When we incorporate PAA into the model, the rate of substrate degradation is much less affected by high substrate concentrations (Fig. 7B). Such “focused” reaction kinetics can also be demonstrated experimentally. In an APC ubiquitylation reaction, the average number of conjugated ubiquitins was more than that obtained in a single encounter assay (using inhibitory concentrations of competitor to prevent substrate‐APC reinteraction), suggesting that ubiquitylated species have multiple encounters with the APC. But this was also the case when only a small fraction (5 to 10%) of substrates had reacted (Fig. 6C and fig. S27). In this case, it is unlikely that the small fraction has multiple encounters and reactions with the APC, while the vast majority of substrates does not, unless the APC “focuses” on the small subset by the process of affinity amplification. A similar ordinary differential equation (ODE) model that incorporates the PAA mechanism to simulate this experiment faithfully reproduces the time evolution of the average conjugated ubiquitins, as well as that of each ubiquitylated species (fig. S28).

Information for recognition by the APC must ultimately be encoded somewhere in the sequence or structure of the substrate molecules. The D and KEN boxes provide the initial interaction with the APC. These motifs are not equally accessible by the APC on all proteins, but because their sequences contain argininyl or lysyl side chains, they are likely to be exposed on a protein surface or in disordered regions (22). Besides the D and KEN boxes, sufficient lysine residues must be present on substrates to initiate the PAA. Indeed, we found 64% more lysine residues within 60 residues of the D box of published APC substrates, compared to random D‐box–containing proteins in the human proteome (23) (Fig. 7C). Lysine residues are not equally ubiquitylatable. Ubiquitylatability must certainly be affected by the distance and orientation of these residues to the ubiquitin thioester bond on E2 molecules, and also affected by the pKa of lysyl side chains, which would be affected by neighboring amino acids (24). There are no known sequence motifs for optimized lysine locations, and we might not expect to find them. Therefore, it appears that what determines substrate specificity in APC reactions, after the complementarity achieved by the simple recognition sequences, is the reactivity of lysine residues, likely near the D and KEN boxes. Ubiquitylatable lysine residues alone may not be sufficient to promote high specificity and efficiency for APC substrates in the absence of PAA. The ubiquitin transfer rate on APC substrates dropped to ~1 ubiquitin/minute on average after the initial phase of reaction (Fig. 2C). Without PAA supporting a stronger interaction with the APC, the very low affinity provided by D and KEN boxes with the E3 should not be able to sustain further ubiquitylation for proteasomal degradation in the presence of active deubiquitylation in the cell.

Finding an optimal trade-off of specificity and efficiency is likely to be a central requirement of all biological systems and, therefore, to be under strong selection. It has long been known that mutations in ribosome and DNA polymerase that increase specificity in protein synthesis and in DNA replication decrease the efficiency (or rate) of these processes, whereas mutations that decrease specificity actually increase efficiency (25, 26). Finding the right substrate is basically a search procedure where high on-rates and weak affinity translate into the speed of the search. If the affinity were strong, the APC would spend too much time sampling related sequences on a large number of decoy substrates in the cell. Results from the SM measurements and mathematical modeling explain how the APC may achieve high specificity without sacrificing too much efficiency. The PAA mechanism only increases the affinity of the APC with real substrates, but not with decoy substrates. The resultant stronger binding of the APC with ubiquitylated substrates will not slow down degradation by interfering with proteasome recognition, because even at submicromolar Kd, the APC still turns over quickly enough on substrates, compared to the rate of degradation. Consistent with PAA, nonubiquitylatable cyclin B with lysine residues mutated to arginine when overexpressed does not cause a cell-cycle block in Schizosaccharomyces pombe (27). This is expected because its interaction with the APC should remain weak, whereas overexpression of a wt–cyclin B, which can be ubiquitylated, inhibits cell-cycle progression. Interaction between APC and K11-linked ubiquitin chains can also be detected in vitro (28); however, most substrate-E3 interactions are not stable enough to be detected in the cell. Besides the positive feedback on reaction rates, deubiquitylation and multiple encounters with the APC are essential components in PAA, similar to the requirements in the kinetic proofreading model (1).

In the PAA model, APC achieves high substrate specificity with short binding motifs, while maintaining reaction efficiency. Similarly, recognition motifs of many important enzymes, such as the cyclin-dependent kinases (CDKs) and the Polo kinase, are short. For these and many other kinases, phosphorylation often occurs on multiple sites. Phosphorylation, like ubiquitylation, is opposed by hydrolases that reduce the efficiency of the process. In the case of CDK, when associated with the Cks protein, the complex has a higher affinity for substrates that have been phosphorylated by CDK. Thus, the initial phosphorylation could facilitate additional phosphorylation, essentially creating a positive feedback like PAA (29). Iterative cycles of binding and reversible modification, found in many systems, may be a common contrivance in biology to increase the specificity of a process, without too much sacrifice in speed.

Materials and methods

Details of the experimental procedures, statistical analysis, and data sets are provided in the supplementary materials.

Protein purification, labeling, and modification

Human full-length cyclin B, securin, securin(KMDBM), single-lysine k48-securin, Ub–k48-securin, UbcH10(K119R), N-terminal cyclin B(1–104) from Xenopus, single-lysine k64–cyclin B, and geminin from Xenopus were expressed in Escherichia coli and purified using intein-mediated purification with an affinity chitin-binding tag. For substrates requiring biotinylation, a biotin-containing peptide was ligated to the C terminus of these proteins through protein ligation, with a ~90% ligation efficiency. Biotinylated proteins were reacted with alexa488-maleimide and then purified by gel filtration. Human ubiquitin with a cysteine residue and a His-tag at the N terminus was purified from E. coli using cation exchange chromatography and was labeled on cysteine with alexa647-maleimide. HeLa cell G1 extract preparation and APC purification were performed as previously described (4). To radioactively label substrates for in vitro ubiquitylation reaction, a PKA site was added to the N terminus and labeled using p33ATP.

Single-molecule experimental procedures

Glass slides were passivated using 5-kD polyethylene glycol (PEG) + 2.5% 5-kD biotin–PEG in a “clouding-point” solution on amino-silanized slides for 4 hours. We passivated the slides further with 1-kD–PEG–NHS for 1 hour and 10% (w/v) bovine serum albumin (BSA) for 30 min just before the experiment. Streptavidin (0.2 mg/ml) was incubated with the passivated slide for 5 min to immobilize the biotinylated substrates. For experiments with cell extracts, endogenous free ubiquitin was first depleted from the cell extract just before the experiment by incubating with His6-UbcH10 for 20 min at room temperature, followed by Ni-agarose depletion of the conjugates. During this time, 500 nM dylight550–anti-APC4 antibody was added to cell extract to label the APC. Ub-depleted extracts were supplemented with 3 µM alexa647-Ub. A desired amount of biotinylated substrate was mixed with cell extract or in vitro reaction buffer and incubated with a functionalized slide to allow binding before data acquisition at room temperature. Time-lapse data were analyzed with custom-built software in Matlab.

Degradation and ubiquitylation assay

For degradation assay, 1/10 volume of radioactive substrate from in vitro translation system was mixed with functional cell extract supplemented with energy mix and 10 µM ubiquitin. Reactions were performed at room temperature. For in vitro ubiquitylation assay, PKA-labeled substrates were mixed with 4 to 20 nM APC, 100 nM E1, 2 µM UbcH10, 10 to 30 µM ubiquitin, BSA (2 mg/ml), and the energy-regenerating system in UBAB buffer [25 mM Tris-HCL (pH 7.5), 50 mM NaCl, 10 mM MgCl2], incubated at 30°C. Results were quantified using a PhosphorImager.

Supplementary Materials

Materials and Methods

Figs. S1 to S38

References (3034)

Movies S1 to S8

References and notes

  1. Acknowledgments: We thank R. King for providing plasmids and W. Ma, L. Bai, M. Rape, and R. King for commenting on the manuscript. We are grateful to the Nikon imaging center at Harvard Medical School and in particular to J. Waters for technical support; Y.L. is a Damon Runyon Cancer Research Fellow and a Lallage Feazel Wall Fellow. We acknowledge the continuing support of the National Institute of General Medical Sciences (grant no. 5R01GM039023-26).
View Abstract

Stay Connected to Science

Navigate This Article