Substrate degradation by the proteasome: A single-molecule kinetic analysis

Most, if not all, APC substrates contain multiple ubiquitylatable lysine residues. To reduce the complexity of these mixed configurations, we sought to control the number, length, and linkage of conjugated ubiquitin chains on wild-type (WT) substrates. Purified APC was used as the ubiquitin ligase to conjugate preformed di- or tetraubiquitin chains of Lys⁴⁸ (K48) linkage, which support most proteasome-mediated degradation in cells (24) (Fig. 1A). The chains were methylated on lysines before conjugation to prevent secondary elongation. We then separated the reaction products having different numbers of conjugated ubiquitins by electrophoresis. This approach limits the heterogeneity of ubiquitin configurations on the substrate to combinations of sites accepting a known number of defined ubiquitin chains. We measured the degradation rate of the ubiquitylated substrate for each electrophoretically resolved species by exposure to purified human 26S proteasomes that were free of reversibly associating ubiquitin receptors, such as Rad23 (Fig. 1A). Success of this strategy required the absence of interconversion of different ubiquitylated species generated by partial deubiquitylation by the proteasome; this was ensured by removing the DUB Usp14 on the proteasome by salt wash (25). Usp14-free proteasomes are known to efficiently degrade substrates, such as cyclin B and Sic1, without generating partially deubiquitylated products (fig. S1) (8, 25, 26). Another DUB on the proteasome, Uch37, does not appreciably deubiquitylate the substrates used here (25). Controls using a general DUB inhibitor (not active against Rpn11) further confirmed that DUB-driven interconversion of ubiquitylated species was unlikely (fig. S2).

We analyzed the rates of degradation of ubiquitylated securin, geminin, and cyclin B–NT (N-terminal fragment from cyclin B), each with a known number of ubiquitin chains of defined length. We incubated these substrates with the 26S proteasome at a nonlimiting concentration (25) (fig. S3). The measured degradation rates were similar to rates observed in cells (26, 27). The degradation rates were substrate-dependent, even for substrates with the same number of conjugated ubiquitins. For example, highly ubiquitylated securin was degraded fastest, followed by cyclin B and geminin (Fig. 1, B to D). Both methylated K48-diubiquitin chains and WT ubiquitins promoted efficient proteasomal degradation. Conjugation with K48-diubiquitin chains supported a higher rate of degradation than K48-tetraubiquitin chains when the same number of total ubiquitins were conjugated to a substrate molecule (Fig. 1, B to D). The slower degradation associated with tetraubiquitin chains was not due to methylation (fig. S4) nor to potential competition from free tetraubiquitin chains at the experimental concentration, as multidiubiquitylated securin was degraded at the same rate, even after free tetraubiquitin chains were added to the degradation reaction (fig. S4). Faster degradation was observed with diubiquitin chains, even when the two chains were closely apposed (fig. S5); this also held for K11-linked chains (fig. S6). In contrast to cyclin B, multi-monoubiquitylated securin and geminin, generated with either methylated ubiquitin or lysine-free ubiquitin (Ub0K) precluding chain formation, were degraded very slowly compared with constructs with the same ubiquitin stoichiometry but containing chains (Fig. 1, B and C, and fig. S7).

The distribution of ubiquitylated lysines on the substrate may set the degradation rate. Because our method to control ubiquitin configurations does not constrain which lysine residues receive the ubiquitin chains, we explicitly tested the contribution of the sites of ubiquitylation to the degradation rate. Multi-monoubiquitylated cyclin B–NT was degraded as efficiently as WT Ub–conjugated cyclin B–NT (8) (Fig. 1D), perhaps because of the specific configurations of lysine targets on cyclin B. We made cyclin B–NT mutants that contain only three or four lysine residues at their original locations, with other lysines mutated to arginines, and studied rates of degradation of the multi-monoubiquitylated products of those mutants. A clear pattern emerged: Given the same number of total conjugated ubiquitins, the degradation rate strongly depended on the position of the most N-terminal, ubiquitylated lysine residue; the closer to the N terminus, the higher the rate of degradation (Fig. 1E). This pattern was not explained by some mutants being inherently nondegradable, because all mutants were degraded similarly when conjugated to WT Ub at high ubiquitin multiplicity (Fig. 1F) (Ub = 7 to 8). Therefore, moving lysine residues away from the N terminus appeared to progressively convert cyclin B into a substrate whose degradation was more similar to that of securin and geminin. SM studies presented below provide mechanistic insights into these properties.

Recombinant, full-length human securin, Ube2S, GST-Emi1 (297-447), Xenopus geminin, Xenopus N-terminal cyclin B (amino acids 1 to 104), human N-terminal cyclin B (amino acids 1 to 88), and mutants were purified from Escherichia coli cells using an IMPACT kit (NEB, E6901S). A PKA (protein kinase A) site (RRASV) was also placed at the N terminus of the substrates used in degradation assays. Human ubiquitin, UbK48, Ub0K, UBK6A, and UbI44A mutants, each with a cysteine residue and a His tag at the N terminus, were purified from E. coli cells using cation-exchange chromatography (GE, 17-1152-01) labeled with DyLight 550 maleimide (Pierce, 62290). After removing unreacted dyes, labeled ubiquitin was subjected to anion-exchange chromatography (GE, 17-1153-01) to separate labeled and unlabeled species. Finally, the N-terminal His tag was cleaved off using thrombin.

Fluorescently labeled ubiquitin chains were synthesized using E2-25K in reactions containing 30 μM DyLight 550–labeled UbK48 and 10 μM DyLight 550–labeled Ub0K; carried out in UBAB buffer [25 mM Tris-HCl (pH 7.5), 50 mM NaCl, 10 mM MgCl₂], 3 mM ATP, 1 μM E1, and 10 μM E2-25K; and incubated at 37°C for 16 hours. The product was purified and separated using anion-exchange chromatography. For binding assays with the proteasome, different fractions of ubiquitin chains were mixed as the sample.

Anti-20S antibody (MCP21) was biotinylated using biotin-NHS (Pierce, 20217) and was purified using desalting columns.

Radioactive ³³P-ATP was used to label substrates with a PKA site at the N terminus for in vitro ubiquitylation and degradation assays.

E1, E2s (UbcH10, E2-25K), WT ubiquitin, biotin ubiquitin, methylated ubiquitin, K48-diubiquitin chains, and K48-tetraubiquitin chains were from Boston Biochem. Purified streptavidin was from Invitrogen.

For biotinylating substrates, a biotin-containing peptide (NEB, Bio-P1) was ligated to the substrate C terminus, which had been activated by intein cleavage during purification, resulting in ~90% ligation efficiency.

Ubiquitin chain samples were buffer-exchanged into a mixture of 50 mM Hepes-Na (pH 7.5), 150 mM NaCl, and 2% glycerol. Dimethylamine borane and formaldehyde were added to the sample to final concentrations of 20 and 40 mM, respectively. After 2-hour incubation on ice, another 20 mM dimethylamine borane and 40 mM formaldehyde were again added, bringing final concentrations to 40 and 80 mM, respectively, and incubation was continued for 16 hours on ice. Methylated proteins were buffer-exchanged to tris-buffered saline [20 mM Tris-HCL (pH 7.5), 150 mM NaCl].

HeLa cell G₁ extract preparation and APC purification were performed as described previously (13). Briefly, 2L HeLa S3 cell spinner culture was synchronized at prometaphase using thymidine/nocodazole double block and was then released for 3 hours into G₁. Cells were harvested and subjected to nitrogen cavitation in 75% volume swelling buffer [20 mM Tris-HCl (pH 7.5), 5 mM KCl, 1.5 mM MgCl₂, 1 mM dithiothreitol (DTT), protease-inhibitor tablet (Roche, 05892953001)]. APC was purified from cell extracts using anti-Cdc27 (AF3.1, custom-made) agarose and eluted using a competitive peptide.

Human proteasomes were affinity-purified on a large scale from a stable human embryonic kidney 293 cell line harboring HTBH-tagged hRPN11 (a gift from L. Huang). The cells were Dounce-homogenized in lysis buffer [50 mM NaH₂PO₄ (pH 7.5), 100 mM NaCl, 10% glycerol, 5 mM MgCl₂, 0.5% NP-40, 5 mM ATP, and 1 mM DTT] containing protease inhibitors. Lysates were cleared and then incubated with NeutrAvidin agarose resin (Thermo Scientific) overnight at 4°C. The beads were then washed with excess lysis buffer, followed by the wash buffer [50 mM Tris-HCl (pH 7.5), 1 mM MgCl₂, and 1 mM ATP]. Usp14 was removed from proteasomes using the wash buffer plus 100 mM NaCl for 30 min. 26S proteasomes were eluted from the beads by cleavage, using TEV protease (Invitrogen).

The APC ubiquitylation reactions were carried out in UBAB buffer [25 mM Tris-HCl (pH 7.5), 50 mM NaCl, 10 mM MgCl₂] containing 5 nM APC, 100 nM E1, 2 μM UbcH10, 2 mg/ml bovine serum albumin (BSA), an energy regenerating system, and either ubiquitin or ubiquitin chains: 10 μM ubiquitin, methylated ubiquitin, or DyLight 550–Ub0K; 5 μM methylated K48-diubiquitin; and 1 μM methylated K48-tetraubiquitin. Reactions were incubated at 30°C. Results were quantified using a phosphorimager. For PKA-labeled substrates, calyculin A (EMD, 19-139), a broad-spectrum phosphatase inhibitor, was added at 10 μg/ml.

Ube2S autoubiquitylation reactions were carried out in UBAB buffer containing 6 μM Ube2S, 0.5 μM E1, and 5 mM ATP. 100 μM ubiquitin, methylated ubiquitin, or 40 μM methylated K48-diubiquitin chain were also added. Reactions were incubated at 37°C for 4 hours.

PKA-labeled substrates (800 nM) were ubiquitylated by the APC using various forms of ubiquitin or ubiquitin chains for 45 min at 30°C. Reaction products were then mixed with proteasome-containing solution (UBAB buffer, 3 nM 26S proteasome, 2 mg/ml BSA, 1 μM GST-Emi1, 3 mM ATP, 10 μg/ml calyculin A) at a ratio of 1:4. The degradation reaction was sampled at 0, 3, 6, and 15 min at 37°C and was quantified using a phosphorimager.

To calculate the degradation rate, the intensity of each ubiquitylated species was quantified using ImageJ. These values were normalized to the intensity of the unmodified substrate (Ub0) to correct for loading error and nonspecific dephosphorylation. Finally, traces of the time course of degradation were fitted with an exponential decay function to obtain the rate constant.

Products of Ube2S autoubiquitylation reaction (6 μM) were mixed with proteasome-containing solution (UBAB buffer, 6 nM 26S proteasome, 2 mg/ml BSA, 3 mM ATP, 30 μM Z-GGL-AMC, 0.4 mM DTT) at a 1:1 ratio. Hydrolysis of GGL-AMC was monitored at 37°C using a fluorescence spectrometer.

Wild-type securin was ubiquitylated in a standard APC reaction with either WT Ub or Ub0K in a total volume of 20 μl for 1.5 hours at 30°C. The product was incubated with GST-magnetic beads conjugated with GST-Rpn10 for 40 min at 4°C (8). After 2× wash with buffer [25 mM Tris-HCl (pH 7.5), 100 mM NaCl, 0.5 mM EDTA, 0.05% Tween-20], the supernatant and bead-bound fraction were assayed for securin using anti-securin antibody by Western blot.

We followed a basic slide passivation protocol using 5-kD PEG plus 2.5% 5-kD biotin-PEG (LaysanBio, MPEG-SVA-5000; Biotin-PEG-SVA-5000) in a “clouding-point” solution on amino-silanized slides for 4 hours, as described previously (55, 56). In addition, we passivated them again using 1-kD PEG-NHS (Nacocs, PG1-SC-1k) for 1 hour, followed by 10% (w/v) BSA for 30 min before use. After passivation, slides were assembled into reaction chambers (55). Streptavidin (0.2 mg/ml) was incubated with passivated slides for 5 min for immobilizing biotinylated substrates. The quality of slide passivation was tested by incubating with HeLa cell extracts supplemented with DyLight 550–Ub and measuring the level of nonspecific binding.

Purified substrates (200 nM) were ubiquitylated by the APC with 5 μM fluorescent ubiquitin (WT Ub, Ub0K, UbK6A, or UbI44A) for 40 min (60 min for K64–cyclin B) at 30°C. The reaction product was diluted 100× into imaging buffer (UBAB, 3 mM ATP, 40 mM imidazole, 5 mg/ml BSA). Imidazole in the imaging buffer serves to reduce the nonspecific binding of fluorescently labeled proteins.

26S proteasome (15 nM) and biotinylated MCP21 antibody (15 nM) were mixed and incubated at room temperature for 15 min. The proteasome-antibody mix was loaded onto passivated slides coated with streptavidin. After a brief incubation, unbound protein was washed off and replaced with imaging buffer containing diluted ubiquitylation product. Image acquisition was started immediately with <15 s delay.

For experiments involving ATP-γ-S, ATP in imaging buffer was replaced with 2.5 mM ADP plus 0.5 mM ATP-γ-S. For experiments involving ADP, proteasome was incubated in 3 mM ADP buffer. ADP buffer and ubiquitylation product were also incubated with 30 mM glucose and 2 mg/ml hexokinase for 60 min at 30°C to remove residual ATP in the solution. For experiments involving 1,10-phenanthroline, 3 mM phenanthroline was added to the imaging buffer and proteasomal mixture; the sample was incubated for 20 min at room temperature before performing the experiment.

For the SM degradation assay, human cyclin B–NT was engineered with two cysteine residues close to its N and C termini and was labeled with DyLight 550–maleimide. Doubly labeled cyclin B was enriched via anion-exchange chromatography (HiTrap Q HP, pH 8.0). After ubiquitylation with unlabeled WT ubiquitin, 2 nM cyclin B was studied in the SM setup, as described above. Data analysis was identical to that described for the study of deubiquitylation.

We used a Nikon Ti TIRF microscope equipped with three laser lines of 491 nM (27 mW, full output from objective), 561 nM (30 mW), and 640 nm (59 mW), as well as an Andor DU-897 EMCCD camera. Time series were acquired at 200 ms per frame for 2 min, unless indicated.

The fluorescence intensity of a single DyLight 550 molecule on ubiquitin was calibrated through photobleaching preformed ubiquitin chains (27). Ubiquitin chains were synthesized in a standard E2-25K reaction (see above) containing 30 μM DyLight 550–labeled ubiquitin and 5 μM biotin-ubiquitin in UBAB buffer supplemented with 3 mM ATP, 1 μM E1, and 5 μM E2-25K. The reaction was incubated at 37°C for 16 hours.

The E2-25K reaction product was diluted 10,000× in imaging buffer (above) and loaded on a passivated slide via streptavidin. Photobleaching was observed with the use of a TIRF microscope, illuminated with a laser-intensity level 2.5× higher than that in proteasome experiments (to achieve faster photobleaching).

The uncertainty of measuring ubiquitin stoichiometry is ~30%, as suggested by the standard deviation of the intensity distribution of single fluorescent ubiquitin. This uncertainty is primarily due to uncorrected uneven illumination.

To test the reproducibility of dwell time data as in fig. S13, the same experiment with slightly varying concentrations (±50%) of ubiquitylated substrates was repeated three times on the same coverslip but at different positions.

Long-term (<1 year) reproducibility tests of the results of binding measurements carry 10 to 20% variation, probably due to batch-to-batch variability of proteasome samples or laser intensity drifts. Therefore, to minimize systematic variation in the measurements, each experiment with its controls was performed on the same slide using the same batch of proteasome and substrates.

The workflow of data analysis is illustrated in fig. S38.

Raw images were first corrected for stage drifting after subtracting a uniform-intensity background. Stage drifting was corrected by subtracting the stage motion, which was derived using successive image registration to calculate the shift (step 1 in fig. S38). A custom-built algorithm was applied to identify binding spots of ubiquitylated molecules to the proteasome on the basis of their absolute intensity and local signal-to-noise ratio (step 2 in fig. S38). The spot-identification algorithm is based on finding the local maxima of ubiquitin fluorescence intensity in a field of view by applying a filter requiring the local signal-to-noise ratio to be larger than 6 and no other spots within four pixels around it. In this way, >90% of substrate-binding events can be identified compared with a blind manual identification (fig. S10). The false-positive rate is less than 5%, as shown by the nonsubstrate control, considering that one ubiquitin generates an average of ~20 photons on the camera per frame. After each substrate-binding spot has been recognized in a time lapse, spots are registered along the time axis according to their coordinate to designate substrate-binding events (step 3 in fig. S38). Specifically, if the coordinates of two substrate spots in two consecutive frames are less than or equal to one pixel apart, they are considered to be the same binding event. The absolute intensity of each spot was obtained by fitting the SM diffraction pattern with a two-dimensional Gaussian function, which generates both the signal intensity I(s) and the local background level I(bg) (step 4 in fig. S38). The signal intensity I(s) is then corrected for inhomogeneous illumination (step 5 in fig. S38). Inhomogeneous illumination was corrected by a self-adaptive algorithm that separates the entire field of view into 15-by-15 identical squares and uses the average fluorescent spot intensity in each square as a surrogate for the relative illumination intensity in the center of that square. Because the illumination intensity varies slowly across the field of view, we interpolated the illumination intensity at a given position from the center values in each square. The corrected signal intensity was then converted to absolute ubiquitin numbers by normalizing with the intensity value of a single ubiquitin obtained in the calibration step. The resulting traces were directly used for subsequent analysis or represented in figures, with no further processing.

A custom-built algorithm was used to measure the duration of substrate-binding events (dwell time), based on the background noise intensity plus a cutoff of 0.7 Ub level. The binding measurement is insensitive to the choice of the cutoff (0.5 to 0.9 Ub) or laser intensity (fig. S14). The maximal fluorescence intensity (converted to the number of ubiquitins using a calibrated single-fluorophore intensity value) during each binding event was measured and plotted versus the dwell time.

A custom-built step-detection algorithm was used to identify stepped events (fig. S39). The formal definition of a stepped event is as follows: If two consecutive drops of fluorescent signal (>0.7 Ub) occur within 5 s, this is designated as a stepped event. Also, it is required that each intermediate state must last longer than one frame. Traces (~80%) can be identified correctly as compared with manual identification. Obvious classification errors were later corrected manually.

For counting the fraction of stepped events, only those events whose ubiquitin number was ≥3 were included in the analysis. We also excluded binding events that lasted longer than 90 s (5 to 10% of total binding events).

The same step-detection algorithm was also used in the analysis of photobleaching traces, to extract single-fluorophore intensity values.

Segments of the curve (indicated in the figure) showing dwell time versus number of ubiquitin relationships were fitted with a linear function or an exponential function. A χ² test $\left({\chi }^2={\displaystyle \sum \frac{{\left(y_i-y_i^0\right)}^2}{e_i^2}}\right)$ (y_i, data point i; $y_i^0$, data point on the fitted line; e_i, standard error) was performed to obtain the P value for each fitting model.

Biotinylated securin was ubiquitylated by APC using DyLight 550–Ub and was immobilized on passivated coverslips via streptavidin. Time-lapse images were acquired under conditions identical to the SM proteasomal assay, but in the absence of proteasome. The total signal intensity in a central region of the field was analyzed to extract photobleaching information (fig. S26).