USP14-regulated allostery of the human proteasome by time-resolved cryo-EM – Nature

Expression and purification of human USP14

Wild-type USP14 and its mutants were cloned into pGEX-4T vector obtained from GenScript (Nanjing, China). For purification of recombinant USP14 and mutants, BL21-CondonPlus (DE3)-RIPL cells (Shanghai Weidi) transformed with plasmids encoding wild-type or mutant USP14 were grown to an OD₆₀₀ of 0.6–0.7 in LB medium supplemented with 100 mg ml⁻¹ ampicillin. Cultures were cooled to 20 °C and induced with 0.2 mM IPTG overnight. Cells were collected by centrifugation at 3,000g for 15 min and resuspended in lysis buffer (25 mM Tris-HCl (pH 8.0), 150 mM NaCl, 0.2% NP-40, 1 mM DTT, 10% glycerol and 1× protease inhibitor cocktail). Cells were lysed by sonication and the lysate was cleared through centrifugation at 20,000g for 30 min at 4 °C. The supernatant was incubated with glutathione Sepharose 4B resin (GE Healthcare) for 3 h at 4 °C. For the purification of wild-type USP14, USP14 UBL domain (USP14-UBL) and USP14 USP domain (USP14-USP), the resin was washed with 20 column volumes of washing buffer (25 mM Tris-HCl (pH 8.0), 150 mM NaCl, 1 mM DTT, 10% glycerol), then incubated with cleavage buffer (20 mM Tris-HCl (pH 8.0), 150 mM NaCl) containing thrombin (Sigma) overnight at 4 °C. The eluted samples were further purified on a gel-filtration column (Superdex 75, GE Healthcare) equilibrated with 25 mM Tris-HCl (pH 8.0), 150 mM NaCl, 1 mM DTT, 10% glycerol. For the purification of USP14 mutants, the resin was washed with 20 column volumes of washing buffer (25 mM Tris-HCl (pH 8.0), 300 mM NaCl, 1 mM DTT), then incubated with cleavage buffer (20 mM Tris-HCl (pH 8.0), 150 mM NaCl) containing thrombin (Sigma) overnight at 4 °C. To remove thrombin, the GST eluent was incubated with Benzamidine-Sepharose (GE Healthcare) for 30 min at 4 °C.

Expression and purification of human 26S proteasome

Hexahistidine, TEV cleavage site, biotin and hexahistidine (HTBH)-tagged human 26S proteasomes were affinity purified as described^8,9,10 from a stable HEK 293 cell line (a gift from L. Huang). Further authentication of cell lines was not performed for this study. Mycoplasma testing was not performed for this study. In brief, HEK 293 cells were Dounce-homogenized in a lysis buffer (50 mM PBS (77.4% Na₂HPO₄, 22.6% NaH₂PO₄, pH 7.4), 5 mM MgCl₂, 5 mM ATP, 0.5% NP-40, 1 mM DTT and 10% glycerol) containing 1× protease inhibitor cocktail. The cleared lysates were incubated with Streptavidin Agarose resin (Yeasen) for 3 h at 4 °C. The resin was washed with 20 bed volumes of lysis buffer to remove endogenous USP14 and UCH37 associated with the proteasome^7,10. The 26S proteasomes were cleaved from the beads by TEV protease (Invitrogen) and further purified by gel filtration on a Superose 6 10/300 GL column. Western blot was used to detect RPN13 and USP14 in the proteasomes using anti-RPN13 antibody (Abcam, 1:10,000 dilution) and anti-USP14 antibody (Novus, 1:1,000 dilution). For ubiquitin–vinyl-sulfone (Ub–VS)-treated human proteasome, 1 µM Ub–VS (Boston Biochem) was added to the proteasome-binding resin and incubated for 2 h at 30 °C. Residual Ub–VS was removed by washing the beads with 30 bed volumes of wash buffer (50 mM Tris-HCl (pH7.5), 1 mM MgCl₂ and 1 mM ATP). The proteasomes were cleaved from the beads using TEV protease (Invitrogen) and used to measure the DUB activity of USP14 using the Ub–AMC hydrolysis assay.

Preparation of polyubiquitylated Sic1^PY

Sic1^PY and WW-HECT were purified as previously described¹⁰. The PY motif (Pro-Pro-Pro-Tyr) is recognized by the WW domains of the Rsp5 family of E3 ligases. In the Sic1^PY construct, a PY motif was inserted to the N-terminal segment (MTPSTPPSRGTRYLA) of the Cdk inhibitor Sic1, resulting in a modified N terminus of MTPSTPPPPYSRGTRYLA^43,44 (the PY motif is underlined). Human UBE1 (plasmid obtained as a gift from C. Tang) and human UBCH5A (obtained from GenScript) were expressed as GST fusion proteins from pGEX-4T vectors. In brief, UBE1-expressing BL21-CondonPlus (DE3)-RIPL cell cultures were induced with 0.2 mM IPTG for 20 h at 16 °C, whereas UBCH5A expression was induced with 0.2 mM IPTG overnight at 18 °C. Cells were collected in lysis buffer (25 mM Tris-HCl (pH 7.5), 150 mM NaCl, 10 mM MgCl₂, 0.2% Triton-X-100, 1 mM DTT) containing 1× protease inhibitor cocktail and lysed by sonication. The cleared lysates were incubated with glutathione Sepharose 4B resin for 3 h at 4 °C and subsequently washed with 20 bed volumes of lysis buffer. The GST tag was removed by thrombin protease (Sigma) in cleavage buffer (20 mM Tris-HCl (pH 8.0), 150 mM NaCl, 1mM DTT) overnight at 4 °C. The eluted samples were further purified by gel-filtration column (Superdex 75, GE Healthcare) equilibrated with 25 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1 mM DTT, 10% glycerol.

To ubiquitylate Sic1^PY, 1.2 μM Sic1^PY, 0.5 μM UBE1, 2 μM UBCH5A, 1.4 μM WW-HECT and 1 mg ml⁻¹ ubiquitin (Boston Biochem) were incubated in reaction buffer (50 mM Tris-HCl (pH 7.5), 100 mM NaCl, 10 mM MgCl₂, 2 mM ATP, 1 mM DTT and 10% glycerol) for 2 h at room temperature. His-tagged Sic1^PY conjugates (polyubiquitylated Sic1^PY, Ub_n–Sic1^PY) were purified by incubating with Ni-NTA resin (Qiagen) at 4 °C for 1 h. Afterwards the resin was washed with 20 column volumes of the wash buffer (50 mM Tris-HCl (pH 7.5), 100 mM NaCl, 10% glycerol). The Ub_n–Sic1^PY was eluted with the same buffer containing 150 mM imidazole, and finally exchanged to the storage buffer (50 mM Tris-HCl (pH 7.5), 100 mM NaCl, 10% glycerol) using an Amicon ultrafiltration device with 30K molecular cut-off (Millipore).

Expression and purification of human RPN13

To purify human RPN13, pGEX-4T-RPN13-transformed BL21-CondonPlus (DE3)-RIPL cells were cultured to an OD₆₀₀ of 0.6 and then induced by 0.2 mM IPTG for 20 h at 16 °C. Cells were resuspended in lysis buffer (25 mM Tris-HCl (pH 7.5), 300 mM NaCl, 1 mM EDTA, 0.2% Triton-X-100, 1 mM DTT) containing 1× protease inhibitor cocktail and lysed by sonication. A 20,000g supernatant was incubated with glutathione Sepharose 4B resin (GE Healthcare) for 3 h at 4 °C. The resin was washed with 20 column volumes of washing buffer (25 mM Tris-HCl (pH 8.0), 150 mM NaCl, 1 mM DTT, 10% glycerol) and 10 column volumes of cleavage buffer (20 mM Tris-HCl (pH 8.0), 150 mM NaCl). The GST tag was cleaved by incubating with thrombin (Sigma) overnight at 4 °C. The eluted samples were further purified by gel-filtration column (Superdex 75, GE Healthcare) equilibrated with 25 mM Tris-HCl (pH 8.0), 150 mM NaCl, 1 mM DTT, 10% glycerol.

In vitro degradation assay

Purified human proteasomes (~30 nM) were incubated with RPN13 (~300 nM), Ub_n–Sic1^PY (~300 nM) in degradation buffer (50 mM Tris-HCl (pH 7.5), 5 mM MgCl₂ and 5 mM ATP) at 37 °C. Purified recombinant USP14 variants (~1.2 μM) were incubated with proteasome for 20 min at room temperature before initiating the degradation reaction. The reaction mixtures were incubated at 37 °C for 0, 0.5, 1.0 and 2.0 min, or 10 °C for 0, 0.5, 1.0, 5.0, 10 and 30 min, then terminated by adding SDS loading buffer and subsequently analysed by western blot using anti-T7 antibody (Abcam, 1:1,000 dilution), which was used to examine fusion protein T7–Sic1^PY.

Ubiquitin–AMC hydrolysis assay

Ubiquitin–AMC (Ub–AMC; Boston Biochem) hydrolysis assay was used to quantify the deubiquitylating activity of wild-type and mutant USP14 in the human proteasome. The reactions were performed in reaction buffer (50 mM Tris-HCl (pH 7.5), 5 mM MgCl₂, 1 mM ATP, 1 mM DTT, 1 mM EDTA and 1 mg ml⁻¹ ovalbumin (Diamond)), containing 1 nM Ub–VS-treated proteasome, 0.2 μM USP14 variants and 10 nM RPN13. The reaction was initiated by adding 1 μM Ub–AMC. Ub–AMC hydrolysis was measured in a Varioskan Flash spectral scanning multimode reader (Thermo Fisher) by monitoring an increase of fluorescence excitation at 345 nm with an emission at 445 nm. For free USP14 activity, the reaction was performed using 1 μM USP14 variants and 1 μM Ub–AMC (BioVision).

ATPase activity assay

ATPase activity was quantified using malachite green phosphate assay kits (Sigma). Human proteasomes (30 nM), RPN13 (300 nM) and USP14 variants (1.2 μM) were incubated in assembly buffer (50 mM Tris-HCl (pH 7.5), 5 mM MgCl₂ and 0.5 mM ATP) for 20 min at room temperature. Ub_n–Sic1^PY (300 nM) was subsequently added, and the sample was incubated for 1 min at 37 °C. The reaction mixtures were mixed with malachite green buffers as described by the manufacturer (Sigma). After 30 min of room temperature incubation, the absorbance at 620 nm was determined using a Varioskan Flash spectral scanning multimode reader (Thermo Fisher).

Peptidase activity assay

Peptide hydrolysis by the human proteasomes was measured using fluorogenic substrate Suc-LLVY-AMC (MCE). Human proteasomes (1 nM) were incubated with USP14 variants (1 μM) in buffer (50 mM Tris-HCl (pH 7.5), 100 mM KCl, 0.5 mM MgCl₂, 0.1 mM ATP and 25 ng μl⁻¹ BSA) for 20 min at room temperature. 10 μM Suc-LLVY-AMC was added to the reaction mixture, which was incubated for 30 min at 37 °C in the dark. Peptide activity was measured in a Varioskan Flash spectral scanning multimode reader (Thermo Fisher) by excitation at 380 nm with an emission at 460 nm.

Microscale thermophoresis

The human proteasomes were labelled with red fluorescent dye NT-650-NHS using the Monolith NT Protein Labeling Kit (NanoTemper). After labelling, excess dye was removed by applying the sample on column B (provided in the kit) equilibrated with reaction buffer (50 mM Tris-HCl (pH 7.5), 5 mM MgCl₂ and 1 mM ATP). 0.05% Tween-20 was added to the sample before MST measurements. For interaction of NT-650-NHS-labelled proteasomes with USP14, USP14-UBL or USP14-USP, concentration series of USP14, USP14-UBL or USP14-USP were prepared using a 1:1 serial dilution of protein in reaction buffer containing 0.05% Tween-20. The range of USP14, USP14-UBL or USP14-USP concentration used began at 8 μM, with 16 serial dilution in 10-μl aliquots. The interaction was initiated by the addition of 10 μl of 30 nM NT-650-NHS-labelled proteasomes to each reaction mixture and measured by Monolith NT.115 (NanoTemper) at 20% LED excitation power and 40% MST power. To evaluate the effect of Ub_n–Sic1^PY on the interaction of USP14 with the proteasome, 30 nM Ub_n–Sic1^PY was added to the reaction mixture. Data were analysed using MO Control software provided by NanoTemper.

Cryo-EM sample preparation

To prepare cryo-EM samples, all purified proteins were exchanged to imaging buffer (50 mM Tris-HCl (pH 7.5), 5 mM MgCl₂ and 1 mM ATP). Human proteasomes (1 μM) were incubated with 10 μM RPN13, 10 μM USP14 in imaging buffer (50 mM Tris-HCl (pH 7.5), 5 mM MgCl₂ and 1 mM ATP) for 20 min at 30 °C, then cooled to 10 °C. 10 μM Ub_n–Sic1^PY was added to the mixture and incubated at 10 °C for 0.5, 1, 5 and 10 min. 0.005% NP-40 was added to the reaction mixture immediately before cryo-plunging. Cryo-grids made without the addition of substrate corresponded to the condition of 0 min of reaction time and were used as a baseline control for time-resolved analysis (Fig. 1e). For ATP-to-ATPγS exchange and ATPase quenching, after the reaction mixture was incubated at 10 °C for 1 min, 1 mM ATPγS was added to the reaction mixture at once, and incubated for another 1 min, then NP-40 was added to the mixture to a final concentration of 0.005% before cryo-plunging.

Cryo-EM data collection

The cryo-grids were initially screened in a 200 kV Tecnai Arctica microscope (Thermo Fisher). Good-quality grids were then transferred to a 300 kV Titan Krios G2 microscope (Thermo Fisher) equipped with the post-column BioQuantum energy filter (Gatan) connected to a K2 Summit direct electron detector (Gatan). Coma-free alignment and parallel illumination were manually optimized prior to each data collection session. Cryo-EM data were acquired automatically using SerialEM software⁴⁵ in a super-resolution counting mode with 20 eV energy slit, with the nominal defocus set in the range of −0.8 to −2.0 μm. A total exposure time of 10 s with 250 ms per frame resulted in a 40-frame movie per exposure with an accumulated dose of ~50 electrons per Ų. The calibrated physical pixel size and the super-resolution pixel size were 1.37 Å and 0.685 Å, respectively. For time-resolved sample conditions, 1,781, 2,298, 15,841, 2,073 and 2,071 movies were collected for cryo-grids made with the reaction time of 0, 0.5, 1, 5, and 10 min, respectively. For the condition of exchanging ATP to ATPγS at 1 min after substrate addition, 21,129 movies were collected.

Reference structures

Comparisons to protein structures from previous publications used the atomic models in the PDB under accession codes: 2AYN (USP domain of USP14 in its isolated form⁵), 2AYO (USP domain of USP14 bound to ubiquitin aldehyde⁵), 6MSB (state E_A1 of substrate-engaged human proteasome¹⁰), 6MSD (state E_A2), 6MSE (state E_B), 6MSG (state E_C1), 6MSJ (state E_D1), 6MSK (state E_D2), 5VFT (state S_B of substrate-free human proteasome^8,9), 5VFU (state S_C), 5VFP (state S_D1) and 5VFR (state S_D3). Cryo-EM maps from previous publications used in comparison are available from EMDB under access codes EMD-9511 (USP14–UbAl-bound proteasome²⁵), EMD-3537 (Ubp6-bound proteasome map²⁶) and EMD-2995 (Ubp6–UbVS-bound proteasome²³).

Cryo-EM data processing

Drift correction and dose weighting were performed using the MotionCor2 program⁴⁶ at a super-resolution pixel size of 0.685 Å. Drift-corrected micrographs were used for the determination of the micrograph CTF parameters with the Gctf program⁴⁷. Particles were automatically picked on micrographs that were fourfold binned to a pixel size of 2.74 Å using an improved version of the DeepEM program⁴⁸. Micrographs screening and auto-picked particles checking were both preformed in the EMAN2 software⁴⁹. A total of 213,901, 106,564, 1,494,869, 212,685, 141,257 and 1,387,530 particles were picked for the 0 min, 0.5 min, 1 min, 5 min, 10 min and ATPγS datasets, respectively. Reference-free 2D classification and 3D classification were carried out in software packages RELION⁵⁰ version 3.1 and ROME⁵¹. Focused 3D classification, CTF and aberration refinement, and high-resolution auto-refinement were mainly done with RELION 3.1, whereas the AlphaCryo4D software²⁷ was used to analyse the conformational changes and conduct the in-depth 3D classification for time-resolved analysis. Particle subtraction and re-centering were performed using RELION 3.1 and SPIDER⁵² software. We applied a hierarchical 3D classification strategy to analyse the data (Extended Data Fig. 2), which were optimized as previously described¹⁰. The entire data-processing procedure consisted of five steps. Datasets of different conditions were processed separately at steps 1 and 2 and combined at steps 3 and 4.

Step 1: doubly capped proteasome particles were separated from singly capped ones through several rounds of 2D and 3D classification. These particles were aligned to the consensus models of the doubly and singly capped proteasome to obtain their approximate shift and angular parameters. With these parameters, each doubly capped particle was split into two pseudo-singly capped particles by re-centring the box onto the RP–CP subcomplex. Then the box sizes of pseudo-singly capped particles and true singly capped particles were both shrunk to 640 × 640 pixels with a pixel size of 0.685 Å, and down-sampled by two-fold to a pixel size of 1.37 Å for the following processing. A total of 3,429,154 particles from all datasets were obtained after this step.

Step 2: particles were aligned to the CP subcomplex through auto-refinement, followed by one round of CTF refinement to correct optical aberration (up to the fourth order), magnification anisotropy, and per-particle defocus together with per-particle astigmatism. After another run of the CP-masked auto-refinement, an alignment-skipped RP-masked 3D classification was performed to separate the S_A-like states from the S_D-like states. Poor 3D classes showing broken 26S proteasome were removed for further analysis at this step. The RP subcomplex of the S_D-like states rotated by a large angle compared to the S_A-like states, and only in S_D-like states was the USP domain of USP14 observed to bind the OB ring of the proteasome. There were 1,774,110 particles in total in S_A-like states and 1,360,329 particles in total in S_D-like states in all datasets after this step.

Step 3: considering the particle number of some datasets were not enough to ensure high accuracy of independent 3D classification, in the following procedure we pooled particles together from all datasets except for the 0-min condition, in which the substrate was not yet added into the reaction system. For the S_D-like state, CP-masked auto-refinement was performed, followed with two rounds of CTF refinement and another run of CP-masked auto-refinement. Alignment-skipped RP-masked 3D classification was then performed, while conformational changes were analysed using AlphaCryo4D²⁷, which yielded three clusters, designated S_B-like, S_D-like, and E_D-like states. These names were correspondingly referred to their similar states in previously published studies^9,10. The S_A-like particles were processed in the same way, resulting in a cluster named E_A-like state; bad classes showed blurred densities in RPN10 and part of the lid. The 0-min dataset was processed independently for the lack of substrate, resulting in three classes, named S_A-like (92.8%), S_B-like (4.3%) and S_D-like (3.0%).

Step 4: particles in different clusters were individually refined with the CP masked. The CP density was then subtracted, and the particle box was recentred to the RP subcomplex and shrunk to 240 × 240 pixels, with a pixel size of 1.37 Å. For each cluster, the CP-subtracted particles were subjected to several rounds of RP-masked auto-refinement and alignment-skipped RP-masked 3D classification followed by AlphaCryo4D analysis²⁷, finally resulting in thirteen major conformational states of the USP14-bound proteasome, named ({{rm{E}}}_{{rm{A1}}}^{{rm{UBL}}}), ({{rm{E}}}_{{rm{A2}}.0}^{{rm{UBL}}}), ({{rm{E}}}_{{rm{A2}}.1}^{{rm{UBL}}}), ({{rm{S}}}_{{rm{B}}}^{{rm{USP14}}}), ({{rm{S}}}_{{rm{C}}}^{{rm{USP14}}}), ({{rm{S}}}_{{rm{D4}}}^{{rm{USP14}}}), ({{rm{S}}}_{{rm{D5}}}^{{rm{USP14}}}), ({{rm{E}}}_{{rm{D0}}}^{{rm{USP14}}}), ({{rm{E}}}_{{rm{D1}}}^{{rm{USP14}}}), ({{rm{E}}}_{{rm{D2}}.0}^{{rm{USP14}}}), ({{rm{E}}}_{{rm{D2}}.1}^{{rm{USP14}}}), ({{rm{E}}}_{{rm{D4}}}^{{rm{USP14}}}) and ({{rm{E}}}_{{rm{D5}}}^{{rm{USP14}}}). For state ({{rm{E}}}_{{rm{A1}}}^{{rm{UBL}}}), particles with blurred RPN1 were excluded for final high-resolution reconstruction. For state ({{rm{E}}}_{{rm{D2}}.1}^{{rm{USP14}}}), particles with blurred RPN2 were excluded for final high-resolution reconstruction. These states exhibit remarkable conformational changes of the AAA ring and the full RP, as well as the interactions of the RP and USP14. Time-resolved analysis of conformational changes and comparison in the presence and absence of ATPγS were both done after this step, by simply separating the particles for each state based on their time labels. Namely, the proportion of particles of each state at a given time point was obtained by summing up the number of particles for each state at the same time point and then calculating the fraction of particles of each state with respect to the total number of particles at this time point^53,54. Similarly, final analysis of state percentage for the ATP-to-ATPγS exchange condition was done by counting the particles of each state under this condition, with the particles of each state used for separate refinement, reconstruction and comparison with those under ATP-only conditions (Extended Data Figs. 2c, 5a, 6i).

Final refinement of each state was performed using pseudo-singly capped particles with the pixel size of 1.37 Å. Two types of local mask were applied for the auto-refinement, one focusing on the complete RP and the other focusing on the CP, resulting in two maps for each state, which were merged in Fourier space into one single map. Based on the in-plane shift and Euler angle of each particle from the auto-refinement, we reconstructed the two half-maps of each state using pseudo-singly capped particles with the pixel size of 0.685 Å. The Fourier shell correlation (FSC) curves of thirteen states were calculated from two separately refined half maps in a gold-standard procedure, yielding the nominal resolution ranging from 3.0 to 3.6 Å, and the local RP resolution ranging from 3.3 to 4.6 Å (Extended Data Figs. 2a, 3b–d). Before visualization, all density maps were sharpened by applying a negative B-factor ranging from −10 to −50 Ų.

In order to further improve the local density quality of USP14 and RPN1, which suffered from local conformational dynamics, another round of RP-masked 3D classification was done using CP-subtracted particles for some states to exclude 3D classes with blurred USP14 and RPN1. These locally improved maps were only used for visualization and adjustment of atomic models of USP14 and RPN1. For states ({{rm{E}}}_{{rm{A1}}}^{{rm{UBL}}}), ({{rm{E}}}_{{rm{A2}}.0}^{{rm{UBL}}}) and ({{rm{E}}}_{{rm{A2}}.1}^{{rm{UBL}}}), 3D classes with unblurred RPN1 and especially visible UBL on the RPN1 T2 site (including 152,802, 66,966 and 61,930 particles, respectively) were selected and refined by applying a mask on the RPN1-UBL component. The resulting RPN1-UBL density in these states were compared with previously reported E_A1 state (Extended Data Fig. 7e). For states ({{rm{E}}}_{{rm{D0}}}^{{rm{USP14}}}) and ({{rm{E}}}_{{rm{D2}}.0}^{{rm{USP14}}}), 3D classes with unblurred RPN1 density (including 61,447 and 53,145 particles, respectively) were selected and refined to 4.1 and 4.2 Å for the RP, respectively. For state ({{rm{S}}}_{{rm{C}}}^{{rm{USP14}}}), ({{rm{E}}}_{{rm{D4}}}^{{rm{USP14}}}) and ({{rm{E}}}_{{rm{D2}}.1}^{{rm{USP14}}}), 3D classes with improved USP14 densities (including 34,659, 54,642, 142,814 particles, respectively) were selected and refined to 4.5, 4.2 and 3.8 Å for the RP component, respectively, for better visualization of the full-length USP14 in the proteasome.

Atomic model building and refinement

Atomic model building was based on the previously published cryo-EM structures of the human proteasome¹⁰. For the CP subcomplex, initial models of the closed-gate CP and open-gate CP were respectively derived from the E_A1 model (PDB 6MSB) and the E_D2 model (PDB 6MSK). For the RP subcomplex, the previous E_D2 model was used as an initial model. All subunits of the initial model were individually fitted as a rigid body into each of the reconstructed maps with UCSF Chimera⁵⁵, followed by further adjustment of the main chain traces using Coot⁵⁶. Initial model of the full-length USP14 was first derived from a predicted one by AlphaFold⁵⁷, which was verified by comparing to a crystal structure⁵ (PDB 2AYO). The USP14 model was then merged with the initial proteasome model by independently fitting models of the USP14 UBL and USP domains as rigid bodies into the cryo-EM maps, and manually fitting the linker between the UBL and USP domains in Coot⁵⁶. Despite the presence of RPN13 in our purified proteasome (Extended Data Fig. 1g) and the addition of excessive RPN13 to saturate the proteasome, no reliable density was observed for RPN13 in all cryo-EM maps, thus precluding the atomic modelling of RPN13 and likely reflecting its highly dynamic association with the proteasome. The atomic models of subunit SEM1 and the N terminus of subunit RPN5 fitted into their corresponding local densities of lower resolution were rebuilt by considering the prediction of AlphaFold⁵⁷. The atomic model of USP14 was first rebuilt and refined against the map of state ({{rm{E}}}_{{rm{D2}}.1}^{{rm{USP14}}}) with improved local USP14 density, the resulting model of which was then used as starting USP14 models to fit into the USP14 densities in other states. For some structures with partially blurred or missing subunit densities, the atomic models were revised by removing these regions, for example, the UBL domain of USP14 was removed in the models of ({{rm{E}}}_{{rm{D5}}}^{{rm{USP14}}}) and ({{rm{S}}}_{{rm{D5}}}^{{rm{USP14}}}). Given that the substrates were not stalled in a homogeneous location during their degradation and that substrate translocation through the proteasome is not sequence-specific, the substrate densities were modelled using polypeptide chains without assignment of amino acid sequence. For states ({{rm{E}}}_{{rm{A1}}}^{{rm{UBL}}}), ({{rm{E}}}_{{rm{A2}}.0}^{{rm{UBL}}}), ({{rm{E}}}_{{rm{A2}}.1}^{{rm{UBL}}}), ({{rm{E}}}_{{rm{D0}}}^{{rm{USP14}}}), ({{rm{E}}}_{{rm{D1}}}^{{rm{USP14}}}), ({{rm{E}}}_{{rm{D2}}.0}^{{rm{USP14}}}), ({{rm{E}}}_{{rm{D2}}.1}^{{rm{USP14}}}) and ({{rm{E}}}_{{rm{D4}}}^{{rm{USP14}}}), the nucleotide densities are of sufficient quality for differentiating ADP from ATP, which enabled us to build the atomic models of ADP and ATP into their densities (Extended Data Fig. 6j). For other states with the local RP resolution worse than 4.0 Å, the nucleotide types or states were hypothetically inferred from their adjacent states at higher resolution with the closest structural similarity, based on the local densities, the openness of corresponding nucleotide-binding pockets as well as their homologous structural models of higher resolution if available.

After manually rebuilding, atomic models were all subjected to the real-space refinement in Phenix⁵⁸. Both stimulated annealing and global minimization were applied with non-crystallographic symmetry (NCS), rotamer and Ramachandran constraints. Partial rebuilding, model correction and density-fitting improvement in Coot⁵⁶ were then iterated after each round of atomic model refinement in Phenix⁵⁸. The refinement and rebuilding cycle were often repeated for three rounds or until the model quality reached expectation (Extended Data Table 1).

Structural analysis and visualization

All structures were analysed in Coot⁵⁶, PyMOL⁵⁹, UCSF Chimera⁵⁵, and ChimeraX⁶⁰. Inter-subunit interactions and interfacial areas were computed and analysed using the PISA server⁶¹ (https://www.ebi.ac.uk/pdbe/prot_int/pistart.html). Local resolution variations were estimated using ResMap⁶². Figures of structures were plotted in PyMOL⁵⁹, ChimeraX⁶⁰, or Coot⁵⁶. Structural alignment and comparison were performed in PyMOL⁵⁹ and ChimeraX⁶⁰.

Data reporting

No statistical methods were used to predetermine sample size. The experiments were not randomized, and investigators were not blinded to allocation during experiments and outcome assessment.

Statistical analysis

Statistical analyses of mutagenesis data were performed using two-tailed unpaired t-tests with SPSS v.27.0 unless otherwise indicated. Statistical significance was assessed with a 95% confidence interval and a P value of < 0.05 was considered significant.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this paper.