Introduction

The proteasome is the primary protease that is responsible for degrading the majority of proteins in eukaryotic cells and plays a key role in metabolism and protein quality control mechanisms1,2,3,4,5,6,7,8. Proteasome function is crucial for maintaining nutrient balance and removing cytotoxic polypeptides that can cause cellular stress and dysfunction. Consequently, decreases in proteasome levels and activity are associated with aging and neurodegenerative diseases such as Alzheimer’s disease and Parkinson’s disease9,10,11,12. To maintain homeostasis, cells must replenish existing proteasomes as well as produce new proteasomes in response to elevated levels of damaged and misfolded proteins5,6,13,14.

The 20S proteasome is the core complex that degrades proteins into short peptides. It consists of four heptameric rings of α and β subunits arranged in a α(1-7):β(1-7):β(1-7):α(1-7) configuration. The α subunits form gates at both ends of the barrel-shaped complex that control substrate entry, while the catalytic β subunits cleave substrates inside the complex. There are three different catalytic β subunits in eukaryotic 20S proteasomes: β1 has caspase-like activity, β2 has trypsin-like activity, and β5 has chymotrypsin-like activity15. The β subunits are expressed with pro-peptides that protect the critical N-terminal threonine of the catalytic β subunits16. These pro-peptides also aid in 20S assembly and are cleaved in the mature 20S complex16,17,18,19,20.

The production of new proteasomes is an intricate process that requires the precise assembly of many different components and involves the action of multiple chaperones. Early in the biogenesis pathway, proteasome assembly chaperones (PAC) 1-4 coordinate the arrangement of 7 different α subunits (α1-7) into a hetero-heptameric α-ring21,22,23. Biochemical studies indicate that the PAC1/PAC2 and PAC3/PAC4 heterodimers bind initially to α4-7, which provides a subcomplex that seeds the addition of α1-3. Structural studies have provided important insights into how PAC1/PAC2 engages with the α subunits24,25,26, which both contain C-terminal HbYX motifs that help to anchor PAC1/PAC2 to the α-ring27,28. However, there is a lack of structural data on how PAC3/PAC4 engages with the α subunits to facilitate α-ring formation. Furthermore, it has been postulated that the N-terminal tail of PAC1 may be involved in α-ring formation24, but how this may play a role remains unclear.

Once the α-ring is formed, the proteasome maturation protein (POMP) chaperone assembles the 7 different β subunits onto the α-ring29,30,31,32 with concomitant unbinding of the PAC3/PAC4 heterodimer. While PAC3/PAC4 helps to assemble the α-ring, they block the completion of the β-ring. Binding of POMP and dissociation of PAC3/PAC4 marks the transition from α-ring formation to β-ring formation, but how this transition occurs remains unclear. The different β subunits are added in a stepwise fashion, with the addition of β7 incorporating last to complete the β-ring. The α-ring/β-ring heterocomplex (half-proteasome) dimerizes to form the pre-20S complex, which activates the autocatalytic cleavage of the β pro-peptides17,18. How dimerization of the half-proteasomes induces cleavage of the β pro-peptides remains unclear. Finally, POMP and the PAC1/PAC2 heterodimer dissociate to yield the mature 20S proteasome16,21,29. The sequential order of these final steps and their structural basis remains largely unexplored.

Here, we provide a broad structural analysis of 20S biogenesis, which clarifies the role of each chaperone across the assembly process and reveals molecular insights into PAC3/PAC4 function and late-stage maturation of the pre-20S complex. The complexity of proteasome assembly has made it challenging to replicate the process in vitro. Knock-down and mutation of some subunits in yeast and mammalian cells have been useful strategies in trapping certain proteasomal intermediates for biochemical and structural analyses16,24,29,33. However, decreases in cellular fitness and changes in complex stoichiometry associated with traditional approaches have complicated the in-depth structural characterization of the proteasome biogenesis pathway21,22,34,35,36,37. We have previously developed efficient strategies of tagging and purifying endogenous proteins for cryo-EM analysis38,39, which we have employed in this study to dissect the stepwise assembly pathway of the human 20S proteasome.

Results

Overall assembly of the 20S proteasome

Assembly of the proteasomal 20S core complex is orchestrated by five dedicated chaperones. To study their function, we individually tagged three of the five 20S chaperones in Expi293F cells using CRISPR/Cas gene editing38,40 and separately purified the endogenous chaperone-associated complexes. Analysis of the samples by single particle cryogenic electron microscopy (cryo-EM) enabled structure determination of intermediate complexes at each major stage of 20S assembly (Fig. 1a, Supplementary Figs. 1–3, Supplementary Table 1). Overall, each α and β subunit interacts with at least one of the 20S chaperones except for β7 (Fig. 1b). Evidence of protein dynamics can be observed throughout the assembly process. In particular, the α subunits undergo substantial movements that appear to correlate with binding of the β subunits (Fig. 1c). On the other hand, the chaperones mainly undergo local structural rearrangements with limited changes in overall architecture (Fig. 1d). In the following, we describe mechanistic insights into α-ring formation, the transition to β-ring assembly, and dimerization and maturation into 20S complexes.

Fig. 1: Stepwise assembly of the 20S proteasome is mediated by multiple chaperone interactions.
figure 1

a Cryo-EM structures of human 20S intermediate complexes depicting the stepwise assembly and maturation of the proteasome. Assembly begins with formation of the α-ring, aided by PAC1-4. The β-ring is formed on the α-ring, aided by POMP. Dimerization of α-ring/β-ring half-proteasomes triggers cleavage of the β pro-peptides (pro-β) and dissociation of POMP and PAC1/PAC2 to yield the mature 20S proteasome. Scale bar: 20 Å. b Interaction map between proteasome chaperones and 20S subunits. Each chaperone makes extensive contacts with multiple 20S subunits, with PAC1-4 interacting primarily with the α4-7 subunits while POMP interactions are spread out across many α and β subunits. c Root mean square deviation (r.m.s.d.) values of the α subunit atomic positions across the PAC1-4/α-ring (I), pre-13S (II), mixed-13S (III), pre-20S (IV), and mature 20S (V, PDB 7NAN) structures. d R.m.s.d. values of proteasome chaperone atomic positions across the PAC1-4/α-ring (I), pre-13S (II), mixed-13S (III), and pre-20S (IV) structures.

Full size image

α-ring formation mediated by PAC1-4

To gain molecular insights into the function of PAC3/PAC4 in α-ring formation, we tagged PAC3 (PSMG3) with the Twinstrep tag in Expi293F cells and purified the PAC3-associated complexes (Supplementary Fig. 1e, f). Cryo-EM analysis of the sample resulted in a structure of the α-ring complex bound by PAC1-4 at ~3.0 Å resolution (Fig. 2a, Supplementary Figs. 2a, 3, and 10). While the PAC1/PAC2 and PAC3/PAC4 heterodimers bind on different sides of the α-ring, these two heterodimers interact with overlapping sets of α subunits. α5 and α6 make the most extensive contacts with PAC1-4, with the N termini of α5 and α6 inserting into a hydrophobic crevice formed between PAC1 and PAC2 (Supplementary Fig. 4a), which was also observed in yeast24,25. Meanwhile, PAC3 and PAC4 are wedged into one side of the α-ring next to α5/α6 and clamp onto one of the α-helices in α5, explaining the tight binding of PAC3 and PAC4 to α5 (Fig. 2a, Supplementary Fig. 4b)35,41,42. PAC3 also interacts with α6 (Supplementary Fig. 4c). In addition, the C terminus of PAC1 inserts into the HbYX-binding pocket formed between α5 and α6 (Supplementary Fig. 4d)27,28. The extensive structural contacts between PAC1-4 and α5-6 suggest that these components may form an initial subcomplex that catalyzes α-ring formation.

Fig. 2: PAC1-4 helps to stabilize the α4-7 subcomplex to initiate proteasome assembly.
figure 2

a Cryo-EM structure of the PAC1-4/α-ring complex showing the distinct binding sites of PAC1/PAC2 and PAC3/PAC4. PAC1/PAC2 bind to α5-7 on the outward facing side of the α-ring while PAC3/PAC4 bind to α4-6 on the concave inward facing side of the α-ring. Scale bar: 15 Å. b PAC4 K113 and R82 form salt bridges with α4 E99 and E90, respectively. Scale bar: 5 Å. c PAC3 D56 forms a salt bridge with α7 R115 and coordinates with α7 Y119. Scale bar: 5 Å. d PAC4 W22 fits into a pocket formed by α3 R96 and Q100.Scale bar: 5 Å. e Cross-section of PAC1-4/α-ring map shows that the N-terminal tail of PAC1 is not inserted into the α-ring pore. Scale bar: 15 Å. f Cross-section of pre-13S assembly intermediate shows that the N-terminal tail of PAC1 is inserted into the α-ring pore in later stages of 20S assembly. Scale bar: 15 Å. g Positions of the α subunit N-terminal tails in a closed gate (mature 20S, PDB 7NAN), partially closed gate (PAC1-4/α-ring), partially open gate (pre-13S), and fully open gate (PA28-20S, PDB 7NAO). The N-terminal tail of PAC1 is observed in the pre-13S structure but not in the PAC1-4/α-ring structure.

Full size image

To gain further insights into α-ring formation, we carried out computational 3D classification analysis of the PAC1-4/α-ring cryo-EM data. The analysis revealed two 3D subclasses with good density for PAC1/2 (Supplementary Fig. 2a, I and IV) and two subclasses with low-quality density for PAC1/2 (Supplementary Fig. 2a, II and III), indicating that PAC1/2 may be dissociating in these subclasses. All subclasses appear to contain PAC1/2, PAC3/4, and α1-7. Interestingly, we observed a 3D subclass with strong map densities for α4-7 and lower map densities for α1-3 (Supplementary Fig. 2a, I). In contrast, similar analysis of β-ring assembly intermediates revealed consistent α subunit map densities (Supplementary Fig. 5a). This result indicates that PAC1-4/α4-7 likely forms a stable subcomplex, consistent with previous biochemical studies43. PAC1-4/α4-7 is stabilized by multiple contacts between α4 and PAC4, including salt bridges between α4 E90 and PAC4 R82, and between α4 E99 and PAC4 K113 (Fig. 2b, Supplementary Fig. 10). Meanwhile, α7 makes contacts with both PAC2 and PAC3, with the C-terminal tail of PAC2 inserting into the HbYX-binding pocket formed between α6 and α7 (Supplementary Fig. 4d). In addition, a loop in PAC3 extends out toward α7 to form a salt bridge between α7 R115 and PAC3 D56 (Fig. 2c, Supplementary Fig. 10). While PAC1-4 are crucial for stabilizing α4-7, they do not appear to play a substantial role in the incorporation of α1-3 with the only contact being PAC4 W22 stacking in between α3 R96 and Q100 (Fig. 2d, Supplementary Fig. 10). α4-7 have the highest map densities in the 3D subclass, followed by α1, α3, and α2 (Supplementary Fig. 2a, 4e). This suggests that α1-3 are incorporated after formation of a stable PAC1-4/α4-7 complex to complete α-ring formation.

Substrate entry into the proteasome is regulated by the N-terminal tails of the α subunits, which form a closed gate in the mature 20S complex. In structures of β-ring assembly intermediates, the N-terminal tail of PAC1 is inserted into the α-ring pore and blocks the open gate (Fig. 2f, g, Supplementary Fig. 5c). It has been postulated that the N-terminal tail of PAC1 may help to coordinate α-ring assembly in yeast24,44. Surprisingly, the N-terminal tail of PAC1 is absent from the α-ring pore in the human PAC1-4/α-ring structure reported here (Fig. 2e, g, Supplementary Fig. 5c). The N-terminal tails of some α-subunits in the PAC1-4/α-ring assembly adopt unique conformations that result in a partially closed gate (Fig. 2g). Specifically, the N-terminal tails of α2-4 are oriented inwards in the PAC1-4/α-ring complex, similar to the mature 20S closed gate conformation. The unique conformations of the α2 and α3 N-terminal tails in the PAC1-4/α-ring complex would sterically clash with the N-terminal tail of PAC1 if it was inserted in the pore (Supplementary Fig. 5c). Consequently, the N-terminal tail of PAC1 is inserted later in the assembly pathway.

Transition from α- to β-ring assembly

The assembly of the β-ring is initiated by the binding of the POMP chaperone and β2 to the completed α-ring. Subsequently, PAC3/PAC4 must dissociate from the α-ring to make room for the incoming β subunits. To gain insight into how this occurs, we tagged POMP with an mNG2(11)-StrepII tag in Expi293F cells and purified the POMP-associated complexes (Supplementary Fig. 1e, f). Cryo-EM analysis of the sample revealed a mixture of β-ring assembly intermediates (Supplementary Fig. 2b), which were separated into four major classes using computational 3D classification of the cryo-EM images. One of the 3D subclasses lacked β subunits and PAC3/4, while another 3D subclass resembled a low-resolution structure of the 13S complex. Two 3D subclasses exhibited high-quality structural features and were refined to high-resolution. These two structures correspond to a pre-13S complex with partial binding of β4 at ~3.0 Å resolution and a mixed-13S complex with partial binding of β1, β5, and β6 at ~3.0 Å resolution (Supplementary Fig. 3)45. Interestingly, comparison of the PAC1-4/α-ring and pre-13S structures shows that POMP would fit into the concave surface on the α-ring next to PAC3/PAC4 with minimal overlap between the structures (Fig. 3a). In addition, β2 binds to the α-ring on the opposite side of where PAC3/PAC4 is bound. Biochemical studies have suggested that POMP could bind independently to the α-ring intermediate46, which is consistent with our structural data showing that binding of β2 would partially block access to the POMP binding pocket. Binding of POMP helps to recruit β2, but β2 also helps to stabilize binding of POMP16. Steric clash would occur in one small region of PAC4 with the C-terminal tail of POMP and the β2 pro-peptide (Fig. 3b, c). The structural arrangement of the POMP C terminus and β2 pro-peptide appears to be stabilized by interactions with β3, which include a salt bridge between β3 R66 with POMP E125. The release of PAC3 is coupled to the binding of β316, and our structural data provides a possible mechanism of how this may occur. POMP and β2 bind to the PAC1-4/α-ring complex and recruit β3, which induces structural changes in the POMP C terminus and β2 pro-peptide (Fig. 3d). These changes displace PAC4, which would cause the concurrent dissociation of PAC3. Comparison of our PAC1-4/α-ring structure with the recently published PAC1-4/α-ring/POMP/β2 structure47 shows that binding of POMP disrupts the salt bridge between PAC3 D56 and α7 R115 (Fig. 3e). Breaking of this salt bridge likely destabilizes the binding of PAC3 to the α-ring and promotes the dissociation of PAC3.

Fig. 3: POMP and β2 displaces PAC3/PAC4 to enable β-ring assembly.
figure 3

a Cryo-EM map of PAC1-4/α-ring overlayed with the pre-13S atomic model. POMP fits into the space between PAC3/PAC4 and α1-2. β2 and β3 bind to the α-ring on the opposite side of PAC3/PAC4. Scale bar: 15 Å. b Isolated views of the overlayed structures show good fit between POMP and PAC3/PAC4 with little overlap between the chaperones. Scale bar: 15 Å. c The PAC4 structure clashes with the C-terminal region of POMP and the N-terminal tail of the β2 pro-peptide (arrowheads), which are stabilized by interactions with β3. Scale bar: 5 Å. d Schematic depicting the initiation of β-ring assembly and dissociation of PAC3/PAC4. e Comparison of our PAC1-4/α-ring complex with the structure of PAC1-4/α-ring/POMP/β2 (PDB 8QYJ) shows that binding of POMP disrupts the interaction between PAC3 and α7. Scale bar: 3 Å.

Full size image

The pre-13S and mixed-13S cryo-EM maps represent averages of multiple β-ring assembly intermediates (Supplementary Fig. 2b). Therefore, we may infer the order of β subunit incorporation from their cryo-EM map density levels. The pre-13S cryo-EM map shows strong density for β2/β3 and lower density for β4 (Supplementary Figs. 2b, 5b), indicating that β4 is added after β3 as expected16. In the mixed-13S map, β2-4 have comparable densities that are higher than for the other β subunits (Supplementary Figs. 2b, 6). β3 has slightly higher average map density compared to β2 and β4, which reflects the flexibility of the terminal β subunits of the incomplete β-ring (Supplementary Figs. 2b, 3e, 6). After β4, β5 is likely incorporated next as has been observed in yeast24, followed by β1 and β6 (Supplementary Fig. 2b). There does not appear to be a clear sequential order for β1 or β6, but this result may be affected by the low density and potential flexibility of the subunits. Interestingly, no density above background is observed for β7 in either the pre-13S or mixed-13S cryo-EM maps. This result is consistent with previous studies showing β7 is the last subunit added to yield the half-proteasome16,48,49 and suggests two half-proteasomes will quickly dimerize to form the pre-20S complex.

Dimerization and autocatalytic activation

The catalytic β subunits (β1, β2, β5) are expressed with N-terminal pro-peptides (Supplementary Fig. 1b, c) that are cleaved after formation of the pre-20S complex. To gain insight into how dimerization activates the catalytic β subunits, we tagged PAC2 (PSMG2) with an mNG2(11)-StrepII tag in Expi293F cells and purified the PAC2-associated complexes (Supplementary Fig. 1e, f). Cryo-EM analysis of the sample and 2D classification revealed 20S complexes mixed with α-ring and β-ring assembly intermediates (Supplementary Fig. 2c). Because samples were prepared on holey carbon grids without graphene oxide, the α-ring and β-ring assembly intermediates adopted strongly biased orientations on the grid that prevented further processing. Classes corresponding to the pre-20S proteasome were processed and resulted in a structure of the pre-20S complex at ~2.8 Å resolution (Fig. 4, Supplementary Figs. 2c,3). The cryo-EM map shows clear density for PAC1/PAC2 and POMP, as well as an unknown strand of density located at the interface between the two β4 subunits (Supplementary Fig. 7). The close proximity of this density to threonine T1 of β5 suggests it may be part of the β5 pro-peptide (Supplementary Fig. 1c), which is crucial for half-proteasome dimerization and can work in trans16,19,20. The recently published structures of half-proteasome and pre-20S intermediates47,50 lack this unassigned strand of density and show the β5 pro-peptide in a different location, which does not support β5 as the unassigned density and the identity of the strand remains unclear. However, we cannot rule out the possibility that the density may also represent the N terminus of POMP30 or an endogenous peptide from the cell. Structures of the 20S proteasome bound by the endogenous inhibitor PI31 have been reported recently51, but the structure of PI31 is distinct from the unassigned strand of density observed in our map, indicating they are likely different molecules. Comparison of the β pro-peptide densities reveals that the different pro-peptides are processed at different rates. The β1 pro-peptide density appears to be lower compared to the β2 pro-peptide (Fig. 4e, Supplementary Fig. 8), although the difference is not statistically significant. Interestingly, little density for the β5 pro-peptide can be observed near β5 T1, indicating that the β5 pro-peptide is mostly cleaved and has dissociated in the pre-20S cryo-EM map reported here (Supplementary Fig. 8). These data indicate that the β5 pro-peptide is likely processed first after dimerization (Fig. 4f), followed by processing of the β1 and β2 pro-peptides.

Fig. 4: Pre-20S formation triggers autocatalysis of the β pro-peptides.
figure 4

a The cryo-EM map of mixed-13S shows poorly resolved density for the loops near the dimerization interface, indicating the region is likely flexible during β-ring assembly. Scale bar: 5 Å. b The sidechain of β2 K33 is not ordered in the mixed-13S structure, preventing cleavage of the β pro-peptide. Scale bar: 3 Å. c Dimerization of two half-proteasomes reorders the loops near the dimer interface into a stable structural arrangement. Scale bar: 5 Å. d Structural rearrangements induced by half-proteasome dimerization positions the sidechain of K33 into a catalytically active conformation. A break in the cryo-EM density (gray) can be observed between β2 T1 and T-2 at higher σ thresholds, indicating cleavage of the pro-peptide. Scale bar: 3 Å. e Quantification of the average cryo-EM map densities for the first six residues of the β pro-peptides after T1. Box plots indicate median (middle line), 25th, 75th percentile (box) and 5th and 95th percentile (whiskers) as well as outliers (single points). n = 6 residues. *p-value < 0.01, evaluated using two-tailed t-test with unequal variance. f Schematic depicting how structural rearrangement of K33 activates autocatalysis of the β pro-peptide. Scale bar for cryo-EM maps: 10 Å. Source data are provided as a Source data file.

Full size image

Each of the catalytic β-subunits contain a catalytic threonine T1, a conserved lysine K33, and a conserved aspartic acid D1715,18,52,53. Cleavage occurs between T1 and the conserved G-1 of the pro-peptide (Supplementary Fig. 1c), which liberates the catalytic T1 residue of the β subunit. Focusing on β2, K33 and the adjacent loop region are disordered prior to half-proteasome dimerization (Fig. 4a, b), disrupting the active site54. Consistent with a catalytically inactive enzyme, strong density for the β2 pro-peptide is observed in the pre-13S and mixed-13S cryo-EM maps (Fig. 4b, Supplementary Fig. 8). Upon dimerization of two half-proteasomes, K33 and the adjacent loops of β2 are stabilized by interactions at the dimer interface (Fig. 4c, d). A recently published study identified one critical interaction as a salt bridge between a conserved arginine in the catalytic β subunit and a C-terminal aspartate in the adjacent midline spanning non-catalytic β subunit50. This structural rearrangement places K33 into the correct position to coordinate T1 and D17, thus activating autocatalytic cleavage of the β2 pro-peptide. Consistent with an active enzyme, a break in the pre-20S cryo-EM map density between β2 T1 and G-1 of the pro-peptide can be observed at higher σ thresholds (Fig. 4d, Supplementary Fig. 8). Map density for the β2 pro-peptide can still be observed at higher σ thresholds, indicating that the pro-peptide can remain bound after cleavage. While some density can be observed for β1 and β5 in the mixed-13S cryo-EM map, the quality of the β1 and β5 density is not sufficient to carry out a similar analysis as was done for β2. However, the conserved nature of the autocatalytic mechanism52 of these β subunits leads us to speculate that stabilization of K33 is a key feature of autocatalysis activation for β1 and β5 as well.

Maturation of pre-20S to 20S

The final steps of 20S assembly involve the removal of the β pro-peptides and dissociation of the bound chaperones, but how this occurs remains unclear. Examination of the pre-13S and mixed-13S cryo-EM maps show that the chaperones and β2 pro-peptide are bound stoichiometrically relative to β2 (Fig. 5, Supplementary Fig. 9). In the pre-20S cryo-EM map, both the chaperones and β pro-peptides have lower densities compared to the α and β subunits, indicating that the pre-20S cryo-EM map likely represents a mixture of maturing 20S complexes. The β2 pro-peptide and the N-terminal region of POMP have substantially lower density in the pre-20S map determined in this study. Interestingly, a recently published structure of the pre-20S complex shows stronger β2 pro-peptide density and an extended POMP N-terminal region that interacts with the β2 pro-peptide47 (Supplementary Fig. 11). Our structure of the pre-20S complex likely represents a later assembly stage compared to the recently published model, which together show that β2 pro-peptide processing is linked to disengagement of the POMP N-terminal region (Fig. 5c, Supplementary Fig. 9). The interaction between the β2 pro-peptide and the POMP N-terminal region likely mediates the concerted maturation of these components. The C-terminal tail of PAC1 is also significantly lower in density relative to the rest of the protein in the pre-20S map. The relative differences in density observed within each component indicate that the final maturation of the pre-20S complex occurs through multiple stages. Because the proteins were prepared at 4 °C, we reasoned that the complexes would mature further at physiological temperatures. Indeed, incubation of purified pre-20S complexes at 37 °C for 1 h prior to cryo-EM grid preparation and analysis revealed a more mature complex, which we designate pre-20S’ (Fig. 5a, Supplementary Figs. 2d, 3, 9). 2D and 3D classes corresponding to α-ring and β-ring assembly intermediates were observed as well, but orientation bias hindered further analysis of these complexes (Supplementary Fig. 2d). The cryo-EM map of pre-20S’ shows lower density for the β2 pro-peptide, PAC1, PAC2, and POMP. Taken together, these results support a 3-stage model of 20S maturation (Fig. 5d). First, the catalytic β subunits are activated through the cleavage and displacement of the β pro-peptides. Second, dissociation and degradation of the β2 pro-peptide causes displacement of the N-terminal region of POMP, which occurs concurrently with the destabilization of the PAC1 C-terminal tail. Finally, the C-terminal region of POMP dissociates and is degraded and PAC1/PAC2 is released to yield the mature 20S proteasome.

Fig. 5: 20S maturation proceeds in three stages.
figure 5

a Cryo-EM maps of pre-20S and pre-20S’ with cutaway section to show POMP density. Scale bar: 10 Å. b Isolated view of the POMP interaction with β2 from the pre-13S atomic model. The β2 pro-peptide (pro-β2) interacts with the N-terminal region of POMP (POMP-NT). c Quantification of the cryo-EM density map values corresponding to β2, POMP, PAC1, and PAC2 shows a decrease in chaperone occupancy over the course of 20S maturation. Examination of specific regions within the proteins reveals a correlation between density changes in pro-β2, POMP-NT, and the C-terminal tail of PAC1 (PAC1-CT). Meanwhile, density changes in the C-terminal portion of POMP are correlated with changes in the N-terminal portion of PAC1 and PAC2. Box plots indicate median (middle line), 25th, 75th percentile (box) and 5th and 95th percentile (whiskers) as well as outliers (single points). pro-β2, n = 36 residues; β2, n = 163 residues; POMP-NT, n = 21 residues; POMP, n = 76 residues; PAC1, n = 264 residues; PAC1-CT, n = 11 residues; PAC2, n = 235 residues; PAC2-CT, n = 10 residues. *p-value < 0.001, evaluated using two-tailed t-test with unequal variance. d Cartoon depicting the 3-stage progression of 20S maturation involving β pro-peptide cleavage, pro-β2 and POMP-NT dissociation, and POMP/PAC1/PAC2 release and degradation. Source data are provided as a Source data file.

Full size image

Discussion

In this study, we present the structures of the human PAC1-4/α-ring and late-stage pre-20S complexes, which together with the β-ring assembly intermediates and previous structures of the yeast homologs24,26 allow us to derive a more complete understanding of the 20S proteasome assembly pathway. Four of the five chaperones (PAC1-4) help to stabilize the initial α4-7 subcomplex through multiple contacts. On the other hand, α1-3 do not interact meaningfully with PAC1-4 until they bind POMP, which coincides with the insertion of the PAC1 N-terminal tail into the α-ring pore. This complex interaction between PAC1, POMP, and the α subunits may serve as a checkpoint to ensure the proper formation of the α-ring. Binding of POMP also marks the start of β-ring assembly through the recruitment of β2. Interestingly, the clash between POMP/β2 and PAC3/PAC4 in our structures of the human proteins is substantially less severe than what has been predicted for the yeast homologs24, suggesting a potential species-level difference in proteasome assembly at this stage. β-ring assembly appears to proceed in a similar fashion between human and yeast with β7 as the last subunit to be incorporated, triggering dimerization of the half-proteasomes. Autocatalytic cleavage of the β pro-peptides occurs prior to the release of POMP and PAC1/PAC2, which is also observed in the yeast preholoproteasome26.

One outstanding question concerns the role of the PAC1 N-terminal tail in 20S assembly, which inserts into the α-ring pore24. It has been suggested that the PAC1 N-terminal tail may be involved in α-ring formation. However, the N-terminal tail of PAC1 is not inserted into the α-ring pore in the PAC1-4/α-ring structure from this study. Therefore, it is unlikely that the PAC1 N terminus plays a role in α-ring formation and its function remains unclear. The N terminus of PAC1 is inserted in the α-ring pore of the pre-13S, mixed-13S, and pre-20S structures, indicating a structural rearrangement after completion of the α-ring and initiation of β-ring formation. It is possible that insertion of the PAC1 N-terminal tail helps to stabilize the binding of PAC1/PAC2 to the correctly formed α-ring. The N-terminal tail of PAC1 also helps to block the open α-ring gate to prevent substrate entry into the immature proteasome24. In addition, the N terminus of PAC1 interacts with POMP, which may be forming a signal relay network that aids in the release of the chaperones from the maturing 20S complex.

The β5 pro-peptide is processed more quickly relative to the pro-peptides of β1 and β2, but the mechanistic basis for this is unclear. β5 contains a conserved histidine at the −2 position while threonine occupies the same position in β1 and β2 (Supplementary Fig. 1c). The conserved histidine may be serving as a second proton acceptor in β5, which is a role that is fulfilled exclusively by K33 in β1 and β2. The two proton acceptors in β5 (K33 and H-2) may help to boost the autocatalysis rate of the pro-peptide. Indeed, mutation of H-2 to phenylalanine reduces but does not abrogate cleavage of the β5 pro-peptide19,52. In contrast, mutation of the conserved histidine to lysine, which also functions as a proton acceptor, preserves autocatalytic activity19,52. The rapid processing of the β5 pro-peptide may help to catalyze the degradation of other β subunit pro-peptides and POMP, which would help promote the maturation of the pre-20S complex into functional 20S proteasomes. While mutations that prevent pro-peptide cleavage of β1 and β2 can be tolerated, cleavage of the β5 pro-peptide is crucial15,20,55,56,57,58. The critical importance of β5 may be the reason for the enhanced efficiency and built-in redundancy in the cleavage of the β5 pro-peptide. Even so, studies in yeast have demonstrated that 20S complexes can still assemble and retain partial function with unprocessed β5 subunits52, highlighting the robustness of 20S proteasome assembly.

Methods

Primers for gRNA synthesis

ML557

TAATACGACTCACTATAG

ML558

AAAAAAAGCACCGACTCGGTGC

ML611

AAAAAAAGCACCGACTCGGTGCCACTTTTTCAAGTTGATAACGGACTAGCCTTATTTAAACTTGCTATGCTGTTTCCAGCATAGCTCTTAAAC

PSMG2 oligo

TAATACGACTCACTATAGGCGACCATGTTCGTTCCCTGGTTTAAGAGCTATGCTGGAA

PSMG3 oligo

TAATACGACTCACTATAGCGCCATGGAAGACACGCCGTGTTTAAGAGCTATGCTGGAA

POMP oligo

TAATACGACTCACTATAGGAGCTGCGGAAGATGGTGAGGTTTAAGAGCTATGCTGGAA

Single-Strand DNA for Homology-Directed Repair (HDR)

PSMG2(PAC2)-nt-mNG2(11)-StrepII

TCTTGCCAGGGCCGCGGTTAGTCCCTGCTGGCCACCCCACTGCGACCATGACCGAGCTCAACTTCAAGGAGTGGCAAAAGGCCTTTACCGATATGATGTGGTCCCACCCTCAATTCGAGAAGGGGAGTGCGTTCGTTCCCTGCGGGGAGTCGGCCCCCGACCTTGCCGGCTTCACCCTCCT

PSMG3(PAC3)-nt-twinStrep

TTTTTTCCCCTTTCTTTTTAACCTAAATTAAAGCTGCCACTGCAGAGCCCCGCCATGTGGTCCCACCCTCAATTCGAGAAGGGAGGGGGGTCCGGTGGAGGTTCTGGGGGCTCTGCTTGGTCACATCCGCAGTTTGAGAAAGAAGACACGCCGTTGGTGATATCGAAGCAGAAGACGGAGGTGGTGTGCGGGGTCCCCA

POMP-nt-mNG2(11)-StrepII

GGCGGGGTCGACTGACGGTAACGGGGCAGAGAGGCTGTTCGCAGAGCTGCGGAAGATGACCGAGCTCAACTTCAAGGAGTGGCAAAAGGCCTTTACCGATATGATGTGGTCCCACCCTCAATTCGAGAAGGTGAGTGGGTACCCGGGCGGCTGGAGTTCCACGCGGGCTCGGGACCGTGGCTCGGCAGAAACAGGCAGTC

Cas9 purification

The pET15-SpCas9-NLS-HIS plasmid was purchased from Addgene (#62731)59. T7 Express cells (New England Biolabs) were transformed with the plasmid and grown in 5 ml of LB media (VWR) supplemented with 100 μg/ml carbenicillin (final concentration) overnight at 37 °C with shaking at 225 RPM. The next day, the cells were added to 100 ml of LB + carbenicillin (100 μg/ml) and incubated at 37 °C with shaking at 225 RPM. When OD600 reached 0.8, the 100 ml culture was added to 1 l of pre-warmed LB + carbenicillin (100 μg/ml) + 0.5% glucose and incubated at 37 °C with shaking at 225 RPM. When OD600 reached 0.4, the temperature was reduced to 16 °C. When OD600 reached 0.6, protein expression was induced with 0.2 mM IPTG (final concentration) overnight. The next day, cells were harvested by centrifugation at 4000 × g for 10 min and resuspended in 20 ml HBS buffer (50 mM HEPES, pH 7, 300 mM NaCl) supplemented with 25 mM imidazole, 100 μl HALT protease inhibitor cocktail (ThermoFisher), and 4 ml of 3 M NaCl. Protein purification was carried out at 4 °C. Cells were lysed by sonication and large debris were removed by centrifugation at 4000 × g for 5 min. The supernatant was centrifuged at ~100,000 × g for 20 min in an ultracentrifuge (Beckman) to remove small insoluble aggregates. The supernatant was passed through a 0.22 μm syringe filter prior to adding to 0.5 ml NEBExpress Ni Resin (New England Biolabs) by gravity flow. The beads were washed with 3 × 1 ml of HBS buffer supplemented with 25 mM imidazole. Protein was eluted with 3 × 1 ml of HBS supplemented with 300 mM imidazole and loaded onto a 1 ml HiTrap SP column equilibrated with buffer A (50 mM HEPES, 400 mM NaCl, pH 7). The column was washed with 3 ml of buffer A and bound proteins were eluted with 3 ml of buffer B (50 mM HEPES, pH 7, 650 mM NaCl). The elution was diluted 3× with buffer C (50 mM HEPES, pH 7, 150 mM NaCl) and supplemented with 1 mM DTT and 10% glycerol (final concentrations). Protein was concentrated in a 100 M.W. centrifugal device (Millipore) to ~13 mg/ml, aliquoted, and stored at −80 °C.

Cas9 HDR knock-ins

The gRNA IVT template was generated by PCR in 20 μl reactions containing 8 μl Milli-Q water, 1 μl ML557 + 558 (50 μM stock), 0.5 μl ML611 (2 μM stock), 0.5 μl gene-specific oligo (2 μM stock), and 10 μl Q5 2X master mix. The PCR products were purified with the DNA clean & concentrator kit (Zymo Research, D4014) and eluted with 10 μl Milli-Q water. The gRNA was synthesized using the HiScribe T7 Quick High Yield RNA Synthesis Kit (New England Biolabs, E2050S) in 20 μl reactions containing 6 μl Milli-Q water, 1 μl DNA template (50–100 ng total), 10 μl NTP buffer mix, 2 μl T7 10X master mix, and 1 μl RNAsin (Promega). The reaction mixture was incubated at 37 °C for 16 h overnight and the gRNA was purified using the RNA clean & concentrator kit (Zymo Research, R1014). The gRNA was eluted in 10 μl Milli-Q water and 1 μl of 1.5 M NaCl (150 mM final concentration) was added to the RNA elution to prevent Cas9 precipitation during RNP reconstitution. The gRNA was incubated at 70 °C for 5 min and stored at −20 °C until use. The Cas9-gRNA RNP was reconstituted by adding 20 μg gRNA (2–5 μl) to 5 μl Cas9 (80 μM; 13 mg/ml) at room temperature. 1.5 μl of ssDNA HDR template (IDT, 100 μM stock) was added to the Cas9-gRNA mixture. For each knock-in, 1 × 106 Expi293F cells (Thermo Fisher, A14528) were spun down at 500 × g for 3 min and resuspended in 15 μl of Expi293 expression medium (Gibco). 10 μl of cells were added to the Cas9 RNP and mixed well. The mixture was gently transferred to a Lonza Nucleofector X unit S 16-well strip without generating air bubbles. Program FS-100 was executed in a Lonza 4D Nucleofector and cells were immediately transferred to a 6-well plate containing 2 ml of Expi293 expression medium with 2.9 μl HDR enhancer (IDT). The plate was incubated at 37 °C, 8% CO2 with shaking at 125 RPM. After 24 h, the media was replaced with fresh media without HDR enhancer.

FACS for PSMG2(PAC2)-mNG2(11)-StrepII

For the PSMG2(PAC2)-mNG2(11)-StrepII knock-in, cells were sorted by fluorescence assisted cell sorting (FACS) to isolate cells with successful incorporation of the tag (Supplementary Fig. 1d). A plasmid expressing the first 10 β strands of fluorescent protein mNeonGreen2, mNG2(1-10), is transfected into the knock-in cells. mNG2(1-10) combines with the 11th β strand of mNG2, mNG2(11), in successful knock-in cells to generate a fluorescent protein that can be used to sort the cells by FACS. When cells have recovered (7–10 days after knock-in), cells were transferred to 10 ml of Expi293 expression medium in a 125 ml flask and incubated at 37 °C, 8% CO2, with shaking at 125 RPM. When cells reached ~2 × 106/ml, cells were diluted to 1 × 106/ml in 20 ml media. Cells were transfected with 20 ug of pcDNA5-mNG2(1-10) plasmid using Expifectamine 293 (Gibco). Two days after transfection, fluorescent cells containing successfully incorporated tags were sorted by FACS at the Sanford Burnham Prebys Flow Cytometry Shared Resource. Sorted cells were expanded, aliquoted, and stored in liquid nitrogen.

Clonal selection of knock-in cell lines

Clonal selection was carried out on PSMG3(PAC3)-twinstrep and POMP-mNG2(11)-StrepII knock-ins to isolate cells with successfully incorporated tags. Cells were seeded at ~100 cells/well in a 96-well plate and incubated at 37 °C, 8% CO2. After the cells grew to large colonies (>100 cells), individual colonies were extracted under a Leica DM IL inverted light microscope using a pipette and seeded into a new well with fresh Expi293 expression medium containing Penicillin-Streptomycin (Cytiva). The plate was incubated at 37 °C, 8% CO2. After 10–14 days, cells were transferred to 2 ml of fresh Expi293 expression medium in a 6-well plate and incubated at 37 °C, 8% CO2. When cells reached 1 × 106/ml, 1 ml of cells were spun down at 1000 × g and used for sequencing. Genomic DNA was extracted from cells using the Zymo Plasmid Miniprep kit (Zymo Research, D4209) and eluted in 25 μl Milli-Q water. DNA 100–700 bp upstream and downstream of the tag incorporation site was amplified by PCR. PCR products were purified using the DNA clean & concentrator kit (Zymo Research, D4014) and sequenced by Genewiz. Positive clones were expanded, aliquoted, and stored in liquid nitrogen.

Cell culture and proteasome purification

2 × 100 ml cultures of each cell line were grown in Expi293 expression medium (Gibco) at 37 °C, 8% CO2, and shaking at 125 RPM. When cells reached a density of 3–4 × 106/ml, cells were harvested by centrifugation at 500 × g for 5 min. Cells were resuspended in 10 ml HBS buffer (50 mM Hepes, 150 mM NaCl, pH 7.2) with the addition of DTT (1 mM final concentration), 100 µl of 100X HALT protease inhibitor cocktail (ThermoFisher), and 200 µl BioLock (iba). Cells were lysed by sonication and insoluble debris was removed by centrifugation at 4000 × g for 10 min. The supernatant was centrifuged at ~100,000 × g for 20 min at 4 °C in an ultracentrifuge. The supernatant was filtered through a 0.2 μm syringe filter and transferred to a column containing 50 μl Strep-Tactin XT Sepharose resin (Cytiva) by gravity flow. The resin was washed with 4 × 500 μl HBS buffer and the bound proteins were eluted with 1 ml HBS buffer with 50 mM biotin (final concentration). The eluted proteins were concentrated in a 50,000 MW centrifugal concentrator (Millipore).

Immunoblotting

1–2 µg of purified and concentrated (50k MW cutoff) proteins were mixed with 6X SDS sample buffer and loaded into a 10-well 4–20% TGX precast polyacrylamide gel (Bio-Rad) along with 1 μl of Dual Color Precision Plus protein ladder (Bio-Rad). The gels were run using a mini-PROTEAN Tetra cell (Bio-Rad) and proteins were transferred to a PVDF membrane using a Trans-Blot Turbo (Bio-Rad). The membrane was washed three times in 10 ml TBST and incubated in 10 ml of 5% milk/TBST for 15 min with rocking. The membrane was washed three times with 10 ml TBST and incubated with 10 ml of 5% BSA/TBST + anti-StrepII primary antibody (Novus Biologicals, NBP2-43735, 1:10,000) for 1 h at 4 °C with rocking. The membrane was washed three times using 10 ml of TBST with 5 min incubation with rocking at room temperature between washes. The membrane was incubated with 10 ml of 5% milk/TBST + anti-mouse secondary antibody (Thermo Scientific Pierce, 31446, 1:10,000) for 1 h at 4 °C with rocking. The membrane was washed three times using 10 ml TBST with 5 min incubation with rocking at room temperature between washes. The blots were imaged by chemiluminescence using Clarity MAX (Bio-Rad) in a ChemiDoc imaging system (Bio-Rad).

Cryo-EM imaging and data collection

For PSMG2(PAC2)-tagged samples, 2.5 μl of purified protein (1.3 mg/ml) were applied to Quantifoil Cu 300 mesh R 2/2 grids (glow-discharged in a Pelco easiGlow at 15 mA for 25 s), blotted for 10 s (4 °C, 100% humidity, blot force 0) using Whatman 1 filter paper, and plunge frozen in liquid ethane using a Vitrobot Mark IV (ThermoFisher). For PSMG3(PAC3) and POMP-tagged samples, 2.5 μl of purified protein (0.5 mg/ml) was applied to Quantifoil Cu 300 mesh R 2/2 grids coated with graphene oxide (Sigma)60, blotted for 1 s (4 °C, 100% humidity, blot force 0) using Whatman 1 filter paper, and plunge frozen in liquid ethane using a Vitrobot Mark IV (ThermoFisher). The vitrified samples were imaged in a Titan Krios electron microscope (FEI/ThermoFisher) in the Cryo-EM facility of the Structural Biology Shared Resource at Sanford Burnham Prebys Medical Discovery Institute. Images were collected on a Gatan K3 detector using SerialEM61 under super-resolution settings62 (Supplementary Table 1).

Cryo-EM image processing

Image processing of the micrographs was carried out in cryoSPARC63,64. Motion correction of the movies was carried out using the Patch Motion Correction tool using default parameters, including a maximum alignment resolution of 5 Å and B-factor of 500. CTF estimation was performed using the Patch CTF tool with an amplitude contrast of 0.1, minimum resolution of 25 Å, maximum resolution of 4 Å, minimum search defocus of 10,000 Å, and maximum search defocus of 30,000 Å. Particle picking was first carried out on the PSMG2-mNG2(11)-StrepII data using the Blob Picker tool with a minimum particle diameter of 100 Å, maximum particle diameter of 150 Å, and low-pass filter of 20 Å. 2D classification was performed using 100 classes and an initial low-pass filter resolution of 30 Å. Good 2D classes were used as templates for particle picking for the PSMG2-mNG2(11)-StrepII, PSMG2-mNG2(11)-StrepII (+37C), and POMP-mNG2(11)-StrepII data using the Template Picker tool with a particle diameter of 150 Å, low-pass filter of 20 Å, and angular sampling of 5 degrees. Good 2D classes from the POMP-mNG2(11)-StrepII data were used as templates for particle picking for the PSMG3-twinStrep data using the Template Picker tool with a particle diameter of 150 Å, low-pass filter of 20 Å, and angular sampling of 5 degrees. Multiple rounds of 2D classification were carried out for each dataset in order to remove images of non-proteins, ice, and other artefacts, as well as to isolate images that align well to produce high-quality 2D classes. Particle images that produce high-quality 2D classes were used to generate initial maps using Ab-Initio Reconstruction with initial resolution of 20 Å and maximum resolution of 12 Å. The map from the Ab-Initio Reconstruction was used as an initial model for 3D refinement using Homogeneous Refinement with an initial lowpass resolution of 20 Å and gold-standard FSC split resolution of 20 Å. For POMP-mNG2(11)-StrepII, 3D Classification was performed using 4 input maps with different β subunits masked out and initial low-pass resolution of 25 Å. 3D classes were then refined using Homogeneous Refinement with initial low-pass resolution of 30 Å. For PSMG3-twinStrep, 3D Classification was performed using 4 input maps with and without PAC1/PAC2 and PAC3/PAC4 masked out and initial low-pass resolution of 25 Å. 3D classes were then refined using Homogeneous Refinement with initial low-pass resolution of 30 Å. For PSMG2-mNG2(11)-StrepII (+37C), particles images were separated using Ab-Initio Reconstruction with 5 classes and initial lowpass resolution of 20 Å and maximum resolution of 12 Å. The Ab-Initio class corresponding to the pre-20S’ was subjected to Heterogeneous Refinement using as initial models a previous map of 20S and two different maps from processing of the PSMG2-mNG2(11)-StrepII dataset, with initial low-pass resolution of 20 Å. The 3D classes corresponding to pre-20S’ were combined and refined using Homogeneous Refinement with one of the 3D classes as the initial model, initial low-pass resolution of 30 Å, and gold-standard FSC split resolution of 20 Å. Resolution of the refined cryo-EM maps were estimated by gold-standard Fourier Shell Correlation at the 0.143 criterion65. Directional FSC of the maps were calculated66. Atomic models were built into the cryo-EM density maps using Coot67,68 and Phenix69,70. Initial models used for model building were Protein Data Bank (PDB) ID codes 7NAN (20S), 6JPT (PAC3), and 5WTQ (PAC4). The resulting cryo-EM maps and models were visualized using UCSF Chimera and UCSF ChimeraX71,72.

Density analysis and statistics

The average cryo-EM map density values corresponding to individual subunits and chaperones were calculated per residue of an atomic model using the ‘Values at Atom Positions’ function in Chimera and analyzed in Microsoft Excel. Only backbone atoms were considered in the analysis. Statistical significance was evaluated using a two-tailed t-test with unequal variance to calculate p-values. The buried surface area of the PAC1 and PAC2 C-terminal tails were calculated in ChimeraX using the measure buriedarea function.

Reporting summary

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