Molecular basis for the stepwise and faithful maturation of the 20S proteasome

Abstract

The proteasome degrades most superfluous and damaged proteins, and its decline is associated with many diseases. As the proteolytic unit, the 20S proteasome is assembled from 28 subunits assisted by chaperones PAC1/2/3/4 and POMP; then, it undergoes the maturation process, in which the proteolytic sites are activated and the assembly chaperones are cleared. However, mechanisms governing the maturation remain elusive. Here, we captured endogenous maturation intermediates of human 20S proteasome, which are low abundance and highly dynamic, and determined their structures by cryo–electron microscopy. Through structure-based functional studies, we identified the key switches that remodel and activate the proteolytic sites. Our results also revealed that the POMP degradation is tightly controlled by a dual-checking mechanism, while the α5 subunit senses POMP degradation to induce PAC1/2 release, achieving the full maturation. These findings elucidate mechanisms directing and safeguarding the proteasome maturation and set basis for building proteasomes to counteract the decline of protein degradation in aging and disease.

The protein degradation machinery uses an elaborate program to activate the proteolytic sites and clear the substrate entrance.

INTRODUCTION

Different types of proteasomes play crucial roles in a variety of key cellular processes. The 26S proteasome is responsible for most targeted protein degradation in cells (1–11) and, thus, participates in numerous physiological events, such as protein quality control, stress response, cell cycle progression, and amino acid supply (4–6, 9, 12, 13). The immunoproteasome and thymoproteasome are involved in the generation of major histocompatibility complex class I peptides for antigen presentation, which is required for immune response and T cell development (14–16). Spermatoproteasome is essential for sperm development (17–19). Proteasome dysfunction is implicated in many pathological conditions including aging, neurodegeneration, and autoimmune disorders (4, 5, 12, 14, 20–24). Although cap sub-complexes of proteasomes, such as 19S regulatory particle, 11S particle, and PA200, have distinct conformations, their core particles, which carry all the proteolytic sites, share a similar barrel-like structure with limited subunit replacement by variants (4, 10–12, 17, 25, 26).

The core particle of 26S proteasome, also known as 20S proteasome or constitutive CP, is a ~700-kDa complex and carries two chymotrypsin-like (β5), two trypsin-like (β2), and two caspase-like (β1) proteolytic sites. It is assembled from 28 subunits into a four-stacked-ring conformation with a composition of α1-7β1-7β1-7α1-7. The assembly is assisted by five dedicated chaperones, including proteasome assembly chaperone 1/2/3/4 (PAC1/2/3/4; Pba1/2/3/4 in yeast) and proteasome maturation protein (POMP; Ump1 in yeast) (12, 25, 27–29).

Current biochemical and genetic evidence suggests the following model for the CP assembly (12, 25, 27–29). The α-ring is first formed with the help of PAC1/2/3/4 (30–34), followed by sequential incorporation of seven β subunits mainly assisted by POMP to form a complex named half-proteasome (35–40). Subsequently, two half-proteasomes dimerize to form preholoproteasome (37, 41–43). Because the catalytic subunits β1, β2, and β5 are incorporated as inactive forms with N-terminal propeptides inhibiting their activity, these propeptides must be cleaved at the CP maturation stage to liberate the catalytic sites. Meanwhile, the assembly chaperones PAC1/2 and POMP, which block the substrate entrance and occupy the proteolytic chamber, must be removed, eventually resulting in a mature CP (12, 25, 28, 29, 35, 44–46).

Although the CP maturation is a crucial step to form an active proteasome, the molecular mechanisms directing this process and coupling it with the subunit installation remain to be learned. One reason for this situation is that the CP maturation intermediates half-proteasome and preholoproteasome are low abundance and highly dynamic in cells, which are hard to acquire for analysis. Here, we established a system to increase and efficiently capture maturation intermediates of human CP, resolved their structures by cryo–electron microscopy (cryo-EM) and revealed an elaborate program governing and safeguarding the CP maturation.

RESULTS

Capture of CP maturation intermediates

CP maturation intermediates are very low abundance and transiently exist in cells. To investigate mechanisms of the CP maturation, we established a system to increase and capture intermediates. FLAG-PAC2 and a StrepTagII-tagged β7 mutant lacking of the last 11 residues (∆W254-E264, β7-∆C) were expressed in human cells (Fig. 1A). FLAG-PAC2 can mark and separate assembly intermediates from mature CPs, the number of which is dominant in cells. The β7 C terminus crosses the middle line to hook the other half-proteasome, and loss of this region was shown to reduce the half-proteasome dimerization and cause its accumulation in yeast (41, 42, 47). We found that the β7-∆C expression in human cells successfully increased the number of half-proteasomes (Fig. 1B). In addition, StrepTagII affinity purification allowed us to enrich the maturation intermediates half-proteasome and preholoproteasome, which contain β7.

Fig. 1. Capture and cryo-EM analysis of CP maturation intermediates.

(A) A diagram showing the truncated region of the β7-∆C mutant. (B) Expression of the StrepTagII-tagged β7-∆C mutant in HEK 293F cells led to the accumulation of half-proteasomes. Proteasome complexes were captured with StrepTagII affinity purification and loaded onto native gels. The α-α7 blot indicates the population of total proteasomes, and the α-PAC1 blot indicates the population of assembly intermediates. (C) Flowchart for the purification of CP maturation intermediates. See Materials and Methods for details. Created with BioRender.com. (D) Western blot results for size exclusion chromatography (SEC) to enrich maturation intermediates. α-β2 and α-POMP blots were used to locate fractions containing CP intermediates. The collected fractions (SEC fractions) are indicated under the blots. (E) Western blot validation of the purification samples. Whole-cell extract (WCE), elute of StrepTagII affinity purification (StrepTagII AP), collected SEC fractions (SEC fractions), and elute of α-FLAG affinity purification (α-FLAG AP, the final sample) were examined. (F) Western blot results for native gel assays verifying the final sample enriching CP maturation intermediates (α-FLAG AP). (G) Cryo-EM maps of the intermediates showing the stepwise maturation of CP. Intermediate compositions are shown with each structure. Main events for each step are indicated on top of or under the arrows.

To acquire maturation intermediates (Fig. 1C), CP complexes containing β7 were first captured through StrepTagII affinity purification and subjected to size exclusion chromatography (SEC; Fig. 1D). Then, anti–FLAG-PAC2 affinity purification was used to separate intermediates from mature CPs (Fig. 1E). Results of native gel assays indicated that the sample enriched half-proteasome and preholoproteasome (Fig. 1F).

To understand the CP maturation mechanisms at atomic levels, we performed cryo-EM analysis of the intermediates and obtained four structures, including half-proteasome (α1-7·β1-7·PAC1/2·POMP), preholoproteasome-1 [(α1-7·β1-7·PAC1/2·POMP)x2], preholoproteasome-2 [(α1-7·β1-7)x2·PAC1/2·POMP], and mature CP [(α1-7·β1-7)x2] (Fig. 1G, figs. S1 and S2, and table S1). They provide an opportunity to study mechanisms underlying the key events of the CP maturation, including the proteolytic-site activation and chaperone removal.

The half-proteasome structure reveals the mechanism warranting β-ring completion

The β-ring carries all the catalytic subunits, and its faithful formation is crucial for the proteolytic function of CP. It was suggested that POMP is the major chaperone assisting the β-ring formation (35–38). Through analysis of the half-proteasome structure (Fig. 2, A to C), we found that POMP uses different regions to bind the β1-β6 subunits (Fig. 2C and fig. S3A), supporting its central role in the β-ring organization. β7 is the last inserted β subunit, and its incorporation initiates the half-proteasome dimerization and CP maturation (36, 37, 41, 42). As the first intermediate that β7 incorporates into, half-proteasome provides an opportunity to study the mechanism for β7 recruitment. We found that both the β5 propeptide and the β1 propeptide strongly interact with β7 (Fig. 2, D to F), suggesting that they cooperate to recruit β7. Because β1 and β5 are among the last β subunits to be installed (36), this mechanism guarantees that β7 incorporates only when the other β subunits are correctly positioned, which acts as a checkpoint to prevent premature transition from the β-ring assembly to the half-proteasome dimerization. Together, the half-proteasome structure demonstrates that POMP and the β subunit propeptides work together to organize the β-ring and safeguard its completion.

Fig. 2. The half-proteasome structure reveals the mechanism warranting β-ring completion.

(A to C) Overview structures of half-proteasome. A side view (A), a bottom view (B), and a top view (C) of the density map are shown. POMP, β1 propeptide, β2 propeptide, and β5 propeptide are shown in red, yellow, orange, and green, respectively. For (C), only the β-ring and POMP are shown for the clarity. (D to F) Structures of β1, β5, β6, and β7 in half-proteasome suggest that the β1 and β5 propeptides involve in the β7 recruitment. An overview structure (D) and zoomed interaction regions of β7 with the β1 propeptide (E) and the β5-propeptide (F) are shown. β7 is shown in surface representation, while the other subunits are shown in atomic model. The β1 and β5 propeptides are shown in yellow and green, respectively.

Key regions on the half-proteasome dimerization interface promote the catalytic-site activation

The catalytic-site activation only occurs when all subunits are installed in place (12, 25, 28, 29). However, mechanisms underlying this temporal regulation and preventing premature activation were not clear before. We resolved the structures of half-proteasome, preholoproteasome-1, and preholoproteasome-2 (Fig. 1G). The main difference between the two preholoproteasomes is that one side of No. 2 is fully mature. The densities of propeptides and POMP are weak in preholoproteasomes and only visible at a lower density threshold, indicating that they are undergoing removal, whereas the PAC1/2 density is relatively stronger, suggesting that their removal likely occurs later (Fig. 3A).

Fig. 3. Key loops on the dimerization interface of two half-proteasomes promote the catalytic-site activation.

(A) Cross sections for density maps of preholoproteasome-1 and preholoproteasome-2 under different thresholds (Thr.). (B) Comparison of three catalytic pockets in half-proteasome and preholoproteasome-1. (C) The β7′ R224-Y229 loop of the bottom half-proteasome approaches and stabilizes the β1 D51-V64 loop of the top half-proteasome. Top: β1 of the top half-proteasome (Half, dark gray) and β7′ of the bottom half-proteasome (Half′, light gray) were aligned with these subunits in preholoproteasome-1 (Preholo-1, blue). Bottom: The zoomed structure for two half-proteasomes (left) and preholoproteasome-1 (right). The β1 catalytic-pocket residues D51 and K67 are highlighted in salmon for half-proteasome and in green for preholoproteasome-1. The β7′-R224 and β1-Y59 residues are labeled in magenta and purple, respectively. (D and E) Western blot results show that the mutations of β7′-R224 (D) and β1-Y59 (E) led to an accumulation of β1 precursors. (F) Western blot results for SEC fractions of cells expressing β7-WT or β7-R224A. (G) The β6′ E212-Y216 loop of the bottom half-proteasome approaches and stabilizes the β2 T61-N73 loop of the top half-proteasome. The β2 catalytic-pocket residues D60 and K76 and the β6′-D214 and β2-V69 residues are highlighted as in (C). (H and I) Western blot results show that the mutations of β6′-D214 (H) and β2-V69 (I) led to an accumulation of catalytic-subunit precursors. (J) The β3′ D176-M183 loop of the bottom half-proteasome approaches and stabilizes the β5 R78-T89 loop of the top half-proteasome. The β5 catalytic-pocket residues D76 and K92 and the β3′-D178 and β5-S87 residues are highlighted as in (C). (K and L) Western blot results show that the mutations of β3′-D178 (K) and β5-S87 (L) led to an accumulation of β5 precursors.

Catalytic sites of β1, β2, and β5 (β1T35, β2T44, and β5T60) are auto-activated, facilitated by lysine (β1K67, β2K76, and β5K92) and aspartic acid (β1D51, β2D60, and β5D76) residues in catalytic pockets (44, 45, 48–50). Through comparing half-proteasome and preholoproteasome-1, we found that all three catalytic pockets become more compact and side-chain densities in the pockets are also clearer in preholoproteasome-1, likely favorable for autolysis (Fig. 3B). This suggests that the half-proteasome dimerization remodels catalytic pockets for activation.

To identify key elements integrating the dimerization signal, two half-proteasomes were docked together and compared to preholoproteasome-1. A series of conformational changes take place at the dimerization interface (fig. S3, B and C). We focused on regions that may remodel catalytic pockets (Fig. 3, C, G, and J). We found that the β7′ R224-Y229 loop of the bottom half-proteasome approaches the β1 D51-V64 loop of the top half-proteasome, which is close to the catalytic-pocket residues D51 and K67 of β1 (Fig. 3C). In preholoproteasome-1, the β1 D51-V64 loop becomes visible along with the conformational change of the β7′ loop, suggesting that they push each other, and this stabilizes the β1 loop (Fig. 3C and fig. S3D). We further identified β7′ R224 as the forefront residue on the loop of the bottom half-proteasome, and it is located closely to the β1 Y59 residue in preholoproteasome-1 (Fig. 3C). Mutations of these two residues caused accumulation of β1 precursors, suggesting an impairment of the catalytic-site activation (Fig. 3, D and E). Our further study showed that, in the β7 R224A mutant, β1 precursors were mainly present in preholoproteasome and, unexpectedly, also in 26S proteasome (Fig. 3F), supporting that this is an activation defect instead of an impairment of early assembly steps. Similar mechanisms also control the β2 and β5 activation (Fig. 3, G to L). It should be noted that, unlike the β7′-β1 and β3′-β5 loops that are mainly involved in their own activation, the mutants of the β6′-β2 loops impaired the activation of all three catalytic sites (Fig. 3, H and I). The results of SEC experiments for the β6 D214G mutant showed that β2 precursors were not only accumulated in the preholoproteasome fractions but also increased in the fractions containing early assembly intermediates (fig. S3E). This suggests that the β6′-β2 loops may have an additional function before the CP maturation, which requires further investigation. Together, these results indicate that the crucial loops on one half-proteasome can act as “switches” to remodel the catalytic pockets of another half-proteasome and, thereby, turn on their activation, which ensures that the catalytic-site activation only occurs in the preholoproteasome stage.

The dual-checking system controls the timing of the POMP degradation

Propeptides of β subunits, including catalytic subunits β1, β2, and β5 and non-catalytic subunits β6 and β7, can guide the subunit incorporation and protect catalytic-site threonines from oxidative damage. However, they also inhibit the proteolytic activity of CP and must be cleaved during maturation (25, 28, 29, 36, 37, 44, 51). It was unclear whether they are cleaved simultaneously and how their cleavage is orchestrated with other assembly events. We found that the β2 and β5 propeptides are visible in both half-proteasome and preholoproteasome-1, although their densities are weaker in the latter (Fig. 4A). In contrast, the β1 propeptide is visible in half-proteasome but completely disappears in preholoproteasome-1 (Fig. 4A). We could not observe the β6 and β7 propeptides in any intermediate structures. This difference in visibility of the propeptides may be due to different flexibility of these regions, but it also strongly suggests a model of stepwise cleavage, and the β2 and β5 propeptides may be the last ones to be processed. This hypothesis was supported by the results of SEC experiments, in which we found that the β2 and β5 precursors were present in both early-intermediate and preholoproteasome fractions; however, the β1 precursor was present in early-intermediate fractions but absent in preholoproteasome fractions, and the β6 and β7 precursors were absent in all fractions (Fig. 4B).

Fig. 4. The POMP degradation is tightly controlled by a dual-checking system of β2 and β5.

(A) Comparison of the β1, β2, and β5 propeptides in half-proteasome and preholoproteasome-1 suggests that the β2 and β5 propeptides are cleaved later than the β1 propeptide. (B) Western blot results for SEC fractions of CP intermediates indicate different cleavage status of propeptides in preholoproteasome. CP intermediates were captured by α-FLAG affinity purification from cells expressing FLAG-PAC2 and β7-∆C-StrepTagII. Peak positions of preholoproteasome/mature CP and half-proteasome/early intermediates are indicated under the blots. The ladder-like bands of β2 are pointed out by arrows. The large number of mature CP was likely due to the maturation during purification. (C) The structures of β2, β5, and POMP in preholoproteasome-1. (D) Western blot results of cells expressing the active-site mutants of β1 (β1-T35A) and β5 (β5-T60A) suggest that these active sites are not responsible for the β2-propeptide trimming. (E) Western blot results of cells expressing the propeptide mutants of β2 show that the F14G and F16G mutants suppressed the β2-propeptide trimming and also inhibited the POMP degradation, without affecting the β1 and β5 activation. (F) Western blot results for SEC fractions of cells, which expressed wild-type subunits or mutants of the active-site threonines, indicate that POMP was only accumulated in preholoproteasome of the β5-T60A mutant. The α-α7 blots were used to locate the preholoproteasome/mature CP peak, and the α-β1 and α-β2 blots showed the normal activation of these two catalytic subunits.

What is the potential function for the β2 and β5 propeptides to be cleaved at last? The N-terminal extension of the β2 propeptide acts as a hook to attach POMP in place (Fig. 4C). We saw ladder-like bands of β2 on Western blots of preholoproteasome fractions (Fig. 4B), suggesting that an N-terminal trimming occurs, which may remove its POMP-binding region and prepare POMP for degradation. To gain insight into the trimming, we expressed the catalytic-site mutants of β1 and β5 and found that neither affected the β2-propeptide processing, suggesting that the catalytic sites are not responsible for the trimming (Fig. 4D). We further performed a mutational scan of the relevant residues on the β2 propeptide. It showed that the F14G and F16G mutants strongly suppressed the β2-propeptide processing, and they also inhibited the POMP degradation without affecting the β1 and β5 activation (Fig. 4E). These results suggest a crucial role of F14/F16 residues and support the involvement of the β2-propeptide processing in the POMP removal. Furthermore, when mutating three catalytic-site threonines, we found that only the β5 T60A mutation blocked POMP degradation in preholoproteasome without affecting the activation of β1 and β2 (Fig. 4F). This indicates that β5 is the major subunit to directly degrade POMP. Together, these results suggest that the processing of β2 propeptide and the activation of β5 are two prerequisites for the POMP degradation, ensuring that the POMP removal only occurs after CP is properly assembled and the catalytic-sites are activated. Once POMP is degraded, the proteolytic chamber of CP is cleared.

Coupling of the POMP degradation and the PAC1/2 release by α5

PAC1/2 block the substrate entrance of CP, and they must be removed to achieve the full maturation of CP. It was unclear whether PAC1/2 are degraded by newly formed CP or released (30). In the intermediates, PAC1/2 and the catalytic sites of β1, β2, and β5 are fully separated by the α-ring. Therefore, it is more likely that PAC1/2 are released instead of being degraded. Furthermore, as mentioned above, analysis of preholoproteasomes suggests that the PAC1/2 removal likely occurs after the POMP degradation (Fig. 3A).

To understand the mechanism triggering the PAC1/2 release, we compared the structures of preholoproteasome-1 and mature CP (Fig. 5A and fig. S4, A and B). The N terminus of α5 deeply inserts into the PAC1/2 dimer (Fig. 5A and fig. S4A), and the α5 L121-G138 loop contacts POMP in preholoproteasome-1 (fig. S4B). Once POMP is degraded, the conformation of the L121-G138 loop changes, which further induces a movement of the α5 N terminus (Fig. 5B). This movement compromises its interaction with PAC1 and also leads to a clash with the PAC1 N terminus (Fig. 5B), which may promote the PAC1/2 release. To examine this model, we mutated α5-D9, a residue that may be involved in the clash between the α5 N terminus and PAC1, and α5-F123, a residue on the L121-G138 loop that may pull the α5 N terminus. Both mutants caused the PAC1 retardation on preholoproteasome (Fig. 5, C to F) without affecting the POMP degradation and the activation of β1, β2, and β5 (Fig. 5G). Thus, α5 may perceive the signal of the POMP degradation to promote the PAC1/2 release. Once PAC1/2 are released, the substrate entrance is cleared, and, eventually, a mature CP is formed.

Fig. 5. α5 senses the POMP degradation to promote the PAC1/2 release.

(A) Structural comparison of the α-ring in preholoproteasome-1 and mature CP [PDB-5LE5; (60)]. The visible parts of the first 12 amino acids of all α subunits are colored magenta for Preholo-1 and cyan for mature CP, and the other parts of α subunits of preholoproteasome-1 and mature CP are colored light gray and dark gray, respectively. The α5 N terminus inserts deeply into the PAC1/2 dimer in preholoproteasome-1 and substantially moves during the CP maturation. (B) Comparison of preholoproteasome-1 and mature CP structures demonstrates that α5 may sense the POMP degradation to promote the PAC1/2 release. Upon the POMP degradation, the L121-G138 loop of α5 moves upward (the bottom arrow) and likely pulls its N terminus toward the PAC1 N terminus (the top arrow), which weakens the interaction of α5 with PAC1/2 and results in a clash between the α5-D9 residue and the PAC1 N terminus. (C and E) Western blot results for native gel assays show that expression of the α5-D9A (C) or α5-F123G (E) mutants caused the PAC1 retardation on preholoproteasome. Proteasomes containing α5-WT-StrepTagII, α5-D9A-StrepTagII, or α5-F123G-StrepTagII were captured by StrepTagII affinity purification. The α-α7 blot was used to mark the position and normalize the number of preholoproteasome/CP. (D and F) Quantitative results of (C) and (E). Individual values and means + SEM of three independent experiments are shown; unpaired two-tailed t test; *P < 0.05. (G) Western blot results for SDS–polyacrylamide gel electrophoresis (PAGE) show the effects of the α5-D9A and α5-F123G mutants on the CP maturation. Proteasomes containing α5-WT-StrepTagII, α5-D9A-StrepTagII, or α5-F123G-StrepTagII were captured by StrepTagII affinity purification and examined for the PAC1 and POMP levels and the propeptide cleavage of β1, β2, and β5. A.U., arbitrary units.

DISCUSSION

On the basis of previous findings in the field (12, 25, 28, 29) and the mechanisms uncovered here, we propose the following working model for the CP maturation (Fig. 6). A half-proteasome is formed upon the recruitment of β7 to the β-ring by the β1 and β5 propeptides. The crucial loops of one half-proteasome can function as switches to remodel the catalytic pockets of another half-proteasome into a more compact conformation and, thereby, initiate their activation. The β2 and β5 propeptides are processed late in preholoproteasome. Trimming of the β2 propeptide changes the POMP conformation, while the activated β5 degrades POMP. The POMP loss is perceived by α5 through a conformational change, which, in turn, promotes the PAC1/2 release. Eventually, a mature CP is formed. This model needs further verification.

Fig. 6. The working model for the stepwise and faithful maturation of CP.

(A) Proteolytic-site activation. Two half-proteasomes dimerize, and the key loops (blue triangles) on one half-proteasome cross the middle line and compact the catalytic pockets of another half-proteasome to facilitate their activation. (B) POMP degradation. Trimming of the β2 propeptide changes POMP conformation, and the activated β5 degrades POMP. (C) PAC1/2 release. α5 senses the POMP degradation and promotes the PAC1/2 release. Last, a mature CP is formed. Created with BioRender.com.

Our structure-directed functional study is a step toward the full understanding of proteasome biogenesis. Recent structural analyses have largely improved our knowledge on this fundamental process (38–40, 43, 46). Nevertheless, connections between structure and function are still lacking; mechanisms for key steps have not been thoroughly investigated and verified; and a method to conveniently study the endogenous formation of the human proteasome is also urgently needed. We established a simple system for expressing a tagged assembly chaperone in human cells to capture endogenous intermediates of the human 20S proteasome. This is different from the yeast system or the recombinant proteasomes produced in insect cells (40, 46), and it can be easily used for further mechanistic studies. Using this system, we captured and resolved structures of half-proteasome, preholoproteasome-1, and preholoproteasome-2, which are consistent with a recent structural study of the recombinant CP intermediates (fig. S5) (40). On the basis of these structures, we studied the proteasome maturation process in detail with extensive mutagenesis and biochemical validation, leading to the identification of several key mechanisms. The catalytic-site activation step has drawn much attention (40, 46); however, the mechanism liberating all three catalytic sites was not uncovered. Our structural and functional studies found the unified “molecular switch” mechanism activating all the catalytic sites. The chaperone removal, which clears the substrate entrance and proteolytic chamber, is the final and crucial step for the full maturation of CP. However, it is the least studied step and its mechanism is largely unknown. Our study revealed a dual-checking mechanism warranting the proper degradation of POMP and elucidated the mediator role of α5 in coupling the PAC1/2 release with the POMP removal.

Our study revealed that the CP maturation is a temporally well-controlled serial event with elaborate mechanisms to safeguard its timing and fidelity. First, the timing for the transition from the β-ring assembly to the CP maturation is warranted by the dependence of β7 incorporation on the β1 and β5 propeptides. This also resolves a debate on the role of the β5 propeptide in half-proteasome dimerization. Previous studies noticed that a ∆propeptide mutant of β5 inhibited the dimerization of half-proteasomes, and overexpression of β7 suppressed lethality caused by this mutant (37, 52). However, the β5 propeptide does not locate on or near the middle line (38, 43). Our results demonstrate that the β5 propeptide is indirectly involved in the half-proteasome dimerization through recruiting β7. Second, one half-proteasome can switch on another half-proteasome. It was unknown how the CP activation is achieved and also restricted in preholoproteasome (12, 25, 28, 29). Our study identified three crucial loops of one half-proteasome can compact and switch on the catalytic pockets of another half-proteasome. This mechanism warrants the timing of the catalytic-site activation. Third, the timing of POMP removal is overseen by the β2-propeptide trimming and the β5 activation synergistically. This dual-checking mechanism prevents the premature removal of POMP. Fourth, the PAC1/2 release is induced by the POMP degradation through a conformational change of α5, which guarantees the timely full maturation of CP.

To gain insight into the conservation of the key maturation mechanisms identified in this study, we compared our structures with the yeast intermediate structures. Because the yeast half-proteasome structure is not available, we compared the preholoproteasome structures (43). Although the amino acid sequences of the assembly chaperones are not highly conserved, their three-dimensional structures are quite similar, especially for PAC1/Pba1 and POMP/Ump1 (fig. S6). The conformations of the human β7′-β1 and β6′-β2 loops that are crucial for the catalytic-site activation are highly conserved in yeast (fig. S7, A and B; the β3′-β5 loop pair is not visible in the yeast preholoproteasome structure). Furthermore, the N terminus of yeast α5 also inserts deeply into the Pba1/2, and the conformational changes of human α5 during the CP maturation also occur in yeast (fig. S7, C to E), suggesting that the PAC1/2 release mechanism identified in this study may be functional in yeast. Collectively, these observations support that the CP maturation mechanisms are likely conserved from yeast to human.

The failure of protein degradation is common in aging and associated diseases, e.g., neurodegenerative disorders (4, 12, 23). The CP maturation mechanisms uncovered in this study establish the basis for building more functional proteasomes to counteract the decline of protein degradation in relevant diseases. Besides constitutive CP, there are variant CPs with subunit replacement (4, 12, 25). For example, β1, β2, and β5 subunits are replaced by β1i, β2i, and β5i in immunoproteasomes and by β1i, β2i, and β5t in thymoproteasomes (14). Although their maturation mechanisms may have differences, all these CPs share most subunits and a barrel-like structure and are assembled with assistance of the same chaperones PAC1/2/3/4 and POMP (29, 53). Therefore, our methods and findings could be extended to studies of variant CPs and associated diseases, e.g., proteasome-associated autoinflammatory syndrome, which is thought to be caused by the defective formation and activation of immunoproteasomes (14, 54). It is interesting to compare the maturation mechanisms of different CPs.

Our study has limitations. We used FLAG-PAC2 to capture CP intermediates. Although it appears that PAC1/2 are the last chaperones to be removed and stay on CP intermediates for the whole assembly and maturation process, we cannot rule out that other PAC2-free intermediates exist. It is thorough to examine intermediates captured by tagging other assembly chaperones or through other purification methods to see whether more intermediates can be acquired. In addition, it will be interesting to apply techniques that can directly visualize the intermediate dynamics, e.g., single-molecule imaging, to further verify the mechanisms uncovered in this study. Furthermore, other mechanisms may be involved synergistically or in parallel to regulate the CP maturation, which requires more investigation.

MATERIALS AND METHODS

Cells

Human embryonic kidney (HEK) 293T (American Type Culture Collection, CRL-3216) was maintained in high-glucose Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum and penicillin-streptomycin. HEK 293F (Thermo Fisher Scientific, R79007) was maintained in SMM 293-TII medium (SinoBiological Inc., M293TII) with penicillin-streptomycin.

Cloning of CP subunits and mutagenesis

Wild-type CP subunits, PAC2, and the β7-∆C truncation mutant were cloned from a HEK 293T cDNA library using primers listed in table S2 and inserted into the pCDH-CMV plasmid (a gift from K. Oka; Addgene, no. 72265) engineered to bear an N-terminal FLAG tag or a C-terminal StrepTagII tag. Point mutations were generated through polymerase chain reaction (PCR)–based site-directed mutagenesis using corresponding wild-type plasmids as templates and primers listed in table S2. PCR products were treated with Dpn I at 37°C for 3 hours to remove template DNAs, and transformed into the Escherichia coli strains DH5α or Stbl3. All constructs were validated by Sanger sequencing.

Lentiviral production

Lentiviruses were packaged by co-transfection of indicated expression plasmid, pMD2.G and psPAX2 (a gift from D. Trono; Addgene, nos. 12259 and 12260) into HEK 293T cells, and collected after 48 hours. They were concentrated with Lenti-X concentrator (Takara Inc., no. 631231), resuspended in DMEM medium and stored at −80°C.

Purification of CP maturation intermediates

HEK 293F cells stably expressing FLAG-PAC2 and β7-∆C-StrepTagII were collected and resuspended in the lysis buffer [25 mM Hepes (pH 7.4), 100 mM NaCl, 5 mM MgCl₂, 2 mM adenosine 5′-triphosphate (ATP), 0.25% NP-40, 10% glycerol, and protease inhibitor cocktail (GLPBio, GK10014); LysB]. The cells were broken with Dounce homogenizer for 40 strokes, and then centrifuged at 15,000g for 20 min at 4°C. The supernatant was incubated with Streptactin beads (Smart-Lifesciences, SA053025) at 4°C for 2 hours. After three washes with LysB and another two washes with the elution buffer 1 [25 mM Hepes (pH 7.4), 100 mM NaCl, 5 mM MgCl₂, 2 mM ATP, 10% glycerol, and 0.1% NP-40; EB1] supplemented with protease inhibitor cocktail, the beads were eluted with 5 mM biotin (Solarbio, D8150) in EB1 at 4°C for 1 hour. The elute was concentrated with 10-kDa cutoff Amicon centrifugal filter unit (Millipore, UFC801024), filtered through a 0.45-μm low-protein-binding filter (Pall, no. 4614), and subjected onto the Superose 6 Increase 10/300 GL column (Cytiva, no. 29091596) in EB1. Flow rate and fraction size were set at 0.25 ml/min and 0.3 ml, respectively. The desired fractions were combined and incubated with α-FLAG beads (Smart-Lifesciences, SA042025) at 4°C for 2 hours. The beads were washed three times with EB1 and another two times with the elution buffer 2 [25 mM Hepes (pH 7.4), 100 mM NaCl, 5 mM MgCl₂, 2 mM ATP, and 0.025% n-dodecyl-β-D-maltoside; EB2] and eluted with 3xFLAG peptides (400 μg/ml; GLPBio, GP10149) in EB2 at 4°C for 45 min. The elute was the final sample containing CP maturation intermediates.

Cryo-EM sample preparation and data collection

The sample was first checked by negative staining using 0.2% (w/v) uranyl acetate on a Tecnai T12 Transmission Electron Microscope (Thermo Fisher Scientific) and then loaded onto homemade thin carbon layer–coated and glow-discharged 200 mesh 2/1 Au grids (Quantifoil, N1-C11nAu30-01) to reduce preferred orientation and improve particle concentration and distribution. After 20 s, grids were blotted for 2 s with a blot force of −2 and plunged into cooled ethane in a Vitrobot Mark IV (Thermo Fisher Scientific) set at 4°C and 100% humidity.

Cryo-EM data were collected using a 300-kV Titan Krios G4 microscope (Thermo Fisher Scientific) equipped with a BioContinuum K3 Direct Electron Detector with Gatan GIF imaging filter. Data were collected at a magnification of ×81,000 with a pixel size of 0.5275 Å in super-resolution mode. Data collection with beam image-shift was performed using EPU, capturing 40 frames per movie stack for a total dose of 49.41 electrons/Ų with defocus range of −1.4 to −2.4 μm.

Cryo-EM data processing and model building

The collected movie stacks underwent gain normalization, binning, dose weighting, and motion correction using Motioncorr2 (55). The contrast transfer function parameters were determined with CTFFIND4 (56). Subsequently, particles were automatically picked with Gautomatch (57), extracted from the micrographs, and subjected to two-dimensional (2D) classification in RELION-3.1 (58). Particles in the 2D classes displaying secondary structure features were selected and further processed through either CryoSPARC (59) ab initio Reconstruction or RELION 3D Classification. The 3D classes exhibiting distinct structural features were chosen and subjected to an additional round of 3D Classification using either CryoSPARC Heterogeneous Refinement, CryoSPARC ab initio Reconstruction, or RELION 3D Classification. Particles contributing to the same class were combined and further refined using Non-uniform Refinement in CryoSPARC. Fourier shell correlation curves were calculated in CryoSPARC. For additional details regarding micrograph numbers, particle numbers, and the processing workflow, please refer to fig. S1 and table S1.

We used the crystal structure of the human 20S proteasome [Protein Data Bank (PDB): 5LE5; (60)] and AlphaFold2 (61)–predicted structures for human PAC1/2 and POMP as initial models. These models were fitted into cryo-EM density maps using UCSF Chimera (62) and manually adjusted in Coot (63). Iterative cycles of real-space refinement in PHENIX (64) and manual adjustments in Coot were used to optimize the models. Detailed statistics of the model-building process can be found in table S1.

Native gel assay

HEK 293T cells stably expressing StrepTagII-tagged wild-type or mutant CP subunits were lysed by five freeze-thaw cycles in LysB and then centrifuged at 21,300g for 15 min at 4°C. The supernatant was incubated with Streptactin beads at 4°C for 2 hours. After six washes with LysB, the beads were eluted with 5 mM biotin (Solarbio, D8150) in LysB at 4°C for 1 hour. The elutes were collected and mixed with the 2× native sample buffer [62.5 mM tris-HCl (pH 6.8), 40% glycerol, and 0.01% bromophenol blue] supplemented with 5 mM MgCl₂ and 2 mM ATP and loaded onto 3 to 8% tris-acetate gels (Bio-Rad, no. 3450131). Purified human CP (Boston Biochem, E-360) was usually loaded as a standard to mark positions of proteasome complexes. Electrophoresis was conducted at 4°C with 50 V for 1 hour followed by 120 V for 5 hours in the running buffer (90 mM tris, 90 mM boric acid, 1 mM EDTA, 2.5 mM MgCl₂, 1 mM ATP, and 0.5 mM dithiothreitol). Then, proteins in gels were transferred onto 0.45-μm Immobilon-P polyvinylidene difluoride (PVDF) membranes (Millipore, IPVH00010) for Western blot analysis. For Fig. 5 (C to F), the samples were normalized by α-α7 SDS–polyacrylamide gel electrophoresis (PAGE)–Western blots before loaded onto native gels.

Size exclusion chromatography

HEK 293T cells stably expressing wild-type or mutant CP subunits were lysed with the same method of native gel assays. After centrifugation, supernatants were filtered through 0.45-μm low-protein-binding filters, and, maximally, 0.5 ml of supernatants were subjected onto Superose 6 Increase 10/300 GL column. Flow rate and fraction size were set at 0.25 ml/min and 0.3 ml, respectively.

Western blot analysis

HEK 293T cells stably expressing wild-type or mutant CP subunits were lysed in a modified radioimmunoprecipitation assay buffer [50 mM tris-HCl (pH 7.5), 150 mM NaCl, 0.5% sodium deoxycholate, 0.5% SDS, and 1% Triton X-100] supplemented with protease inhibitor cocktail and Benzonase (20 U/ml; Sigma-Aldrich, E1014). Cell lysates were cleared by centrifugation at 21,300g for 15 min at 4°C. Supernatants were mixed with 5× SDS-PAGE loading buffer and subjected to 10 or 12.5% tris-glycine SDS-PAGE gels. Proteins were transferred onto 0.45-μm Immobilon-P PVDF membranes. Membranes were blocked with 5% nonfat milk in TBST [20 mM tris-HCl (pH 7.5), 150 mM NaCl, and 0.1% Tween 20] at room temperature for 1 hour, followed by incubation with primary antibodies (listed in table S3) diluted in 5% bovine serum albumin/TBST at 4°C overnight. After three washes with TBST, membranes were incubated with horseradish peroxidase–conjugated secondary antibodies (Thermo Fisher Scientific) diluted in 5% nonfat milk/TBST at room temperature for 1 hour. After four washes with TBST, immunoreactivity was developed using ECL reagent and detected with Tanon-5200 Chemiluminescent Imaging System (Tanon Science & Technology).

Sequence alignment

The alignments of amino acid sequences shown in fig. S6 (A to C) were generated with the Pairwise Sequence Alignment using the Stretcher algorithm in the European Bioinformatics Institute (EMBL-EBI) (65). The substitution scoring matrix of BLOSUM62 was used, with the GAP OPEN and GAP EXTEND settings of 12 and 2, respectively.

Statistical analysis

Western blot results were quantified by ImageJ (66), and statistical analyses were performed using Prism9 (GraphPad). Statistic details are shown in the figure legends.

Supplementary Materials

This PDF file includes:

Figs. S1 to S7

Tables S1 to S3