Structure of the 26S proteasome from Schizosaccharomyces pombe at subnanometer resolution (original) (raw)

Abstract

The structure of the 26S proteasome from Schizosaccharomyces pombe has been determined to a resolution of 9.1 Å by cryoelectron microscopy and single particle analysis. In addition, chemical cross-linking in conjunction with mass spectrometry has been used to identify numerous residue pairs in close proximity to each other, providing an array of spatial restraints. Taken together these data clarify the topology of the AAA-ATPase module in the 19S regulatory particle and its spatial relationship to the α-ring of the 20S core particle. Image classification and variance analysis reveal a belt of high “activity” surrounding the AAA-ATPase module which is tentatively assigned to the reversible association of proteasome interacting proteins and the conformational heterogeneity among the particles. An integrated model is presented which sheds light on the early steps of protein degradation by the 26S complex.

Keywords: deubiquitylating enzymes, macromolecular complex, ubiquitin-proteasome pathway, ubiquitin receptor, single particle classification

In eukaryotic cells, most proteins in the cytosol and the nucleus are regulated via the ubiquitin-proteasome pathway and malfunctions of this pathway have been implicated in a wide variety of diseases (1). The 26S proteasome is the most downstream element of this pathway, executing protein degradation (2–4). Unlike constitutively active proteases, the proteasome has the capacity to degrade almost any protein, yet it acts with exquisite specificity. The key stratagem is self-compartmentalization: The active sites of the proteolytic 20S core particles (CPs) are sequestered from the cellular environment in the interior of this barrel-shaped subcomplex (5). Proteins destined for degradation are marked by a polyubiquitin chain, a degradation signal that is recognized by the 19S regulatory particles (RPs) that bind to either one or both ends of the CP to form the 26S holocomplex. The RPs (i) recognize the polyubiquitylated substrates, (ii) trim and recycle the polyubiquitin chains, (iii) unfold substrates to be degraded, and (iv) open the gate to the CP and assist in substrate translocation into the interior of the CP. These tasks are performed by a complex machinery involving at least 19 different subunits, 6 AAA-ATPases (Rpt1–6), and 13 non-ATPases (Rpn1–3, Rpn5–13, Rpn15/Sem1p).

Although the structure of the CP has been elucidated in great detail by X-ray crystallography (6, 7), the structure of the RP is only dimly understood at present. Best characterized are the AAA-ATPases which form a heterohexameric subcomplex situated at the base of the RP in close proximity to the α-rings of the CP (8, 9). The C-terminal residues of Rpt2 and Rpt5 were shown to be involved in opening the gate in the α-rings, allowing substrates to enter the CP. A similar mechanism has been postulated for proteasome-activating nucleotidase (PAN), the archaeal homohexameric homolog of the eukaryotic AAA-ATPase module (10). Crystal structures of the two major fragments of PAN suggest that the N- and C-terminal domains form two stacked concentric rings (N ring and AAA ring) (11, 12); the N ring is implicated in substrate unfolding and the AAA ring in gate opening and substrate translocation.

Among the non-ATPases, the functions of Rpn10, Rpn11, and Rpn13 are well known. Rpn11 is a deubiquitylating enzyme and responsible for ubiquitin (Ub) removal from substrates and Ub recycling (13, 14), whereas Rpn10 and Rpn13 are Ub receptors (15). Their localization within the RP would be particularly informative in terms of understanding the sequence of events between the initial binding of substrates and their translocation into the CP.

Given the complexity of the 26S proteasome and its fragile nature, its dynamics and the association-dissociation of proteasome interacting proteins (PIPs) and, therefore, their presence in variable amounts, it has been impossible so far to obtain crystals suitable for a high-resolution structural analysis by X-ray crystallography. For cryoelectron microscopy (cryo-EM) and for protein–protein interaction studies, the requirements for sample homogeneity are less stringent. A ∼25-Å resolution structure of the Drosophila melanogaster 26S proteasome has been reported recently (9), revealing the basic organization of the RP and defining, in particular, the localization and the boundaries of the AAA-ATPase module.

Here we present a structure of the 26S complex isolated from Schizosaccharomyces pombe at much higher resolution [9.1 Å at Fourier shell correlation (FSC) of 0.5 and 6.7 Å at FSC of 0.3]. Moreover, we have used chemical cross-linking in conjunction with MS to identify numerous residue pairs in close proximity to each other providing us with an array of spatial restraints (16–18). The integration of these data into the medium resolution EM maps allowed us to generate a model providing insights into the structural organization of the 26S holocomplex.

Results and Discussion

Purification and MS Analysis.

We analyzed the purified S. pombe 26S proteasomes by quantitative MS (Fig. S1). The canonical 26S proteasome subunits α1–7, β1–7, Rpt1–6, Rpn1–3, and Rpn5–12 are all present in equimolar amounts. Rpn13 is present in two different orthologs, Rpn13a (Uniprot ID Q9Y7Y6) and Rpn13b (Uniprot ID Q9USM1), each in a ∼0.2∶1 ratio. Expression of organ-specific Rpn13 orthologs was previously reported for D. melanogaster testis (19), but simultaneous expression of different orthologs has not been described in other species. The abundance of the 9 kDa peptide Rpn15 could only be determined with low accuracy by our MS approach because only one unique peptide was identified. The ubiquitin C-terminal hydrolases Uch2 and Ubp6, as well as Ub were detected in significant amounts.

When comparing the relative stoichiometry of the S. pombe 26S proteasome to that determined for Drosophila (9) we found the following differences: Rpn10 is substantially more abundant in S. pombe (∼0.8∶1) than in the Drosophila (∼0.25∶1). Moreover, Rpn13a/Rpn13b and Uch2 are present in substoichiometric amounts in S. pombe, whereas Rpn13 and Uch37 were found in stoichiometric amounts in Drosophila proteasomes. Interestingly, in 26S proteasomes purified from Homo sapiens, none of these subunits (Rpn10, Rpn13, Uch2/Uch37) are present in significant amounts (20). Thus, quantitative analysis indicates that α1–7, β1–7, Rpt1–6, Rpn1–3, Rpn5–9, and Rpn11–12 constitute the canonical core of the 26S proteasome, whereas the subunits Rpn10 and Rpn13, albeit always present, are found in variable amounts.

The 26S Proteasome at Subnanometer Resolution.

From vitrified samples of our 26S preparation, micrographs were acquired in a semiautomated fashion. Micrographs typically showed complete double-capped 26S proteasome particles, but some single capped particles, as well as isolated CPs were also observed (Fig. S2). From 20,000 micrographs, we selected 270,000 holocomplexes for single particle analysis.

Due to the fragility and dynamics of the 26S proteasome, we anticipated substantial structural variability among individual particles. To “purify” the particle ensemble in silico we used a maximum likelihood-based classification method (ML3D) (21) and classified the particles into two different groups (Fig. S3). One class showed all the features of 26S holocomplexes, whereas the other clearly corresponded to partially (dis)assembled 26S proteasomes (∼85,000 particles). In a second ML3D classification step, we were able to further separate the holocomplexes into 26S proteasomes which differed by a distinct mass in the cap region (“class D1” with one and “class D2” with two “extra masses” containing ∼85,000 and ∼100,000 particles, respectively).

The variable mass was at approximately the same position where a variable mass was observed before with the D. melanogaster proteasome (9). However, in the present study occupancy with the extra mass was much higher than in D. melanogaster. Whereas in S. pombe the extra mass was present at either one or both RPs, the extra mass was observed only at one RP or not at all in the D. melanogaster proteasome (Fig. S4). In D. melanogaster, the relative abundance of the extra mass compared to the remaining RP density (∼25%) correlated to the abundance of Rpn10 as determined by MS. Therefore, Rpn10 was tentatively localized to the extra mass. Interestingly, in the present study, the relative abundance of the extra mass (∼75%) also correlates to the Rpn10 stoichiometry (0.8∶1). Thus, our EM and MS data support the putative localization of Rpn10.

The 3D densities corresponding to intact holocomplexes displayed C2 symmetry (apart from the extra mass). Therefore, to obtain higher resolution, we imposed C2 symmetry for the reconstruction of the 26S proteasome density from all holocomplexes (resolution of 9.1 and 6.7 Å at FSC of 0.5 and 0.3, respectively; Fig. S5). In the final symmetrical reconstruction, several helical motifs became discernible (Fig. 1A, Fig. S6, and Movie S1).

Fig. 1.

Fig. 1.

Structure of the 26S proteasome from S. pombe. Model of the 26S proteasome and two different views rotated around the pseudo-sevenfold axis of the CP by 90°. The isosurface threshold was set to include a protein mass of 755 kDa for the CP. Secondary structure elements like bihelical repeats can be clearly discerned (A, *). B shows the isosurface representations of the model (lower half). The transparent view through the reconstruction shows the fitted atomic models of the CP (red, **: α-helix of subunit α4) and the AAA-ATPase (blue). A mesh representation of the reconstruction with an overlay isosurface in green highlighting the main variances among the particles is shown in C.

We fitted a comparative model of the CP into the EM map (Fig. 1B). The prominent N-terminal α-helix, a hallmark of the α4 subunit, allows unambiguous positioning of the CP. The fit of the atomic model and the EM map is excellent, as indicated by the high cross-correlation coefficient (CCC = 0.78, both models filtered to 9 Å) and the visual correspondence of helices in model and map. The density at the gate of the CP is in good agreement with the closed-state of the crystal structure.

Adjacent to the CP, we positioned a comparative model of the AAA-ATPases Rpt1–6 into the EM map as described previously (22) (Fig. 1). The C-terminal AAA-ATPase domains form the AAA ring that binds to the CP, whereas the N-terminal segments form a smaller ring positioned atop of the AAA ring (11, 12). In the EM map, density corresponding to the N-terminal coiled coils is clearly discernible and correlates well to the atomic model. The atomic model of the AAA ring is in good agreement with the EM density, albeit some helical densities do not colocalize precisely with the helices in the atomic model indicative of dynamic conformation changes (see below). The center of the hexameric AAA-ATPase is shifted by ∼20 Å from the pseudo-sevenfold rotational symmetry axis of the CP, similar as observed for the D. melanogaster 26S proteasome (30 Å). The hexamer is less inclined with respect to the CP axis; whereas the pseudo-threefold symmetry axis of the AAA-ATPase hexamer was tilted by ∼10° with respect to the CP axis in the D. melanogaster density, it is tilted by ∼4° in the map from S. pombe.

In the remaining density of the RP, features reminiscent of short bihelical repeats in supercoiled quaternary structures are recognizable (Fig. 1A, *). Indeed, such repetitive motifs, probably similar to tetratricopeptide repeats (TPRs), have been predicted before for Rpn1 and Rpn2, as well as the proteasome, COP9 signalosome, eIF3 module containing subunits Rpn3, Rpn5, Rpn6, Rpn7, Rpn9, and Rpn12 (8, 23–26). The width of the helical ribbons is ∼25 Å, which is consistent with the typical helix length observed in TPRs (Fig. S6).

The stoichiometries determined by our MS analysis suggest configurational variation among individual particles beyond those that can be resolved by ML3D classification. To visualize the spatial distribution of the major variations in the 26S proteasome, we calculated a 3D variance map (27) (Fig. 1C). As expected, the variance is low in the density corresponding to the CP. The lid region of the RP also shows a low variance level. The variance is highest in a belt surrounding the AAA-ATPase in the RP. It is known from structural studies that AAA-ATPases typically undergo large conformational changes as part of their ATP cycle (28). However, the structural variation extends to areas adjacent to the hexameric AAA-ATPase. These structural changes can probably be mostly attributed to the substoichiometrically bound proteins Rpn10, Rpn13a/b, and Uch2. Indeed, when comparing a reconstruction of 26S proteasomes from an Uch2 deletion strain to the wild-type map, we could not observe significant differences, which is probably due to the fact that the region of Uch2 binding is highly variable.

Classification of the RP.

To obtain further insights into the different 26S proteasome conformations we grouped the particles into different classes to localize structural variations specifically in the belt around the AAA-ATPases (Fig. S7). Classes derived from focused classification (3D) are consistent with results obtained from 2D analysis (Fig. S8). The analysis further resolves the extra mass: This mass adopts a variety of different shapes, e.g., it protrudes to the lid in one class (Fig. 2). This result would be consistent with Rpn10 positioned in the extra mass because the ubiquitin interacting motif is highly flexible (29). Other classes show significant extra densities (of up to 30–40 kDa) in a region around the AAA-ATPases (e.g., Fig. 2, Fig. S9, Movie S2). The most likely candidates for these extra densities are Rpn13a, Rpn13b, Ub, and Uch2. Taken together, the variance map and the classification results suggest that the distal part of the lid provides a mostly invariable frame, whereas the base and the base–lid interface are sites of high variability. The structural variation (activity) is caused by both structural flexibility of constitutive subunits as well as the reversible association of additional densities, tentatively assigned to PIPs.

Fig. 2.

Fig. 2.

Classification of S. pombe 26S proteasomes reveals variations in the RP. Five class averages obtained by focused classification (CP in red, AAA-ATPase-ring in blue, remaining RP in brown) are shown in A. Major differences of each class to the overall average correlate to the localization of the variance map shown in Fig. 1 (B) Additional density in green; (C) missing density in yellow. As expected from ML3D classification, a subset of 26S proteasomes is lacking the extra mass (class 4). Further classes show additional densities in the region between base and lid, close to the upper AAA-ATPase ring (classes 2, 3, and 5).

Localization of Rpn11.

Using the C-terminal 3x-FLAG-epitope of Rpn11, we labeled Rpn11 with an anti-FLAG antibody (Ab) (Fig. 3). We acquired images of 26S proteasome incubated with the anti-FLAG Ab by cryo-EM. Micrographs displayed typical 26S proteasome “twin” configurations (i.e., two holocomplexes aligned with their long axes), as well as isolated 26S proteasomes (Fig. S2). The 2D class averages clearly indicate a distinct density above the lid region, which is not present in the unlabeled proteasome (Fig. 3A). Because the twins adopt a single spatial orientation in ice, we reconstructed single 26S proteasomes to map the Ab density in 3D. Comparison of the 17.4-Å resolution map with the map of the unlabeled 26S proteasome reveals a rather blurred additional density indicative of a high degree of flexibility (Fig. 3B). This density can be mainly attributed to the structural variability of the 17-residue C-terminal linker to the 3x-FLAG epitope, the presence of three FLAG epitopes, and the intrinsic variability of the Ab. Thus, the Ab density maps the C terminus of Rpn11 with relative low resolution: We estimate that the Rpn11 C terminus is located within a radius of ∼80 Å around the center of the Ab mass (Fig. 3C).

Fig. 3.

Fig. 3.

Monoclonal anti-FLAG antibodies bind to the Rpn11–C-terminal 3xFLAG-tag in intact 26S proteasomes. After incubation of proteasomes with anti-FLAG-antibody, extra densities are clearly visible in 2D class averages (A, Top Left: control). After classification and 3D reconstruction (B) the center of mass of the antibody can be clearly seen on the top of the RP lid (green). The cross-link between Lys49 in the coiled-coil region of Rpt3 (C, yellow) and Lys281 of Rpn11 has an estimated length of 20 Å and allows mapping Rpn11:Lys281 within a sphere of 20-Å radius (yellow sphere). A sphere of ∼80-Å radius (dotted line) with its central point in the center of mass of the antibody depicts the possible localization of the IgG bindings site.

To obtain further data on protein–protein proximity, we subjected the sample to cross-linking with disuccinimidyl suberate and identified the cross-linked lysine residues by MS (16–18). Among the identified cross-links, one intersubunit interaction involved Rpn11: Lys49 located in the coiled-coil region of Rpt3 is in proximity to Lys281 in the C-terminal domain of Rpn11 (Table 1). Previous tandem affinity purification experiments revealed Rpt3/Rpt6/Rpn8/Rpn11 forming a subcomplex (30), consistent with the proximity of Rpt3 and Rpn11. The placement of Rpn11 close to the AAA-ATPases is not unexpected: The deubiquitylating activity of Rpn11 is ATP dependent, but Rpn11 is not an ATPase itself, which implies cooperativity of substrate translocation and deubiquitylation (13, 14). Thus, the cross-linking data suggest that part of the Rpn11 C-terminal domain is accessible from inside the cavity surrounding the AAA-ATPase, whereas the C terminus itself projects to the periphery of the RP. There are 29 residues between Lys281 and the C terminus, which are sufficient to bridge the ∼30-Å space between the proteasome mouth and the periphery. Interestingly, it is exactly these ∼30 C-terminal residues that have been shown to be essential for the role of Rpn11 in mitochondrial biogenesis (31).

Table 1.

Crosslinked lysines in the RP of S.pombe proteasomes

Residue 1 Residue 2 C_α_-C_α_, Å
Intra-ATPase
Rpt4:71 Rpt5:124 15
Rpt4:318 Rpt5:217 9
Rpt5:77 Rpt4:38 15
Rpt3:299 Rpt6:357 16
Rpt4:71 Rpt5:124 15
Rpt3:186 Rpt4:164 24
Rpt5:72 Rpt4:41 13
Rpt4:37 Rpt4:41 6
Rpt3:242 Rpt3:299 17
Rpt5:72 Rpt4:38 10
Intra-ATPase cavity
Rpt5:124 Rpt3:207 34
Rpt5:124 Rpt4:205 31
Rpt6:84 Rpt6:218 33
Rpt1:120 Rpt4:205 35
Rpt1:124 Rpt3:207 39
ATPase—20S
alpha4:168 Rpt1:422 17
alpha4:168 Rpt1:426 22
alpha2:176 Rpt6:390 17
alpha2:176 Rpt6:347 16
ATPase—lid
Rpt3:49 Rpn11:281 NA

Based on the Ab-labeling and the cross-linking data, we can approximately map the C-terminal domain of Rpn11. From the cross-link Rpt3:Lys49-Rpn11:Lys281, we conclude that these two lysine residues are ∼20 Å apart from each other; the C-terminal domain of Rpn11 is placed at the intersection of the sphere centered on the FLAG-Ab and a sphere around Lys49 of Rpt3 (Fig. 3C). At this point, we cannot precisely map the deubiquitylating site, which is located in the N-terminal MPN (Mpr1/Pad1 N-terminal) domain. However, it appears reasonable that it is accessible from the “pharynx” between base and lid, such that the placement of Rpn11 near the mouth of the AAA-ATPase is ideally positioned for Ub removal from substrates immediately prior to substrate translocation into the CP.

Quaternary Structure of the CP–AAA-ATPase Subcomplex.

We previously suggested a model for the quaternary structure of the AAA-ATPase hexamer and its position on the CP based on the cryo-EM map of the D. melanogaster 26S proteasome and protein–protein interactions from the literature (22). The predicted AAA-ATPase topology, Rpt1/Rpt2/Rpt6/Rpt3/Rpt4/Rpt5 was recently confirmed by disulfide engineering (32). Our suggested positioning of the AAA-ATPase hexamer is however not in accordance with refs. 33 and 34: From interactions of the isolated C-terminal peptides of Rpt2 and Rpt5 with the CP subunits, a different CP-AAA-ATPase topology was suggested, which is not compatible with our model and the protein–protein interactions underlying it.

The cross-linking and MS data revealed several interactions within the AAA-ATPase and between CP and the AAA-ATPase subunits (Table 1). We could identify four different cross-links between Rpt and CP subunits: two different residue pairs link Rpt1-α4 and Rpt6-α2, respectively (Fig. 4A). All four cross-links are in excellent agreement with our model. In fact, no other AAA-ATPase rotation (Fig. 4B) can fulfill these restraints. Thus, our cross-linking data derived from the fully assembled 26S proteasome comply with the published protein–protein interaction data, such as Rpt2 binding to α4 (35, 36), as well as genetic data suggesting Rpt2-α3 interaction (37), but they do conflict with the interactions reported for the synthetic C-terminal Rpt peptides (e.g., Rpt5-α3 and Rpt5-α4) (33, 34). This discrepancy might be explained by different binding specificities of the full-length proteins as compared to the short C-terminal peptides.

Fig. 4.

Fig. 4.

Structural and mechanistic depiction of the regulatory particle. Rpt1-α4 and Rpt6-α2 cross-links (A, yellow) between AAA-ATPase (blue) and CP (red) corroborate the previously suggested CP-AAA topology (32). The cross-links between AAA-ATPase subunits (A, green) confirm the Rpt1/Rpt2/Rpt6/Rpt3/Rpt4/Rpt5 topology of the AAA-ATPase hexamer. Cross-links in the AAA cavity are indicative of inducible conformational changes of central rings (A, magenta). (B) Top-view visualizing the positioning of the AAA-ATPase ring (t1–t6, center of AAA ring: white asterisk) relative to the α-ring of the CP (α1–α7, center: black asterisk). (C) Integrated model of the RP and α-ring of the CP: Ubiquitylated substrates are recognized at the RP (e.g., Rpn10 or Rpn13) of the 26S proteasome, deubiquitylated (e.g., Rpn11 or Uch2) for Ub recycling and translocated (Rpt1–6) through the AAA-ATPase cavity and the opened gate (Rpt2, Rpt5) of the CP for degradation (extra mass, cyan; Rpt1–6, as in B; Rpt3–Rpn11 cross-link, as in Fig. 3).

Cross-Linking Data Suggest a Similar Translocation Mechanism as in HslU.

We identified five cross-links connecting residues located in the AAA-ATPase channel. All of these cross-links connected residues in the proximal ring of the N-terminal domains and the highly conserved Ar-Φ loop located in the AAA ring (Fig. 4A). The distances of these residues in our AAA-ATPase model (∼35 Å) exceeded the distance that is typically bridged by disuccinimidyl suberate (< 20 Å). Indeed, it has been suggested that PAN and the proteasomal AAA-ATPases possess a translocation mechanism similar to that of heat shock locus U (HslU) (38). In HslU, the Ar-Φ loop undergoes substantial conformational changes upon ATP hydrolysis, which is believed to enable it to pull the substrate through the pore (28). The identified cross-links suggest that the Ar-Φ loop of the AAA-ATPase of the 26S proteasome also displays a structural flexibility similar to that observed with HslU.

Conclusions

The 26S proteasome from S. pombe is constituted of the canonical subunits α1–7, β1–7, Rpt1–6, Rpn1–9, and Rpn11–12, whereas the subunits Rpn10, Rpn13a/b, Uch2, and Ubp6 are present in significant, yet varying, amounts. In D. melanogaster and in S. pombe, the occupancy of the extra mass derived from the class averages correlated to the abundance of Rpn10 as determined by MS. Three-dimensional reconstruction of vitrified proteasomes yields an EM map with a resolution of 9.1 Å. Patches with the characteristic features of short α-helical repeats become discernible, supporting the prediction for large portions of many RP subunits to adopt α-solenoidal folds. Ab labeling and the cross-linking data suggest that the C-terminal domain of Rpn11 is placed near the opening of the AAA-ATPase cavity, allowing efficient access to its substrate and release of Ub. At the base–lid interface, a belt of high variance surrounds the AAA-ATPase. The distal part of the lid complex, however, is rather invariable, thus providing a rigid frame or roof. Classification of particles according to the belt of major variance is indicative of the reversible binding of PIPs, Ub, or substrate. With the availability of much larger datasets than used in this study, allowing an exhaustive classification and/or the use of different biochemically defined proteasome-substrate states, it will be possible to analyze this belt “high activity” in more detail.

Materials and Methods

Protein Purification and Analysis.

Purification of intact 26S proteasomes containing a 3xFLAG-tag at the C terminus of Rpn11 were purified from S. pombe (972 h+ rpn11∷RPN11-3xFLAG-His3) essentially as described (39) (SI Text). In short, cleared cell lysates were incubated with anti-FLAG M2 agarose beads (Sigma), washed, and the bound proteins eluted with 3xFLAG peptide (Sigma). The elute was applied to a sucrose gradient and fractions were tested for protease activity. The fraction from sucrose gradients with the highest proteolytic activity was analyzed by MS as described (9), used for cryo-EM and chemical cross-linking (SI Text). In brief, purified proteasomes were concentrated, cross-linked with disuccinimidyl suberate d0/d12 (Creativemolecules, Inc.), and trypsin digested (Promega). Liquid chromatography-MS/MS analysis was carried out on an LTQ Orbitrap XL mass spectrometer (Thermo Electron). Data were searched using xQuest (40) and further analyzed manually. For Ab labeling, purified proteasomes were incubated with M2 Anti-FLAG Ab (Sigma) for 1 h at 4 °C.

EM.

Focal pairs with a nominal defocus of 1–3 μm were recorded on an Eagle CCD camera using an FEI Tecnai F20 microscope at 200 kV (final magnification, 63,500×; object pixel size, 2.2 Å) in a semiautomated manner using the SerialEM software package (41). The contrast transfer function was determined and micrographs were deconvoluted by phase flipping and compensation for the modulation transfer function of the CCD camera (42). From ∼21,000 micrographs, ∼320,000 particles were selected for 3D reconstruction (26S ∼270,000, 26S-Ab ∼50,000).

Reconstruction, 3D Variance, and Classification.

The initial 3D model for refinement was the D. melanogaster 26S proteasome density filtered to 4 nm. Data were sorted using ML3D classification (Fig. S3) (21) and densities were further refined as described in ref. 43. The temperature factor was determined (44) and applied to the final reconstruction (∼-250 _Å_2). The variance map was calculated according to ref. 27. For focused classification, angularly refined particles were C2 symmetrized, masked, and split into 30 iteratively optimized groups (Fig. S7). The position and radius of the mask was determined by evaluating the variance map.

Modeling.

Comparative models of the S. pombe 20S proteasome and the AAA-ATPases were built as described previously (22). The positions of CP and AAA-ATPase hexamer were locally refined in the experimental map using the University of California, San Francisco Chimera (45).

Supplementary Material

Supporting Information

Acknowledgments.

We thank Y. Saeki (Tokyo Metropolitan Institute of Medical Science) for providing a plasmid encoding the 3xFLAG-His-Tag and M. Shravan for providing S. pombe support. E.S. is a recipient of a Japan Society for the Promotion of Science Research Fellowship for Young Scientists. F.F. thanks Human Frontier Science Project Organization for a career development award. This work was supported in part by funding from the European Union Seventh Framework Program PROSPECTS (Proteomics Specification in Space and Time Grant HEALTH-F4-2008-201648).

Footnotes

The authors declare no conflict of interest.

References

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supporting Information