A molecular census of 26S proteasomes in intact neurons

See allHide authors and affiliations

Science  23 Jan 2015:
Vol. 347, Issue 6220, pp. 439-442
DOI: 10.1126/science.1261197

A detailed look at proteasomes in situ

The 26S proteasome is a protein machine that degrades intracellular proteins in the cytosol. The proteasome is critical for protein quality control and for the regulation of numerous cellular processes in eukaryotic cells. The structure of isolated proteasomes is well established, but how intact proteasomes look within the cell is less clear. Asano et al. used an improved approach to electron cryotomography to look at proteasomes in intact hippocampal neurons. Their analysis suggests that these cells only use about 20% of their proteasomes in an unstressed state, which leaves significant spare capacity to deal with proteotoxic stress.

Science, this issue p. 439


The 26S proteasome is a key player in eukaryotic protein quality control and in the regulation of numerous cellular processes. Here, we describe quantitative in situ structural studies of this highly dynamic molecular machine in intact hippocampal neurons. We used electron cryotomography with the Volta phase plate, which allowed high fidelity and nanometer precision localization of 26S proteasomes. We undertook a molecular census of single- and double-capped proteasomes and assessed the conformational states of individual complexes. Under the conditions of the experiment—that is, in the absence of proteotoxic stress—only 20% of the 26S proteasomes were engaged in substrate processing. The remainder was in the substrate-accepting ground state. These findings suggest that in the absence of stress, the capacity of the proteasome system is not fully used.

The ubiquitin proteasome system (UPS) plays a key role in many cellular processes, including cell cycle control and proteostasis (1). The 26S proteasome executes the processive degradation of proteins marked for destruction by the attachment of polyubiquitin chains (2, 3). Malfunctions of the UPS result in the accumulation of damaged or misfolded proteins and are implicated in a number of diseases (4). The 26S proteasome is involved in many fundamental processes in neurons (5), including the regulation of synaptic strength and of the presynaptic (6) as well as postsynaptic proteome (7). However, the exact molecular mechanisms underlying these phenomena remain to be elucidated, and quantitative structural studies performed in situ should contribute to a deeper understanding.

Cellular electron cryotomography combines the power of high-resolution three-dimensional (3D) imaging with the best possible structural preservation (8). This is particularly important for fragile structures such as the 26S proteasome. Electron cryotomography performed with the recently developed “Volta” phase plate (9) allows the identification and localization of molecular structures in the cellular environment with high fidelity.

The 26S proteasomes could readily be identified by visual inspection (Fig. 1A, insets). Nevertheless, for a systematic and comprehensive statistical analysis of the 26S proteasomes, an automated template-based search was performed (10). To enable detection of both single-capped and double-capped 26S proteasomes, a double-capped 26S particle [Electron Microscopy Data Bank (EMDB), 2165] (11) with one regulatory particle (RP) removed computationally was low-pass filtered to a resolution of 40 Å and used as a template for an initial search of cytosolic volumes. Particles identified by cross-correlation coefficients (CCC) above a generous cut-off value (Fig. 1B) showed an isotropic orientation distribution (fig. S1) and were validated by visual inspection. Both single- and double-capped particles yielded almost identical CCCs (mean CCC ~ 0.28 for both populations) (Fig. 1B), indicating that the template-based search did not bias the statistics toward one or the other assembly state. In addition, we performed a search with a 20S-Cdc48 template (12); we were unable to distinguish a class populated by such a complex.

Fig. 1 Raw image and quality assessment.

(A) Slice from a representative cryotomogram of a cultured hippocampal neuronal cell. One single-capped (middle) and two double-capped 26S proteasomes (left and right) are indicated by a red frame. The insets display magnified 1.7-nm Z slices through each proteasome volume. Scale bars: overview, 100 nm; insets, 25 nm. (B) Histogram depicting the CCCs of all single-capped (blue bars) and double-capped (red bars) 26S proteasomes. The histograms were fitted with a Gaussian distribution (blue: single-capped 26S proteasomes, mean CCC ~0.28; red: double-capped 26S proteasomes, mean CCC ~0.28). The dashed gray line indicates the average threshold (~0.12), below which template matching results have been discarded. (C) Histogram showing the FCR of each single-capped 26S proteasome [blue in (B)] and the atomic model of the S. cerevisiae single-capped 26S proteasome (PDB, 4cr2). The mean cross resolution is at 47 Å.

The CCC values had an approximately Gaussian distribution (Fig. 1B), suggesting that almost all (>3σ) 26S proteasome particles present in the tomograms were detected. In contrast, we could not identify 26S proteasome particles reliably in cryotomograms acquired without the Volta phase plate (9). To quantify the signal quality of particles acquired with the phase plate, we compared the single-capped 26S proteasomes to a high-resolution map obtained by cryoelectron microscopy (cryo-EM) single-particle analysis [Protein Data Bank (PDB), 4cr2] (13) by Fourier cross-resolution (FCR). The resolution of the majority of the subtomograms, each depicting an individual single-capped 26S proteasome particle, ranged from 35 to 50 Å (Fig. 1C).

Previous analyses of isolated proteasomes suggest that double-capped proteasomes, single-capped proteasomes, uncapped proteasomes, and different assembly intermediates all coexist in the cell (14, 15). However, the relative abundance of the different assemblies is unclear because cell lysis and the subsequent purification steps may cause partial disassembly. To analyze the structural heterogeneity of our in situ particle ensemble, we subjected it to statistical analysis by subtomogram classification and averaging. First, we coherently aligned and averaged the subtomograms without using any external starting model to avoid template or reference bias (16) (figs. S2 and S3). The resulting subtomogram average showed a single-capped 26S proteasome. However, the global average, as well as the corresponding variance map, exhibited considerable heterogeneity at the uncapped end of the core particle (CP) (Fig. 2A). Both the faint density at the CP end and the variance maps suggested the coexistence of single- and double-capped 26S proteasomes in our data set, consistent with in vitro observations (15, 17). We used a variance-based 3D focused classification (18) approach to divide the particle set into two structurally homogeneous classes, as well as a background bin comprising hypervariable subtomograms (fig. S4). The background bin comprised 20S core particles, (dis-)assembly intermediates and, possibly, a few remaining false positives. The particles assigned to the background bin were discarded after each classification step because the various molecular species contained in this bin were too few to yield meaningful averages. The main difference between the two homogeneous classes was the presence or absence of the second RP (Fig. 2B). Thus, the first class depicts an average of double-capped 26S proteasomes, whereas the second represents the single-capped population. Double-capped 26S proteasomes were far less abundant (425 particles, 27%) than the single-capped complexes (1156 particles, 73%), very similar to the ratio observed in native gel densitometry of rat cortex (19). On average, 23 single- or double-capped 26S proteasomes were found per tomogram, which translates into a mean cytosolic concentration of ~190 nM, a value close to the 26S proteasome subunit concentration previously reported for Saccharomyces cerevisiae (140 to 200 nM) (20).

Fig. 2 Subtomogram averaging and classification workflow.

(A) Global average from all particles displayed as center slice (left) and isosurface view (middle), as well as isosurface representation of the variance map (right; variance in violet). (B) First classification round separating single-capped from double-capped 26S proteasomes. The difference map between the two classes (right; difference shown in red) shows the additional RP. The in silico cutting was performed along the blue dashed line between the two 20S β rings. (C) Resulting average displayed as slice (left) and isosurface (middle), as well as corresponding variance map (right). (D) Second classification of the cut particles into two classes. Class 2 (left, light blue isosurface) and Class 1 (middle, light green isosurface) mainly differ at Rpn1, Rpn6, and the substrate entry location (right, red isosurface). Scale bars for the slices, 25 nm.

To reveal differences in the two RPs of double-capped 26S proteasomes, all proteasome particles were subjected to in silico cutting between the two β rings of the 20S core particle (Fig. 2C). The two halves of double-capped particles were then treated as separate subtomograms. Because of the larger effective number of subtomograms, the resolution of the subtomogram average depicting a single-capped 26S proteasome improved to 31 Å. The corresponding variance map indicated a high variability for distinct RP subunits, such as Rpn6 and Rpn1 (Fig. 2C). Moreover, the variance was high in a region surrounded by Rpn1, Rpn2, Rpn10, and Rpn11, where there is no density in single-particle reconstructions of purified 26S proteasomes (11, 13, 21, 22). We then separated the truncated particles into two major classes, class 1 (at 27 Å resolution) and class 2 (31 Å), which were found in both single- and double-capped 26S proteasomes (Fig. 2D). The size of the two classes was unbalanced: Class 1 accounted for 80% of all RPs (1367 particles) and class 2 for only 20% (339 particles). The difference map between the two classes (Fig. 2D) was in good agreement with the variance map, which was most intense for Rpn1, Rpn6, and the area enclosed by Rpn1, Rpn2, Rpn10, and Rpn11 (Fig. 2D).

A very large cryo-EM data set of isolated S. cerevisiae 26S proteasomes has recently revealed coexisting conformations of purified 26S proteasomes referred to as s1, s2, and s3 (13). Substrates most likely primarily bind to the most abundant low-energy state s1; s2 mediates their tighter binding (“commitment”) and activates the Rpn11 deubiquitylating module; and s3 finally enables translocation of substrates into the CP (13). To compare the mammalian in situ classes to the S. cerevisiae in vitro conformations, atomic models of the S. cerevisiae 26S proteasome subunits were rigidly fitted into the densities of class 1 and class 2 (Fig. 3A and fig. S5). Class 1 was most similar to s1 (substrate accepting) [root mean square deviation (RMSD) ~ 8.9 Å], whereas class 2 most closely resembled s3 (translocating) (RMSD ~ 7.4 Å) (fig. S5). In vitro, the 26S proteasome in complex with a polyubiquitylated substrate, GFP-I27-Ub4, adopts a conformation that is essentially identical to s3 and hence similar to class 2 (23). Thus, class 2 represents proteasomes that are engaged in substrate processing. Accordingly, we refer to the major conformational states represented in class 1 and class 2 as “ground state” (GS) and “substrate-processing state” (SPS), respectively. Only 20% of 26S proteasomes in the analyzed neurons were in a SPS. We generated a 3D atlas of the 26S proteasome within a cell with their individual orientations and conformation states (Fig. 3B).

Fig. 3 Rigid body fitting of atomic subunit models into EM densities.

(A) Fitted atomic models of the S. cerevisiae 26S proteasome subunits in the GS and SPS EM densities (Fig. 2D). The subunits Rpn9/5/6/7/3/12 are colored in different shades of green, Rpn8/Rpn11 in light/dark magenta, Rpn10 and Rpn13 in purple, Rpn1 in brown, Rpn2 in yellow, the AAA-ATPase hexamer in blue, and the CP in red (13). Red arrows indicate selected Rpns and the unassigned density in the SPS. (B) Single-capped and double-capped 26S proteasomes displayed as green (GS) and blue (SPS) isosurfaces overlaid on a slice of a representative tomogram. Scale bar, 500 nm. (C) Chart showing the overall distribution of single-capped and double-capped 26S proteasomes (left) and of the different states within single-capped (middle) and double-capped (right) 26S proteasomes.

Next, we analyzed whether the conformations of the two RPs in each double-capped 26S proteasomes correlated with one another. This analysis matched the observed distribution with the values predicted for uncorrelated states: GS-GS, 64% (measured) versus 63% (predicted); GS-SPS, 32% versus 33%; and SPS-SPS, 4% versus 4% (Fig. 3C), similar to previous single-particle cryo-EM studies (13, 23), favoring the notion that both RPs can act independently.

The GS density (Fig. 4A) showed a high variance in the vicinity of the Rpn1 and Rpn13 subunits. The corresponding subtomograms were classified into four classes to deconvolute the underlying major structural differences (fig. S6). The first class (GS1, 24%) essentially did not differ from the density simulated with the fitted atomic model of a complex lacking the ubiquitin receptor Rpn13 (24) (Fig. 4A). Rpn13 was also absent in class GS2 (18%) and was found only in ~58% of all proteasome particles, in agreement with a dynamic binding of Rpn13 (25). Compared with GS1, GS2 exhibits a prominent additional density associated with Rpn1, which functions as a hub for a number of proteasome-interacting proteins (PIPs) (2). The remaining two ground-state classes, GS3 (34%) and GS4 (24%), showed density adjacent to Rpn2 that colocalizes with Rpn13 in S. cerevisiae proteasomes (11, 13, 21, 22). This density was less prominent in GS3 but substantially larger in GS4, most likely due to the presence of PIPs.

Fig. 4 Subclassification within GS and SPS.

(A) Classification of GS class into four subclasses. (Left) Variance map (top), the GS average (middle) and corresponding down-sampled atomic model (bottom). (Center) Subclasses GS1 to GS4 displayed as isosurfaces; their relative abundances are indicated. (Right) Differences between GS1 to GS4 and the down-sampled atomic model (red). (B) Classification of SPS class into three subclasses, SPS1 to SPS3, displayed as in (A).

The subtomograms contributing to the SPS were separated into three classes (Fig. 4B). The differences of all three classes, when compared to the atomic model, are located in a region above the “mouth” of the proteasomal ATPase Associated with diverse cellular Activities (AAA)–adenosine triphosphatase (AAA-ATPase). In single-particle studies, substrate was localized where we observed the additional masses in situ (23). However, the additional densities in each of the three in situ classes correspond to ~150 kD to 250 kD, which is substantially larger than a typical proteasomal substrate or the one used in the in vitro studies (23). In fact, one would anticipate a high degree of structural variability of the proteasome-associated densities if they corresponded to substrates only, given their heterogeneous nature and different stages of processing. Most likely, the proteasome-associated densities in the SPS classes depict mostly substrate-processing cofactors of the 26S proteasomes, such as deubiquitylating enzymes and E3 ubiquitin ligases, which are frequently found in association with 26S proteasomes (2).

Advances in technology, such as direct detectors and the contrast-enhancing Volta phase plate open up opportunities for structural studies in situ and visual proteomics (26). Fragile and highly dynamic macromolecular complexes can be studied in their functional and unperturbed cellular environments, providing quantitative information about their states of assembly and conformation. A challenge for the future is to correlate this information in a systematic manner with topographic information about the cellular environments they inhabit.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S6

References (2741)

References and Notes

  1. Acknowledgments: We thank Y. Chen for technical assistance and troubleshooting and P. Unverdorben for support in modeling and discussions. The research leading to these results has received funding from the European Commission under FP7 GA no. ERC-2012-SyG_318987-ToPAG. The work was additionally supported by the Deutsche Forschungsgemeinschaft Excellence Cluster CIPSM and SFB 1035 (both to W.B.) and FO 716/3-1 (to F.F.). Data availability: The cryo-EM maps for the GS and SPS were deposited into the EMDB with the accession codes EMD-2830 and EMD-2831, respectively. FEI Company has submitted a patent for the Volta phase plate, which is currently pending. Information on materials and methods is available on Science Online
View Abstract

Navigate This Article