Direct visualization of degradation microcompartments at the ER membrane

Significance Endoplasmic reticulum-associated degradation (ERAD) is an essential process that removes misfolded proteins from the ER, preventing cellular dysfunction and disease. While most of the key components of ERAD are known, their specific localization remains a mystery. This study uses in situ cryo-electron tomography to directly visualize the ERAD machinery within the native cellular environment. Proteasomes and Cdc48, the complexes that extract and degrade ER proteins, cluster together in non–membrane-bound cytosolic microcompartments that contact ribosome-free patches on the ER membrane. This discrete molecular organization may facilitate efficient ERAD. Structural analysis reveals that proteasomes directly engage ER-localized substrates, providing evidence for a noncanonical “direct ERAD” pathway. In addition, live-cell fluorescence microscopy suggests that these ER-associated proteasome clusters form by liquid–liquid phase separation.

. Functional domain organization of ERAD components in Saccharomyces cerevisiae, Homo sapiens, and Chlamydomonas reinhardtii. All protein sequences were retrieved from NBCI (for yeast and human proteins) or Phytozome (1) (for Chlamydomonas) and submitted to SMART (Simple Modular Architecture Research Tool, http://smart.embl-heidelberg.de) for domain detection. Chlamydomonas proteins with gene names highlighted in color have large stretches of low complexity, which are not supported by GRAbB-generated in silico cDNAs (SI Appendix, Fig. S2) and may denote an incorrect gene model. All orthologous proteins are shown in the order: yeast, human, Chlamydomonas.

Figure S1 (continued). Functional domain organization of ERAD components in
Saccharomyces cerevisiae, Homo sapiens, and Chlamydomonas reinhardtii. All protein sequences were retrieved from NBCI (for yeast and human proteins) or Phytozome (1) (for Chlamydomonas) and submitted to SMART (Simple Modular Architecture Research Tool, http://smart.embl-heidelberg.de), for domain detection. Chlamydomonas proteins with gene names highlighted in color have large stretches of low complexity, which are not supported by GRAbBgenerated in silico cDNAs (SI Appendix, Fig. S2) and may denote an incorrect gene model. All orthologous proteins are shown in the order: yeast, human, Chlamydomonas. Figure S2. Resolving incorrect gene models for the Chlamydomonas NPL4, HRD1, Cue1 and EDEM1 orthologs. We used GRAbB (2) to assemble the full coding sequences corresponding to the Chlamydomonas NPL4, HRD1, Cue1 and EDEM1 genes. The sequences were assembled from a paired-end RNAseq experiment with high sequence coverage (run SRR2132411 from the Short Read Archive SRA at NCBI), using only the parts of the genes that code for expected functional domains as seeds. The final GRAbB sequences and deduced protein sequences (with functional domains) are shown in the grey boxes below each GRAbB seed.

Figure S3. Cytosolic proteasome puncta dynamically assemble. (A-D) Time series of live
Chlamydomonas mat3-4 cells expressing Rpn11-mVenus. 3D Z-stacks were acquired once per minute by widefield deconvolution fluorescence microscopy. The first image of each time series is an overlay of Rpn11-mVenus (green) and chlorophyll autofluorescence (magenta) at 0 min, displaying the apical (api) and basal (bas) sides of the cell. The greyscale images in each timeseries track Rpn11-mVenus over 14 minutes, showing maximum intensity projections through all Z-slices containing puncta. Small yellow arrowheads indicate newly assembled puncta, with clear signal that appears and can be tracked over time, despite significant photobleaching during the time series. The large red arrowhead in D is a fusion event between the newly assembled punctum and a preexisting punctum. See Movie 1.     Comparing the distribution of measured Cdc48 positions (red) with randomly simulated Cdc48 positions (grey) reveals a highly significant non-random accumulation of Cdc48 at the cluster proteasomes. Kolmogorov-Smirnov test for cluster proteasomes: h=1 (different from random) with p=0.001 (***); for non-cluster proteasome: h=0 (same as random) with p=0.0849. (C) Distances of cluster-associated proteasomes (red) and Cdc48 (yellow) to the ER membrane, showing a similar distribution. (D) The cumulative distribution of distances between non-cluster Cdc48 and the ER membrane is indistinguishable from the distribution with randomly positioned Cdc48. Thus, while 80% of ER-proximal Cdc48 is found outside the degradation microcompartments, it appears to be randomly distributed along the ER membrane instead of clustered. (E) Distance analysis of each proteasome to all cytosolic ribosomes was performed for cluster proteasomes only (left) and for cytosolic non-cluster proteasomes (right), showing no correlation.     Left column: ER-proximal proteasomes within the degradation microcompartments, with 19S caps that are in the substrate-processing conformational state and are bound to an orange extra density that originates from the ER membrane ( Fig. 5B-C). The extra density is bound to the proteasome's Rpt AAA-ATPase ring near the cap's Rpn1 subunit. Middle-left column: substrateprocessing state proteasomes bound to poly-GA neurodegenerative aggregates, with the blue extra density corresponding to the engaged aggregate, a known substrate (EMD-3915) (7). This extra density is also bound to the AAA-ATPase ring near Rpn1. Middle-right column: proteasomes localized to the nuclear pore complex, with pink extra density corresponding to a linker protein and the inner nuclear membrane (EMD-3936) (8). The linker density binds the Rpn9 subunit of the proteasome's 19S cap. Right column: overlay of the three averages, fitted to each other in UCSF Chimera software (9). The density originating from the ER (orange) occupies the same binding position on the proteasome as the neurodegenerative aggregate substrate (blue). Thus, the orange density is very likely substrate that is engaged by the proteasome. As the orange density emanates from the ER membrane (Fig. 5C), we conclude that these proteasomes are likely engaged with ER-localized substrates.

Figure S14. Effects of pharmacological inhibitors on proteasome localization. (A-D) Live
Chlamydomonas mat3-4 cells expressing the tagged proteasome subunit Rpn11-mVenus, imaged in 3D by widefield deconvolution fluorescence microscopy. Prior to imaging, cells were treated for 2 hr with either (A) 5 µg/mL tunicamycin to induce acute ER stress, (B, D) 50 µM MG132 to inhibit proteasome function, or (C) 5 µg/mL NMS-873 to inhibit Cdc48 function. Panel D shows enlarged views of the MG132 treatment (top: greyscale, bottom: two-color overlay) to more clearly illustrate the accumulation of bright proteasome puncta at the nuclear envelope. All images are maximum intensity projections. White arrowheads mark cells where the nucleus has been displaced from its central position by swelling of the vacuole, indicating toxic effects of tunicamycin and NMS-873. Scale bars in A and D: 5 µm. (E) Histogram of the number of cytosolic puncta per cell. N= 213, 149, and 192 cells for control, tunicamycin, and MG132, respectively. (E) The intensity of Rpn11-mVenus fluorescence within the cytosolic puncta and at the nuclear envelope, normalized by fold-change over each cell's cytosolic background. Quantification of MG132 was separated into regions of the nuclear envelope that are not occupied by bright puncta (Non-Puncta) and regions that do contain puncta (Puncta). The y-axis is plotted in log2 scale. Error bars show standard deviation. Figure S15. Hybrid Cdc48-proteasome complexes are not found within the cellular tomograms by template matching. Left: the tomograms where searched with templates of a 20S proteasome core particle attached on its ends to one (single-hybrid) or two (double-hybrid) Cdc48 complexes. Right: the particles with the top scoring correlation coefficients were extracted and subjected to reference-free alignment in PyTom software (10). Rather than recovering a hybrid structure, these particles aligned to produce a structure of a normal 26S proteasome. Fig. 5B). The probability of drawing a group of proteasomes with >50% substrate-processing caps from the pool of clustered proteasomes >20 nm from the membrane, evaluated over different sample sizes. In blue, the pool is comprised of ground, processing, and unclassified caps, and the probability was determined by numerical simulations (1000 draws). In red, the probability was analytically calculated from a pool of only ground and processing caps, using an urn model without replacement and order. The arrow indicates the number of caps <20 nm from the membrane in our dataset (sample size= 14). The numerical and analytical results overlap, suggesting that the unclassified caps can be discarded without affecting the probability. Table S1. Homologs to yeast, human, and Arabidopsis thaliana ERAD components identified in the Chlamydomonas reinhardtii genome (11,12). Chlamydomonas and Arabidopsis gene accessions numbers are from the Phytozome platform (1). Accompanies Figs. S1-S2.

S. cerevisiae H. sapiens
A. thaliana C. reinhardtii Description

Movie 3. Proteasomes and Cdc48 form ribosome-excluding degradation microcompartments
that contact a specialized patch on the ER membrane. Sequential sections back and forth through the tomographic volume shown in Fig. 2A (orthographic view), followed by reveal of the 3D segmentation from Fig. 2D (perspective view). Golgi: dark grey, ER: light grey, other organelles: white, proteasomes: red, Cdc48: yellow, free cytosolic ribosomes: light blue, membrane-bound ribosomes: dark blue. A close-up view rotates around the microcompartment, then the proteasomes are removed to show the ribosome-free patch on the ER membrane, then free cytosolic ribosomes are removed to show the clustering behavior of the proteasomes.

Movie 4. Accurate identification of individual proteasome and
Cdc48 structures within the cell. Z-slices through the tomogram region shown in Fig. 2B-C, with transparent red and yellow silhouettes corresponding to the mapped in subtomogram averages of proteasomes and Cdc48, as shown in Fig. 2D. A clipping plane has been used to only show the averages that are just above each tomogram Z-slice, allowing comparison between the placed averages and the tomogram. Accompanies Fig. S5.