Several proteasome subunits contain disordered regions
Proteasomal degradation begins at disordered regions within proteins. Regions within several subunits of the regulatory particle are not defined in the recent high-resolution structures of yeast proteasome (e.g., PDB 6FVT, and 6FVU), suggesting that these regions are disordered18. For example, six of the non-ATPase subunits end in stretches of 20 residues or more that lack a defined structure (Table 1). We first asked which of these tails, if any, were long enough to reach the entrance of the translocation channel and thus could make the proteasome susceptible to accidental self-digestion and destruction.
We built a complete atomic model of the yeast proteasome based on recent high- resolution structures (PDB 6FVT and 6FVU) and added the unresolved residues in the tails using homology modeling. The tail structures were built by assembling amino acids in random order from di-peptide conformations found in a library of non-redundant high- resolution protein structures19,20 (see Methods, Fig. 1A, and Supplementary Data 1). The length of the assembled tails was calculated as the distance between the Cα atom of the last resolved residue at one end of the sequence and the Cα atom at the other end of the tail (Supplementary Fig. 1A). We also calculated the direct through-space distance as the shortest possible connection between the Cα atom of the last resolved residue and entrance to the translocation channel (Supplementary Fig. 1B). The ratio of these two distances provides some measure of the likelihood of a tail reaching the entrance to the translocation channel: the larger the ratio, the more likely the tail can reach the channel entrance, with 1.0 the theoretical threshold, though in reality, steric constraints imposed by the rest of the proteasome structure could make the true threshold somewhat higher than this (Fig. 1B). The C-terminal disordered tails of Rpn1, Rpn2, and Rpn13 are either too short or too distantly located to reach the channel entrance (ratio of max tail length to distance from the translocation channel < 0.5 for Rpn2, Rpn13, and 0.7 for Rpn1, Fig 1B). On the other hand, the disordered tails of some subunits were either long enough (Rpn10 ratio ≥ 1) or almost so (Rpn3, and Rpn8, ratio ≥ 0.8) to reach the entrance of the translocation channel (Fig. 1B and Table 1).
We chose three subunits with disordered C-terminal tails to investigate further. We chose Rpn10 and 13 because they are substrate receptors and their binding partners become degraded routinely, suggesting that they themselves might be susceptible to degradation. We chose Rpn3 because it is an essential structural component of the proteasome complex and its disordered tail is potentially within reach of the entrance channel. Finally, we focused on C-terminal tails because our experimental tools are better suited for their analysis (see below).
Rpn13 escapes degradation because it does not access the translocation channel
The Rpn13 subunit is one of the three ubiquitin receptors; substrates bound to Rpn13 are degraded while the receptor itself escapes. Although the last 23 amino acids of Rpn13 appear to be disordered, the subunit is located apically, near the tip of the 19S regulatory particle, and unlikely to reach the entrance to the substrate channel and the protein translocation motor (Figs. 1, 2A and Table 1). If Rpn13 escaped degradation simply because of its location and failure to engage the motor, extending its C-terminus by an artificial tail should lead to its destruction (Fig. 2B). To determine the abundance of Rpn13 and derivatives, we attached a hemagglutinin (HA) tag to its N-terminus and expressed the protein in yeast under the control of a GAL1 promoter from a CEN/ARS plasmid (Fig. 2B). As expected, HA-Rpn13 was easily detected by western blotting (Fig. 2C). Indeed, attaching an artificial tail with a sequence that can be recognized by the proteasome (Su9,21,22) to the C-terminus of HA-Rpn13 led to its degradation (HA- Rpn13-Su9; Fig. 2C), whereas attaching a tail with a sequence that escapes proteasome recognition (HA-Rpn13-SP2,21,22 ) allowed HA-Rpn13 to accumulate just as HA-Rpn13 without the additional tail (HA-Rpn13; Fig. 2C).
Rpn13 with a suitable extended tail was degraded by the proteasome as it was stabilized by treatment with the proteasome inhibitor MG132. Ubiquitination was not required for degradation, presumably because as a proteasome subunit, Rpn13 is by default localized to the proteasome, even without a ubiquitin tag (Fig. 2C, and Supplementary Fig 2A). Attenuating the first enzyme in the ubiquitination pathway (ubiquitin-activating enzyme Uba1), by using a temperature-sensitive strain23, did not affect the abundance of HA-Rpn13-Su9 (Supplementary Fig. 2A), whereas a substrate with the classic N-end degron24,25, which depends on ubiquitination, was stabilized (Supplementary Fig. 2B).
The HA-tagged Rpn13 mutants were incorporated into the proteasome, at least in part, as they were detectable by western blotting after immunoprecipitation of proteasome particle through a FLAG epitope fused to Rpn11 (Supplementary Fig. 3). Degradation of Rpn13 with a tail (Rpn13-Su9) was not due to its over-expression because integration of Rpn13-Su9 at its genomic locus (FLAG-Rpn13-Su9) and expression from its native promoter also led to its degradation, whereas FLAG-Rpn13 without a tail as well as with a tail that is not recognized by the proteasome accumulated (Fig. 2D).
Thus, Rpn13 escapes degradation despite its location on the proteasome in the vicinity of the entrance to the translocation channel because its unstructured region is too short to reach the entrance to the channel and thus cannot be engaged by the proteasome.
Rpn10 and Rpn3 are protected by their amino acid sequence
The Rpn10 is another ubiquitin receptor and again substrates bound to Rpn10 are degraded while Rpn10 itself escapes. Rpn10 also contains a disordered region at its C- terminus and this tail is long enough to reach the entrance of the translocation channel (Fig. 3A, Table 1). This raises the question how Rpn10 escapes degradation.
The proteasome has distinct preferences for the amino acid sequence of the polypeptide at which it initiates degradation. If Rpn10 escaped degradation because its amino acid sequence masks it from recognition, then replacing the relevant region with a sequence that can be engaged by the proteasome should lead to Rpn10’s destruction. To test this model, we attached an HA-tag to the N-terminus of Rpn10 and constructed two further hybrid proteins by replacing Rpn10’s C-terminal tail (amino acids 243-268) with a sequence recognized by the proteasome (Su921,22) or with a sequence that is ignored (SP221,22) (Fig. 3B). Rpn10 and its derivatives were again expressed from a GAL1 promoter on a CEN/ARS plasmid and their abundance determined by western blotting. HA-Rpn10 with its native tail accumulated but replacing the tail with the Su9 sequence led to its degradation (Fig. 3C). It is unlikely that the transplantation of the tail in itself caused Rpn10’s destabilization because the Rpn10 hybrid with a SP2 tail was stable (Fig. 3C). The Rpn10 degradation was by the proteasome and did not require ubiquitination (Fig 3C, and Supplementary Fig. 2A). We observed similar protein levels of HA-Rpn10 mutants when expressed from Rpn10’s native promoter at its genomic locus (Fig. 3D). Thus, Rpn10 escapes degradation even though it is physically accessible to the proteasome’s proteolytic machinery because it is constructed from a sequence that is not recognized by the proteasome.
Is the same mechanism used to ensure the stability of other proteasome subunits? Rpn3 is a structural subunit of the regulatory particle and, just as Rpn10, it contains a disordered C-terminal tail that appears to have the potential to reach the entrance to the translocation channel but remains stable (Fig. 4A). Thus, Rpn3 may also escape degradation because the amino acid sequence of the tail is not recognized by the proteasome. Indeed, HA-tagged Rpn3 was degraded by the proteasome when its C- terminal tail (amino acids 478-523) is replaced with an effective initiation sequence (Su9; Fig. 4B and C) without the need for ubiquitination (Supplementary Fig. 2A). HA- tagged Rpn3 with its native tail or with a tail that is not recognized by the proteasome (SP2; Fig. 4B and C) remain stable. The Rpn3 and Rpn10 derivatives were incorporated into the proteasome, at least to some extent, because they could be detected by western blotting after immunoprecipitation of proteasome particle through a FLAG-tag Rpn11 (Supplementary Fig. 3). Thus, Rpn3 too is protected from degradation by the amino acid sequence of its C-terminal tail.
Unstructured tails of proteasome subunits are not efficient initiation sequences:
As Rpn10 and Rpn3 are protected from degradation because the relevant part of their polypeptide chain is not recognized by the proteasome’s translocation machinery, we asked whether other subunits are also constructed from sequences with similarly stealthy properties? To address this question, we fused the disordered regions from the C-termini of regulatory particle subunits that were at least 20 amino acids long to a model proteasome substrate and tested its degradation by the proteasome in cells. The model proteins were built on a yellow fluorescent protein (YFP) scaffold to simplify their detection in yeast and an N-terminal ubiquitin-like domain (UBL) of Rad23 to target to the proteasome. Rad23 is a non-stoichiometric proteasome subunit that associates with the regulatory particle through its UBL domain22. The UBL-YFP fusion proteins were not degraded in yeast unless a disordered region was also present in the proteins to allow the proteasome to initiate degradation (Fig. 5A)22. We attached the disordered regions derived from regulatory particle subunits to the C-terminus of the UBL-YFP fusion proteins and expressed these proteins together with a red reference protein from a CEN/ARS plasmid in yeast driven by TPI1 and PGK1 promoters, respectively. The yellow fluorescence of each cell, relative to its red fluorescence, determined by flow cytometry, served as a reliable measure of model substrate’s proteasomal degradation22.
A UBL-YFP protein with a tail that is recognized by the proteasome was degraded so efficiently that YFP fluorescence was as low as background (Su9 in Fig. 5A). Degradation is due to the proteasome and is attenuated by proteasome inhibitor (Supplementary Figure 4A). In contrast, a UBL-YFP protein with a tail that is not recognized by the proteasome accumulates so that cells fluoresce many-fold over background (SP2 in Fig. 5A). The tails derived from the C-termini of regulatory particle subunits accumulated to more than 30-fold over background, suggesting that these polypeptide sequences avoid recognition by the proteasome (Fig. 5A). The constructs with the C-terminal disordered regions of Rpn3, Rpn10 and Rpn13 were the most stable proteins, and as stable as the test substrates without a tail (Fig. 5A). Interestingly, the C-terminal tail of Rpn2, and Rpn13, which does not seem to be able to reach the entrance pore, also escaped proteasome recognition. Moreover, even two copies of the disordered region of Rpn13 (or Rpn8, the shortest of the tails) escape recognition by the proteasome (Supplementary Fig. 4B). Thus, these proteins appear to be protected by two mechanisms: physical (the tails are too short to reach the channel) and chemical (even when the sequences can reach the entrance to the degradation channel, their amino acid sequence prevents recognition).
In addition to terminal-mediated degradation, the proteasome can initiate degradation from an internal region in a protein21,26. In fact, the proteasome subunit Rpn1 contains a large disordered loop (amino acids 626-735). To mimic a proteasome substrate with an internal disordered loop with both ends anchored in a folded domain, we attached the internal Rpn1 loop sequence to UBL-YFP and fused a BFP domain to the C-terminal end of the polypeptide. The proteasome was not able to initiate degradation at the Rpn1 loop or a negative control (SRR), even though a control sequence derived from a natural proteasome substrate (Spt2327 ) allowed degradation (Fig. 5B). In summary, all the disordered region of subunits in the proteasome cap tested here have sequences that resist degradation.
Impact of degradation of essential proteasomal subunit Rpn3
Finally, we asked whether the protective proteasome sequences we discovered are important physiologically. Rpn3 is a structural component of the proteasome complex and essential for proteasome function and thus cellular viability. Therefore, it may be necessary for Rpn3 to have a disordered sequence that is proteasome resistant in vivo. If so, expression of Rpn3 in which the degradation resistant sequence is replaced by a sequence that allows recognition should compromise the integrity of the proteasome and thus cell viability.
We tested this hypothesis by inserting either an efficient proteasome engagement sequence (HA-Rpn3ΔC-Su9) or an inefficient engagement sequence (HA- Rpn3ΔC-SP2) at the C-terminus of Rpn3 and predicted that the Su9 tail would cause loss of cell viability when the mutant subunit is not covered by a wildtype copy in the genome, whereas the SP2 tail would support viability. The constructs were carried on centromeric plasmids in cells with a heterozygous deletion of RPN3 at the chromosomal locus (strain CMY 3749, Supplementary Table 1) and cell viability was determined by sporulation followed by tetrad analysis (see Methods). Deletion of RPN3 was confirmed to be lethal, with the exception of a single viable spore, (see comments below). Viability was restored by a plasmid expressing HA-Rpn3 with the native C-terminus (strain CMY 3751) and we recovered 38 viable rpn3Δ spores carrying the plasmid (Table 3). Neither the empty vector (strain CMY 3750) nor the vector expressing HA-Rpn3ΔC-Su9 supported cell viability, although we again recovered a single viable rpn3Δ spore, this time carrying the HA-Rpn3ΔC-Su9 plasmid. In contrast, we recovered 20 viable rpn3Δ spores carrying the plasmid expressing HA-Rpn3ΔC-SP2. The unexpected observation of viable spores whose nominal genotype should not allow the accumulation of Rpn3 is probably the result of chromosome mis-segregation events. It is known that the proteasome is responsible for the degradation of chromosomal cohesins and therefore has a direct role in faithful chromosome segregation28. Therefore, proteasome impairment leads to higher rates of mis-segregation. In summary, we conclude that the native C-terminus of Rpn3 seems to have evolved to resist proteasomal degradation and ensure proteasome integrity and thus function.