Mapping interaction of assembly factor Rpn14 with the proteasome base reveals a bipartite interface and implies ordered remodeling of intersubunit contacts during proteasome biogenesis

Abstract

The 26S proteasome is the largest known protease and an essential mediator of targeted protein degradation, a transformative therapeutic modality for human diseases. Assembly of the 26S proteasome from its 66 cognate subunits depends on nine dedicated assembly chaperones. These chaperones generally function by stabilizing fragile assembly intermediates and/or by regulating the order of subunit association. Whereas the basic functional mechanisms of eight of these nine dedicated chaperones have been at least partially elucidated, the function of Rpn14 (PAAF1 in humans) has remained fully enigmatic. Here, we use a combination of genetics, engineered crosslinking coupled with mass spectrometry, and structural modeling to reveal how Rpn14 interacts with the assembling proteasomal ATPase ring. This model refutes previous Rpn14 binding models and identifies several points of inter-protein steric clash that must undergo remodeling during proteasomal regulatory particle subcomplex maturation. We further show that Rpn14 cooperates with nucleotide to stabilize a known assembly intermediate of the proteasomal base subcomplex. Together, our results illuminate the first known function of Rpn14 during proteasome biogenesis, and provide a framework for detailed mechanistic analyses of how specific interfaces within and between proteasomal subcomplexes are remodeled during their assembly.

Introduction

The 26S proteasome (hereafter proteasome) is a 66-subunit ATP-dependent protease complex that is responsible for most regulated protein degradation in eukaryotes [1, 2]. Proteasomal proteolysis contributes to most physiological processes, and thus the proteasome serves as a key regulatory hub for numerous biological signaling pathways. Recently, targeted delivery of otherwise “undruggable” proteins to the proteasome for destruction has emerged as a powerful and highly promising strategy for pharmacotherapy in diverse human diseases. Further, deregulation of proteasome activity, composition, or assembly contributes to numerous human diseases [3].

The proteasome is composed of a barrel-shaped core particle (CP) that can be capped on one or both barrel ends by the 19S regulatory particle (RP). The CP mediates degradation of substrates, whereas the RP is responsible for capturing substrates, removing their proteasomal targeting signal, unfolding them, and translocating them through a central pore into the CP peptidase cavity. The RP can be divided structurally into lid and base subcomplexes. The lid consists of subunits Rpn3, Rpn5-9, Rpn11, Rpn12, and Sem1/Rpn15. Rpn11 serves as the sole intrinsic deubiquitinating enzyme within the proteasome. The core of the base is a heterohexameric ring of AAA+ family ATPase subunits arranged Rpt1-Rpt2-Rpt6-Rpt3-Rpt4-Rpt5 [4] Each Rpt subunit consists of an N-terminal coiled coil domain that pairs with a neighboring Rpt subunit, followed by an N-terminal oligonucleotide/oligosaccharide-binding (OB) domain. Sitting beneath these domains are the C-terminal AAA+ domains, each consisting of a large and small subdomain, which serve as the ATP-dependent motor of the proteasome. The ATPase ring is decorated with the non-ATPase subunits Rpn1, Rpn2, and Rpn13, which act as substrate receptors [5-11]. One additional substrate receptor, Rpn10, interacts with both the lid and base during catalysis [12, 13].

Assembly of the proteasome from its individual subunits is a highly regulated process that occurs with exceptional fidelity and speed in cells. During RP assembly, the lid and base subcomplexes are thought to assemble independently of one another [2, 14]. Lid assembly occurs autonomously and is mediated by the hierarchical assembly of subunits, driven by avid interactions [15-19]. In contrast, formation of the base relies on four chaperones, Rpn14, Nas6, Nas2, and Hsm3 (PAAF1, p28/gankyrin, p27, and S5b/PSMD5 in humans) for efficient assembly [20-26]. These chaperones associate with non-overlapping sets of base subunits to yield Nas2-Rpt4-Rpt5 (Nas2 module), Hsm3-Rpt1-Rpt2-Rpn1 (Hsm3 module) and Rpn14-Rpt6-Rpt3-Nas6-Rpn2-Rpn13 (Rpn14/Nas6 module) subcomplexes. These modules form via co-translational association of the N-terminal coiled coil and OB fold domains of Rpt1-2, Rpt4-5, and Rpt6-3 [27], followed by chaperone and non-ATPase subunit association. Importantly, changes to the sequence or abundance of human Nas6, human Nas2, and human Hsm3 are associated with cancers [28-33] and type 2 diabetes [34]. More recently, Rpn14 was linked to Parkinson’s disease [35].

The four base chaperones appear to serve distinct roles during RP base biogenesis. For example, Hsm3 interacts primarily with Rpt1, and in addition to serving as a scaffold for the Rpt1-Rpt2-Rpn1 intermediate [36], it also occupies a surface on the underside of the base ATPase ring to prevent the base from docking prematurely onto the CP [36-38]. A similar role has been observed for Nas6 via its interaction with the small AAA+ subdomain of Rpt3 [23, 24, 38-40], whereas Nas2 serves to enforce an ordered sequential association of the three base assembly intermediates by binding Rpt5 and regulating entry of the Hsm3-Rpt1-Rpt2-Rpn1 complex into the assembling base [4, 41, 42].

Understanding of Rpn14 function has lagged substantially behind that of the other base assembly chaperones. Although the structure of Rpn14 has been known for some time [43, 44], how it interacts with the assembling base, as well as its mechanistic function(s) during proteasome biogenesis have remained enigmatic. Rpn14 is composed of an N-terminal globular domain (NTD), followed by a C-terminal WD40 β-propeller domain (CTD) that packs closely against the NTD. Rpn14 was proposed to use its β-propeller domain to bind the small AAA+ subdomain of Rpt6 based on bacterial co-expression experiments and NMR analyses [44, 45]. However, its relative positioning on Rpt6 was not determined. Several largescale crosslinking-mass spectrometry studies investigating the proteasome interaction network have collectively identified a single crosslink between Rpn14 and Rpt6 in humans. However, the long spacer arms utilized in these studies and the paucity of crosslinks detected have precluded inference of interaction surfaces or orientations of these proteins relative to one another [46, 47]. Our unpublished attempts at largescale crosslinking-mass spectrometry using purified recombinant yeast base and Rpn14 similarly failed to yield an extensive network of Rpn14 crosslinks, even with long-reach crosslinkers. Thus, understanding exactly how Rpn14 contacts Rpt6 and any other base subunits or assembly chaperones would greatly advance understanding of chaperone function during proteasome base biogenesis, but likely will require alternative approaches.

Using a combination of genetically engineered photocrosslinking, engineered disulfide crosslinking, mass spectrometry, and integrative modeling, we report a biochemically validated structural model for interaction of Rpn14 with base subunit Rpt6. In contrast to previous proposals, Rpn14 contacts both the Rpt6 small and large AAA+ subdomains with its N-terminal domain and C-terminal β-propeller domain, respectively. Both interactions are important for its function in living cells. Further, we demonstrate that the binding site of Rpn14 clashes sterically with base subunits Rpt2 and Rpn1 as well as lid subunit Rpn7, suggesting an ordered remodeling of base subunit, base-chaperone, and lid-base interactions during RP biogenesis. Together, our work provides some of the first insights into Rpn14 function during proteasome biogenesis, and will guide future mechanistic analyses of RP remodeling during assembly.

Results

Multi-domain interactions between assembly chaperone Rpn14 and base ATPase subunit Rpt6

Previous reports suggested that Rpn14 uses its CTD to bind the small AAA+ subdomain of Rpt6 [44, 45]. However, no direct evidence of this mode of interaction exists, and no role for the globular N-terminal domain (NTD) has yet been proposed. We employed genetically engineered site-specific photocrosslinking [16, 48] to investigate exactly how Rpn14 binds to Rpt6 or any other proteasomal subunits or assembly chaperones. The non-natural photocrosslinking amino acid p-benzoyl-L-phenylalanine (pBF) was introduced via amber suppression into FLAG-tagged Rpn14 expressed from a plasmid in rpn14Δ cells (Fig. 1A). The benzophenone side chain of pBF forms a short-lived ketyl radical upon UV irradiation that induces bond formation between molecules within ~2-3 Å of one another [49].

Figure 1. Both the N- and C-terminal domains of Rpn14 crosslink to base ATPase subunit Rpt6.

A, Site-specific photocrosslinking strategy. The non-natural amino acid para-benzoylphenylalanine (pBF, red sphere) is inserted into Rpn14 via an amber codon by an evolved tRNA and aminoacyl-tRNA synthetase (purple). Illumination with 365 nm UV light promotes formation of a covalent crosslink with interaction partners. B, Rpn14 structure and location of pBF crosslinks detected. On the left, the N-terminal domain (NTD) and C-terminal (CTD) β-propeller domain of Rpn14 are shown in cyan and green, respectively. On the right, the native amino acids that, when mutated to pBF, yielded crosslinks are shown in magenta space-filling mode. C, Both the Rpn14 NTD and CTD crosslink to Rpt6. Whole cell extracts from V5-RPT6 yeast expressing the indicated Rpn14 mutants (where X indicates pBF substitution) were irradiated with UV light or not as shown, before SDS-PAGE immunoblotting with antibodies against FLAG (Rpn14) or V5 (Rpt6). G6PD is shown as a loading control. Faint crosslinks in the V5 blot are indicated with arrowheads. See also Supp. Fig. S1C.

We identified seven positions on Rpn14 that formed crosslinks in a manner dependent both on the presence of pBF and on UV irradiation (Fig. 1B, Fig. 1C top panels). Interestingly, we detected six crosslinks across the surface of the CTD β-propeller (Q156X, T180X, V264X, E308X, F351X, and N406X, where X = pBF substitution), suggesting this is a major interaction surface. We also identified a crosslink of ~100 kDa formed with D25X within the NTD. D25 is located highly distal to the β-propeller CTD, suggesting that the NTD forms discrete contacts with one or more proteins as well.

Previous genetic studies indicate that Rpn14 is closely linked to base subunits Rpt6, Rpt3, and assembly chaperone Nas6 [20-24]. To investigate whether Rpn14 was crosslinking to these proteins, we repeated the crosslinking experiment in nas6Δ cells, and in cells expressing HA-tagged Rpt3 or V5-tagged Rpt6, respectively. All seven crosslinks persisted in the nas6Δ cells (Supp. Fig. S1A), and no UV-dependent bands of ~100-150 kDa were observed in anti-HA (Rpt3) blots (Supp. Fig S1B), suggesting these proteins were not the crosslink partners. In contrast, at least six of the seven pBF- and UV-dependent crosslinks within both the NTD and CTD were observed by V5 (Rpt6) immunoblotting (Fig. 1C middle panels), although some were rather faint (Supp. Fig. S1C). Because Rpn14 interacts only with proteasomal base assembly intermediates and is not found in mature proteasomes under normal conditions [20-24], these interactions must occur specifically in proteasomal base assembly intermediate(s). Thus, Rpn14 likely contacts Rpt6 with both its NTD and CTD during base assembly.

A validated pseudoatomic model of Rpn14-Rpt6 interaction

The relatively small fraction of Rpn14 that could be crosslinked to Rpt6 in yeast (Fig. 1C) suggested that approaches to map the crosslinking sites via purification from yeast cells would be unsuccessful. We thus instead attempted to coexpress Rpn14 with Rpt6 in E. coli. However, we found that upon lysis the proteins reliably partitioned into the insoluble fraction. We next attempted to coexpress increasing numbers of subunits and chaperones present in the Rpn14/Nas6 module to yield a soluble complex. We successfully expressed and purified a minimal complex akin to a previously reported assembly intermediate [20] consisting of Rpn14, Rpt6, Rpt3, and Nas6-6His, which migrated via gel filtration with an elution peak consistent with a heterotetramer (Fig. 2A, red arrow). Hereafter, we refer to this minimal Rpn14- and Rpt6-containing complex as the Rpn14 intermediate (Rpn14int).

Figure 2. Mapping of the Rpt6-Rpn14 interaction.

A, Size exclusion analysis of Ni-NTA eluates from E. coli expressing a recombinant tetrameric Rpn14-Rpt6-Rpt3-Nas6-6His complex (Rpn14int). Rpn14int is centered on fraction 21. Free Nas6-6His appears primarily in fraction 25. Elution profile molecular weight markers are indicated above the gel with black arrowheads. Rpn14int and free Nas6-6His are marked with red and blue arrows, respectively. Asterisk, a contaminating E. coli protein. B, UV photocrosslinking of Rpn14(D25X) present in Rpn14int followed by anti-Rpn14 immunoblotting reveals an ~100 kDa crosslink as observed in yeast extracts. C, An AlphaFold model of Rpn14int places all seven observed pBF crosslinks (shown in pink space-filling mode) at the Rpt6-Rpn14 interface. Nas6, Rpt3, Rpt6, and Rpn14 are shown in ribbon mode and are colored pale goldenrod, sky blue, blue, and green, respectively. D, Engineered disulfide crosslink reporter residue pairs used to verify the AlphaFold model are shown in red space-filling mode. Proteins are colored as in (C). E and F, Engineered disulfide crosslink reporters validate the Rpn14-Rpt6 interaction model. Extracts of yeast cells expressing the indicated cysteine substitutions were treated for 10 minutes at 25°C with 240 μM CuCl₂ to induce disulfide formation prior to non-reducing SDS-PAGE and immunoblotting. DTT was added prior to electrophoresis as shown.

We next incorporated pBF into Rpn14int at the positions identified to yield Rpt6 crosslinks in vivo. In most cases, introduction of pBF disrupted formation of the complex, suggesting that the interaction between Rpn14 and Rpt6 is weakened somewhat by substitution of these residues that lie directly at the Rpn14-Rpt6 interface. However, introduction of pBF in place of Rpn14 D25 was well tolerated and yielded a crosslink via SDS-PAGE migrating at ~100 kDa, similar to that observed in yeast extracts (Fig. 2B). Using mass spectrometry, we mapped the crosslink with high confidence to peptides consisting of Rpn14 amino acids 16-34 and Rpt6 amino acids 335-344, which is in the small AAA+ subdomain. The results suggested that the region of amino acids 341-343 in Rpt6 were cross-linked to Rpn14 residue 25 (Supp. Fig. S2A-C).

We next predicted the Rpn14int structure using AlphaFold [50]. Depending on the order in which the proteins were entered, we reproducibly obtained 8 clusters of models with distinct Rpn14 positions (representatives shown in Supp. Fig. S2D). Of these, most failed to satisfy the constraints imposed by our crosslinking results (Table S3). However, a single model cluster placed all seven observed pBF crosslinks at the Rpn14-Rpt6 interface (Fig. 2C), in addition to placing the Rpn14 D25 residue adjacent to Rpt6(341-343) peptide (Supp. Fig. S2E). In this model, Rpn14 contacts the small and large AAA+ subdomains of Rpt6 with its NTD and CTD, respectively, with the observed ~100 kDa crosslinks (Fig. 1C) occurring between Rpn14 and the Rpt6 small AAA+ subdomain, and with the ~150 kDa crosslinks occurring with the Rpt6 large subdomain. We [4, 16, 51, 52] and others [53, 54] have previously observed migrations of crosslinked proteins that are dependent on the site(s) of crosslinking.

To test the accuracy of the Rpn14int model, we identified amino acids in Rpn14 and Rpt6 that, when replaced with cysteines, would be appropriately positioned for disulfide bond formation if the model were accurate. Addition of a mild oxidant such as Cu²⁺ would then promote a diagnostic disulfide, visualized as a DTT-sensitive band shift, by non-reducing SDS-PAGE [55]. Because disulfide formation is highly dependent on the proper spacing and orientation of cysteine thiols [55], this approach reports on Rpn14-Rpt6 interaction with near-atomic precision. Toward this goal, we first introduced plasmids encoding Rpn14(D25C)-FLAG and V5-Rpt6(R339C) cysteine mutants into rpn14Δ rpt6Δ cells by plasmid shuffling [56]. These residues are juxtaposed at the interface of the Rpn14 NTD and the small ATPase domain of Rpt6 (Fig. 2D). Cells harboring these plasmids displayed no obvious growth defects compared to the WT alleles, suggesting the substitutions were well tolerated (Supp. Fig. S2F).

Upon treatment of extracts from these cells with CuCl₂, a prominent high molecular mass species was evident by anti-V5 immunoblotting (Fig. 2E). This species was also evident in anti-FLAG blots (Supp. Fig. S2G). Formation of this species was dependent on the presence of both engineered cysteine residues and on provision of CuCl₂; the species could also be eliminated by the reducing agent DTT, consistent with a disulfide crosslink. Similar results were obtained with the Rpn14(V136C)-FLAG and V5-Rpt6(H170C) mutants (Fig. 2F and Supp. Fig. S2F and S2H) in the Rpn14 CTD and the Rpt6 large AAA+ subdomain, respectively (Fig. 2D). Together, these data confirm the placement of Rpn14 on Rpt6 and suggest an extensive and largely inflexible interaction between the two.

Both the N- and C-terminal domains of Rpn14 are important for its function in proteasome biogenesis

The crystal structure of Rpn14 suggests that the NTD and CTD pack rigidly against one another. The bipartite interaction of Rpn14 with both the large and small AAA+ subdomains of Rpt6 thus suggests it may restrict conformational movements within this subunit triggered by ATP binding and hydrolysis, or alternatively, to help enforce a particular nucleotide state within Rpt6 during proteasome biogenesis. We hypothesized that, if both Rpn14 domains were important for its function, then expression of only one domain would fail to rescue the strong temperature sensitivity of rpn14Δ nas6Δ cells [21-23]. To permit detection of the individual domains, we generated FLAG-tagged forms of Rpn14 or the isolated NTD and CTD (Fig. 3A). Incorporation of a FLAG tag did not disrupt Rpn14 function, because the FLAG-tagged form fully rescued the temperature sensitivity of the rpn14Δ nas6Δ double mutant (Supp. Fig. S3A).

Figure 3. Both the N- and C-terminal domains of Rpn14 are required for its function.

A, Expression of full-length, NTD, CTD, or NTD + CTD from high-copy plasmids in WT yeast cells. The positions of the FLAG-tagged proteins are shown. Asterisk, Rpn14 truncation fragment. B, C, Expression of the isolated CTD impairs cell growth. The indicated FLAG-tagged or untagged proteins were expressed from a high-copy plasmid in rpt6(cim3-1) (B) or nas6Δ (C) cells, followed by growth analysis at the indicated temperatures. D, The NTD and CTD operate in trans to restore Rpn14 function. WT or rpn14Δ nas6Δ cells expressing the indicated proteins from high-copy plasmids were grown at the indicated temperatures.

We reasoned that if both domains of Rpn14 were important for its function, then expression of just the NTD or CTD might function as a dominant negative by blocking recruitment of full-length endogenous Rpn14 to assembly intermediates. We previously found that the rpt6(cim3-1) mutation showed a strong genetic linkage to RPN14 deletion [22]. We thus expressed untagged or FLAG-tagged NTD and CTD in rpt6(cim3-1) cells, with the expectation that it would potentially sicken them. Whereas full-length Rpn14 had no impact on rpt6(cim3-1) cell health at 30°C, expression of CTD-FLAG substantially impaired cell growth (Fig. 3B). This effect was not due to the FLAG tag, as an untagged version of the CTD yielded the same growth defect. Similar results were obtained upon expression of CTD or CTD-FLAG in nas6Δ cells, with a temperature sensitivity reminiscent of rpn14Δ nas6Δ observed at 37°C (Fig. 3C). Notably, we did not observe an impact of the NTD or NTD-FLAG. This may result from either the slightly lower NTD expression compared to the CTD (Fig. 3A), or possibly from failure of the NTD to associate with Rpt6 in the absence of the CTD.

We next tested the ability of the NTD and the CTD to complement Rpn14 function in rpn14Δ nas6Δ cells using the temperature sensitivity of this mutant as a reporter. As expected, expression of these proteins in WT cells generally had no impact on the ability of the cells to grow at 37°C, although a small but reproducible growth defect was observed at 30°C when only the CTD was expressed (Fig. 3D, left panels). Similarly, expression of full-length Rpn14 in rpn14Δ nas6Δ cells fully rescued the lethality observed at 35 and 37°C. In contrast, expression of the NTD or CTD alone failed to provide detectable rescue. Importantly, however, co-expression of the NTD and CTD together fully rescued growth even at 37°C, suggesting that Rpn14 relies on interactions with both domains of Rpt6 for its assembly function in vivo. Together, these observations support a requirement for both the NTD and CTD for Rpn14 function.

Rpn14 and nucleotide cooperate to stabilize the Rpn14 intermediate

The bipartite interaction of Rpn14 with both the Rpt6 large and small domains suggested that it may enforce a unique configuration of the Rpt6 large and small AAA+ subdomains important for assembly. However, there was an RMSD of 1.05 Å or less across six different conformations of the mature 26S proteasome when Rpt6 from the Rpn14int model was superimposed (Supp. Fig. S4A). We observed similar RMSDs of ~1.4 Å or less when the model was superimposed on Rpt6 from structures of the human RP [57] (Supp. Fig. S4B). We thus instead considered that Rpn14 may help to stabilize this module during assembly. The Rpn14/Nas6 module cannot be readily visualized in yeast extracts, presumably due to instability or low abundance [21, 22]. We thus exploited our ability to reconstitute Rpn14int in E. coli (Fig. 2A) to investigate how Rpn14 contributes to the formation or stability of this intermediate. Notably, expression of Nas6, Rpt3, and Rpt6 without Rpn14 failed to yield a soluble Nas6-Rpt3-Rpt6 complex (Supp. Fig. S4C), suggesting that Rpn14 helps to stabilize or solubilize the module.

We next considered the possibility that nucleotide binding by the pocket formed at the Rpt6 and Rpt3 interface cooperated with Rpn14 to stabilize Rpn14int during its assembly and incorporation into the base. We attempted to purify Rpn14int from identical E. coli cultures lysed in the presence of ATP, ADP, or no added nucleotide, and assessed the formation and abundance of the complex by gel filtration as in Fig. 2A above. ATP and ADP supported the purification of approximately equivalent amounts of Rpn14int. However, withholding nucleotide from the purification buffers yielded little to no complex, evidenced by the apparent absence of Rpt3, Rpt6, and Rpn14 in the SEC eluate (Fig. 4A), and a small amount of Nas6-6His corresponding to the leading edge of the monomeric, free protein eluting from the column. This suggests that nucleotide binding by Rpn14int is important for its formation and/or stability.

Figure 4. Nucleotide stabilizes the Rpn14 intermediate.

A, Lysates were prepared from identical cell pellets in the presence of 2 mM ATP, 2 mM ADP, or in the absence of exogenous nucleotide. After Ni-NTA purification and size exclusion chromatography (SEC), a fraction of the whole cell extract (WCE) or of the peak elution fraction of Rpn14int were separated by SDS-PAGE. B, A cartoon of Rpn14int. The triangles and rounded rectangles correspond to the large and small AAA+ subdomains, respectively. The approximate positions of ATP, the Walker A motif, the Walker B motif, and the arginine fingers are shown. Mutations used to disrupt the functions of these motifs are shown in parentheses next to the motif names. C, Mutation of ATP-interacting residues in Rpt3 and Rpt6 disrupts the formation of Rpn14int. The WCEs and SEC eluates prepared in the presence of 2 mM ATP as in (A) above were separated by SDS-PAGE. The nucleotide pocket to which each residue contributes is listed above the cognate mutation. D, The rpt6-EQ mutation partially suppresses the rpn14Δ assembly defect. Extracts of cells expressing the indicated alleles of RPT6, RPN14, and NAS6 were separated by native PAGE before immunoblotting as shown. The positions of 26S proteasomes and subcomplexes are indicated. Asterisk, a CP capped on both ends by Blm10.

In AAA-family ATPases such as those in the proteasome base, the nucleotide-binding pocket is formed at the interface of adjacent ATPase protomers, with each protomer contributing key residues for nucleotide binding, sensing of the nucleotide state, or coordination of Mg²⁺ and water necessary for catalysis (Fig. 4B) [58]. In the Rpt3-Rpt6 pocket, Rpt3 supplies conserved arginine fingers that sense the γ-phosphate of ATP and position it for hydrolysis, whereas Rpt6 supplies the conserved Walker A and Walker B motifs important for nucleotide binding and hydrolysis, respectively. Mutation of the arginine fingers to alanine (RR/AA), the Walker A motif lysine to serine (KS), or the Walker B glutamate to glutamine (EQ) typically renders these elements nonfunctional within proteasomal ATPases [59, 60].

We investigated the importance of these key elements by systematically mutating residues contributing to the Rpt6 nucleotide pocket (rpt6-K195S, rpt6-E249Q, rpt3-R330A,R333A), and testing whether Rpn14int could be purified in the presence of ATP as above. Although their pockets were not completely formed due to the absence of an adjacent ATPase subunit (Fig. 4B), we also investigated the impact of disrupting residues present in Rpn14int that contribute to the Rpt3 nucleotide pocket (rpt3-K219S, rpt3-E273Q) and the Rpt2 nucleotide pocket (rpt6-R306A,R309A). Disruption of the Rpt3 arginine fingers or the Rpt6 Walker A lysine residue that directly contact ATP resulted in loss of the complex, whereas mutation of the Rpt6 Walker B glutamate, which coordinates nucleotide hydrolysis, supported formation and purification of the complex (Fig. 4C).

As might be anticipated, mutation of the Rpt6 arginine fingers that contribute to the Rpt2 pocket had no obvious impact on formation of Rpn14int; in contrast, mutation of the Rpt3 Walker A lysine or Walker B glutamate caused a substantial reduction in the amount of Rpn14int that could be purified (Fig. 4C), with the Walker A mutation resulting in complete loss of Rpn14int. Although it is unclear if Rpt3 is interacting with nucleotide via these motifs in this context, it is possible that transient interactions with nucleotide facilitate the folding or stability of the subunit to permit its incorporation into Rpn14int. Regardless, these results agree with the requirement for nucleotide observed above, and suggest that Rpn14 cooperates with nucleotide binding at the Rpt6 binding pocket to stabilize Rpn14int.

We hypothesized that, if nucleotide binding by Rpt6 cooperates with Rpn14 during base assembly, then enforcing an ATP-bound state in Rpt6 may partially suppress the assembly defect imparted by deletion of RPN14. We performed these experiments in a nas6Δ background, which accentuates the impact of RPN14 deletion, evident as a reduction in the levels of doubly- and singly-capped proteasomes [21-23]. As we previously observed [51], introduction of the rpt6-EQ allele in an otherwise WT background resulted in enhanced abundance of 26S proteasomes, presumably due to compensatory new subunit synthesis in response to the catalytic defect induced by rpt6-EQ (Fig. 4D, compare RP₂CP and RP₁CP levels in lane 1 to lane 3, 9 to 11, and 17 to 19). Deletion of RPN14 alone had no obvious additional effect in rpt6-EQ cells (Fig. 4D, compare lane 3 to lane 4, 11 to 12, and 17 to 18). Deletion of RPN14 in nas6Δ cells expressing WT RPT6 caused a loss of fully assembled proteasomes and accumulation of free CP and lid subcomplexes compared to nas6Δ cells alone (Fig. 4D, compare lane 5 to lane 6, 13 to 14, and 21 to 22). Importantly, when the rpt6-EQ mutation was introduced into rpn14Δ nas6Δ cells, a modest but reproducible rescue of the assembly defect was observed, evident as an increased abundance of 26S proteasomes and reduced abundances of free CP and lid (Fig. 4D, compare lane 6 to lane 8, 14 to 16, and 22 to 24). Taken in sum with our coexpression experiments, these data support a cooperative stabilization of Rpn14int by Rpn14 and nucleotide.

Modeling of Rpn14 onto mature proteasomes reveals conformation-dependent steric clash with lid subunit Rpn7 and base subunit Rpn1

Rpn14 has previously been observed bound to both the isolated base as well as the RP [21, 23]. However, immunoprecipitations of Rpn14 enriched the base substantially over the RP, suggesting some preference between these two complexes [21]. Further, the molecular mechanism(s) and timing of Rpn14 release from base or RP assembly intermediates remains unknown. To gain further insight into Rpn14 function and regulation during proteasome biogenesis, we modeled Rpn14 from our AlphaFold model onto published structures of mature proteasomes in resting and activated conformations [51] via superimposition of the Rpt6 AAA+ domain (RMSD ~1 Å). We observed similar RMSDs (≤ 1.4 Å) when Rpt6 from our model was superimposed on Rpt6 in several available structures of the human RP [57].

Nas6, Hsm3, and Rpn14 have each been suggested to restrict base-CP interaction during assembly [23, 37-40]. However, Rpn14 did not clash appreciably with the CP when modeled onto resting (represented by the s1 conformational state, Fig. 5A) or activated (represented by the s4 state, Fig. 5B, see also Supp. Fig. S5A) proteasomes. Further, overexpression of Rpn14 from low or high-copy plasmids had no appreciable impact on the assembly status of 26S proteasomes, as gauged by the abundances of singly-(RP₁CP) and doubly-capped (RP₂CP) proteasomes upon native polyacrylamide gel electrophoresis (PAGE)-immunoblotting against lid, base, and CP subunits (Fig. 5C). In contrast, overexpression of Nas6 and Hsm3 caused significant loss of 26S proteasomes and accumulation of free RP and CP (Fig. 5C), consistent with roles restricting premature association of the base or RP with the CP. Nas2, which dissociates prior to or immediately after completion of base assembly [4, 42], had no impact on 26S structure as would be anticipated.

Figure 5. Modeling of Rpn14 onto structures of mature proteasomes suggests mechanisms of chaperone function and eviction.

A, B, Modeling of Rpn14 onto resting (s1) state (A, PDB: 6FVT) or active (s4) state (B, PDB: 6FVW) proteasomes reveals no significant clash with the CP. Rpt6 from the AlphaFold model shown in Fig. 2C was superimposed onto Rpt6 in the s1 or s4 states are shown, with Rpn14 colored green, Rpt6 in blue, and the CP in grey. Most RP subunits are omitted for clarity. C, Overproduction of Nas6 and Hsm3, but not Rpn14 or Nas2, disrupt RP-CP interaction. Native PAGE-immunoblotting analysis of WT cells expressing the indicated base assembly chaperones from low- or high-copy plasmids. The positions of proteasomal subcomplexes and assembly intermediates are shown. D,E, As in A, but highlighting conformation-dependent clash with Rpn1. Rpn1 is shown in light blue, Rpt1-6 in blue, and with CP and Rpn14 colored as in A above. Clashes between Rpn1 and Rpn14 are shown in red. Most RP subunits are omitted for clarity. F,G, As in A, but highlighting conformation-dependent clash with Rpn7. Rpn7 is shown in goldenrod, with Rpt1-6, CP, and Rpn14 shown as in F above. Clashes between Rpn7 and Rpn14 are shown in red. Most RP subunits are omitted for clarity. See also Supp. Fig. S5.

A modest steric clash between Rpn14 and the small AAA+ subdomain of Rpt2 was observed when Rpn14 was modeled onto all six states of the yeast proteasome examined (Supp. Fig. S5B). Given the rigid and extensive interaction surface between Rpn14 and Rpt6, and considering that Rpt2 and Rpn14 coexist in RP assembly intermediates [21, 23], this implies that Rpt2 may adopt a distinct conformation during base or RP assembly not observed in mature 26S proteasomes. The base non-ATPase subunit Rpn1, which associates with the peripheral surface of the ATPase ring via interactions with Rpt1-2, readily accommodated Rpn14 in the resting s1 state (Fig. 5D) and the s3 state (Supp. Fig. S5C), but exhibited substantial clash with the C-terminal β-propeller domain of Rpn14 when modeled onto the s4 (Fig. 5E), s2, s5, and s6 states (Supp. Fig. S5C). Since steric clash would remain between Rpn14 and Rpt2 even in states devoid of Rpn1-Rpn14 clash, we believe the Rpn14-bound base must adopt an as-yet unobserved conformation of the ATPase ring that would relieve clash with Rpt2 and position Rpn1 compatibly.

Interestingly, we observed a substantial steric clash between Rpn14 and lid subunit Rpn7 only in the s1 conformation of the proteasome (Fig. 5F), with minimal clash in all other conformations (Fig. 5G and Supp. Fig. 5D). This clash was particularly notable because in the s1 state of the mature proteasome, Rpn7 makes direct contact with the same surface on the large AAA+ subdomain of Rpt6 contacted by Rpn14, centered on Rpt6 Lys169 and His170. Considering that Rpn14 can co-purify the RP [21, 23], this suggests first that release of Rpn14 is critical for adoption of an s1-like interaction of Rpn7 with the base, and second, the lid may initially associate with the base and adopt a non-s1-like conformation in which Rpn7 is positioned away from the large Rpt6 AAA+ subdomain.

Taken together with previous observations that Rpn14 cohabitates fully assembled base and RP complexes, these data suggest an ordered remodeling of inter-subunit contacts within the base during RP biogenesis that is predicated, at least in part, on dissociation of Rpn14. Further, release of Rpn14, either induced by docking of the lid and displacement by Rpn7-ATPase interactions, or via spontaneous dissociation prior to lid docking, would in turn permit docking of the Rpn7 N-terminal domain against the ATPase ring in an s1-like conformation.

Discussion

Our data indicate that Rpn14 binds its cognate ATPase subunit Rpt6 in a manner that is substantially different than that originally proposed, and distinct from that observed for the other three dedicated base chaperones. Surprisingly, we find that the C-terminal β-propeller domain of Rpn14 docks across the outer surface of the Rpt6 large AAA+ subdomain, rather than across the small subdomain as previously proposed based on mutagenesis and NMR analyses [44, 45]. Rpn14 instead uses its N-terminal globular domain to simultaneously contact the small AAA+ subdomain of Rpt6, thereby reinforcing a precise geometry of the AAA+ subdomains relative to one another. Based on our mutagenesis and cell-based data (Fig. 4), we suggest this geometry presents a favorable conformation for interaction with Rpt3 that, in cooperation with nucleotide binding at the Rpt3-Rpt6 interface, stabilizes Rpn14int for subsequent assembly steps (Fig. 6A). Such conformational matching is likely important for the stepwise association of ATPase subunit neighbors within base assembly intermediates.

Figure 6. Model for Rpn14 function during RP biogenesis.

A, Base assembly functions of Rpn14 suggested by this study. Rpn14, ATPase subunits, Rpn1, and Nas6 are shown in green, blue, light blue, and tan, respectively. ATP is shown as a red circle. The identities of nucleotides in the chaperone-bound base are unknown, and the positions shown are speculative. B, Possible release mechanisms for Rpn14: spontaneous dissociation leading to ATPase ring closure (i); nucleotide-dependent Rpt6 remodeling leading to Rpn14 eviction and ring closure (ii), and steric displacement of Rpn14 by ATPase ring closure (iii). See the discussion for details.

Although Rpn2 contacts the N-terminal helical coils of Rpt3-Rpt6 in mature proteasomes and presumably also during base assembly, the highly distal position of Rpn2 relative to the AAA+ domains of Rpt3-6 suggests it is unlikely to be involved in stabilizing the Rpt3-Rpt6 AAA+ domain interaction. Our mutagenesis of Rpn14int suggests that Rpn14 cooperates with nucleotide binding at the Rpt3-Rpt6 interface to stabilize Rpn14int. Our finding that both the N- and C-terminal domains of Rpn14 are required to support its assembly function in yeast (Fig. 3; Fig. 4D) and that the isolated CTD can act as a dominant negative supports this view. Such a role is further supported by our finding that trapping Rpt6 in an ATP-bound state partially suppressed the assembly defect in rpn14Δ nas6Δ cells (Fig. 4D). This proposed role in restricting the conformational flexibility of the ATPase large and small AAA+ subdomains may be somewhat unique to Rpn14 among the base assembly chaperones; the manner in which Nas6, Hsm3, and Nas2 interact with fragments of their cognate ATPase subunits [37, 38, 42, 61] argues against such a role for these chaperones.

The extensive surfaces buried between Rpn14 and Rpt6 suggests a highly rigid mode of interaction, and permits high-confidence modeling of Rpn14 onto available structures of the human RP and the yeast 26S proteasome. The observation that the small subdomain of Rpt2 would clash with Rpn14 in all observed conformations of the yeast proteasome suggests there exists an as-yet undiscovered conformation of the assembling base that permits Rpn14 and Rpt2 to coexist in the same complex. In support of this possibility, the human homolog of yeast base assembly chaperone Hsm3 was recently proposed to weaken interleaving of Rpt2 with Rpt6 via interactions of its C-terminal tail with the interior of the ATPase ring [62]. Thus, an exciting possibility is that Hsm3 functions, either alone or with Rpn14, to open this seam to accommodate both Rpn14 and Rpn2. In fact, while this manuscript was being finalized, a preprint was published describing an open conformation of the ATPase ring within a chaperone-bound base subcomplex [63]. In this structure, such a seam was observed that permitted accommodation of both Rpt2 and Rpn14. Interestingly, this structure suggested direct contact between the exterior surface of the Rpn14 β-propeller and Rpn1 that may help to rigidify the base prior to lateral closure of the AAA+ domain ring.

The trigger(s) for release of Rpn14 from the base or RP are not yet known, and are not directly revealed by our own work. However, at least three possibilities could be conceived based on pre-existing studies and our work presented here (Fig. 6B). In the first, Rpn14 may spontaneously release from the base (or the RP) upon completion of its assembly function (Fig. 6B, scheme i). Dissociation of Rpn14 would likely then relieve steric conflict with Rpt2 (and possibly Rpn1) observed when Rpn14 is superimposed onto mature proteasomes, leading to lateral closure of the ATPase ring. In this model, Rpn14 would help both to control the timing or sequence of Rpt1-Rpt2-Rpn1-Hsm3 module incorporation and would also serve as a negative regulator of premature ATPase ring closure.

In the second, remodeling of Rpt6 due either to stimulation of its ATPase activity or to assembly-coupled conformational changes may promote release of Rpn14 (Fig. 6B, scheme ii). Notably, assembly of the mammalian base from its cognate intermediates in vitro leads to stimulation of its ATPase activity [25], and we found that Rpn14 cooperates with nucleotide to stabilize the Rpt6-Rpt3 interface. Thus, induction of an empty state in the Rpt6 nucleotide pocket either by release of hydrolyzed ATP upon completion of base assembly or via some other remodeling event within the base may alter the geometry between the Rpt6 AAA+ subdomains, weakening Rpn14 interaction and leading to its release. As above, relief of this steric conflict with Rpt2 would then presumably permit ring closure. However, we are not aware of any evidence that the small and large AAA+ subdomains of Rpt6 undergo substantial interdomain movements (Supp. Fig. S4A,B) [12, 13, 51, 64], which seems to render this possibility less likely.

A third possibility is the induction of steric conflict between proteasome subunits and Rpn14 during closure of the ATPase ring (Fig. 6B, scheme iii) and/or repositioning of the lid subcomplex (in the context of the RP; not shown) leads to eviction of Rpn14. Our modeling data indicates conflict between Rpt2 and Rpn14 in all conformations of mature proteasomes. Thus, induced closure of the ATPase ring of the base could lead to eviction of Rpn14 via steric conflict with Rpt2. Given that Rpn14 has been frequently observed bound to the RP, this model would seemingly require that the lid could bind to an immature base subcomplex that has not yet undergone ring closure. Alternatively, remodeling of lid-base contacts during assembly may serve to bring Rpn7 in steric conflict with Rpn14, causing its release. In such a model, the lid subcomplex would presumably first dock in an s4-like conformation which would position Rpn7 away from Rpn14 (Fig. 5G). Conformational shifting, presumably triggered by ATPase activity within the base, would then serve to rotate the lid, forcibly bringing Rpn7 into conflict with Rpn14 and promoting its ejection (Fig. 6A). We previously proposed a similar model for release of the Nas6 chaperone, in which it was effectively shoehorned from the nascent 26S proteasome by conformational changes that brought lid subunit Rpn5 and the CP into simultaneous steric conflict with Nas6 [40]. Our validated model of Rpn14-Rpt6 interaction will likely help to guide future studies to distinguish between these possibilities and to more fully understand the dynamics governing base assembly, coordination between assembly factors, and signals for chaperone release.

Experimental Procedures

Yeast strains and media

Yeast strains used in this study are listed in Table S1. All yeast strains were created using standard genetic approaches. Cells were grown at 30°C in YPD or, if harboring nonessential plasmids, synthetic medium lacking the appropriate auxotrophic amino acid(s) unless otherwise indicated. For growth assays, yeast strains were grown at 30°C overnight to saturation. The following morning, the cultures were diluted in deionized water to an OD₆₀₀ of 0.1. Six-fold serial dilutions were then made in sterile deionized water. These dilutions were then plated on growth media at the temperatures indicated in the figure legends before imaging.

Plasmids

Plasmids used in this study are shown in Table S2. Plasmids were constructed using standard molecular cloning approaches with TOP10 F’ (Thermo-Fisher) as the host strain. Site-directed mutagenesis was conducted via Quikchange (Agilent). All coding sequences were verified by DNA sequencing prior to use. Complete plasmid sequences and construction details are available upon request.

Yeast site-specific photocrosslinking

Yeast strains were co-transformed with pRT3104 encoding the amber suppressor machinery for pBF and a plasmid expressing C-terminally FLAG-tagged Rpn14 with or without amber codons at selected locations from a GPD promoter. Cells were then inoculated to OD₆₀₀ = 0.3 in SC-Trp-Ura + 1 mM pBF and grown at 30°C until OD₆₀₀ ≈ 1.5 (typically about 8 hours). Cells were harvested by centrifugation and lysed by beat-beating in Buffer A (50 mM Tris-Cl, pH 7.5, 150 mM NaCl, 5 mM MgCl₂, 10% glycerol, 2 mM ATP, and 0.1% v/v NP-40). Lysates were clarified by centrifugation, and diluted to ~ 1 mg/mL protein content based on Bradford assay. The lysates were then irradiated for 60 minutes on ice with a Blak-Ray B-100AP high-intensity UV lamp (100 W, 365 nm) before immunoblotting as described below.

Engineered disulfide crosslinking

We performed engineered disulfide crosslinking according to an established protocol [55]. Yeast strains were co-transformed with plasmids encoding WT or cysteine mutants of the indicated proteins. Cells were then inoculated to OD₆₀₀ = 0.2 in SC-Trp media and grown at 30°C until OD₆₀₀ = 1.0. Cells were harvested as above and lysed by cryogenic grinding followed by protein extraction into Lysis Buffer (50 mM HEPES•NaOH, pH 7.5, 150 mM NaCl, and 5 mM MgCl₂ + 2 mM ATP). Lysates were cleared by centrifugation, diluted to 4 mg/mL protein content using Lysis Buffer, and then treated or not with 240 μM CuCl₂ at 25°C for 10 minutes. Crosslinking was terminated with 10 mM EDTA and 10 mM N-ethylmaleimide before immunoblotting as described below. Where indicated, 20 mM DTT was added to the crosslinking reaction for 10 minutes prior to termination as above.

Non-denaturing polyacrylamide gel electrophoresis

Cell extracts were prepared and separated by native PAGE essentially as described previously [65, 66]. Cells were grown in 500 mL of YPD media at 30°C to log phase (OD₆₀₀ ≈ 1.0 – 2.0). Cultures were centrifuged at 5,000 x g for 5 minutes, 4°C. Cell pellets were then washed with 25 mL of ice-cold deionized water and centrifuged again at 4,122 x g, 4°C for 2 minutes. The supernatant was removed and the cell pellets were snap-frozen in liquid nitrogen. Cell pellets were crushed using a mortar and pestle pre-cooled in liquid nitrogen. Cell powder was then hydrated with 1 volume of extraction buffer (50 mM Tris-HCl, pH 7.5, 5 mM MgCl₂, 10% glycerol, 1 mM ATP, 0.015% w/v xylene cyanol) and vortexing frequently for 10 minutes on ice. Insoluble material was pelleted at 21,000 x g for 10 minutes at 4°C, and 60 μg (measured by BCA assay) of the supernatant was applied to a 4% non-denaturing polyacrylamide gel containing 0.5 mM ATP and separated at 100 V, 4°C.

Denaturing polyacrylamide gel electrophoresis (SDS-PAGE)

Cell extracts were prepared via the same method as described for native PAGE. 5x SDS loading buffer was added to the supernatants, applied to 10% SDS polyacrylamide gels and separated at 200 V at room temperature.

Immunoblot analyses

Polyacrylamide gels were transferred to PVDF membranes at 100 V for one hour at 4°C. Membranes were then incubated in 5% milk with antibodies against the CP (Enzo Life Sciences cat# PW9355, 1:5000), Rpt1 (1:5000), , FLAG (Sigma, cat#F3165, 1:5000), Rpn12[66] (1:5000), Rpt3 (Enzo Life Sciences, cat#PW8245, 1:4000), G6PD (Sigma, cat#A9521, 1:20,000), HA (Covance, HA.11, 1:1000). Blots were imaged on a Bio-Rad ChemiDoc MP using HRP-conjugated secondary antibodies (GE Healthcare) and ECL reagent.

Purification of the recombinant Rpn14 intermediate

The Rpn14 intermediate was expressed from a single polycistronic plasmid in bacterial strain LOBSTR (DE3) containing pRARE2. Transformants were grown in 2xYT and the appropriate antibiotics at 37°C, 250 rpm shaking until OD₆₀₀ = 1.0, at which point the temperature was reduced to 16°C and IPTG was added to 0.5 mM. After overnight induction, cultures were centrifuged at 8000 x g for 5 minutes, 4°C, the supernatant was poured off, and cells were frozen at −80°C until purification. The day of purification, cells were thawed in Buffer A (50 mM Tris-Cl, pH 7.5, 150 mM NaCl, 5 mM MgCl₂, 10% glycerol) supplemented with 20 mM imidazole, 1 mM PMSF, and 2 mM ATP. In some experiments, 2 mM ATP was withheld or replaced with 2 mM ADP. The cell suspension was lysed with an Avestin Emulsiflex C-5 and clarified via centrifugation at 30,000 x g, 4°C for 20 minutes. The supernatant was incubated with Ni-NTA resin for 30 minutes at 4°C. After two washes with Buffer A supplemented with 50 mM imidazole, the resin was poured into a disposable Bio-Rad Econo-column, washed with Buffer A + 50 mM imidazole, and eluted with Buffer A containing 250 mM imidazole. Eluates were concentrated using 30,000 Da MWCO filters (Amicon), and Rpn14int was separated from free Nas6-6His by gel filtration on a Superose 6 Increase 10-30 column equilibrated in Buffer A containing 0.5 mM TCEP and 1 mM ATP, and 500 μL fractions were collected. For experiments comparing the abundance of Rpn14int, fractions 21-23 were pooled before analysis.

Photocrosslinking and digestion of the recombinant Rpn14 intermediate

Rpn14int containing pBF was expressed exactly as above, with the following exceptions. In place of the single polycistronic plasmid, cells were cotransformed with a pET42b-based plasmid encoding Rpt6, Rpt3, and Nas6-6His and a second pCDFDuet-1-based plasmid encoding the Rpn14(D25X) amber mutant. When the culture reached OD₆₀₀ ≈ 1.0, cells were harvested by centrifugation and resuspended in 1/10^th volume of Unnatural amino acid media [67] supplemented with 1 mM pBF and 0.05% arabinose, and incubated for 15 minutes at 30°C. The temperature was then adjusted to 16°C and 1 mM IPTG was added, and the culture was incubated overnight before harvest. Purification was performed exactly as described above, with the complex shielded from direct light exposure to prevent premature photocrosslinking.

For biochemical analysis, photocrosslinking of Rpn14int was performed exactly as for yeast extracts. An aliquot of the crosslinking reaction was reserved for validation by anti-Rpn14 immunoblotting. For crosslink identification, the complex was subjected to UV-induced cross-linking on ice by irradiation at 365 nm for 50 min in a light chamber (Analytikjena UVP Cross-linker CL-1000L) [68]. Proteins were then transferred onto a 30 kDa FASP centrifugal filter for enzymatic digestion. After washing with 8M urea buffer, proteins were reduced with 2 mM TCEP for 30 min and alkylated with 10 mM iodoacetamide for 30 minutes in the dark at room temperature. Next, after replacing the buffer with 25 mM ammonium bicarbonate, proteins were reconstituted in 85 μL of 5.6 M urea in 25 mM ammonium bicarbonate and digested with Lys-C (enzyme to protein ratio of 1:100) at 37°C for 4 h. Then, the concentration of urea was reduced to 1.5 M for trypsin digestion (enzyme to protein ratio of 1:50) at 37°C overnight.

LC-MS/MS analysis and database searching

Peptide digests were acidified with trifluoroacetic acid to a final concentration of 0.1% and subsequently desalted using a Sep-Pak C18 cartridge. Following vacuum drying, the eluents were fractionated on a Superdex^™ Peptide 3.2/300 SEC column. Three fractions enriched in cross-linked peptides were collected, dried under vacuum, and subjected to LC-MS/MS analysis [69, 70].

LC-MS/MS was carried out on a Thermo Scientific^™ Dionex UltiMate 3000 system coupled online to an Orbitrap Fusion Lumos^™ mass spectrometer [70]. Peptide separation was achieved using a 50 cm × 75 μm Acclaim^™ PepMap^™ C18 column with a linear gradient from 1% to 25% ACN over 90 minutes at a flow rate of 300 nL/min. Data acquisition was performed in data-dependent mode, where each cycle began with a full MS scan in the Orbitrap (resolution: 120,000; scan range: 375–1800 m/z; AGC target: 4 × 10⁵; maximum injection time: 50 ms), followed by MS/MS fragmentation of multiply charged precursor ions (3+ to 6+) for 3 seconds using top-speed mode. Precursor ions were fragmented by HCD at a normalized collision energy of 30%, and MS/MS spectra were acquired in the Orbitrap at 30,000 resolution with an AGC target of 1 × 10⁵ and a maximum injection time of 54 ms.

Cross-linked peptide pairs were identified using the MeroX(Version 2.0.1.4) [71]. The target protein sequence was uploaded as a FASTA file with the pBF incorporation site substituted by "x" to account for this non-natural amino acid. pBF was defined in the amino acid table using its elemental composition (C₁₆H₁₃NO₂) and monoisotopic mass (251.09462863 Da), and added as a cross-linker with a maximum Cα–Cα distance of 12.0 Å. Specificity site 1 was set to "x" and specificity site 2 was set to include all amino acids and the peptide N-terminus. The maximum number of modifications was set to 2. Mass tolerances were set to 10 ppm for MS1 and 15 ppm for MS2, with a signal-to-noise ratio of 1. Scoring parameters were set to 5% intensity threshold, 5% FDR cutoff, and a score cutoff of 2. The mzML output files from artMS were loaded as input, and the analysis was initiated to identify cross-link candidates, unique crosslink sites, and corresponding fragmentation spectra.

Molecular modeling

AlphaFold predictions were generated with AlphaFold 3 [50] release 2026.03.19, accessed via the AlphaFold server (http://www.alphafoldserver.com/), and using the default settings. Four different protein entry orders were tested: i) Nas6, Rpt6, Rpn14, Rpt3; ii) Rpn14, Nas6, Rpt3, Rpt6; iii) Rpt3, Nas6, Rpt6, Rpn14; and iv) Rpt6, Rpt3, Rpn14, Nas6. Each prediction generated five models from five different random seeds using ten recycling steps. Key metrics for the twenty resultant models are listed in Table S3. The models were then visually inspected and manually assigned to clusters based on the coarse positioning of Rpn14 within the tetramer. All ten models from entry orders (i) and (ii) closely agreed, whereas those from entry orders (iii) and (iv) diverged substantially both within and between entry orders. After clustering, representative models from each cluster were inspected for satisfaction of experimentally observed pBF crosslinks. Only structures from cluster iv satisfied all seven pBF crosslink restraints (Table S4). Model 0 from order (i) above is displayed in the manuscript main figures. Protein interfaces were identified in ChimeraX (version 1.9) using a buried solvent-accessible surface area of ≥ 15.0 Ų as criteria. The pairwise predicted aligned errors for each interface pair were then extracted from the complete JSON model file and are reported in Table S5.

For modeling of Rpn14 onto mature proteasomes, Rpt6 residues 131-381 (encompassing the large and small ATPase domains) from Rpn14int AlphaFold model were superimposed onto the same residues of the yeast 26S proteasome in the s1 (PDB: 6FVT), s2 (PDB: 6FVU), s3 (PDB: 6FVV), s4 (PDB: 6FVW), s5 (PDB: 6FVX), and s6 (PDB: 6FVY) conformations using ChimeraX. RMSD over all backbone atoms was ~1 Å for each superimposition. Steric clashes were evaluated using a minimum van der Waals overlap of 0.6 Å, which likely overestimates true clash given the resolution of the EM models and ~1 Å RMSD arising from superimposition of the AlphaFold model. For assessing clash with human RP subunits, the T1 (PDB: 5VHF), T2 (PDB: 5VHH), and T3 (PDB: 5VHI) structures of the gankyrin-bound human RP [57] were used exactly as above.

Reproducibility

All experiments were repeated at least twice, with most being performed a minimum of 3 times and by multiple researchers.

Data availability

Data is available upon request.