Minimal mechanistic component of HbYX-dependent proteasome activation

SUMMARY The implication of reduced proteasomal function in neurodegenerative diseases combined with numerous studies showing the protective effects of increasing proteasome activity in animal models justify the need to understand how the proteasome is activated for protein degradation. The C-terminal HbYX motif is present on many proteasome binding proteins and functions to tether activators to the 20S core particle. Peptides with a HbYX motif can also autonomously activate 20S gate-opening to allow protein degradation, but the underlying allosteric molecular mechanism is not clear. We designed a HbYX-like dipeptide mimetic that represents only the fundamental components of the HbYX motif to allow rigorous elucidation of the underlying molecular mechanisms of HbYX induced 20S gate-opening in the archaeal and mamalian proteasome. By generating several high-resolution cryo-EM structures (e.g. 1.9Å) we identified multiple proteasome α subunit residues involved in HbYX-dependent activation and the conformational changes involved in gate-opening. In addition, we generated mutants probing these structural findings and identified specific point mutations that strongly activate the proteasome by partially mimicking a HbYX-bound state. These structures resolve 3 novel mechanistic features that are critical for allosteric α subunit conformational changes that ultimately trigger gate-opening: 1) rearrangement of the loop adjacent to K66, 2) inter- and intra- α subunit conformational changes and 3) a pair of IT residues on the α N-terminus in the 20S channel that alternate binding sites to stabilize the open and closed states. All gate-opening mechanisms appear to converge on this “IT switch”. When stimulated by the mimetic, the human 20S can degrade unfolded proteins such as tau, and prevent proteasomal inhibition by toxic soluble oligomers. Collectively, the results presented here provide a mechanistic model of HbYX-dependent 20S gate-opening and offer proof of concept for the robust potential of HbYX-like small molecules to stimulate proteasome function, which could be useful to treat neurodegenerative diseases.


SUMMARY 13
The implication of reduced proteasomal function in neurodegenerative diseases combined with 14 numerous studies showing the protective effects of increasing proteasome activity in animal 15 models justify the need to understand how the proteasome is activated for protein degradation. 16 The C-terminal HbYX motif is present on many proteasome binding proteins and functions to 17 tether activators to the 20S core particle. Peptides with a HbYX motif can also autonomously 18 activate 20S gate-opening to allow protein degradation, but the underlying allosteric molecular 19 mechanism is not clear. We designed a HbYX-like dipeptide mimetic that represents only the 20 fundamental components of the HbYX motif to allow rigorous elucidation of the underlying 21 molecular mechanisms of HbYX induced 20S gate-opening in the archaeal and mamalian 22 proteasome. By generating several high-resolution cryo-EM structures (e.g. 1.9Å) we identified 23 multiple proteasome α subunit residues involved in HbYX-dependent activation and the 24 conformational changes involved in gate-opening. In addition, we generated mutants probing 25 these structural findings and identified specific point mutations that strongly activate the 26 proteasome by partially mimicking a HbYX-bound state. These structures resolve 3 novel 27 mechanistic features that are critical for allosteric α subunit conformational changes that ultimately 28 trigger gate-opening: 1) rearrangement of the loop adjacent to K66, 2) inter-and intra-α subunit 29 conformational changes and 3) a pair of IT residues on the α N-terminus in the 20S channel that 30 alternate binding sites to stabilize the open and closed states. All gate-opening mechanisms 31 appear to converge on this "IT switch". When stimulated by the mimetic, the human 20S can 32 degrade unfolded proteins such as tau, and prevent proteasomal inhibition by toxic soluble 33 oligomers. Collectively, the results presented here provide a mechanistic model of HbYX-34 dependent 20S gate-opening and offer proof of concept for the robust potential of HbYX-like small 35 molecules to stimulate proteasome function, which could be useful to treat neurodegenerative 36 diseases. 37

INTRODUCTION 38
The proteasome is a key component of the ubiquitin-proteasome system (UPS), responsible for 39 removing damaged or unneeded proteins and regulating major cellular processes 1

. Regulation by 40
proteasome activators (PAs) are critical to ensure that only proper proteins are degraded. 41 Dysregulation of the proteasome has been implicated in several neurodegenerative diseases 42 (NDs), often characterized by impairment of proteasome function 2-7 . In this study, we elucidate a 43 minimal mechanistic model that describes how HbYX(hydrophobic-tyrosine-variable C-terminal 44 residue)-motif-containing PAs activate the core particle of the proteasome. We also designed a 45 small molecule that functionally emulates this mechanism of proteasomal gate-opening and 46 robustly activates the archaeal, yeast, mammalian, and human proteasomes. We also generated 47 a high-resolution cryo-EM structure of the archaeal proteasome in complex with this small 48 molecule activator that elucidates new mechanistic understanding of proteasome activation that 49 is conserved from archaea to humans. Moreover, this proteasome-activating small molecule can 50 reverse inhibition of the proteasome by toxic soluble protein oligomers implicated in 51 neurodegenerative disease, such as amyloid-β, α-synuclein, and huntingtin exon1. 52 The core particle of the eukaryotic proteasome, also referred to as 20S (Fig. 1A), consists of four 53 stacked heteroheptameric rings (α-β-β-α) with a central pore for substrate entry. The β rings 54 consist of seven subunits (β1-7), three of which harbor protease sites. The two α rings also consist 55 of seven subunits (α1-7). Substrate entry in the 20S is regulated by the gate, which primarily 56 consists of the N-terminus of α 2, 3, and 4 extending over the central pore thus closing off this 57 barrel shaped structure 8 . The closed gate conformation blocks the central pore and prevents 58 proteins from entering the 20S to be degraded. The N-terminus of each α subunit carries a YDR 59 (tyrosine-aspartic acid-arginine) motif that interacts with neighboring N-termini to stabilize the 60 closed state of the gate 8 . These N-termini extensions can also change their conformation to an 61 "open" state, whereby they point up and outwards from the α ring pore, which is stabilized by an 62 control; even the 3-residue peptide (CT3) had (~2-fold) gate-opening activity (Fig 1I). 212 Based on the successful gate-activation of PAN CT3 (Ac-LYR) after a single modification 213 (acetylated N-terminus), we investigated whether additional modifications would further improve 214 the efficacy of the peptide. Our previous study showed that the substitution of alanine for arginine 215 does not affect PAN-CT8 peptide efficacy to bind α intersubunit pockets and open the gate 16 . We 216 therefore substituted an alanine in place of the bulky arginine residue, which resulted in greater 217 activating efficacy, compared to Ac-LYR (Fig. 1J). Attempting to further minimize the size of the 218 molecule, we chose carboxybenzyl (Z) as an N-terminal blocking group that could eliminate the 219 N-terminal charge and mimic the Hb group (of the HbYX motif). Combining these modifications 220 resulted in a dipeptide with a hydrophobic group preceding the N-terminus of tyrosine (Z-Tyr-Ala 221 or ZYA) (Fig. 1K). We compared ZYA against Ac-LYR and Ac-LYA and observed significant 222 improvements in T20S activation (Fig. 1J). ZYA could substantially activate T20S activity, but its 223 affinity was poor (see below). While ZYA could activate T20S ~13 fold, it could not stimulate gate-224 opening in T20S-K66A to any extent (Fig. 1L), demonstrating that ZYA requires K66 in the 225 intersubunit pocket as expected. This suggests that activation of T20S by ZYA occurs through 226 interactions similar to those responsible for activation by PAN. To rule out the possibility that ZYA 227 might be activating the protease sites rather than inducing gate opening, we also tested the 228 capacity of ZYA to stimulate the N-T20S, which lacks gate-residues and is constitutively open. 229 ZYA did not stimulate the activity of N-T20S to any extent (Fig. 1M). Together, these results 230 demonstrate that ZYA activates the 20S by inducing gate-opening, similar to the HbYX motif on 231 which it was based. 232 Elucidating ZYA's mechanism of activation using cryo-EM 233 Using cryo-EM, we generated a 1.9Å structure of ZYA bound to T20S (ZYA-T20S) (SFig. 1&2) 234 to the open-gate state (Fig. 2C), with the N-termini pointing up and a lack of density in the central 237 channel. This structure also resolves the N-termini of the  subunits up to Gly4, which include 238 three N-terminal residues that have not been previously resolved. The WT T20S map, as 239 expected, does not contain these open state densities but does show clear pore-central densities 240 that are expected for the closed gate (Fig. 2A&D). In addition, our new WT map also resolves the 241 N-termini up to Ala11, which includes 2 additional residues not previously resolved in the closed 242 state. In the ZYA-T20S map, densities corresponding to ZYA bound to the α intersubunit pockets 243 were clearly visible (Fig. 2F) and resembled the expected structure of this dipeptide. To determine 244 the intersubunit (around the ring) and intrasubunit (within a single subunit) conformational 245 changes induced by the binding of ZYA, we compared the ZYA-T20S structure to our 2.1Å 246 structure of the WT T20S (PDB:) (Fig. 2C, G, H; SMovie 1). For the intersubunit changes we first 247 aligned the subunits in the β-ring (Fig. 2H), to allow for visualization of changes in the α ring (no 248 significant changes were seen in the β subunits). These intersubunit conformation changes are 249 presented as a rotation of the α-ring (SMovie 1). This apparent rotation in the α-ring is primarily 250 due to individual rigid body movements of each α-subunit (Fig. 2C&G), centered around the length 251 of Helix 2 which acts as the pivot. Helix's 3, 4, and 5, which are most distant from the pivot move 252 by ~2Å. 253 In contrast, intrasubunit conformational changes were assessed by aligning a single α subunit 254 from WT T20S and T20S ZYA structures. Though subtle, these intrasubunit changes were 255 present. The Pro17 loop and connected N-terminal extension were shifted by ~1.0Å in a direction 256 perpendicular to Helix 0 (Fig. 2I), for comparison, the intersubunit change of Pro17 is 1.3Å so 257 most of the Pro17 shift comes from intrasubunit changes. Even Helix 0 shifted by about ~0.6Å 258 (1.0A for intersubunit changes) moving in the direction pointed towards the Pro17 loop (Fig. 2I). 259 These subtle intersubunit changes were also clearly visible in the electron density map (Fig 2C). 260 In addition, intersubunit conformational change was observed in the loop (S50-E65; back-loop), 261 which is in the outer portion of the intersubunit pocket, adjacent to K66 (Fig. 2J). Though the local 262 resolution of this loop is around 2.7Å, it clearly changes conformation, with the bottom of the loop 263 from I59 to K66 moving in the direction of K66 anywhere from ~1-2Å. This motion appears to be 264 causing or accommodating the rotation of the α subunits described above. We conclude that 265 binding of the HbYX-like ZYA molecule causes unique inter-and intra-subunit conformational 266 changes that allosterically switch the T20S from the closed to the open state. 267 Next, we analyzed the HbYX dipeptide's interactions with the intersubunit pocket to deduce its 268 mechanism for activating proteasome gate-opening (Fig. 2M). The carboxybenzyl group (i.e., Hb 269 group) docks in a hydrophobic pocket, interacting with V24, L21, and A154, 3.7Å, 4.3 Å and 3.6Å In addition to the above-mentioned ionic interactions, we also noticed the K66 and ZYA carboxy 279 group participate in a highly coordinated network of H-bonds due to the new position that K66 280 takes after ZYA binding. In fact, in addition to the K66 interaction the carboxy group of ZYA also 281 interacts with the backbone of G80 and S35, flanking both sides of the carboxy-K66 salt bridge 282 (Fig. 2N). Moreover, Lys66 is also H-bonds to the backbone of S35 and T78, also flanking both 283 sides of the salt bridge. It appears that this network of 6 distinct ionic interaction stabilizes the K66 284 in this new position, likely stabilizing the rearrangement of the K66 adjacent back loop. 285 The high resolution of these structures also allows us to model waters into the T20S. We therefore 286 analyzed how the binding of ZYA might affect water molecules in the intersubunit pockets, that 287 could potentially affect the conformation of the α subunits and the gate. Based on the modeled 288 waters, we found that ZYA's tyrosine displaced a water molecule that hydrogen bonds with the 289 backbone of A30 and another water that further hydrogen bonds with G19 (Fig. 2S). Concurrently, 290 we noted a new water positioned to hydrogen bond with the hydroxyl group of tyrosine, and the 291 backbone nitrogen of L21 and the side chain of E25, residues belonging to the neighboring α 292 subunit Helix 0. These water molecules are well positioned to potentially be important for 293 stabilizing the different states of the T20S gate (more below). 294 The ZYA-T20S model also showed that the binding of ZYA to the intersubunit pocket shortens 295 the distance between G19 on one subunit and K66 on the neighboring subunit by ~1Å (essentially 296 the walls of the intersubunit pocket are pulled together in the range of 1-2Å), primarily due to a 297 shift of K66 α carbon ( Fig 2J&M). Based on these interactions and changes in the pocket, we 298 deduce that ZYA binding acts as a "cable" that bridges across the intersubunit pocket connecting 299 Helix 0 of an α subunit to the K66 loop in the neighboring subunit. The length of this "cable" is just 300 short enough to "pull" or shift the K66 towards its new position (Fig 2J and N). The result of 301 displacing this K66 appears to be a rearrangement of the adjacent back-loop, resulting in the end 302 of the loop (e.g., L57) moving in a direction away from the neighboring subunit and towards K66 303 by ~1.2Å. In addition, the neighboring subunit Helix 3 follows the K66 loop to cause or 304 accommodate the rigid body rotation of the α subunit (SMovie 1). Thus, the intersubunit "bridging" 305 role ZYA plays leads to a K66 shift that promotes a rigid body rotation of the α subunits. 306 Interestingly, the rigid body rotation and intrasubunit shifts in Helix 0 combine to shift Helix 0 away 307 from the Pro17 that is in the neighboring subunit ( Fig 2L and SMovie 1). Since the base of Pro17 308 packs against Helix 0 in the neighbor, and since Pro17 is on a flexible loop, it is able to move with 309 its neighboring Helix 0 causing it to shift ~1.1Å, which is known to be associated with gate-opening. 310 We further tested these conformations induced by ZYA binding and new mechanistic insights via 311

mutagenesis. 312
To further elucidate ZYA's mechanism of activation, we mutated multiple residues to emulate or 313 perturb the conformational changes previously discussed and observed their effect on gate-314 opening by HbYX-dependent and -independent mechanisms. ZYA binding shifted the back-loop 315 proximal to K66 (Fig. 2J) towards K66, so we asked whether shortening the loop by a single 316 residue deletion affect gating. The deletion of I59 (∆I59) (Fig. 2K) resulted in slightly higher 20S 317 activity (p-value: 0.0013), e.g., a more open gate (Fig. 2K). In addition, neither PAN (a HbYX-318 dependent activator) nor PA26 (non-HbYX-dependent activator) could stimulate the activity of 319 T20S-∆I59 (Fig. 2K). This mutation suggests that the loop proximal to K66 does affect gating as 320 expected, demonstrating that this K66 back-loop is important for regulating gate-function. 321 Next, we asked what role the Hb (i.e., Z) group on ZYA plays in gate opening, does it contribute 322 to affinity, or does it actively play a role in inducing gate opening. To ask these questions, we 323 mutated V24 and A154 (both in the Hb binding pocket) to phenylalanine, emulating the binding of 324 the Z (benzene) in the pocket. Mutagenesis of these two residues to phenylalanine in Pymol 325 shows they would occupy overlapping space with the Z group of ZYA ( Fig 2O). Both T20S variants, 326 V24F and A154F did in fact have a far higher activity than the WT control ( Fig. 2P&Q), with V24F 327 being the most activating (~14fold). Additionally, PAN and P26 could neither further stimulate 328 V24F and A154F mutations (Fig. 2P&Q). This could be because the gate could not be further 329 opened, or it could be that altering the Hb binding sites prevents them from binding to the 20S. 330 Since V24F stimulated gate opening so strongly, even stronger than WT plus saturating PA26, 331 this suggests that introducing a large aromatic group in the Hb binding pocket by itself is sufficient 332 to cause maximal gate opening. To test this hypothesis, we generated a V24Y variant, mutating 333 V24 to another residue with a large aromatic group that is also observed to be in the Hb position 334 of the HbYX motif on some proteasome activators (e.g., human Rpt5, yeast Blm10, mammalian 335 PA200). As hypothesized, V24Y T20S had higher activity than the WT control and PAN and PA26 336 could not further stimulate (Fig. 2R). These indicate that the Z group of ZYA likely plays an 337 important and direct role in ZYA's mechanism of action. 338 Recognizing the possibility of water being involved in ZYA mechanism of activation, we sought to 339 further elucidate how they contribute to the conformation we observed. We mutated E25, a 340 residue described above to interact with a water molecule that further interacts with L21 and ZYA's 341 tyrosine. If E25's role in interacting with waters is critical, we predict that E25A would perturb ZYA 342 activation. Surprisingly, E25A T20S exhibits higher basal activity compared to WT T20S (Fig. 2T gating. However, the confidence of accurately identifying water molecules using cryo-EM, e.g., 348 compared to crystallography, is not high (even at 1.9Å), and thus other explanations could always 349 be possible. Despite this caveat, we did consistently see the same water densities in different 350 structures presented here (see below), and their expected displacement by ligand binding. 351 Realizing the pivotal role tyrosine played in our ZYA-T20S structure, and the partial activation 352 induced by the penultimate tyrosine mimicking L81Y mutation (Fig. 1G), we generated a cryo-EM 353 structure of T20S-αL81Y (2.3Å) to more rigorously evaluate how tyrosine contributes to gate-354 opening (SFig. 5&6) (Fig. 3A). Our EM map indicated expected densities corresponding to the 355 open state of the YDR region in our map ( Fig. 3C-red circle), which were not visible in our WT 356 T20S map (Fig. 3D). However, compared to the ZYA-T20S map which we presumed to have a 357 fully opened gate, the YDR densities in the L81Y map were weaker (compare Fig. 2E to 3C), 358 suggesting a partially opened gate, as previously indicated by our biochemical data (Fig. 1G). 359 To determine conformational changes induced by the single mutation mimicking the penultimate 360 tyrosine, we compared the T20S-αL81Y model against our WT T20S (Fig. 3B, E, F & G; SMovie 361 2) and ZYA-T20S models. Unlike the ZYA-T20S model, the conformational changes caused by 362 L81Y were different and subtle. We did not observe a substantial α-ring rotation or a 363 conformational change at the loop proximal to K66 (Fig. 3B, E, &H); however, we did observe a 364 slight rise of Helix 0 in a direction parallel with the 7-fold axis mostly clearly seen in Fig 3B. In fact, 365 we noted a slight rise of the entire surface of the α subunits away from the β ring, with the most 366 prominent changes in Helix 0 (Fig. 3B, E and F). Interestingly, alignment of individual α subunits 367 revealed minimal intrasubunit conformational changes at this resolution. We presume the subtler 368 effect is due to the slightly different placement of tyrosine in the L81 position compared to the 369 tyrosine in ZYA when bound to the T20S (compared Fig. 2M to 3I). Similar to ZYA-T20S, the L81Y 370 mutation caused minimal or no conformational changes in the β subunits ( Fig 3F&G). 371 The T20S-αL81Y model shows that the hydroxyl group of tyrosine is within proximity to hydrogen 372 bond with the backbone of G19, similar to ZYA's tyrosine. Additionally, we noted that tyrosine 373 hydrogen bonded with a water molecule that is also hydrogen bonding to the backbone of L21 374 but not the side chain of E25 (Fig. 3I). We also noted the water which hydrogen bonds with the 375 backbone of A30 is not displaced by L81Y, reinforcing the point that L81Y is not oriented like 376 ZYA's tyrosine. Collectively, our structure suggests that the interactions with G19 and L21 are 377 sufficient to cause a Helix 0 shift upward, which consequently at least partially opens the gate.  Our new cryoEM structure of the WT T20S now resolves more of the N-termini in the closed state, 391 showing clear density beyond the prior resolved T13 to also show the location of I12 and A11. In 392 this new T20S model, the T13 side chain occupies the space between Helices 0 and 2 (Fig. 4A, 393 E, and I) and I12 is seen binding to a hydrophobic pocket created by the neighboring subunits N-394 termini comprising A11, I12 and V14 (Fig. 4N). Interestingly, the ZYA-T20S model shows that T13 395 is pulled out of the pocket from under Helices 0 and now I12 binds into this same pocket, which 396 is mostly hydrophobic, containing A11, I12, and V14 ( Fig. 4 A

versus B, E versus F, I versus J; 397
SMovie 3). Simply put, I12 and T13 switch binding locations under Helix 0 to switch from the 398 closed to the open state. It appears that the ZYA-binding induced rotation of the α-subunit, that is 399 associated with the movement of Helix 0 and displacement of P17, "pulls" T13 out of the Helix 0 400 pocket, and I12 closer to this pocket, allowing it to bind in this position to stabilize gate opening 401 ( Fig 4Q and SMovie 3). Going forward, we refer to the I12, T13 motif and this switching 402 mechanism as the "IT switch". 403 We next looked at I12/T13 in the structure of the T20S-αL81Y and found that it was similar to the up to residue 11 with residues 1-10 being disordered, though the extent of this disorder is not 436

known. 437
We first mutated I12 to alanine (A), phenylalanine (F) or threonine (T) (Fig. 4R). The I12A mutant 438 is expected to reduce hydrophobic interactions with the neighboring subunits A11/V14 pocket in 439 the closed state, and indeed this mutant was about 3-fold more active than WT (Fig. 4R), which 440 is consistent with I12's role stabilizing the closed state. The I12F had less of an effect with ~2-fold 441 activation (Fig. 4R), which is consistent with more retention of hydrophobic interaction with its 442 neighbor despite the added mass. The I12T mutation showed 6-fold activation (Fig. 4R), 443 consistent with this polar residue not being supportive of the required hydrophobic interactions 444 with the neighbors A11 and V14 residues. We next mutated the T13 residue to alanine or 445 isoleucine (Fig. 4S). The T13A mutation again increased the basal activity of this mutant (~3.5 446 fold) (Fig. 4S), which would be expected with a loss of interactions with the IT switch pocket under 447 Helix 0 due to substitution with the much smaller alanine sidechain. Similarly, the T13I mutation 448 resulted in ~5-fold activation (Fig. 4S)

Conservation of the IT Switch in human 20S proteasome 479
While the IT switch is clearly important in archaea proteasome, does it play a role in regulating 480 the H20S? To answer this question, we compared the sequence of the T20S to the 7 different 481 H20S α subunits and determined if similar IT switch function was conserved in the H20S and 482 H26S proteasome. We found that indeed there is a high conservation of the IT switch motif (e.g., 483 a hydrophobic residue paired with a polar residue) in all α subunits except α3, which instead has 484 a conserved pair of threonine residues (i.e., "TT" instead of "IT"; SFig 7B). In addition, this IT motif 485 is separated from the critical YDR motif by exactly one residue in T20S and all seven human α 486 subunits. Additionally, the critical Pro17 is 3 residues away from the IT switch in all cases. 487 Interestingly, we found that α4's N-terminus is far more conserved with the N-termini of the T20S 488 α subunit then any of the other human α N-termini (SFig 7C&D), and correspondingly that α4's IT 489 switch is identically conserved with I and T residues. This is noteworthy because the T20S α N-490 termini all participate in gate closing and opening, since it is a homoheptamer, while primarily the it's T14 IT switch residue (at least in these models) stays in place under Helix 0 in both states. In 517 conclusion, the IT switch function uncovered in the T20S is well conserved in sequence and 518 presumably function in the H20S for 6 of the 7 αsubunits, especially when focusing on the role of 519 the hydrophobic residue switching to the Helix 0 pocket, as we found in the T20S. 520

Mechanistic differences between HbYX-dependent (ZYA) and HbYX-independent (PA26) 521
induced gate opening 522 To elucidate how HbYX (i.e., ZYA)-induced gate opening differed from PA26 induced gate-523 opening, we compared our ZYA-T20S model against the PA26-T20S model (PDB:1YA7) (Fig.  524   5A&B). Both activators show key commonalities and six significant differences in the way they 525 appear to induce gate-opening. Both ZYA and the C-termini interact with the base of the 526 intersubunit pockets via β-sheet like H-bonding (Fig 2M and 5D) 9 . More importantly, both ZYA 527 and PA26 trigger conformational change in the IT switch to stabilize gate opening (Fig 4), and 528 both cause the Pro17 to move away from the central pore "pulling" on the IT switch. The primary 529 difference in mechanism appears to be: 1) how the Pro17 gets moved away from the pore. When 530 PA26 binds to T20S, its activation loop displaces Pro17, without inducing any rotation in the α 531 subunits 9 . However, when ZYA binds, a different conformation is seen in the α ring. We overlayed 532 the T20S α ring from the PA26-bound and ZYA-bound states to see differences in the "gate-open" 533 states induced by these two activators (Fig 5A&B). Notice that Pro17 is, as expected, in similar 534 positions, however the areas of non-overlap show differences in conformation of the 20S activated 535 states. For example, relative to PA26 binding, ZYA-binding causes the α subunits to rotate around 536 the radial axis (SMovie 1), which is clearly visible in Fig. 5A (blue arrows). Interestingly, this ZYA-537 induced α rotation (combine with intrasubunit conformational changes) causes the Pro17 538 displacement that appears to trigger the IT switch to the open state (Fig 2). 2) ZYA-binding also 539 rearranges K66 (Fig 2J versus Fig 5C), but PA26 binding does not appear to do this. This 540 intriguing K66 rearrangement appears to be due to, 3) the limited length of the intersubunit 541 "bridging" of the YX residues in the HbYX motif described above (Fig 2M), combine with the 542 backbone H-bond network stabilizing this position (Fig 2N). We imagine that this lack of 543 intersubunit "bridging" across the pocket explains why PA26 does not cause the α ring rotation 544 that we observed in HbYX-dependent gate-opening. 4) The K66 adjacent back-loop gets 545 rearranged by ZYA binding, but not by PA26-binding (Fig 5A, see asterisks), which is also clearly 546 seen by comparing Fig 2J and Fig 5C (presumably due to K66 reorganization (Fig2N). 5) ZYA-547 binding rearranges water molecules in the intersubunit pockets differently than does PA26-binding 548 ( Fig 2S versus Fig 5E). In fact, the lack of a tyrosine residue in the penultimate position meant 549 that the C-terminus of PA26 did not displace or interact with the water molecules hydrogen 550 bonding with the backbone of G19 and A30 (Fig. 5E). Additionally, the lack of tyrosine meant that 551 there was no hydroxyl group present to interact with the water already hydrogen-bonding to the 552 side chain of E25 and the backbone of L21 (Fig. 5E). Thus, the binding of the HbYX motif appears 553 to rearrange waters in the intersubunit pocket differently than does PA26's C-termini. These 5 554 features we observe in the ZYA-bound T20S are not seen in the PA26-bound structure, leading 555 us to conclude that while both mechanisms converge on IT switch activation, they each trigger 556 the IT switch in separate ways. Collectively, these structures clearly distinguish the HbYX-557 dependent mechanism of activation from the PA26 mechanism, and highlight the causal 558 interactions involved in HbYX-specific gate-opening and the associated conformational changes. 559

ZYA activates mammalian proteasomes 560
We sought to determine if our findings with ZYA on the Thermoplasma 20S would translate to the 561 mammalian system; therefore, we tested Ac-LYR, Ac-LYA, and ZYA on M20S. In agreement with 562 the T20S outcomes, we observed significant improvements in gate-opening efficacy with each 563 modification (SFig. 8A). ZYA exhibited the greatest effect on the M20S, and a dose response with 564 ZYA showed nearly 50-fold activation at saturating levels ( Fig. 6A vs SFig. 8B). Although ZYA 565 can stimulate the M20S activity robustly, it has low affinity with a k obs of ~1mM (Fig. 6A), which is 566 highly similar to its affinity for the T20S (Fig. 1J). To confirm that ZYA activates via gate opening 567 in eukaryotic proteasomes, we asked if ZYA could activate the yeast open-channel mutant (α3∆N) 568 20S. We found that ZYA could not activate it at all but did activate WT yeast 20S as expected 569 (Fig. 6B), consistent with analogous experiments in the archaeal system. Interestingly, the 570 saturation curve for ZYA-induced proteasome activity for the M20S gave a cooperative binding 571 curve with a significant hill coefficient of 1.5 +/-0.1 (Fig. 6A) This indicates that ZYA binding is 572 cooperative and that binding to more than one site occurs during the allosteric induction of gate-573 opening. This is consistent with published cryo-EM structures of the 26S 18,19 that showed multiple 574 C-termini binding before the open-gate state is fully stabilized. While the minimal number of HbYX 575 motifs required to induce gate opening is unknown, the hill coefficient of 1.5 suggests that a 576 minimum binding of two molecules is involved. 577 We next tested the ability of ZYA to stimulate the M20S to hydrolyze three different peptide 578 substrates that are preferentially cleaved by 20S's three different protease sites (LLVY-amc, β5; 579 nLPnLD-amc, β1; LRR-amc, β2). We observed a significant. 10-to 50-fold increase, in the 580 hydrolysis rate of all three peptide substrates (Fig. 6C, compared to DMSO), which is expected 581 for a gate opener. In addition, to determine how well ZYA was able to induce gate-opening, we 582 compared it to M20S activation by PA26. Saturating PA26 concentrations typically stimulate 583 M20S activity by 30-100-fold, depending on the basal activity of the 20S preparation. We found 584 that ZYA was able to stimulate the M20S similar to that of PA26: ~50-fold activation by ZYA and 585 ~90-fold for PA26 (for nLPnLD-AMC). In addition, when we combined ZYA and PA26 in a single 586 reaction there was no synergy between these activators. Instead, there was a slight decrease in 587 activity, relative to PA26 alone, which is likely due to expected competition between ZYA and 588 PA26 binding to the intersubunit pockets. Collectively, these results demonstrate that ZYA is a 589 highly effective and robust gate-opening activator of the M20S proteasome from archaea, yeast 590 and mammals. These results support the hypothesis that this HbYX peptide mimetic functions 591 analogously to the highly conserved HbYX motif. 592

ZYA stimulates protein degradation by the mammalian 20S proteasome 593
We have demonstrated that ZYA robustly activates peptide hydrolysis via gate-opening but what 594 about that 20S's capacity to degrade unstructured proteins? To answer this question, we asked 595 if ZYA could stimulate the M20S proteasome to degrade tau23 (a truncated tau protein that is 596 found in brain) and the model unstructured protein-casein. ZYA significantly increased the 597 degradation of both proteins, as visualized by SDS-PAGE with Coomassie stain (Fig. 6D). Tau 598 and casein appeared to be completely degraded within the first 15 mins and 30 mins, respectively. 599 In support of these results, we also measured the peptides generated from the degradation of 600 14 C-casein by the 20S in solution (Fig. 6E) as measured by acid-soluble counts. In agreement 601 with the gel-based protein degradation assay, ZYA significantly increased the number of soluble 602 peptides from the degradation of 14 C-casein. These results clearly demonstrate that ZYA can 603 robustly stimulate the degradation of unfolded proteins. 604

Probing structure activity relationships of the ZYA gate-opening compound 605
To elucidate how ZYA might be binding to the human proteasome, we computationally docked 606 ZYA in the human α5/6 intersubunit pocket, where the HbYX motif of Rpt5 binds. Our docking 607 results (Fig. 6F) suggest that ZYA binds in a similar configuration to the HbYX motif of various 608 PAs (Fig. 2M), and as reflected in our ZYA-T20S structure. To evaluate the specificity and HbYX 609 motif-like requirements of ZYA for the M20S, we proceeded to probe ZYA's structure/function 610 relationships and efficacy through chemical modifications (Fig. 6G&I). First, we asked if additional 611 negative charges in the "X" position of the HbYX motif could be tolerated by replacing alanine with 612 acidic residues. We found that ZYE and ZYD failed to activate the M20S (Fig. 6G). We also tested 613 if causing a backbone torsion constraint (ZYP) and a polar group (ZYQ) might stabilize the 614 peptide's structure and binding, but they too abrogated M20S stimulation activity (Fig. 6G). Prior 615 studies and sequence conservation showed that small, medium, and large aliphatic, and basic 616 residues could all be tolerated in the "X" position, thus the "X-variable" designation 16 . Our data 617 with the ZYA also suggests there are some limitations (i.e., negative charges) in this position. 618 Prior studies 9,16,27 also showed the HbYX motif's terminal carboxy group forms an ionic bond with 619 K66. To validate that this was important for ZYA function in M20S, we blocked its C-termini 620 carboxyl with a NOH2 group (ZYA-[NOH2]). Carboxy blocking completely abrogated ZYA activity 621 (Fig. 6G), as expected for canonical HbYX motif function. Our ZYA-T20S structure indicated that 622 bridging of the intersubunit pocket by ZYA contributes to inducing conformational changes that 623 cause gate opening. To test this hypothesis in the M20S, we modified the tyrosine of ZYA to 624 lengthen the "bridging" distance by adding a nitro or phospho group to the tyrosine hydroxyl, which 625 would lengthen the bridging distance by 1 or 2 bonds respectively (Fig. 6H). Both ZpYA and 626 Z(nitro-Tyr]A failed to activate the 20S (Fig. 6I), supporting the bridging length requirement 627 between the tyrosine hydroxyl and carboxy C-termini of ZYA. This conclusion is further supported 628 by the fact that Z(4-amino-Phe)A still activates the 20S to a similar extent as ZYA (Fig. 6I)

ZYA abrogates the proteasome-inhibiting activity of three different A11+ oligomers 646
We recently published a study demonstrating that soluble oligomers of Aβ, α-synuclein, and 647 huntingtin(Q53), which share a common tertiary conformation (A11+), allosterically inhibit 20S 648 function by stabilizing the closed gate conformation 20 . Since ZYA stimulates gate opening, we 649 investigated the possibility that it could also rescue the M20S from inhibition by these 650 neurodegenerative-associated oligomers. Remarkably, when M20S activity is measured in the 651 presence of oligomers, the addition of 1mM ZYA completely blocks inhibition by A11+ oligomers 652 (Fig. 7B). Furthermore, even at 50M, ZYA stimulates oligomer inhibited M20S enough to bring 653 its basal activity back to WT levels (compare control with no ZYA to Aβ oligomers at 50M ZYA). 654 This is an unequivocal demonstration for the potential of small molecules to restore proteasome 655 activity in conditions of M20S impairment. In addition, compounds that function like ZYA could 656 also enhance degradation of intrinsically disordered proteins (IDPs) (Fig. 6D&E), which are often 657 found to be the aggregation prone proteins involved in neurodegenerative diseases. 658

DISCUSSION 659
Proteinopathies are associated with many diverse human diseases (e.g., Alzheimer's disease, 660 cardiomyopathies, and type II diabetes) and characterized by the accumulation of intracellular 661 constitutively open 20S proteasome, that could not be inhibited by these toxic oligomers 36 . These 672 worms with "hyperactive" 20S proteasomes have increased longevity and resistance to various 673 proteotoxicities, including heat shock and oxidative damage. Moreover, increasing proteasome 674 amount or activity in mammalian cell culture and multicellular organisms by various means has 675 also proven feasible, and to increase degradation of endogenous and ND related proteins [36][37][38][39][40][41] . 676 Together, these results motivated the current study to further elucidate the mechanism of how 677 HbYX-dependent proteasome activators induce 20S gate-opening and determine the 678 pharmacological tractability of this mechanism to potentially treat the aforementioned pathological 679

proteinopathies. 680
Several cryo-EM structures of the 26S 18,19,23,24 have been generated in the closed and open-gate 681 states but these structures alone could not explain how the HbYX motif induces gate-opening. 682 This is in part due to the complexity and dynamics of the interaction interface (described in the 683 introduction) that encompasses 7 different 20S αsubunits and six different 19S ATPases 684 subunits (RPT 1-6). To circumvent this mechanistic complexity, we first focused on the T20S from 685 archaea, which is homoheptameric (i.e. mechanistically simpler), and contains the same 686 conserved structural gating elements (e.g. N-terminal YDR motif, Pro17-reverse turn, closed and 687 open states). Based on the structure of various HbYX motif's bound to the proteasome (Fig 1), 688 the results of various T20S mutations, and a systematically developed peptides, we sought to 689 develop a minimal component of the HbYX motif-the peptide mimetic Z-YA. ZYA could only 690 activate T20S with a functioning gate, validating it as a bona fide gate-opening compound (Fig 1). 691 The 1.9Å cryo-EM structure of ZYA bound to the T20S also showed that it bound to the 692 intersubunit pockets precisely the same way as the HbYX motif, establishing ZYA as a minimal 693 HbYX mimicking compound. We further probed all three sub-binding pockets of ZYA to determine 694 the potential mechanistic contributions of the Hb, Y, and X positions, in inducing gate opening. 695 Combining the results from the mutagenesis and structures allowed us to develop a mechanistic 696 model for how the HbYX motif (or ZYA) is able to induce gate opening and uncover a new 697 important component of gating function-the IT switch, whose function is summarized in Figure  698

7C. 699
By comparing the structures of WT T20S (with newly resolved N-terminal residues), with ZYA-700 T20S, T20S-αL81Y, and previously published PA26-T20S, we develop a detailed molecular 701 model of how the HbYX motif induce gate opening without contacting Pro17 (as PA26 does with 702 its activation loops 9 ). In this model, the HbYX Tyrosine-Alanine connects two neighboring α 703 subunits by its tyrosine H-bonding to G19 and alanine carboxy salt bridging with K66 in the 704 neighboring α subunit. While this interaction has been noted previously, at this higher resolution 705 we observe a shift in the K66 side chain (Fig. 2J), it's stabilization by a network of ionic interactions 706 (Fig.2N), and a rearrangement of this loop at the back of the intersubunit pockets ( Fig. 2J&N; 707 D57-I67 or "back loop") which is naturally more mobile (SFig. 2 and 4). The length of this "bridge" 708 appears to be critical for gate-opening (Fig 6I), which supports the model that this back-loop 709 reconfiguration is important for gate opening. Moreover, K66 and a back-loop of proper length 710 (Fig. 2K) is required for gate-opening (Fig 1F), as the tyrosine by itself, T20S-αL81Y, did not 711 induce full gate-opening or α subunit rotations. Combined, these indicate an important role of the 712 "bridge" and the "back loop" for inducing α subunit rotations that cause gate-opening in the ZYA-713 bound state. 714 It is our working model that the reorganization of the back loop, which allows for the neighboring 715 Helix 0 to rotate towards its neighbor, sets off an allosteric chain reaction around the α ring. When 716 Helix 0 rotates away from its' neighbor's N-termini, it causes the neighboring Pro17 to also move 717 towards this helix to maintain its packing interaction, leading to both inter and intra subunit induced especially V24F. These mutants suggest that HbYX motifs with a bulky Hb position may be strong 728 activators of T20S gate-opening. However, a high-resolution structure of one of these mutants 729 will be needed to determine the mechanism of gate-opening by aromatic ring occupancy in the 730 Hb binding pocket. V24F could induce gate-opening by changing the Helix 0 position, as we saw 731 with the αL81Y mutation, or it could somehow induce α subunit rotations like ZYA does. Further 732 study is needed to confirm contribution to gating by the Hb pocket, though comparing activation 733 by LYA versus ZYA provides further insight. Both small peptides could induce gate opening in the 734 T20S (Fig 1J), but LYA presents a non-bulky aliphatic side chain to the Hb binding pocket, while 735 ZYA presents a bulky aromatic. The fact that LYA could induce gate-opening similar to ZYA 736 (though at lower affinity) indicates that perhaps intersubunit bridging by the YA is sufficient to 737 induce gate opening. However, providing a bulkier Hb group increases affinity and likely efficacy, 738 which was also a topic in a recent structural study from the Gestwicki group that used a chimeric 739 PA26-HbYX complex 28 . Taken all together, the Hb group likely plays an important role in 740 increasing affinity for the intersubunit pocket but likely also plays a direct mechanistic role in 741 helping induce gate-opening. Therefore, the effects of the Y-A bridge that reorganize the back-  (α1 and α5). What about Rpt2, which interacts with 756 α3? α3 has a modified IT switch (TT instead of IT) and the I12T mutation in T20S (TT) was more 757 open than WT but couldn't be activated by PA26 or PAN. This is consistent with α3 whose IT 758 switch is not engaged under Helix 0 stabilizing its closed state; instead, α3's N-termini is stabilized 759 by sitting on top of and interacting with the closed N-termini of α2 and α4 (SFig 7B) 18 . Based on 760 this unique α3 IT switch, and N-terminal position in the closed state, it's perhaps not surprising 761 that α3 appears to be desensitized to HbYX binding. Conversely, Rpt1 binding to α4 in the 762 activated H26S state (SFig 7) has been linked with gate opening 18 , and it carries a partial HbYX 763 motif (-TYN). Interestingly, α4 has perfectly conserved IT switch, and its N-terminal 1-34 residues 764 is uncannily highly conserved relative to the T20S α subunit, more so than any other H20S α 765 subunit's N-termini (SFig. 8B). These two conserved elements of α4s gating region suggest a 766 unique mechanistic importance for α4 and Rpt1 in controlling gate-opening in the eukaryotic 20S. 767 Perhaps Rpt1s effect on the α4 subunit alone is sufficient to trigger α4s IT switch, also affecting 768 α3 and 2, leading to gate opening, or perhaps contributions from other HbYX motifs are needed 769 to trigger an allosteric system. Further study is needed to test this hypothesis based on the 770 findings presented here. Nevertheless, it is apparent that the eukaryotic 26S gating system 771 evolved a spectrum of Rpt C-termini sequences, IT switches, and α N-termini to fine tune how 772 gate opening is controlled by substrate binding to the 26S proteasome. We expect that the 773 identification and function of the IT switch and the other HbYX relevant mechanism defined here 774 for the T20S will guide new understanding of how these mechanisms regulate the more 775 complicated 26S proteasome as suggested here. 776 Interestingly the mechanism uncovered here also shed light on how oligomers could impair the 777 20S proteasome, without being able to enter the internal chamber of the proteasome 20 . If the 778 HbYX mechanism requires a functioning back loop for activation, which is supported by our 779 structure and the activity of the T20S-I59, then this back loop could be a target for toxic 780 oligomers that impair proteasome function by blocking HbYX dependent gate-opening but not 781 PA26 induced gate-opening 20 . Our results here show that even at concentrations much less than 782 saturating, ZYA is effective at reversing proteasome inhibition by three different ND-related 783 proteins (Fig. 7B). We expect compounds that could similarly induce gate opening at more 784 physiologically relevant affinities could potentially treat ND (Fig. 7D). Such compounds could both 785 stimulate degradation of most ND-related proteins, which are typically IDPs, and simultaneously 786 reverse proteasome impairment, which has been observed in aging. Advancements in 787 understanding the proteasomes gate-regulatory mechanisms provide a framework for the 788 development of small molecules that antagonize proteasome impairment, especially by oligomers, 789 or activate its ability to degrade unstructured proteins. Considering the importance of the The final concentration of DMSO in activity assays was 2%. Oligomers of Aβ*56, α-synuclein, and 817 huntingtin-Q53 were prepared as described 20 . All oligomers used were recognized by the α-

Cryo-EM Sample Preparation and Data Collection 840
Copper Quantifoil R 1.2/1.3 300 mesh (EMS) grids were cleaned using a PELCO easiGlow Glow 841 Discharge cleaning system. A volume of 3 uL of 0.5mg/mL WT T20S, T20S-αL81Y or T20S with 842 4mM ZYA (suspended in 50mM Tris pH 7.4, 150mM NaCl) sample was placed onto a grid, and 843 then flash frozen in liquid ethane using a manual plunge freeze apparatus. Data collection was 844 done using a Titan Krios transmission electron microscope (Thermo Fisher) operated at 300kW 845 and a magnification of x81,000, which resulted in 0.503Å/px. Images were collected using a 846 Falcon IIIEC direct electron detector camera equipped with a K3/GIF operating in counting and 847 super resolution modes. Electron dose per pixel of 50 e-/Å2 was saved as 40 frame movies within 848 a target defocus range of -2.5 to -1.25. All the data was collected using cryoSPARC software 849 (Structura Biotechnology Inc.) 45 . 850 851

Cryo-EM Single Particle Analysis 852
Cryo-EM images of the WT, T20S-αL81Y, and ZYA-T20S proteasome were analyzed using 853 cryoSPARC. Schematic for cryo-EM single-particle data processing available in supplement. 854 WT T20S: From 1744 movies collected, we picked 444,678 particles after four rounds of 2D 855 classification, which were used to generate an Ab-initio model and processed through 856 heterogenous refinement then homogenous refinement (using D7 symmetry). 857 T20S-αL81Y: From the 2850 movies collected, we used 2847 in analysis and picked 889,069 858 particles after three rounds of 2D classification to obtain the best particle sets. The particles 859 chosen from 2D classification were used to generate an Ab-initio model, which was used for 860 homogeneous refinement (using D7 symmetry). 861 ZYA-T20S: From movies collect, we used 830,572 particles after two rounds of 2D classification. 862 Particles isolated were used to generate an Ab-initio model, processed in a heterogenous 863 refinement, then homogenous refinement (using D7 symmetry). 864 Final map was imported into Phenix 46 to run density modification (DenMod) from two half maps. 865 All representations (figures and movies) of the T20S proteasome complex were created using 866 PyMol 2.5.2, WinCoot 0.9.6 EL, and UCSF ChimeraX v1.3 47,48 . 867 868

Atomic model building 869
The atomic models were built using a modified version of the T20S from PDB: 1YA7 as a template, 870 rigid body fitting into the electron density map using PHENIX 1.19.2-4158. The docked models 871 were subjected to a cycle of morphing and simulated annealing, five real-space refinement 872 macrocycles with atomic displacement parameters, secondary structure restraints, and local grid 873 searched in PHENIX. Consequently, the models were refined by oscillating between manual real-874 space refinement in WinCoot 0.9.6 EL and real-space refinement in PHENIX (five macrocycles, 875 without morphing and simulated annealing     A. Top view overlay of α subunits only from PA26-T20S (green) and T20S-ZYA (blue). Red arrows show Pro17 loop. Misalignments show that different conformational changes are present in ZYA bound versus PA26 bound states, e.g. alpha helixes (blue arrow), and differences in back loops (*).
C. View of loop proximal to K66, including overlay of PA26-T20S (green) with WT T20S (yellow) with key residues shown in sticks.