Structural and mechanistic insights into Hsp104 function revealed by synchrotron X-ray footprinting

Hsp104 is a hexameric AAA+ ring translocase, which drives protein disaggregation in nonmetazoan eukaryotes. Cryo-EM structures of Hsp104 have suggested potential mechanisms of substrate translocation, but precisely how Hsp104 hexamers disaggregate proteins remains incompletely understood. Here, we employed synchrotron X-ray footprinting to probe the solution-state structures of Hsp104 monomers in the absence of nucleotide and Hsp104 hexamers in the presence of ADP or ATPγS (adenosine 5′-O-(thiotriphosphate)). Comparing side-chain solvent accessibilities between these three states illuminated aspects of Hsp104 structure and guided design of Hsp104 variants to probe the disaggregase mechanism in vitro and in vivo. We established that Hsp104 hexamers switch from a more-solvated state in ADP to a less-solvated state in ATPγS, consistent with switching from an open spiral to a closed ring visualized by cryo-EM. We pinpointed critical N-terminal domain (NTD), NTD-nucleotide–binding domain 1 (NBD1) linker, NBD1, and middle domain (MD) residues that enable intrinsic disaggregase activity and Hsp70 collaboration. We uncovered NTD residues in the loop between helices A1 and A2 that can be substituted to enhance disaggregase activity. We elucidated a novel potentiated Hsp104 MD variant, Hsp104–RYD, which suppresses α-synuclein, fused in sarcoma (FUS), and TDP-43 toxicity. We disambiguated a secondary pore-loop in NBD1, which collaborates with the NTD and NBD1 tyrosine-bearing pore-loop to drive protein disaggregation. Finally, we defined Leu-601 in NBD2 as crucial for Hsp104 hexamerization. Collectively, our findings unveil new facets of Hsp104 structure and mechanism. They also connect regions undergoing large changes in solvation to functionality, which could have profound implications for protein engineering.

Each Hsp104 monomer is composed of an N-terminal domain (NTD), 4 two AAA ϩ nucleotide-binding domains (NBD1 and NBD2) that bind and hydrolyze ATP, a coiled-coil middle domain (MD) inserted in NBD1, and a short C-terminal region (15). The functional Hsp104 hexamer is large and dynamic with a solvent-filled channel running the length of the assembly (16 -20). In response to ATP binding and hydrolysis, the hexamer undergoes substantial changes in conformation (9,16,17,19). Indeed, cryo-electron microscopy (cryo-EM) has revealed that Hsp104 hexamers populate an open "lock-washer" spiral state in the presence of ADP, but switch to closed ring structures that surround polypeptide substrate inside the channel in the presence of ATP␥S (16,17). These conformational changes are proposed to drive polypeptide translocation across the central channel, which enables the remodeling of diverse substrates, from thermally-denatured aggregates (20 -22) to amyloid conformers (20,(23)(24)(25)(26)(27). However, to fully understand and validate the details of these conformational changes, and to understand the mechanisms that underpin substrate remodeling, Hsp104 structure and dynamics need to be probed using diverse and complementary methodologies (28 -30).

Peptide overview
Samples of the Hsp104 monomer (hexamer dissociated by high salt) and the Hsp104 hexamer in the presence of either ADP or ATP␥S were exposed to millisecond bursts of synchrotron X-rays and analyzed using MS as described under "Experimental procedures." The monomer, hexamer with ADP, and hexamer with ATP␥S have 339, 359, and 342 peptides found in all three replicates of all six time points, respectively (Table 1). Moreover, 122 (monomer), 135 (ADP), and 78 (ATP␥S) of those peptides showed time-dependent modification, although some were omitted from our analysis due to poor signal to noise. After omitting these peptides, we were left with 81 (monomer), 117 (ADP), and 54 (ATP␥S) peptides that showed time-dependent modification that could be fit to a first-order decay curve to yield rate information (Table 1, example of MS data and curve fitting in Fig. 1). Tables S1-S3 list all of the filtered modified and unmodified peptides for the monomer, hexamer with ADP, and hexamer with ATP␥S, respectively. In instances where the modified peptide was selected for fragmentation and MS 2 analysis, the identity of the specific site of modification was determined. The majority of identified peptides are unmodified. For a direct comparison of rates, we need to examine the rate information for the exact same peptide for all the states. Table 2 lists identical peptides found in two or three of the states, which cover some of the regions of interest.
To begin our analysis of the XF solvation data of the Hsp104 hexamer and to extract relevant mechanistic implications, we assessed each Hsp104 domain in turn. We examined the following: 1) regions of known functional importance; 2) regions that are heavily modified, and therefore solvated; and 3) regions that undergo large changes in solvation between states. We then used the XF data to make mechanistic predictions about regions in each domain. These predictions were then tested via mutational analyses. An overview of the Hsp104 variants created and tested is shown in Table 3.

NTD is involved in substrate binding and productive interaction with Hsp70
We have established that the NTD of Hsp104 is crucial for hexamer cooperativity (19). The presence of the NTD enables Hsp104 to dissolve amyloid and is critical for potentiated Hsp104 variants to rescue TDP-43, FUS, and ␣-synuclein toxicity in yeast (19). Nonetheless, how the NTD confers these functions remains unclear. The crystal structures of the NTDs of various Hsp100 proteins, including Hsp104 from S. cerevisiae (38) and Calcarisporiella thermophila (18) and the bacterial Hsp100 homologues ClpA (39), ClpB (40), and ClpC (41), have been solved. The NTDs are highly structurally conserved despite low sequence identity (e.g. 28% sequence identity between the ClpA and ClpB NTDs). The NTDs are stable globular domains composed of two imperfect repeats of four helical bundles (R1 and R2 each contain four helices, A1-A4 and A5-A8) and are connected to NBD1 by a highly-mobile linker (18,39).
A conserved feature of the Hsp100 NTDs is a hydrophobic patch between helices A1 and A5 (39). Interestingly, this hydrophobic patch region is solvated in all three states (peptide 86 -110, Fig. 2, A and B, Table 2, and Tables S1-S3). In fact, four hydrophobic residues in this hydrophobic patch region were identified by MS 2 analysis as being specific sites of hydroxyl radical modification (residues Ala-91, Leu-92, Val-95, and Leu-96, Table 2, and Table S3, Fig. 2C shown as spheres and labeled in purple). This finding indicates that this conserved region is available to interact with substrate in both the ADP and ATP␥S states.
To test whether the NTD of Hsp104 is involved in substrate binding, we measured the affinity for a fluorescent substrate, FITC-casein, using a fluorescence polarization assay. We used the purified wildtype protein (WT) and a truncation mutant (Hsp104 ⌬N ) missing the NTD as well as the NTD-NBD1 linker (19). Under these conditions, in the presence of ATP␥S, Hsp104 binds FITC-casein robustly with a K d of ϳ55 Ϯ 7 nM (Fig. 2D) (16,19). In the same conditions, Hsp104 ⌬N binds FITC-casein with a K d of ϳ194 Ϯ 60 nM, showing an ϳ3-4-fold decrease in affinity (Fig. 2D) (19). This result supports a role for the Hsp104 NTD in binding certain substrates.

Synchrotron X-ray footprinting of Hsp104
Additional residues that were identified as oxidatively modified by MS 2 analysis fall into two clusters: (a) in the beginning of the A5 helix and in the loop between helices A6 and A7, (residues Thr-87, Tyr-90, Lys-94, Ser-124, and Ser-125 high-lighted in light blue in Fig. 2, B and C), or (b) at the end of helix A5 and in the helix A5-A6 loop (residues Ile-102, Gln-103, Lys-107, and Ser-109 highlighted in green in Fig. 2, B and C). These two clusters of residues are in the same regions that were Figure 1. X-ray exposed Hsp104 mass spectra and curve fitting. Unmodified and modified versions of the Hsp104 peptides were identified by ExMS-CL in each time point. The intensities of the singly-modified peptides were pooled, and the fraction unmodified was calculated for each time point. A pseudo-firstorder decay curve extrapolated to zero was then fit to the data.

Table 2
Selected peptide list Peptides that are found in two or more states, which cover important regions of Hsp104, are shown. Normalized modification rates and modified residues are listed side by side for ease of comparison. ND means not determined and pnf means peptide not found.

Monomer
Hexamer with ADP Hexamer with ATP␥S

Synchrotron X-ray footprinting of Hsp104
identified in ClpA (which lacks a MD) to interact with an adapter protein ClpS (39), which targets ClpA to aggregated substrates (42). In another Hsp100 protein, ClpC, which has a truncated version of the MD (motif 1 only), the crystal structure of the hexamer revealed that the adapter protein MecA made extensive contacts with both the MD as well as the NTD (41). Thus, we wondered whether Hsp104 might also interact with Hsp70 via the NTD in addition to proposed contacts with the MD (43)(44)(45)(46). Indeed, one region of the Hsp104 MD found to interact with Hsp70, namely residues within motif 2 near the C-terminal end of helix L2 (47), is adjacent to the NTD in hexameric models of Hsp104 (16 -18, 34, 48). Additionally, residues 100 -102 (which contain the oxidatively modified Ile-102, Fig.  2, A and C) were identified as a potential Hsp70-binding site using peptide array technology (47), although this region was not investigated beyond an initial identification. NTD/Hsp70 interactions may facilitate substrate release from Hsp70 for hand-off to Hsp104. Indeed, Hsp104 ⌬N has reduced ability to disaggregate and reactivate chemically-denatured luciferase in the presence of Hsc70 and Hdj2 (ϳ40% of WT) or Hsp72 and Hdj2 (ϳ86% of WT) (19). Thus, the NTD enables productive interactions with Hsp70. The difference in activity between Hsp104 ⌬N in the presence of Hsc70 versus Hsp72 indicates that the Hsp104 hexamer may interact differently with different Hsp70 proteins. This distinction is important because Hsc70 is constitutively expressed and involved in maintaining protein homeostasis, whereas Hsp72 is induced by heat shock and facilitates response to acute cellular stress. Next, we used the residue-specific information from the XF solvation to design several Hsp104 NTD mutants to assess contributions of specific NTD residues to interactions with substrate or Hsp70. We built five Hsp104 variants with clusters of mutations in NTD regions suggested to be important by the XF data and measured their ability to reactivate luciferase trapped in denatured aggregates. We assessed Hsp104 disaggregase activity under three conditions: (a) in the absence of Hsp70 and Hsp40, but in the presence of 1:1 ATP/ATP␥S; (b) in the pres-ence of ATP and Hsp40/Hsc70; and (c) in the presence of ATP and Hsp40/Hsp72 (19). In the presence of ATP, Hsp104 is inactive in disaggregating luciferase in the absence of Hsp70 and Hsp40 in vitro (20,22). However, Hsp104 can be activated in the absence of Hsp70 and Hsp40 by the addition of permissive ratios of ATP/ATP␥S (19,20,25,49). Hsp104 has optimal activity at an ϳ1:1 ratio of ATP/ATP␥S (20,49). Thus, a 1:1 ratio of ATP/ATP␥S allows assessment of intrinsic Hsp104 disaggregase activity in the absence of Hsp70 and Hsp40. Hence, we could distinguish whether particular NTD mutations disrupted intrinsic Hsp104 disaggregase activity, collaboration with Hsp70, or both.
Hsp104-LVL contained mutations to hydrophobic residues found to be modified within the conserved hydrophobic patch (L92A/V95A/L96A; Fig. 2, A and C). In the presence of ATP and ATP␥S, Hsp104-LVL displayed a significant defect in rescuing aggregated luciferase as expected if the region is involved in substrate binding (Fig. 2E). Hsp104-LVL ATPase activity was only mildly decreased from WT Hsp104 (ϳ70% WT activity), which might also contribute to reduced disaggregase activity (Fig. 2F). However, this defect in luciferase reactivation was mitigated by the presence of either Hsp70 variant (Fig. 2E). Thus, substrate-binding defects, mild ATPase defects, or both in Hsp104 can be buffered by the presence of Hsp70 and Hsp40.
Hsp104-TYK (T87A/Y90A/K94A), Hsp104-QS (Q103A/ S109A), Hsp104-IK (I102A/K107A), and Hsp104-SS (S124A/ S125A) harbor mutations to residues found to be modified and within regions that might be involved in interactions with Hsp70 (Fig. 2, B and C). Generally, Hsp104-SS exhibited similar disaggregase activity to WT Hsp104 although in the presence of Hdj2/Hsc70 the modest decrease was statistically significant (Fig. 2E). Hsp104-SS exhibited similar ATPase activity to WT Hsp104 (Fig. 2F). Thus, Ser-124 and Ser-125 are unlikely to play a major role in interactions with substrate or Hsp70. Thr-87, identified as oxidatively modified in the hexamer with ATP␥S (Fig. 2, A and C), was an interesting target. This residue is just N-terminal to helix A5 (Fig. 2C). In ClpA, an equivalent residue, Table 3 Overview of Hsp104 variants Based on the X-ray footprinting solvation data, Hsp104 variants were designed and purified. Their biochemical properties were then assessed by measuring ATPase activity, intrinsic disaggregase activity (ATP/ATP␥S), and disaggregase activity in collaboration with Hsp40/Hsp70 pairs (Hdj2/Hsc70 or Hdj2/Hsp72). In this table Ϯ indicates no significant difference from WT Hsp104; -denotes a statistically significant decrease in activity down to more than 60% WT Hsp104 activity; -denotes a statistically significant decrease down to 30 -60% WT Hsp104 activity, and -denotes a statistically significant decrease to less than 30% of WT Hsp104 activity. Similarly, ϩ signifies a statistically significant increase up to less than 150% WT Hsp104 activity; and ϩϩ signifies a statistically significant increase greater than 200% WT Hsp104 activity. ND indicates not determined.

Figure 2. Hsp104 NTD is involved in substrate and Hsp70 interactions.
A, peptide sequences and modification rates for a region of the Hsp104 NTD that is highly modified. Residues shown in bold have been identified as modified by MS 2 and are colored by the state in which they were found: yellow, monomer; blue, hexamer with ADP; red, hexamer with ATP␥S; purple, both hexameric states; green, monomer and hexamer with ADP; orange, monomer and hexamer with ATP␥S; black, modified in all states. B, surface rendering of Hsp104 NTD from the crystal structure of the S. cerevisiae Hsp104 NTD (PDB code 5u2u). Acidic residues are shown in red, basic residues in blue, and hydrophobic residues in yellow. Circled regions are important in the structurally-conserved ClpA (PDB code 1r6c) NTD for ClpS binding as follows: site A in green, site C in light blue, and the conserved hydrophobic patch for substrate binding in purple. C, Hsp104 NTD with residues identified by MS 2 as modified in a hexameric state are shown as spheres and labeled when visible. Spheres were colored by the states in which they were found modified, as in A, and labeled with colors corresponding to the regions outlined in B. D, FITC-casein, 2 mM ATP␥S, and either Hsp104 or Hsp104 ⌬N were incubated for 20 min at 25°C. Fluorescence polarization was measured at increasing concentrations of Hsp104, and a single site K d was calculated using a least-squares fit to the data. Each data point represents mean

Synchrotron X-ray footprinting of Hsp104
Thr-81, resides within the main ClpS-binding site (39), and in ClpC, the residue Thr-31 is essential for ClpC activity by mediating the interaction with MecA (41). Hsp104-TYK, Hsp104-QS, and Hsp104-IK all display disaggregation defects in the presence of ATP/ATP␥S and Hsp40/Hsc70 (Fig. 2E). Although the defect in the presence of ATP/ATP␥S was unexpected, it is more severe in the presence of Hsp40/Hsc70 (Fig. 2E). Additionally, Hsp104-TYK displays a defect in activity in the presence of Hsp40/Hsp72 (Fig. 2E). Hsp104 ATPase activity was not grossly affected by these mutations (Fig. 2F). Thus, these regions of the Hsp104 NTD play important roles in the substrate-binding and Hsp70-collaboration steps necessary to disaggregate substrate.

NTD/NBD1/MD interactions that include the NTD-NBD1 linker contribute to Hsp104 activity
To understand how the NTD may be communicating with neighboring domains and how these interactions may change depending on the identity of the nucleotide, we first looked for regions that undergo changes in solvation between different states. These regions include the beginning of helix A1 (residues 8 -15), the loop between helices A2 and A3 (residues 41-51), the loop between helices A3 and A4 (residues 58 -61), and the linker between the NTD and NBD1 (residues 148 -166) (Fig. 3, A and B). A peptide that covers the beginning of helix A1 starting at residue 8 appears to be unmodified in the hexamer with ADP, but it has a rate of 1.85 Ϯ 0.33 s Ϫ1 in the hexamer with ATP␥S (Table 2 and Fig. 3A). The loop between helices A2 and A3, residues 41-51, is covered by the peptide 39 -53, which has rates for the ADP and ATP␥S hexamers of 25 Ϯ 0.005 and 4.3 Ϯ 1.5 s Ϫ1 , respectively ( Table 2 and Fig. 3A). The loop between helices A3 and A4, residues 58 -61, is covered by the peptide 55-64. This peptide is unmodified in the monomer, and it has rates for the ADP and ATP␥S hexamers of 3.6 Ϯ 0.009 and 0.85 Ϯ 0.18 s Ϫ1 , respectively (Table 2 and Fig. 3A). The linker between the NTD and NBD1 is covered by the peptide 148 -166 in all three states (Table 2 and Fig. 3A). It is unmodified in the hexamer with ATP␥S, has a rate of 0.101 Ϯ 0.002 s Ϫ1 in the hexamer with ADP, and 0.215 Ϯ 0.014 s Ϫ1 in the monomer (Table 2 and Fig. 3A).
We then compared these regions from XF to the three tClpB monomer models (37). In the crystal structure of tClpB, each monomer of the spiral trimer (designated models A-C in the PDB) has a different orientation of the NTD (37). By examining the different orientations of the NTD, we discovered regions of the NTD, NBD1, and MD that appeared capable of interacting in one or all of the three models. The homologous Hsp104 NTD residues include 8 -20 found in models A and B, NBD1 residues 264 -274 found in models A and C, MD residues 496 -498 found in models A and C, and NTD residues 39 -52, NTD/ NBD1 residues 155-169, and NBD1 residues 230 -240 found in all three of the models. Many of these regions display changes in solvation, as noted above. Additionally, in a hexameric model of Hsp104 generated using cryo-EM data, which models in the NTD and different MD orientations, these regions are found in close proximity (Fig. 3B) (16). These include the beginning of helix A1 residues 8 -15 in particular (Fig. 3, A and B, yellow), which are protected from solvation in the hexamer with ADP, the loop between helices A2 and A3, residues 39 -53 (Fig. 3, A and B, orange), which shows a decrease of solvation in the hexamer with ATP␥S, and the NTD-NBD1 linker region, 148 -166, which shows a protection from solvation in the hexamer with ATP␥S. Consistent with this observation, formation of Hsp104 hexamers slows hydrogen-deuterium exchange (HX) of the NTD-NBD1 linker (30).
If we investigate the regions in NBD1 and the MD, we find that NBD1 residues 230 -240 (Fig. 3, A and B, blue) are covered by a number of peptides, which are unmodified in both hexameric states (Tables S2 and S3). In fact, the NBD1 peptide 225-243 is unmodified in the hexamers, whereas the monomer has a modification rate of 15.9 Ϯ 11.2 s Ϫ1 ( Table 2 and Fig. 3A). NBD1 residues 262-274 display no clear trend in modification rate (Tables S1-S3), and MD residues 496 -498 (Fig. 3B, green), covered by the peptide 494 -504, are found unmodified in the monomer and the hexamer ATP␥S, but the peptide was not found in the hexamer with ADP (Tables S1-S3). These findings indicate that the NTD, specifically the beginning of helix A1 in the hexamer with ADP, and the A2-A3 loop and NTD-NBD1 linker in the hexamer with ATP␥S may be making a stable interaction with NBD1 residues 230 -240 and potentially the MD, in the loop between MD helices L2 and L3 in motif 2 ( Fig.  3B). Thus, different regions of the NTD can contact the same regions of NBD1 and the MD. The NTD is involved in regulating ATPase activity and is essential for global cooperativity of the hexamer (19). Thus, we suggest that these interactions may relay signals from the NTD to the rest of the hexamer that are elicited by substrate interactions, Hsp70 interactions, or both.
Further support of this model is provided by the mutant, T160M, which resides in a region of the NTD-NBD1 linker proposed to interact with NBD1, the MD, or both. T160M phenocopies the effects of NTD deletion in vivo (50). Additionally, other regions identified by peptide array as potential Hsp70binding sites include NTD residues 34 -39 (47), just N-terminal of the A2-A3 loop, and 241-249 (47), just C-terminal of the unmodified 230 -240 NBD1 region, indicating that there may be a complex network of NTD-NBD1-MD/Hsp70 interactions (Fig. 3B). Mutations in this region of the MD, specifically D498V (51), T499D (52), and T499E (52), and also in the relevant region of NBD1 I230N (53), lead to potentiated Hsp104 variants, and deletion of the NTD precludes this potentiation (19). These findings suggest that this interface is important for hexamer regulation and activity. To test our proposed network of interactions, we made several Hsp104 variants with mutations in the regions suggested by the XF data, as well the T160M mutant (50).
Next, we tested Hsp104-ED (E44R/D45R), which bears mutations in the loop between helices A2 and A3 in the NTD Synchrotron X-ray footprinting of Hsp104 ( Fig. 3, A and B). Unexpectedly, Hsp104-ED exhibited slightly elevated disaggregase activity in the presence of ATP/ATP␥S or Hdj2/Hsc70 and WT disaggregase activity in the presence of Hdj2/Hsp72 (Fig. 3D). Hsp104-ED also displayed mildly reduced ATPase activity (Fig. 3E). Thus, the loop between heli-ces A2 and A3 can be targeted to increase disaggregase activity under some conditions.
Hsp104-TM (T160M) and Hsp104-RRRD (R148D/R152D/ R156D) are both found in the linker between the NTD and NBD1. Like WT Hsp104, Hsp104-TM and Hsp104-RRRD are A and B, based on the solvation data from XF, there may be regions of the Hsp104 NTD, NBD1, and MD that make contacts that are nucleotide-dependent. In the NTD, these regions are the beginning of helix A1 in the hexamer with ADP (residues 8 -15 in yellow), the A2-A3 loop (residues 39 -53 in orange), and NTD-NBD1 linker (not pictured) in the hexamer with ATP␥S. Based on: (1) the XF data, (2) the three monomeric tClpB crystal structures (PDB code 1qvr), and (3) the hexameric cryo-EM reconstruction (PDB code 5vy9, shown in B), we suggest that these NTD regions may be making a stable interaction with NBD1 (specifically residues 230 -240 in blue) and in the hexamer with ATP␥S the MD in the loop between MD helices L2 and L3 in motif 2 (residues 496 -498 in green). Shifts in these contacts in response to stimuli such as Hsp70 binding, substrate binding, or ATP hydrolysis could transmit signals through the hexamer and facilitate cooperativity. In the table with rate information, pnf denotes "peptide not found."

Synchrotron X-ray footprinting of Hsp104
inactive in luciferase disaggregation in the presence of ATP alone (Fig. 3C). They both display reduced disaggregase activity in the presence of ATP/ATP␥S, Hdj2/Hsc70, and Hdj2/Hsp72 (Fig. 3D). Hsp104-TM exhibits reduced ATPase activity, whereas Hsp104-RRD does not (Fig. 3E). Thus, the NTD-NBD1 linker is crucial for optimal disaggregase activity, and the Thr-160 residue is involved in regulating the ATPase activity of the hexamer.
We next assessed Hsp104-DDD (D231A/D232A/D233A), which bears three mutations in NBD1 (Fig. 3, A and B). Like WT Hsp104, Hsp104-DDD is inactive in luciferase reactivation in the presence of ATP alone (Fig. 3C) and exhibits normal ATPase activity (Fig. 3E). Hsp104-DDD displays reduced disaggregase activity in the presence of ATP/ATP␥S (Fig. 3D). Most strikingly, however, Hsp104-DDD is completely inactive in the presence of Hsp70 (Fig. 3D). By mutating these three acidic residues, we have created an Hsp104 variant that retains intrinsic disaggregase activity but is unable to collaborate with Hsp70, establishing a critical role for Asp-231, Asp-232, and Asp-233 in Hsp70 collaboration.
Remarkably, however, Hsp104 -RYD rescued ␣-synuclein (which is connected to Parkinson's disease), FUS (which is connected to ALS and frontotemporal dementia (FTD)), and TDP-43 (which is also connected to ALS and FTD) toxicity in yeast, unlike WT Hsp104 (Fig. 4A). Hsp104 -RYD rescued ␣-synuclein and FUS toxicity just as well as Hsp104 A503V (51,54), whereas rescue of TDP-43 toxicity was more robust for Yeast were serially diluted 5-fold and spotted in duplicate onto galactose (inducing) and glucose (noninducing) media. B, samples were also processed for immunoblotting to assess Hsp104, PGK1 (loading control), and ␣-synuclein-YFP, FUS, or TDP-43 expression. Molecular mass markers are indicated (left). C, quantification of immunoblots from B. Hsp104 levels (red bars) were normalized to PGK (loading control), and the relative expression level compared with Hsp104 WT was determined. ␣-Synuclein-YFP, FUS, and TDP-43 levels (blue bars) were normalized to PGK (loading control), and the relative expression level compared with vector control was determined. Data are displayed as scatterplot with bar representing mean Ϯ S.D. (n ϭ 3). D, Hsp104 -RYD is not toxic to yeast at 30°C. Empty vectors, Hsp104 (WT), Hsp104 A503V , or Hsp104 -RYD plasmid were transformed into W303a⌬hsp104 yeast strains. Yeast were serially diluted 5-fold and spotted in duplicate onto galactose (inducing) and glucose (noninducing) media.

Synchrotron X-ray footprinting of Hsp104
Hsp104 A503V (Fig. 4A). Hsp104 -RYD rescued the toxicity of these disease proteins, without grossly reducing their expression level (Fig. 4B). Indeed, quantification of the immunoblots revealed that Hsp104 -RYD only slightly reduced ␣-synuclein levels to ϳ86% of the vector control (Fig. 4C). Likewise, Hsp104 -RYD only slightly reduced FUS levels to ϳ82% of the vector control (Fig. 4C). Hsp104 -RYD had no effect on TDP-43 expression (Fig. 4C). These slight reductions in ␣-synuclein and FUS expression are unlikely to explain the strong suppression of toxicity conferred by Hsp104 -RYD. Indeed, we have previously shown that reduced FUS expression is not required for enhanced Hsp104 variants to rescue FUS toxicity (53). Moreover, Hsp104 -RYD was typically expressed at a lower level of ϳ61-77% of the WT Hsp104 control. Thus, even though ␣-synuclein and FUS expression was slightly lower, so too was expression of Hsp104 -RYD. Importantly, expression of Hsp104 A503V or Hsp104 -RYD was not toxic to yeast in the absence of disease protein (Fig. 4D). Thus, Hsp104 -RYD is a potentiated Hsp104 variant (53). This finding confirms that enhanced disaggregase activity in the presence of ATP and absence of Hsp70 is a strong indicator of enhanced activity against neurodegenerative disease proteins in yeast (51,53,55,56). Collectively, these observations suggest that the proposed NTD/NBD1/MD interactions that include the NTD-NBD1 linker make important contributions to Hsp104 functionality.

NBD1 AAA ؉ motifs are more solvent-exposed in hexamers with ADP than ATP␥S
NBD1 is a high-affinity, high-turnover site for ATP binding and hydrolysis (57). It is the main site of ATP hydrolysis for the Hsp104 hexamer, but it does not drive hexamerization, which is controlled by NBD2 (57,58). ATP binding and hydrolysis are required for Hsp104 disaggregase activity (22,23,59,60). Thus, we examined changes in solvation of the conserved AAA ϩ motifs between the hexamer with ADP and hexamer with ATP␥S. Overall, NBD1 is more modified in the hexamer with ADP than the hexamer with ATP␥S (Fig. 5A, Table 2, and  Tables S2 and S3), particularly in the regions involved in the conserved ATP-binding pocket. This finding is consistent with a switch from an open spiral hexamer in the presence of ADP to a closed ring hexamer in the presence of ATP␥S suggested by cryo-EM (16 -18) and corroborated by HX studies (30).
The Walker A motif, residues 212-220, the Walker B motif, residues 280 -285, the sensor-1 residue, Thr-317, and the arginine finger, Arg-334, are covered by peptides that suggest that in the hexamer with ATP␥S the ATP-binding pocket is protected from solvation (Fig. 5A). The peptide 211-223 shows no modification in the ATP␥S hexamer, but a rate of 32.6 Ϯ 1.9 s Ϫ1 in the hexamer with ADP (Table 2 and Fig. 5A). Consistent with this observation, HX shows that ATP␥S induces a tight restructuring of the Walker A motif (30). The peptides 282-299, 283-299, and 286 -310 also show no modification in the hexamer with ATP␥S but rates of 75 Ϯ 0.58, 28 Ϯ 0.42, and 27 Ϯ 0.3 s Ϫ1 , respectively, in the hexamer with ADP (Table 2 and Fig. 5A). The peptides 300 -320 and 311-320, which cover the sensor-1 residue, Thr-317, show that the hexamer with ADP is much more solvated than the hexamer with ATP␥S, with rates of 145.6 Ϯ 3.4 and 2.8 Ϯ 0.03 s Ϫ1 for ADP and 0 and 0.37 Ϯ 0.07 s Ϫ1 for ATP␥S, respectively ( Table 2 and Fig. 5A). Finally, the peptides 331-350 and 332-349 that cover the arginine finger, Arg-334, are unmodified in the hexamer with ATP␥S and have rates of 133.5 Ϯ 3.4 and 3.8 Ϯ 0.3 s Ϫ1 in the hexamer with ADP, respectively (Table 2 and Fig. 5A). Additionally, residues in these regions were identified as modified by MS 2 in the hexamer with ADP, including the essential Walker B glutamate, residue Glu-285, as well as Arg-333, the residue directly preceding the arginine finger Arg-334 (Table 2 and Fig. 5A), which is proposed to act like a sensor 2 residue in NBD1 (NBD1 lacks a canonical sensor 2 residue unlike NBD2) (9,16,61). These differences in solvation (Fig. 5A) indicate that the six conserved ATP-binding pockets of NBD1 in the hexamer with ATP␥S are either all occupied by nucleotide consistent with cryo-EM studies (16) or closed (i.e. unoccupied, but shielded from solvent). The situation is different for the hexamer with ADP. In the presence of ADP, a large degree of solvation is detected in NBD1, which is consistent with the open spiral ring of Hsp104 hexamers with ADP observed by cryo-EM and HX where NBD1 from protomer 1 is highly exposed to solvent (16,30).

NBD1 contains a conserved loop with a role in substrate interaction
Although two substrate-binding loops have been identified in Hsp104 (16,62), the partial defect in thermotolerance when the NBD1 loop is mutated (62) indicates that there may be additional regions of NBD1 involved in substrate binding and translocation across the channel. The canonical NBD1 substrate-binding loop, 256 KYKG 259 (62), is oxidatively modified in both hexameric states, indicating that the loop is solvated and capable of substrate interaction (Tables S2 and S3). HX shows that this pore-loop becomes more protected in the presence of ATP␥S, indicating a gain in local structure in preparation for substrate binding (30). Although rate information for identical peptides is unavailable for both states, peptides that cover the loop have much faster rates of modification in the hexamer with ATP␥S than for the hexamer with ADP: compare 10.6 Ϯ 5.3 and 18.2 Ϯ 0.37 s Ϫ1 for peptides 249 -277 and 248 -261 from the hexamer with ATP␥S to 0.51 Ϯ 0.12 and 1.15 Ϯ 0.08 s Ϫ1 for peptides 249 -278 and 249 -271 from the hexamer with ADP (Tables S2 and S3). As the Hsp104 hexamer primarily binds substrate in the ATP state (16,29,63,64), it is anticipated that the loop would be more exposed in the hexamer with ATP␥S than in the hexamer with ADP. Indeed, this loop is not visible by cryo-EM in the presence of ADP, but becomes visible in the presence of ATP␥S and engages casein (16).
Although Tyr-257 has been identified as one of two substrate-binding residues in Hsp104, thermotolerance assays show that mutation of the tyrosine to alanine results in only minor defects in Hsp104-mediated survival (62). This finding is in contrast to mutation of the NBD2 tyrosine, Tyr-662 to alanine, which results in large defects in survival after heat shock (62). This finding led us to hypothesize that regions of the NTD as well as other sites within NBD1 may contribute to substrate binding. Indeed, the NTD is involved in substrate binding (Fig.  2D). This finding is further supported by thermotolerance data showing that although either deletion of the NTD or mutation of Tyr-257 displays only mild defects in thermotolerance,

Synchrotron X-ray footprinting of Hsp104
the combined truncation and mutation variant, ⌬N/Y257A Hsp104, is severely compromised in its ability to mediate survival after heat shock (Fig. 5C).
To determine whether there are other substrate-binding regions of NBD1, we looked for regions that change in solvation in the different hexameric states. A striking example stood out immediately. In the homology model of the Hsp104 NBD1, there are two loops that point into the interior of the hexameric Hsp104 channel (Fig. 5B). One is the canonical 256 KYKG 259 loop and the second is the 291 GNGKD 295 loop. Peptides cover- Figure 5. Second substrate-binding loop in NBD1 is essential for Hsp104 disaggregase activity. A, peptide sequences and modification rates for regions of interest of Hsp104 NBD1. Residues shown in bold have been identified as modified by MS 2 and are colored by the state in which they were found: blue, hexamer with ADP; purple, both hexameric states. B, NBD1 homology modeled on the tClpB crystal structure (PDB code 1qvr) shown in context of a rigid body fit Hsp104 monomer. NBD1 is shown in blue, with the canonical substrate-binding loop in blue spheres and the secondary substrate-binding loop in red spheres. C, after incubation at 37°C for 30 min to induce Hsp104 expression, W303a⌬hsp104 yeast carrying either empty vector or a plasmid encoding the indicated Hsp104 variant were heat-shocked for 20 min at 50°C, immediately transferred to ice for 2 min, plated on SD-ura plates, and after a 2-day incubation at 30°C, colonies were counted using an acolyte automated colony counter. Data are displayed as scatterplot with bar representing mean Ϯ S.D. (n ϭ 3-14). Equal expression levels were confirmed by immunoblot. A one-way ANOVA with Dunnett's test was used to compare WT Hsp104 to the variants with ‡ denoting p ϭ 0.05 and ** denoting p Ͻ 0.01. D, urea-denatured firefly luciferase aggregates were incubated with either WT or an Hsp104 substrate-binding loop 2 (291GNGKD295) variant, for 90 min at 25°C in the presence of 2.6 mM ATP, 2.5 mM ATP␥S, and an ATP-regenerating system (1 mM creatine phosphate and 0.25 M creatine kinase). Reactivation of luciferase was then determined by measuring luminescence and converted to fraction WT activity for each condition. Data are displayed as scatterplot with bar representing mean Ϯ S. D. (n ϭ 3). A one-way ANOVA with Dunnett's test comparing variants to Hsp104-WT was performed with *** denoting p Ͻ 0.0001.

Synchrotron X-ray footprinting of Hsp104
ing the 291 GNGKD 295 loop are unmodified in the hexamer with ATP␥S, but highly modified in the hexamer with ADP (the peptides that cover the Walker B site also cover this loop) ( Fig.  5A and Table 2). In addition to the peptides (282-299, 283-299, and 286 -310, Table 2 and Fig. 5A) displaying high rates of modification in the hexamer with ADP (75 Ϯ 0.58, 28 Ϯ 0.42, and 27 Ϯ 0.3 s Ϫ1 , respectively, Table 2 and Fig. 5A), three residues within and near the 291 GNGKD 295 loop were identified as modified by MS 2 , including Asn-292 and Lys-294 ( Fig. 5A and Table  2). This loop is homologous to the ClpC 286 GAGGA 290 (41) loop, as well as the ClpA 292 GAGAA 296 (65) loop, which are secondary substrate-binding loops. Mutation or deletion of this loop results in severe defects in ClpC (41) and ClpA (65) activity. Cryo-EM has revealed that the 291 GNGKD 295 loop becomes ordered and in close apposition to translocating substrate in the presence of ATP␥S (16). Likewise, HX has revealed that this loop becomes more tightly structured in the presence of ATP␥S (30).
In vivo thermotolerance assays revealed that the K294A Hsp104 variant displayed a slight but statistically significant defect in mediating survival after heat shock compared with Hsp104 (Fig. 5C). When the K294A mutation was combined with the NTD truncation, survival after heat shock was greatly reduced and similar to the ⌬N/Y257A variant (Fig. 5C). The double mutant Y257A/K294A variant was less active than the single Y257A, K294A, or ⌬N mutants, but more active than the single mutants combined with the NTD truncation (Fig.  5C). These data support a role for the 291 GNGKD 295 loop in substrate translocation.
Next, we tested the 291 GNGKD 295 loop mutants in an in vitro assay with minimal components. Thus, we performed luciferase reactivation assays in the presence of a 1:1 ratio of ATP/ ATP␥S. Here, the loop mutants N292A, K294A, and D295A were defective in luciferase reactivation, retaining only 20, 39, and 19% of WT Hsp104, respectively (Fig. 5D). This finding confirms the importance of this loop for productive protein disaggregation.

Solvation of the MD changes dramatically in the presence of different nucleotides
Next, we utilized our solvation data from the XF experiments to understand how local and global dynamics of the MD determine its role in regulating the hexamer (Fig. 6). The homology model of the Hsp104 anti-parallel coiled-coil MD, based on the tClpB structure, retains the leucine zipper-like interactions between helices L1-L4 (Fig. 6B). The MD has been implicated in regulating the ATPase and disaggregase activity of the hexamer (44 -46, 51, 52, 66 -74), transmitting signals between NBD1 and NBD2 (66,67,73), and mediating the interaction with Hsp70 (34,(43)(44)(45)(46)(47). Interactions with Hsp70 proteins may take place in motif 2, specifically residues 480 KKK 482 and residue Arg-496 (47). We have discussed in the preceding sections how these regions are poised to interact with the NTD as well as NBD1 (Fig. 3). A number of interactions between motif 2 of the MD and NBD1 have been identified (37,45,72,74), and stabilization of these interactions leads to repression of the disaggregase (37, 45, 72), whereas destabilization elicits hyperactivity (37,51,72,74). XF data as well as published cross-linking (34,72) and structural data (19,72) support a model in which the domain is highly mobile, undergoing interactions with both NBD1 (37,45,72,74) and NBD2 (34). Additionally, it is likely that MDs within a given Hsp104 hexamer occupy a variety of positions (16 -18, 72).
Interestingly, the XF data of the MD suggested large changes in solvation between the hexamer bound to ADP and the hexamer bound to ATP␥S (Fig. 6A). This finding likely reflects that MD is highly dynamic, and restricting its movement through disulfide cross-linking greatly impairs Hsp104 activity (34,37,45). In general, the hexamer with ADP displays greater solvation than the hexamer with ATP␥S (Fig. 5A, Table 2, and Tables S2 and S3). One part of the MD where this is not the case is the first three residues of the MD, 412-414, found on the peptide 405-414, which mainly covers a portion of NBD1 (Table 2 and Fig. 6A). This peptide is solvated in both hexameric states, but more so in the presence of ATP␥S (compare 21.5 Ϯ 0.4 with 52.8 Ϯ 9.5 s Ϫ1 , Table 2 and Fig. 6A). Modification rates change at the adjacent peptide. The peptide 415-423 shows that the hexamer with ATP␥S is unmodified, whereas the hexamer with ADP has a modification rate of 35.5 Ϯ 0.8 s Ϫ1 and contains several residues identified as modified by MS 2 : Asp-415, Ser-416, Lys-417, Glu-418, Leu-421, and Leu-423 (Table 2 and Fig. 6A).
The 415-423 peptide covers the first half of helix L1 and is adjacent to a region of helix L2, residues 456 -466, which is also unmodified in the hexamer with ATP␥S (Fig. 6A). This peptide, 456 -466, has a modification rate of 1.8 Ϯ 0.5 s Ϫ1 in the hexamer with ADP and contains residues identified as modified by MS 2 , Glu-457 and Glu-460 (Table 2 and Fig. 6A). A region more C-terminal in helix L2, residues 467-474, is unmodified in the hexamer with ADP, but is modified in the hexamer with ATP␥S. A peptide that covers this region, 467-493, is unmodified in the hexamer with ADP and has a modification rate of 0.84 Ϯ 0.1 s Ϫ1 in the hexamer with ATP␥S (Table 2 and Fig.  6A). The changes in solvation indicate that transitions in this region of helix L2 may play a role in transmission of conformational changes from motif 1 to motif 2 in response to stimuli, e.g. interaction with Hsp70, binding of substrate in the NTD and NBD1, and ATP hydrolysis at NBD1 or NBD2. Indeed, the MD is crucial for communication between NBD1 and NBD2 (66,67,73). Additionally, a mutation in this region, L462R, disrupts intersubunit collaboration (20), suggesting that this region is critical for communication within the Hsp104 hexamer.
Several cross-links between motif 2 of the MD and NBD1 have been found in ClpB (37,45,72), and in Hsp104 there may be corresponding salt bridges (74). Models proposing that this motif 2/NBD1 interaction is dynamic are supported by the XF solvation data, which show large increases in solvation in the hexamer with ADP compared with the hexamer with ATP␥S. These increases in solvation include the C-terminal end of helix L2 (Tables S2 and S3) as well as the C-terminal end of helix L3 through the end of the MD: peptides 505-526, 506 -522, and 508 -536 with rates for the hexamer with ADP of 350 Ϯ 0.9, 20 Ϯ 1.2, and 89 Ϯ 3.4 s Ϫ1 , respectively, versus 0 and 1.5 Ϯ 1 s Ϫ1 , and undetermined for the hexamer with ATP␥S (Table 2 and Fig. 6A). Residues 494 -504 are not found in the hexamer with ADP but are unmodified in the hexamer with ATP␥S. This Synchrotron X-ray footprinting of Hsp104  (n ϭ 3). One-way ANOVA with Dunnett's test comparing the variants to Hsp104-WT were performed with ** denoting p Ͻ 0.01.

Synchrotron X-ray footprinting of Hsp104
finding might indicate that helices L3 and L4 of motif 2 are partially shielded from solvent in hexamer bound to ATP␥S. If this protection from solvent is representative of the repressive motif 2-NBD1 contacts, it would support a model where this repressed state of the hexamer is populated in the hexamer with ATP␥S. Additionally, several potentiating mutations are found in helices L3 and L4 (37,51,72,74) indicating that large changes in solvation between nucleotide-bound states may indicate mechanistically important regions of the Hsp104 hexamer.
To explore how conformational changes might be transmitted through the MD, we focused on a region of helix L2 that undergoes large changes in solvation. Within this region there is a break in the helix, and within this break there is a proline residue. We predicted that this residue, Pro-461, may serve as a "hinge" allowing large changes in the position of the MD that are important for functionality (Fig. 6B). Thus, we made Hsp104 -P461A, which displayed reduced disaggregase activity in the presence of ATP/ATP␥S, Hsc70 and Hdj2, and Hsp72 and Hdj2 (Fig. 6C), whereas ATPase activity was only modestly reduced (Fig. 6D). Reduced disaggregase activity was most severe in the presence of Hsc70 and Hdj2 (Fig. 6C), indicating that Pro-461 is critical for collaboration with specific Hsp70 chaperones.
We also assessed Hsp104 -P557L, a mutant found just outside the MD at the end of the NBD1 small subdomain. Hsp104 -P557L conveys normal thermotolerance in vivo but is unable to propagate [PSI ϩ ] prions (75,76). Hsp104 -P557L also showed decreased levels of luciferase reactivation in the presence of ATP/ATP␥S, Hdj2/Hsc70, and Hdj2/Hsp72 (Fig. 6C), while retaining WT levels of ATPase activity (Fig. 6D). Here, the defect in disaggregase activity was most severe in the presence of ATP/ATP␥S, indicating that intrinsic disaggregase activity was affected in a way that could be mitigated by Hsp70 and Hsp40.

NBD2: AAA ؉ motifs are more solvent-exposed in hexamers with ADP than ATP␥S
Next, we examined modification of the conserved NBD2 ATP-binding motifs as well as the region that contains the nuclear localization signal (NLS) (77). There is no coverage of either the arginine finger (Arg-765) or the sensor 1 (Asn-728) residue. Similar to our findings in NBD1, a peptide 684 -697 that covers the Walker B motif is unmodified in the hexamer with ATP␥S but has a rate of 49 Ϯ 1.3 s Ϫ1 in the hexamer with ADP (Table 2). Indeed, Leu-684 and the conserved Walker B Glu-687 are identified by MS 2 as being modified in the hexamer with ADP (Table 2). Additionally, the peptide 813-831 that covers the sensor 2 residue (Arg-826) is unmodified in the hexamer with ATP␥S and has a rate of 42 Ϯ 0.2 s Ϫ1 in the hexamer with ADP (Table 2). These findings are consistent with the transition from open spiral hexamers in the presence of ADP to closed rings in the presence of ATP␥S observed by cryo-EM (16) and HX (30). Indeed, the open spiral ring of Hsp104 hexamers with ADP would leave NBD2 from protomer 6 highly exposed to solvent (16).
Just C-terminal of the Walker B motif is a peptide 688 -696, which in the hexamer with ATP␥S has a modification rate of 57 Ϯ 9.4 s Ϫ1 , and a peptide 613-623, which covers the Walker A motif and has a rate of 9.5 Ϯ 0.3 s Ϫ1 in the hexamer with ATP␥S with residues 616 and 618 having been identified by MS 2 as modified ( Table 2 and Table S3). These data indicate that in contrast to NBD1, ATP␥S binding in NBD2 does not fully shield the nucleotide-binding pockets from solvent. This difference may be due to the presence of empty, incompletely closed sites, and/or less stable ATP␥S binding, which would result in partial solvation of some of the sites in the Hsp104 hexamer population. Differences between NBD1 and NBD2 might reflect that these NBDs originate from different AAA ϩ clades (78) and are responsible for divergent functions within the Hsp104 hexamer (57).

NBD2 NLS is generally solvent-exposed
The NLS in Hsp104, residues 773-789 (77), appears to be solvated in the monomer as well as both of the hexameric states, with the hexamer with ATP␥S the least modified ( Table 2). The NLS also includes the NBD2 hinge 773 NKLS 776 , which connects the large and small subdomains of NBD2. Peptides that cover this regions include 769 -785 and 786 -806 with rates for the monomer, the hexamer with ADP, and the hexamer with ATP␥S of 1.15 Ϯ 0.1, 0.64 Ϯ 0.003, and 0.14 Ϯ 0.02 s Ϫ1 , respectively, for 769 -785 and 0, 17.6 Ϯ 6.3, and 0 s Ϫ1 , respectively, for 786 -806 (Table 2). There are many MS 2 -identified and modified residues in this region, including lysine, which when all mutated to alanine result in a loss of nuclear Hsp104 (77), as well as residues that are in the NBD2 hinge. Lys-774 may or may not be necessary for nuclear localization; because it is in the NBD2 hinge mutating it to glutamate disrupts Hsp104 stability (79). The  (Table 2). It would appear that Lys-782 is always exposed to solvent, and in the monomeric state the hinge is highly exposed (three of four NBD2 hinge residues were identified as modified, compared with the hexamer with ATP␥S (two of four) and the hexamer with ADP (one of four)). The third essential lysine of the NLS, Lys-789 (77) is not on the peptide 769 -785 and therefore we have not confirmed its solvation state. Nonetheless, these data suggest that the Hsp104 NLS is generally exposed to solvent and thus can likely be engaged by nuclear-import receptors for nuclear import.

Cryo-EM hexameric models of NBD2 are consistent with the XF solvation data
NBD2 is critical for Hsp104 hexamerization (80). We have used the hexameric NBD2 model from cryo-EM studies (16) and the XF solvation data to identify a key region involved in the protomer/protomer interface (Figs. 7, A and B, and 8). We first used the XF modification rate information to assess the validity of the proposed hexameric NBD2 structures from cryo-EM studies (16). The NBD2 hexameric models have been colored based on the XF modification rates for the hexamer with ADP as well as the hexamer with ATP␥S in both the closed and extended conformations (Fig. 8). Both of the states agree well Synchrotron X-ray footprinting of Hsp104 overall with the models (Fig. 8); the hexamer with ADP has more regions accessible to solvent, which is consistent with an open spiral hexamer in ADP and a closed ring in the presence of ATP␥S (16).
Using the cryo-EM models and the XF solvation data, we proposed a crucial hexameric interface. This interface includes helix D3 and part of its C-terminal loop in the large NBD2 subdomain (residues 586 AIKAVSNAVRLSRSGL 601 , with the conserved hydrophobic residues in bold), and the last helix, E3, and part of its C-terminal loop in the neighboring NBD2 small subdomain (residues 836 ILNKLALRILKNEI 849 , with conserved hydrophobic residues in bold) (Fig. 7B) (16). Additionally, HX reveals that there is a large increase in protection in helix D3 upon hexamerization and a smaller increase in helix E3 (30).
There is a large number of conserved, hydrophobic residues in these regions indicating that they may be mediating the subunit/subunit interaction. If the subunit/subunit interaction was Residues shown in bold have been identified as modified by MS 2 and are colored by the state in which they were found: blue, hexamer with ADP; red, hexamer with ATP␥S; purple, both hexameric states. B, hexameric models (hexamer with ADP, PDB code 5vy8; hexamer with ATP␥S closed, PDB code 5vjh, and extended, PDB code 5vya, conformations) of NBD2 generated from cryo-EM studies (16) showing the location of hydrophobic residues found on helices D3 and E3. On the right-hand side are higher magnification views of the interface between residues 586 AIKAVSNAVRLSRSGL 601 of the large subdomain of subunit 1 (helix D3) and residues 836 ILNKLALRILKNEI 849 of the small domain of subunit 2 (helix E3). Residues Leu-601 (large domain D3, teal), Leu-596 (large domain D3, blue), and Leu-837 (small domain E3, purple) are shown as spheres. C, glutaraldehyde cross-linking. Hsp104 shows a robust ability to form hexamers even in the presence of EDTA and absence of ATP. The double Walker A (DWA) mutant displays defects in hexamerization even in the presence of ATP. The mutant L601K, which resides in the proposed hexamer interface but outside of any secondary structure, also displays defects in the ability to form hexamers.

Synchrotron X-ray footprinting of Hsp104
driven by hydrophobic interactions, it would explain a number of previously published phenomena. The protomer interface undergoes large changes in conformation during the functional cycle of the hexamer (16 -18, 30), and cross-linking the small and large subdomains results in stable hexamers with diminished function (81). If the interface was composed of rigid salt bridges, it may hinder rapid and smooth movements of the domains. If, instead, the region were a network of hydrophobic residues, then the interface could easily change register upon conformational changes of nearby regions. The proposed interface is well-suited to respond to changes in nucleotide. Indeed, the C-terminal end of the small NBD2 helix E3 contains a number of conserved hydrophobic residues that we propose to be involved in the hexamer interface, whereas the N-terminal end contains a region called the sensor and substrate discrimination motif, which contains the sensor 2 arginine (Arg-826) (37,61). Additionally, helix D3 and loop in the NBD2 large domain that we propose to be involved in the interface is just N-terminal to the short, highly-conserved ␤-strand d1 that immediately precedes the conserved Walker A motif (37).
Upon cursory inspection, this finding would make it seem unlikely that the interface was mediated by hydrophobic interactions; however, Hsp104 undergoes continual and rapid monomer exchange (20). Upon mixing Hsp104 hexamers with different (measurable) properties, monomer exchange occurs on a timescale of just a few minutes to yield an ensemble of hexamers containing subunits of different Hsp104 variants explained by the binomial distribution (20). This phenomenon of monomer exchange explains why high-salt conditions would disrupt hexamer integrity. As a monomer is released from the hexamer into high-salt conditions, there will be a propensity to make intramolecular contacts to bury the newly-exposed hydrophobic hexamer interface. We find support for this hypothesis by examining the solvation information from the XF modification data. In the monomeric sample, which was created by adding 500 mM NaCl to the buffer, there are two regions near the proposed hydrophobic interface that undergo changes in solvation. Although there is no coverage of 569 -583 (just N-terminal to the D3 helix) in the hexamer with ATP␥S, it is highly modified in the hexamer with ADP (Tables S2 and S3), likely due to exposure to solvent upon shifting to the "open" lock-washer conformation (Figs. 7B and 8) (16). Peptide 597-611, which covers the C-terminal end of the proposed D3-loop  (16) and colored based on the XF modification rates with blue as unmodified, yellow less than 1 s Ϫ1 , dark yellow 1-10 s Ϫ1 , orange 10 -50 s Ϫ1 , red more than 50 s Ϫ1 , and gray is no coverage for the hexamer with ADP (A, PDB code 5vy8) and ATP␥S in the closed (B, PDB code 5vjh) and extended conformations (C, PDB code 5vya).

Synchrotron X-ray footprinting of Hsp104
portion of the interface, has rates of 1.26 Ϯ 0.45, 6.15 Ϯ 0.10, and 0 s Ϫ1 for the monomer, hexamer with ADP, and hexamer with ATP␥S, respectively ( Fig. 7A and Tables S1-S3). Peptide 843-857, near the E3 portion of the proposed interface, is unmodified in the monomer and modified in both of the hexamers, with a number of residues identified as modified by MS 2 (3.8 Ϯ 0.14 s Ϫ1 with residues Arg-843, Lys-850, Glu-853, and Val-855 for the hexamer with ADP; and 6.1 Ϯ 4.5 s Ϫ1 with residues Arg-843, Leu-845, Lys-846, Glu-848, Ile-849, Lys-850, Lys-852, and Val-857 for the hexamer with ATP␥S, Table 2 and Fig. 7A). Both of these regions also contain highly-conserved hydrophobic residues (37), supporting the possibility that they may be involved in intramolecular interactions that block hexamerization in high salt. These residues in helix D3 and helix E3 are protected in the closed state but become solvated in the hexamer with ADP and in the extended conformation hexamer with ATP␥S (Fig. 7B) (16).
Next, we tested whether we could disrupt the hexameric interactions by mutating a conserved hydrophobic residue in the proposed protomer interface to a charged residue. We chose Leu-601 because it is in a loop region, so it was unlikely to grossly perturb secondary structure, and it is conserved in Hsp100 proteins from plants to bacteria to yeast (37). Using a glutaraldehyde cross-linking assay (58,80), we assessed the ability of low concentrations of Hsp104, the double Walker A mutant (DWA; bearing K218/:K620T mutations that disrupt hexamerization (58,80)), and an L601K variant to form hexamers under two conditions, adding either EDTA or ATP. Addition of EDTA chelates the magnesium essential for ATP binding and therefore disrupts the ability of Hsp104 to form hexamers, whereas addition of ATP promotes the formation of hexamers (58). After glutaraldehyde cross-linking, samples were fractionated by SDS-PAGE to assess hexamerization (58,80). As expected, Hsp104 exists as a mixture of monomer, dimer, trimer, and hexamer in the presence of EDTA, and upon addition of ATP shifts entirely to the hexamer (Fig. 7C) (58,80). Hsp104 -DWA, our negative control, shows very little hexamer in the presence of EDTA, and upon addition of ATP remains predominantly a mixture of monomer, dimer, and trimer ( Fig.  7C) (58,80). The L601K Hsp104 mutant appears to strongly resemble the Hsp104 -DWA variant, and even upon addition of ATP, there is very little hexamer but instead mostly smaller oligomers and monomer (Fig. 7C). These data support the placement of Leu-601 near a hydrophobic hexamer interface and strongly suggest that Leu-601 enables hexamerization.

Discussion
Using XF oxidative modification data in combination with XF-guided mutational analysis, we have uncovered regions of Hsp104 crucial for its activity. Our methodology and findings are summarized in Fig. 9, and establish the following. 1) Hsp104 hexamers switch from a more solvated state in ADP to a less solvated state in ATP␥S, consistent with switching from an open spiral to a closed ring visualized by cryo-EM and HX (16,30). 2) Specific regions of the Hsp104 NTD contribute to substrate binding and Hsp70 collaboration. 3) NTD residues (Glu-44 and Asp-45) can be mutated to enhance Hsp104 disaggregase activity. 4) Novel NTD/NBD1/MD interaction sites enable optimal Hsp104 disaggregase activity. 5) The NTD-NBD1 linker enables optimal Hsp104 disaggregase activity. 6) Critical residues in NBD1 (Asp-231, Asp-232, and Asp-233) enable Hsp104 to collaborate with Hsp70. 7) Hsp104 -RYD (R496A/Y497A/D498A) is a novel potentiated Hsp104 MD variant that rescues ␣-synuclein, TDP-43, and FUS toxicity in yeast. 8) An additional pore-loop in NBD1 ( 291 GNGKD 295 ) likely engages substrate during translocation and collaborates with the NTD and the primary pore-loop in NBD1 ( 256 KYKG 259 ) to enable optimal Hsp104 disaggregase activity. Figure 9. Summary of methodology and major findings. Hsp104 monomer and hexamer with ADP or ATP␥S were subjected to increasing exposure times of synchrotron-generated X-rays. Mass spectrometer analysis of the samples yielded oxidative modification rates for pepsin-derived peptides, and MS 2 data identified specific sites of modification. Using the modification rates as a proxy for solvation, Hsp104 variants were designed with substitutions in residues found to be highly solvated or in regions that displayed large changes in solvation across the different states. Activity of these Hsp104 variants was then assessed using in vitro and in vivo assays to determine their role in Hsp104 function. Synchrotron X-ray footprinting of Hsp104 9) An important region of the MD (the hinge region of helix L2) enables optimal Hsp104 disaggregase activity. Substitution of Pro-461 in this hinge region specifically ablates Hsp104 disaggregation activity in the presence of Hsc70 (versus Hsp72). This finding indicates a crucial role in the MD for interactions with specific Hsp70 variants that may mediate the response of Hsp104 to diverse substrate classes. 10) A hydrophobic interaction in NBD2 that involves Leu-601 mediates Hsp104 hexamer formation.
We have also shown that our XF solvation data fit well with published cryo-EM structures (16,17), HX data (30), and can explain a number of published Hsp104 variant phenotypes. Indeed, by combining information from static cryo-EM structures, backbone and secondary structure information from HX-MS, and side-chain solvation information from XF-MS, we are able to piece together a comprehensive picture of how the Hsp104 hexamer changes and the regions involved in its autoregulation and activity. Our findings suggest a model in which the energy of ATP binding to Hsp104 induces the formation of stable secondary structure as well as minimizing large dynamic movements, as seen by decreased solvation by XF. This decrease in conformational dynamics in response to ATP binding can be thought of as a pre-payment of the entropic cost of substrate binding and may have important implications for discrimination between soluble and aggregated substrates. When binding a large, stable aggregate, the main entropic penalty would come from decreases in mobility of the Hsp104 hexamer. Conversely, when binding soluble, potentially properly folded substrates, the main entropic penalty would come from the substrate. This mechanism of limiting the conformational dynamics of the Hsp104 hexamer in preparation for substrate binding could contribute to discriminating between aggregated substrates and soluble proteins that could, but should not, be remodeled.
The large size and dynamic behavior of Hsp104 has made it difficult to determine structural details of the hexamer and to pin down regions involved in its protein-remodeling activities.
Here, we have shown that using solvation data from XF experiments to focus our mutational analyses has dramatically decreased the number of variants we need to screen to unveil new regions crucial to Hsp104 function. Our findings will help inform rational design of Hsp104 variants with specific activities. Importantly, out of the 17 Hsp104 variants we designed, built, and tested based on the XF solvation data, only one of them, Hsp104-SS (S124A/S125A), did not display an interesting phenotype. Thus, there is a strong positive correlation between the regions undergoing large changes in solvation and important functional regions of a protein, which may have profound implications for protein engineering.

Protein expression and purification
Hsp104 variants were generated using QuikChange lightning mutagenesis (Agilent Technologies). Hsp104 variants were expressed and purified as N-terminally His 6 -tagged constructs in a modified pPROEX HTb vector as described (82) or as untagged constructs in a pNOTAG vector (61,73). Unless otherwise stated, Hsp104 concentrations refer to the hexamer.

Synchrotron X-ray footprinting of Hsp104
Immunoblotting and quantification Yeast were grown in raffinose media to saturation, diluted to an OD 600 of 0.6 (A 600 nm ϭ 0.6), and induced for 5 h (FUS and TDP-43 samples) or 8 h (␣-synuclein samples). Samples were normalized to an OD 600 of 0.6 (A 600 nm ϭ 0.6) post-induction, and 6 ml of cells were pelleted and resuspended with 0.1 M NaOH and then with 1ϫ SDS sample buffer. Boiled lysates were separated by SDS-PAGE (4 -20% gradient, Bio-Rad) and transferred to a polyvinylidene difluoride membrane. Primary antibodies used were anti-Hsp104 polyclonal (Enzo Life Sciences), anti-PGK monoclonal (Invitrogen), anti-GFP polyclonal (Sigma), anti-FUS polyclonal (Bethyl Laboratories), and anti-TDP-43 polyclonal (Proteintech). Secondary antibodies used were anti-rabbit (Li-Cor) and anti-mouse (Li-Cor). LI-COR Odyssey FC Imaging system was used to image blots. Western blottings were quantified using ImageJ software. Hsp104 levels were normalized to PGK (loading control) and Hsp104 WT. ␣-Synuclein-YFP, FUS, and TDP-43 levels were normalized to PGK (loading control) and strains not expressing Hsp104 (empty vector).

Fluorescence polarization
Fluorescence polarization experiments with Hsp104 variants and FITC-casein were performed as described (51). Briefly, Hsp104 variants in increasing concentrations (10 nM to 4 M) were added to FITC-casein (6 nM) and ATP␥S (2 mM) in LRB. After a 20-min incubation, fluorescence polarization of the FITC-casein was measured using a Tecan Infinite M1000 plate reader.

Glutaraldehyde cross-linking
Glutaraldehyde cross-linking experiments were carried out as described (58). Briefly, Hsp104 variants were diluted to 0.04 mg/ml in 40 mM HEPES-KOH, pH 7.4, 140 mM KCl, 10 mM MgCl 2 , 1 mM DTT plus either 20 mM EDTA or 5 mM ATP. The samples were incubated with glutaraldehyde (0.1%) for 12 min. The cross-linking reaction was quenched by addition of 1 M glycine, pH 6, and the proteins were precipitated using TCA. After two washes with chilled acetone, the samples were run on an SDS-polyacrylamide gel, and reaction products were visualized using a silver stain kit (Invitrogen).

Synchrotron hydroxyl radical footprinting (XF)
WT Hsp104 was purified as in Ref. 82, with a final step of gel filtration using a Superdex 200 column (GE Healthcare) equilibrated in 50 mM cacodylate, pH 7.0, 140 mM KCl, 10 mM MgCl 2 (for the hexamer) or 50 mM cacodylate, pH 7.0, 500 mM KCl, and 20 mM MgCl 2 (for the monomer, as confirmed by sizeexclusion chromatography followed by multiangle light scattering (Wyatt)). Samples were kept on ice and transported to Brookhaven National Laboratory (Upton, NY). All samples were kept on ice and diluted at the beamline 10 min prior to use to 10 M Hsp104 in 50 mM cacodylate, pH 7.0, 140 mM KCl, 10 mM MgCl 2 , and 1 mM ADP or ATP␥S (for the hexamer) or 50 mM cacodylate, pH 7.0, 500 mM KCl, and 20 mM MgCl 2 (for the monomer). X-ray dose-dependent oxidation of Alexa488 (Invitrogen) in the three buffer conditions was used to determine optimal exposure times and provide normalization constants. The presence of nucleotide in our buffer affects the effective hydroxyl radical population (i.e. nucleotide quenches the oxidation of the sample). Therefore, all of the rates have been normalized to the monomer sample based on the Alexa488 decay data (87,88) with normalization factors of 2.35 and 2.04 for the hexamers with ADP and ATP␥S, respectively. The samples were exposed to a mirror-focused synchrotron X-ray beam (5.5 mrad angle, focus value of 6 mm) at the X28C beamline of the National Synchrotron Light Source at Brookhaven National Laboratory for 0 -20 ms. The exposure time of the samples was controlled by flow rate through the flow cell of a KinTek (Austin, TX) stopped-flow apparatus (87). Oxidation was immediately quenched by the addition of 10 mM methioninamide, and samples were frozen with dry ice, transported to the University of Pennsylvania, and stored at Ϫ80°C.

Mass spectrometry on the XF samples
The irradiated samples were thawed on ice and diluted to 1 M in 5 mM HEPES-NaOH, pH 7.0, 140 mM KCl, 10 mM MgCl 2 , 1 mM DTT, 1% TFA and immediately injected onto an in-line fragmentation-separation/MS analysis system. Samples were first passed through an immobilized pepsin column onto a C8 trap column and washed for 3 min (2% acetonitrile, 1% formic acid, pH 2, 0°C). Samples were then flowed onto an analytical C18 column, eluted using a gradient optimized for the highlycharged Hsp104 peptides (6 l/min linear 2-30% acetonitrile gradient over 20 min, followed by a linear 30 -50% acetonitrile gradient over 10 min, 0.1% formic acid, pH 2, and 0°C), and injected by electron spray ionization into an linear quadrupole ion trap (LTQ) Orbitrap XL mass spectrometer (Thermo Fisher Scientific). The four most abundant peptides from each scan were selected for fragmentation by collision-induced dissociation and measured in the LTQ stage.

Identification of oxidatively modified peptides
A modified version of the ExMS (89) program, called ExMS-CL (which stands for "covalent labeling"), was used to process and analyze the X-ray footprinting MS/MS data to identify oxidative modifications on the protein samples. ExMS-CL makes a peptide list of the target protein, including unmodified and modified peptides. The unmodified peptides are obtained from searching MS/MS runs (similar to SEQUEST), with a user-set score threshold (in this case a p score cut-off of 0.01) to build a "peptide pool." From here, ExMS-CL identified modified and unmodified peptides, and their intensities were assigned and calculated as described in Ref. 90. Importantly, the user-set shift in the RT search window for modified peptides (Ϫ3 min to ϩ5 min) was determined by manually searching the datasets for modified peptides using SEQUEST. The ExMS-CL output contains the peptide se-Synchrotron X-ray footprinting of Hsp104 quence, all the modifications found with their individual position on the peptide, the peptide intensity quantification, and the number of modification possibilities.

Analysis of modification rate of peptides using Matlab
Output of ExMS-CL was read into Matlab (Mathworks) and analyzed using custom scripts. Modified and unmodified versions of peptides and their intensities were read from the ExMS-CL output files and matched up. Fraction unmodified was calculated for each MS/MS sample using the peptide intensities (the sum of unmodified versions of the peptide divided by the sum of all modified and unmodified versions of the peptide). For each time point, the MS/MS run outputs were averaged together, and standard deviation was calculated. For each peptide, the average fraction unmodified was reported for each time point. Once the time-dependent modification was calculated, each peptide was binned into one of three categories by the Matlab program: no time-dependent modification (0 -1 time points after time 0 were modified with respect to time 0); sporadic modification (2 time points after time 0 were modified with respect to time 0); and time-dependent modification (greater than 2 time points after time 0 were modified with respect to time 0). For the peptides that displayed time-dependent modification, the program used the first four time points (0, 2.5, 5, and 7.5 ms) to determine the rate of modification by fitting to a first-order exponential decay. For those peptides that did display time-dependent modification but were classified as sporadically modified by the Matlab program, Origin 8.1 was used to manually fit the curve. In a few cases, there does appear to be time-dependent modification; however, a good fit to the data was impossible. In those cases, the rate of modification is listed as not determined.
Because of the presence of sporadic modifications on certain peptides in the unexposed samples, we took care to remove peptides that displayed modification in the absence of X-ray exposure. Additionally, we limited our analysis to singly-modified peptides. This eliminated the possibility that a preexisting modification (independent of the X-ray footprinting experiment) was causing local conformational changes or unfolding events that would affect the analysis. Deviations from a linear response curve that could be extrapolated to zero indicated a buildup of multiple modified species that may or may not be solvation-dependent in the later time points of some peptides. Additionally, the amount of singly-modified peptides began to decrease in the 20-ms time point, again indicating a shift to multiple modifications. Therefore, we limited our analysis to the 0 -7.5-ms time points, which did not display these deviations.
Comparisons of the modification rates of peptides in different states (monomer, hexamer with ADP, and hexamer with ATP␥S) were done by the following: 1) comparing the rates of identical peptides, and 2) comparing the rates of overlapping peptides. Because oxidative modification can theoretically alter the cleavage preference of pepsin, it was important to take into account the modification rate of all overlapping peptides and to use these values as a relative guide of solvation rather than an absolute value.

Homology modeling and cryo-EM model assessment
Hsp104 was homology-modeled domain by domain using SWISS-MODEL (84,91,92) based on the S. cerevisiae Hsp104 NTD (PDB code 5u2u) (38) and ClpB crystal structure PDB code 1qvr (NBD1, MD, and NBD2) (37). To assess large-scale changes in solvation between the hexameric states, the cryo-EM models, PDB codes 5vy8, 5vjh, and 5vya (16), were colored based on the peptide with the highest rate of modification for each residue.