Role of the D1-D2 Linker of Human VCP/p97 in the Asymmetry and ATPase Activity of the D1-domain

Human AAA+ protein p97 consists of an N-domain and two tandem ATPase domains D1 and D2, which are connected by the N-D1 and the D1-D2 linkers. Inclusion of the D1-D2 linker, a 22-amino acid peptide, at the end of p97 N-D1 truncate has been shown to activate ATP hydrolysis of its D1-domain, although the mechanism of activation remains unclear. Here, we identify the N-terminal half of this linker, highly conserved from human to fungi, is essential for the ATPase activation. By analyzing available crystal structures, we observed that the D1-D2 linker is capable of inducing asymmetry in subunit association into a p97 hexamer. This observation is reinforced by two new crystal structures, determined in the present work. The effect of D1-D2 linker on the ATPase activity of the D1-domain is correlated to the side-chain conformation of residue R359, a trans-acting arginine-finger residue essential for ATP hydrolysis of the D1-domain. The activation in D1-domain ATPase activity by breaking perfect six-fold symmetry implies functional importance of asymmetric association of p97 subunits, the extent of which can be determined quantitatively by the metric Asymmetric Index.

The N-terminal half of the D1-D2 linker is highly conserved from eukaryota to archaea. Amino acid sequences of p97 and its eukaryotic homologues are highly conserved. Pair-wise sequence alignments between human p97 and its homologues including green plants and fungi revealed a sequence identity between 69% and 79%. Multiple sequence alignment across the 16 homologues of p97 listed in Fig. 2, only 51% of the residues (409 residues out of 806 of human p97) are identical. Among the three main domains of p97, the D1-domain (residues 211-459) is the most conserved, with a sequence identity of 69%, followed by the D2-domain (residues 482-762) and the N-domain (residues 1-184) with 51% and 32% sequence identities, respectively. Whereas the N-D1 linker (residues 185-210) shows a 50% sequence identity across different species, the D1-D2 linker (residues 460-481) has a higher sequence identity of 64%. The most striking revelation of the sequence alignment is the degree of conservation of the first 11 residues (residues 460-470) in the D1-D2 linker, which is invariable in our sequence alignment (Fig. 2) and coincides with the maximal activation of ATPase activity of various N-D1 truncates (Fig. 1B,C). This correlation between high sequence conservation and critical role in activation of D1-domain ATPase activity of the D1-D2 linker suggests its indispensable function in p97 over the course of evolution. The sequence conservation is perhaps not limited to eukaryotic organisms because VCP-like ATPase (VAT) in archaeal T. acidophilum has an overall sequence identity of 45% to human p97, but the sequence identity for the first 11 residues of the D1-D2 linker remains to be 64% (Fig. 2).
Location and conformation of the D1-D2 linker. Among all the published structures of wild-type FL p97, the nucleotide states at the D2-domains vary, bound with either ADP, ADP-AlFx or AMP-PNP, whereas the nucleotides found at the D1-domains are invariably ADP 3,4 . In the full-length structures, the D1-D2 linker begins immediately after the last helix (α 14) of the D1-domain at residue S459, and has been modeled as a random loop with no well-defined secondary structural elements. The beginning of the D1-D2 linker is found below helix α 12A (residues 409-426) of the D1-domain, the loop extends towards the entrance of the D1 nucleotide-binding site and it begins to turn away from the site at residues L464 (Fig. 3A). The D1-D2 linker in a similar conformation can also be observed in the structure of isolated D2-domain that includes the D1-D2 linker (residues 463-806) and forms a heptamer 3 .
In the structures of ND1 p97 Lng (residues 1-481) with ADP bound (Table 1), the polypeptide chains could only be traced up to residue S462, indicating that the D1-D2 linker is largely disordered, which is consistent with the random loop observed in the structures of FL p97. By contrast, when ATPγ S is bound to the D1-domain, a few more residues from the D1-D2 linker could be traced in the electron density but the majority remains disordered. However, in the highest resolution structure containing the R155H mutation and with bound ATPγ S (PDB:4KO8 at 1.98 Å resolution), one out of two subunits in the asymmetric unit has the D1-D2 linker traced up to residue V469 (Fig. 3B). Interestingly, instead of a random loop, this D1-D2 linker folds into an α -helix starting from P461. This result is consistent with the secondary structure prediction, suggesting this 10-residue peptide (P461-E470) has a high propensity to form a helix, even though the exact circumstances under which the helix may form remain unclear. Similar to the full-length structures, this helical linker extends below helix α 12A of the D1-domain towards the entrance of the ATP-binding pocket of the D1-domain (Fig. 3B).
The D1-D2 linker prompts asymmetric association of p97 subunits. Reviewing all crystal structures of hexameric p97 available in the Protein Databank (PDB) determined in various lengths, nucleotide states and in the absence of adaptor proteins, we found it interesting that structures of p97 containing the D1-D2 linker were usually determined from crystals with lower symmetry (Table 1). In other words, these structures have multiple subunits in a crystallographic asymmetric unit, which are related by non-crystallographic symmetry (NCS). For example, wild-type FL p97 was crystallized in three different space groups. Two of them contain NCS symmetry (PDB: 3CF1, 3CF2, and 3CF3), except for the structure with the PDB code 1R7R. The latter structure was poorly determined, as was acknowledged by the authors 4 . Likewise, all ND1 p97 Lng with the D1-D2 linker included were crystallized in crystals of lower symmetry (PDB: 4KOD, 3HU1, 3HU2, 4KLN, and 4KO8). By contrast, the wild-type ND1 p97 Shrt without the D1-D2 linker (PDB:1E32) was crystallized in a high symmetry crystal system, having a space group of P622 with one subunit per crystallographic asymmetric unit. Since all structures of ND1 p97 Lng also bear single-site pathogenic mutations, this survey seems to indicate two possibilities: one is that when the D1-D2 linker is present, the subunits within a hexameric p97 arrange themselves in a way that deviates from perfect or proper six-fold symmetry, as the linker is absent. We term this asymmetric subunit association or simply asymmetry. Another possibility is disease-associated mutations cause the asymmetry.
To test the above two possibilities, we crystallized wild-type ND1 p97 Lng with the D1-D2 linker (residues 1-481) in the presence of ADP. This wild-type ND1 p97 Lng construct crystallized in the low symmetry space group P2 1 2 1 2 1 with 12 subunits per asymmetric unit forming two hexamers (Tables 1 and 2), which is very different from the corresponding short form ND1 p97 Shrt that crystallized in the high symmetry space group P622 (PDB:1E32) ( Table 1). Since all previous structures of ND1 p97 Lng proteins were solved using pathogenic p97 mutants, the solution of this wild-type ND1 p97 Lng structure has confirmed that the deviation from a perfect hexameric association of the ND1 p97 Lng is not a consequence of pathogenic mutations but is caused by the D1-D2 linker.
The notion that pathogenic mutations are not the cause of lower symmetry crystals gained further support by the structure of ND1 p97 Shrt bearing the L198W mutation, determined in the present study with ADP bound ( Table 2). During the crystallization of L198W ND1 p97 Shrt (residues 1-460), AMP-PNP was added. However, due to the instability of AMP-PNP under acidic crystallization conditions, ADP was found instead at the D1-domain in the structure. Consistent with our hypothesis, this pathogenic mutant without the D1-D2 linker crystallized in the space group P622, containing a 6-fold crystallographic axis that, just like in the wild-type ND1 p97 Shrt , coincides with the molecular 6-fold axis (Tables 1 and 2).
To visualize this asymmetry, we used two ADP-bound wild-type p97 N-D1 truncates: ND1 p97 Shrt (PDB:1E32), determined previously 10 and ND1 p97 Lng (PDB:5DYI), determined in this work. A hexameric ring of ND1 p97 Shrt was constructed by applying crystallographic symmetry operations to the ND1 p97 Shrt subunit and this hexamer was used to compare with the ND1 p97 Lng hexamer. We superposed one subunit (subunit A) of the hexameric wild-type ND1 p97 Shrt with the equivalent subunit in the wild-type hexameric ND1 p97 Lng . Whereas subunits A of the two hexamers align well (rms deviation over 426 residues = 0.65 Å), all the other subunits are out of alignment, as the ND1 p97 Lng hexamer ring is out of the plane of the ND1 p97 Shrt ring (Fig. 4A). Although this misalignment does not necessarily mean a lack of proper six-fold symmetry for the ND1 p97 Lng hexamer, which we shall define more precisely in the next section, it is an indication that the D1-D2 linker alters interactions between neighboring subunits.
The Asymmetric Index: a measurement for the asymmetric association of p97 subunits. To obtain a more quantitative measurement of the amount of deviation in p97 subunit association from perfect six-fold symmetry in each structure, we devised a metric, asymmetric index (Asym Index in degree, Table 3), that measures the average deviation between near parallel cross vectors that are constructed from vectors connecting equivalent atoms in the structure of a p97 hexamer (see Fig. S1 and Experimental Procedures section for details). Based on this metric, a proper or perfect hexamer should have an Asym Index of 0 degree (0°), because all such cross vectors are parallel to each other and to the six-fold molecular axis. For instance, the structure of the wild-type ND1 p97 Shrt that crystallized into the space group of P622 (PDB:1E32) has an Asym Index of 0°. For structures that crystallized into lower symmetry crystal forms, we selected sequence ranges for residues that have well-defined electron densities from each domain, which were used to calculate the Asym Index for that domain (Table 3). We calculated the Asym Index for five regions: the N-domain, and two each for the AAA ATPase domains (the RecA domain and the helical domain) ( Table 3). Based on the calculations, we conclude that the more subunits present in a crystallographic asymmetric unit (more NCS-related subunits), the larger the Asym Index is. This conclusion is consistent with our initial intuitive observation that when the D1-D2 linker is present, there are lower symmetry crystal forms.
Although different domains or sub-domains have different Asym Index values, on average, the ADP-bound forms, for both full-length and N-D1 fragments with the D1-D2 linker, display the greatest deviation from proper six-fold symmetry (Table 3), which is in agreement with previous observations on full-length p97 2,3 . The ATPγ S-bound forms of ND1 p97 Lng have, in general, lower Asym Index values but the binding of ATPγ S induces more deviation to the N-domain than to other domains. Clearly, this metric provides a quantitative measurement for the observed role of the D1-D2 linker in offsetting each p97 subunit from a proper hexamer.
Correlation between D1-domain ATPase activity and side-chain mobility of trans-acting arginine-finger residue R359. Interactions between neighboring subunits when the D1-D2 linker is present are quite different (Fig. 4B). To illustrate the differences, we chose the structure of L198W ND1 p97 Shrt (PDB:5DYG), determined in this work, and that of R155H ND1 p97 Lng (PDB: 4KOD) for comparison. These two structures have the highest resolutions among the structures with the same N-D1 length; both have ADP bound; they feature the same N-domain conformation; and as we have shown, disease-related mutations do not interfere with subunit association (Table 1). We superposed subunit A of L198W ND1 p97 Shrt hexamer with equivalent subunit of R155H ND1 p97 Lng . While the structural alignment for subunit A is quite good (rms deviation = 0.78 over 429 residues), there is a significant misalignment observed for the two adjacent subunits they interact with. Here, we focus on the ADP-binding environment for the two hexamers. The C-terminus of ND1 p97 Shrt has its ends at N460, but a few more residues from the D1-D2 linker can be seen in the ND1 p97 Lng structure, ending at residue S462 (Fig. 4B). This D1-D2 linker points toward the nucleotide binding pocket of the same subunit and is adjacent to the in trans SRH (second region of homology) motif of the neighboring subunit, which is one of the defining  (a, b, c, α, β, γ)  features of the family of AAA + proteins, in addition to the Walker A and B motifs 15 . Within this SRH motif, the highly conserved arginine finger (Arg-finger) residue R359 that protrudes from the neighboring subunit and reaches into the nucleotide-binding site to interact, in trans, with the bound nucleotide was observed to undergo significant movement. The guanidine group moves on average more than 3 Å when the D1-D2 linker is present (Fig. 4B). This in trans Arg-finger residue R359 is known to undergo large conformational change during the ATP hydrolysis cycle. When the D1-domain is bound with ADP, the guanidine group of R359 is pointing towards the β -phosphate of ADP from the adjacent subunit. When ATPγ S occupies the D1-domain, the side-chain of R359 moves out of the way to accommodate the γ -phosphate of the ATPγ S (Fig. S2). The side chain mobility of R359 is also clearly influenced by the D1-D2 linker. When the D1-D2 linker is missing, the side chain of R359 is well ordered, interacting with bound ADP in the neighboring subunit, as shown in both ND1 p97 Shrt structures (Table 1) by well-defined electron density for the side chains (Fig. S3A,B). By contrast, in all structures containing the D1-D2 linker, the electron densities for the side-chain of R359 is absent in all subunits in the asymmetric unit, indicating this residue becomes disordered in the presence of the D1-D2 linker (Fig. S3C-E).

Discussion
The phenomenon of asymmetry in structures of AAA + proteins and its relation to function has been extensively documented, especially for those with asymmetry induced by binding of nucleotides. A well-studied example is bacterial ClpX, in which it is essential for each subunit in the hexameric assembly to have a different affinity for the nucleotide in order to carry out ATP hydrolysis 16,17 . The full-length bacterial unfoldase ClpA with bound ADP assembles into a hexameric spiral structure, deviating from a hexameric ring in solution, suggesting a mixed  Table 2. Statistics on the qualities of diffraction data sets and atomic models. a R merge is defined as Σ |I h,i − < I h > |/Σ I h,i , where I h,i is the intensity for i th observation of a reflection with Miller index h, and < I h > is the mean intensity for all measured I h s and Friedel pairs. b Values in parentheses are for the highest resolution shells. c Data was diffracted anisotrpically to a resolution of 3.7 Å in the a* and c* directions and 5.2 Å in the b* direction according to an I/σ I > 3 criterion. Data was anisotropically truncated and scaled with the diffraction anisotropy server (http://services.mbi.ucla.edu/anisoscale/). The resulting data set is 68.1% complete overall to 3.7 Å resolution.
nucleotide state in solution 18 , and a similar observation was made for bacterial disaggregase ClpB 19 . Dramatic asymmetry has been induced in the bacterial enhancer-binding protein NtrC1 upon partial ATP occupancy 20 . The structural asymmetry of p97 has been noticed in both full-length p97 and disease-related N-D1 truncates upon interaction with nucleotides 3,5,6,8,21 . For wild-type p97, the nucleotide state of each D1-domain controls the movement of its cognate N-domain and is tightly regulated to coordinate multiple nucleotide states among the six chemically identical subunits, resulting in sub-stoichiometric amounts of pre-bound ADP at the D1-ring 6,8,22 . Even in the presence of saturating amounts of ATPγ S, a subset of D1-domains in wild-type p97 is still bound with ADP, leading to the hypothesis that there is an asymmetric conformational movement of N-domains within the hexameric p97 5,6,23 . This asymmetrical N-domain movement in wild-type p97 seems to be important for its cellular functions, as pathogenic mutants that showed symmetric, or uniformed, movement are functionally defective 5,6 . However, how the asymmetry in the D1-domain nucleotide states is established within a homo-hexamer is unknown. In other words, it is not clear what sequence or structural elements in p97 subunits allow each subunit to exist in different nucleotide states.
In this paper, we show that one such element might be the linker connecting the D1-and D2-domain. This linker, when present, activates D1 ATPase activity and simultaneously induces asymmetrical arrangement of the p97 subunits, even though structurally the linker appears to be either disordered in N-D1 fragments or a loop with no defined secondary structures in the full-length protein. Thus, we surmise that asymmetry is required for the activation of the N-D1 ATPase activity and the increase in ATPase activity by breaking the 6-fold symmetry supports the hypothesis that p97 works asymmetrically 5 . In addition to the relationship revealed between the linker and asymmetry, our study has set the stage to understand the role of this linker in regulating D1-domain activity and how the linker may affect the function of full-length p97.
Aside from the difference in ATPase activity, both ND1 p97 Shrt and ND1 p97 Lng have similar (within 2-fold) nucleotide-binding affinities, as determined by surface plasmon resonance (SPR) 13 , indicating the low ATPase activity in the ND1 p97 Shrt is not likely to be due to its inability to bind nucleotides. Indeed, structure superposition  1E32) and wild-type ND1 p97 Lng (PDB:5DYI) hexamers. Only subunit A from each hexamer was superposed. Subunits of the ND1 p97 Lng structure are in different colors and labeled. All subunits of ND1 p97 Shrt are in gray. (B) A close-up on the nucleotide-binding interface between subunits A and B from superposition of subunit A of the hexameric structure of L198W ND1 p97 Shrt in gray (PDB:5DYG) and that of R155H ND1 p97 Lng in yellow (PDB:4KOD). The adjacent subunit for R155H ND1 p97 Lng is colored magenta and that for L198W ND1 p97 Shrt is darker gray. The bound ADP molecules are shown as stick models. Motifs necessary for the ATP cycle such as the Walker A (K251), Walker B (E305Q) and SRH (R359) and the D1-D2 linker observed in R155H ND1 p97 Lng are shown as stick models and labeled. of short and long p97 subunits did not show noticeable differences with respect to nucleotide-binding elements such as the Walker A and B motifs (Fig. 4B). In other words, the lack of ND1 p97 Shrt ATPase activity cannot be attributed to changes in these structural motifs.
Besides the Walker A and B motifs, another highly conserved nucleotide-binding element in AAA + proteins is the SRH (second region of homology) motif. In a classic study on the bacterial AAA + protein FtsH by Karata et al., mutations were introduced to this motif, which led to the loss of ATPase activity without significant impact on the binding of nucleotide 15 . In p97, the SRH motif in the D1-domain is 346-ATNRPNSIDPALRRFGRFD-364, which contains two perfectly conserved arginine residues, R359 and R362. Crystal structures of both full-length p97 and N-D1 fragments showed that R359 extends its side chain into the nucleotide-binding site of an adjacent subunit, interacting directly with the bound nucleotide, which qualifies it to be an in trans Arg-finger residue. Alanine substitution of R359 in N-D1 truncated p97 altered the hexameric status of the protein, resulting in complete loss of ATP hydrolysis 24 . In full-length p97, the R359 mutation did not have any effect on the nucleotide binding in the D2-domain, as determined by the tryptophan fluorescence spectrum and by isothermal titration calorimetry (ITC), but displayed a severe reduction (90%) in ATPase activity 24 . These mutational studies demonstrated that changes to the R359 residue led to loss of ATPase activity in the D1-domain.
In the present study, we have provided structural evidence to show that conformational alterations to the side chain of the trans-acting Arg-finger residue R359 forms the structural basis for the observed activation of D1 ATPase activity by the D1-D2 linker. To reiterate, the multi-structure alignment of ND1 p97 Shrt with ND1 p97 Lng and FL p97, all in the ADP bound form, showed significant misalignment of the adjacent subunits leading to a side chain displacement of R359 that cannot be ignored (> 3 Å) (Fig. 4B). This displacement appears to stabilize the side-chain of R359 in ND1 p97 Shrt , which is disordered in both ND1 p97 Lng and FL p97 (Fig. S3), by interacting strongly with the bound nucleotide, resulting in a symmetric arrangement of p97 that is defective in hydrolyzing ATP.
Although able to participate in various cellular pathways, p97 has been considered, by a emerging consensus, to perform the function of protein segregase, disrupting protein complexes or extracting proteins from the membrane environment 1 . With 12 ATPase domains stacked in two separate rings, it is conceivable that there are communication channels among these ATPase domains, in particular between the two rings. Indeed, evidence of communication between the D1 and D2 rings of p97 has been accumulating. Binding of nucleotide at the D1-domain is required for the ATPase activity of the D2-domain 6,12 ; and a motion transmission from D2-domain as a result of nucleotide or ligand binding to the D1-domain is mediated through the D1-D2 linker 2,3,21,25 . A conformational change in the D1-D2 linker may therefore serve as a means of communication between the two ATPase domains of p97. However, our current finding does not provide evidence if such communication channel is inter-protomer as proposed by Li et al. 25 , or intra-protomer. Since the D1-D2 linker bridges the two domains, its structure could be changed in response to changes in both domains, such as changes resulting from nucleotide binding. Although the linker is proposed to serve as a "lever" to pry changes in the D1-domain, changes in the nucleotide state at the D2-domain do not seem to affect the secondary structure of the D1-D2 linker, as it always appears to be a loop regardless of the nucleotide state at the D2-domain 2,3 . How changes in the D1-domain might affect the conformation of the D1-D2 linker in the context of full-length p97 is not clear, because in all available full-length p97 structures the D1-domains are always bound with ADP. Nevertheless, studies with N-D1 fragments did provide a hint that the D1-D2 linker may undergo nucleotide-dependent secondary structure change going from random loop in the presence of ADP to a helical segment in the presence of ATPγ S (PDB: 4KO8), which is consistent with the high helical propensity of this linker from secondary structure prediction. Although  the exact circumstances under which the helical secondary structure may form for the D1-D2 linker remain to be seen, nucleotide-induced loop-to-helix conformational change has been observed in the N-D1 linker of p97 5 . The N-D1 linker is a random loop when the D1-domain is occupied by ADP and becomes a helix when bound with ATPγ S, thus driving the N-domain to undergo both rotational and translational conformational change.

Experimental Procedures
Materials. All chemicals were sourced from Sigma Aldrich (St. Lois, MO) unless otherwise stated.

Plasmids construction, protein expression and purification of p97 N-D1 variants. Expression
and purification of p97 was done as previously described 5 . Variants of p97 N-D1 containing different lengths of the D1-D2 linker were generated using a QuikChange Site-Directed Mutagenesis kit (Agilent Technologies, Santa Clara, CA).
Determination of ATPase activity. ATPase activity of p97 was determined by measuring the amount of inorganic phosphate released from ATP hydrolysis, which reacts with a complex of molybdate and malachite green 26,27 . The activity assay was performed in an assay buffer containing 50 mM Tris-HCl, pH 8.0, 20 mM MgCl 2 , 1 mM EDTA, and 1 mM DTT. A total of 50 μ l reaction mix containing 2-5 μ g of protein and 4 mM ATP in the assay buffer was incubated at 37 °C for 15 min. The reaction was immediately stopped by the addition of 800 μ l dye buffer (a fresh mixture of 0.045% malachite green and 1.4% ammonium molybdate tetrahydrate in 4 N HCl in a 1:3 ratio) followed by the addition of 100 μ l of 34% sodium citrate solution after 1 min. After 10 min incubation at room temperature, 16 μ l of 10% Tween-20 was added to dissolve any precipitate. Absorbance was then measured at 660 nm. The amount of inorganic phosphate released was calculated based on the standard curve established by a known amount of KH 2 PO 4 (50-300 μ M) in assay buffer.
Crystallization and structure determination. Crystals of p97 N-D1 truncates were grown using the sitting-drop vapor diffusion method at 16 °C. A protein solution of 7 mg/ml was mixed with MgCl 2 and ADP or AMP-PNP to a final concentration of 40 mM and 4 mM, respectively, and incubated on ice for 30 min before spun at 14,000 rpm for 30 m 28 in at 4 °C. The admixture of 2 μ l wild-type ND1 p97 Lng and an equal volume of a well solution containing 0.1 M sodium citrate, pH 6.0, 0.3 M NaCl, 15.2% PEG3350, 20% glycerol, 2% benzamidine was set up for crystallization. Crystals were cryo-protected with the well solution supplemented with 20% glycerol and flash-frozen in liquid propane. Similarly, crystals of L198W ND1 p97 Shrt were grown by mixing 1:1 with a well solution containing 3.7 M sodium formate, pH 6.0, 8% glycerol. Crystals were cryo-protected with the well solution supplemented with 25% 7 M sodium formate and flash-frozen in liquid propane. X-ray diffraction experiments were carried out at 100 K at the SER-CAT beam lines of the Advanced Photon Source at Argonne National Laboratory. Diffraction images were collected with MarCCD detectors, and processed and scaled with the HKL200 package 29 . The diffraction data for wild-type ND1 p97 Lng was anisotropic, the scaled data was truncated and scaled with the diffraction anisotropy server (http://services.mbi.ucla.edu/anisoscale) 30 . The resulting data was then used for structural determination. The structures were determined by molecular replacement using PDB:4KOD 5 as a search model using the program MOLREP 31 in CCP4 program package 32 . The structures were refined using Refmac 33 in the CCP4 package. All structure models were manually built using the program COOT 34 .