Engineered Interfaces of an AAA+ ATPase Reveal a New Nucleotide-dependent Coordination Mechanism*

Homohexameric ring AAA+ ATPases are found in all kingdoms of life and are involved in all cellular processes. To accommodate the large spectrum of substrates, the conserved AAA+ core has become specialized through the insertion of specific substrate-binding motifs. Given their critical roles in cellular function, understanding the nucleotide-driven mechanisms of action is of wide importance. For one type of member AAA+ protein (phage shock protein F, PspF), we identified and established the functional significance of strategically placed arginine and glutamate residues that form interacting pairs in response to nucleotide binding. We show that these interactions are critical for “cis” and “trans” subunit communication, which support coordination between subunits for nucleotide-dependent substrate remodeling. Using an allele-specific suppression approach for ATPase and substrate remodeling, we demonstrate that the targeted residues directly interact and are unlikely to make any other pairwise critical interactions. We then propose a mechanistic rationale by which the nucleotide-bound state of adjacent subunits can be sensed without direct involvement of R-finger residues. As the structural AAA+ core is conserved, we propose that the functional networks established here could serve as a template to identify similar residue pairs in other AAA+ proteins.

Hexameric AAA ϩ ATPases (ATPases associated with various cellular activities) are involved in multiple cellular processes and are present in all kingdoms of life. Mutations in these proteins have been reported in several human diseases, including fronto-temporal dementia with inclusion body myopathy and Paget disease of bone, hereditary spastic paraparesis, dystonia, and prostate cancer (1)(2)(3)(4)(5)(6)(7)(8). However, the underlying mechanisms by which these mutations affect the AAA ϩ ATPase (resulting in these diseases) are largely unknown, reducing drastically the possibility of treatment. Understanding the conserved mechanism by which AAA ϩ ATPases achieve their functionality is then of prime importance in the search for therapeutics.
AAA ϩ ATPase molecular machines require nucleotide hydrolysis to remodel their substrate. They share a structural core (containing Walker A and B motifs, the sensor I sequence and the second region of homology) usually made more elaborate by inserted motif(s) (defining them as subclade) to establish the specialized functionality needed for substrate interaction specificity (reviewed in Refs. 9 -11). Some AAA ϩ ATPases have evolved complexity through formation of heteromeric assemblies (e.g. m-AAA protease Yta10 -12 or AAA helicase MCM2-7), strongly suggesting differential contributions of subunits to hexamer activity (12)(13)(14)(15).
Sensing the adjacent subunit functional state is expected to be critical for hexameric ring activity. Conserved arginine residues, so-called arginine finger residues (R-finger), have been shown to play a crucial role in the hexameric ring ATPase activity by acting in trans, stabilizing the nucleotide in the adjacent nucleotide binding pocket (16 -18). Nevertheless, the mechanism(s) allowing coordination between subunits remains poorly understood. We speculate that other residues sensitive to the nucleotide-bound state will support the coordinated conformational changes. Defining the functional networks between residues of the common structural AAA ϩ core is crucial to fully understand how these molecular machines work.
Studying the molecular communication pathway operating between the different subunits of the AAA ϩ oligomeric ring requires use of an established tractable system. Here, we used the AAA ϩ domain of PspF (PspF 1-275 ), a bacterial enhancer binding protein involved in bacterial pathogenicity (19) and an archetypal member of subclade 6 of the AAA ϩ protein family (which includes HslU, ClpX, MCM, RuvB, and Ltag) as a model of the conserved AAA ϩ core (9,20). Extensive studies using PspF as an 54 -transcriptional activator, functionally analogous to eukaryotic RNA polymerase II TFIIH ATPase (21,22), allow us to use a range of assays to measure the different activities needed for full AAA ϩ ring functionality, including the following: (i) its homo-oligomerization, (ii) substrate interaction, and (iii) nucleotide hydrolysis-dependent substrate remodeling.
Previous studies with clade 6 AAA ϩ ATPases, including MCM, ClpX, and PspF, indicated that functional cross-talk operates between the subunits of the hexamer (25, 28 -33). Indeed, the mixed nucleotide occupancy and stimulation of ATPase activity by ADP described for PspF suggested that com-munication and coordination between subunits were required for optimal activity (33). In addition, structural studies showed that all six of the substrate-interacting motifs (L1) do not contact 54 (the target) at the same time, further inferring coordination between subunits (25,34).
Here, we provide functional evidence for new communication networks used to establish substrate binding and remodeling activities, suggesting how subunits communicate the local effects of nucleotide binding and hydrolysis to support coordination of subunit activities in the hexamer.
ATPase Activity-The ATPase activity assays were performed in buffer containing final concentrations of 35 mM Tris acetate, pH 8.0, 70 mM potassium acetate, 15 mM magnesium acetate, 19 mM ammonium acetate, 0.7 mM DTT, and 3 M PspF 1-275 . The mixture was preincubated at 37°C for 10 min, and the reaction was started by adding 3 l of an ATP solution containing 0.6 Ci/l [␣-32 P]ATP (3000 Ci/mmol) plus 3.33 mM ATP and incubated for varying times at 37°C. Reactions were quenched by addition of 5 volumes of 2 M formic acid. The [␣-32 P]ADP was separated from [␣-32 P]ATP by thin layer chromatography, and radiolabeled ADP and ATP were measured by PhosphorImager (Fuji Bas-1500) and analyzed using Aida software. Activity is expressed as a percentage of PspF 1-275 WT turnover value. All experiments were performed at least in triplicate, and fluctuations of turnover values were maximally 10%.
In Vitro Open Complex and Abortive Initiation Assay-Open complex formation assays were performed in a 10-l volume containing 10 mM Tris acetate, pH 8.0, 50 mM potassium acetate, 8 mM magnesium acetate, 0.1 mM DTT, 4 mM dATP, 0.1 M core RNA polymerase enzyme, 0.4 M 54 , and 20 nM promoter DNA. The mixture was preincubated at 37°C for 5 min, and the reaction was started by addition of 5 M of PspF 1-275 WT or variants and incubated for 5 min at 37°C. Open complex formation was monitored following the synthesis of the short transcript ( Ϫ1 UpGGG ϩ3 ) started by simultaneous addition of heparin (100 g/ml), initiating dinucleotide UpG (0.5 mM), GTP (0.01 mM), and 4 Ci of [␣-32 P]GTP for a further 10 min. The reaction was quenched by addition of loading buffer and analyzed on a 20% denaturating gel. Radiolabeled RNA products were measured by PhosphorImager (Fuji Bas1500) and analyzed using Aida software.
In Vitro Full-length Transcription Assays-Full-length or abortive transcription assays were performed in a 10-l volume containing 10 mM Tris acetate, pH 8.0, 50 mM potassium acetate, 8 mM magnesium acetate, 0.1 mM DTT, 4 mM dATP, 0.1 M core RNA polymerase enzyme, 0.4 M 54 , and 20 nM promoter DNA. The mixture was preincubated at 37°C for 5 min, and the reaction was started by addition of 5 M of PspF 1-275 WT or variants and incubated for varying times at 37°C. Fulllength transcription (from the supercoiled Sinorhizobium meliloti nifH promoter) was initiated by adding a mixture containing 100 g/ml heparin, 1 mM ATP, CTP, GTP, 0.5 mM UTP, and 3 Ci of [␣-32 P]UTP for a further 10 min. The reaction was stopped by addition of loading buffer and analyzed on 6% sequencing gels. Radiolabeled RNA products were measured by PhosphorImager (Fuji Bas-1500) and analyzed using Aida software.

RESULTS
Structural data from PspF, NtrC, RuvB, and Ltag in the presence of different bound nucleotides suggest that the side chains of several strategically placed and physically paired residues could move in a nucleotide-dependent manner (Fig. 1). The PspF "dimer" presented in Fig. 1 has some uncertainties as follows: (i) the crystal structure of PspF 1-275 only defines a PspF 1-275 monomer, and the interface structure of the dimer is therefore modeled; and (ii) the nucleotide-bound form of the monomeric structure was obtained by soaking the preformed apo-crystal with nucleotide, so movements observed for the residue side chains could be restricted due to crystal packing constraints. The structural model of the PspF dimer interface requires experimental validation. Based on this model, we identified two putative "networks" that might allow the coupling of the nucleotide-bound state to functional activities in the PspF hexamer, probably via the formation of differential salt bridges between compatible pair residues. The first network includes residues located in the same subunit as follows: Glu 81 , Arg 91 , Glu 97 , and Arg 131 , potentially associated with coordination between L1 and L2 in response to the nucleotide bound in the subunit (Fig. 1, A and B). The second is composed of residues located in two adjacent subunits (at their interface), trans Arg 95 , trans Arg 98 , and Glu 130 , potentially to establish coordination between subunits in the hexameric ring ( Fig. 1, C and D). Residues of the two putative pathways are sufficiently close to each other to directly support a cross-talk between the two networks based on mutually exclusive but dynamic pairwise interactions.
Rationale for Choice of Targeted Residues, Covariance and Conservation-We examined whether the two putative networks described above were conserved among bacterial enhancer binding proteins (supplemental Fig. 1S). The positions corresponding to Glu 81 /Arg 131 and Glu 97 /Arg 131 pairs in PspF are clearly conserved (supplemental Fig. 1S, A and B). The Arg 95 /Glu 130 equivalent pair is most frequently Arg-Glu (salt bridge) and then Leu-Arg (hydrogen-bound) followed by Lys-Glu (salt bridge), suggesting a high degree of functional conservation (supplemental Fig. 1SC). In the case of Arg 98 /Glu 130 , two pairs are well represented, Leu-Glu and Arg-Glu, whereas a multitude of additional combinations are observed (supplemental Fig. 1SD).
We reasoned that if the paired amino acids are indeed simply directly interacting for hexameric ring functionality and are not directly involved in any other activities, we should be able to invert the residues of the same pair without major effects on protein activity (e.g. Arg-Glu is substituted by Glu-Arg, and allele specific suppression is assessed). To do so, we measured the effect of double substitutions on protein activity compared with WT and single substitution variants. A second substitution compensating for a negative effect of a single substitution would provide clear evidence that the residues act as a pair. We constructed single, double, and triple variants using alanine substitutions as a control for Glu to Arg and Arg to Glu substitutions (supplemental Table 1S) to address the following contributions: (i) Glu 81 , Arg 91 , Glu 97 , and Arg 131 for intrasubunit communication, and (ii) Arg 95 , Arg 98 , and Glu 130 for intersubunit communication.
Single Substitution Effects on AAA ϩ Activator Activities-After purification, we tested whether the above PspF 1-275 variants were functional (as in Ref. 38) in the following: (i) substrate remodeling; (ii) substrate interaction; (iii) ATP hydrolysis; and (iv) oligomerization.
Each single substitution greatly reduced the ability of PspF 1-275 to remodel its substrate, measured as a reduction in 54 -dependent transcription activation (Figs. 2A and 3A). Indeed, we were not able to detect any significant signal for the single substituted variants except for R91A (2% of WT, Fig. 2A, lane 4) and E130A (6% of WT, Fig. 3A, lane 6), suggesting that the substitutions used are targeting critical residues.
We next addressed whether the lack of substrate remodeling activity was due to a lack of interaction between PspF and its primary target, 54 . Using the transition state nucleotide analogue ADP-AlF x , a stable PspF-54 complex can be formed (39). In the presence of radiolabeled 54 , we only observed (Figs. 2B and 3B) complexes between 54 and PspF 1-275 WT (100%), R91A (99%), R91E (70%), R131A (73%), and E130A (8%). As expected, the variants able to remodel their substrate (R91A and E130A) can interact with 54 . Strikingly, R91E and R131A, which are inactive for substrate remodeling can interact stably with 54 at a similar level to WT. These results suggest that Arg 91 and Arg 131 are directly involved in a pathway supporting substrate remodeling after the initial nucleotide-dependent substrate interaction.
To investigate whether the lack of remodeling activity observed with the variants able to interact with 54 was due to defective energy coupling or a loss of ATPase activity, we performed ATPase assays. Variants R91A and R91E retained 100% activity (Fig. 2C, lanes 4 and 5), establishing that R91E ATPase activity is uncoupled from substrate remodeling. All the other single substitutions caused drastic reductions in ATPase activity (Figs. 2C and 3C). These residues are remote from the site of ATP hydrolysis and not predicted to be directly involved in ATPase activity, rather they may indirectly impact on the active site architecture.
In conclusion, single substitutions of these residues affect to a similar extent the final output (i.e. decrease the substrate remodeling activity) but have distinct defects as follows: decreased ATPase and/or substrate binding and/or signal coupling between ATPase and substrate remodeling activities.
Single Residue Substitutions at the Interface Affect Oligomer Organization-Many of the targeted residues chosen for this study are anticipated to be involved in intrasubunit communication or to be at the interface between subunits. Hence, we addressed whether these substitutions affected formation of the hexamer and subsequently the organization of the catalytic interface (to potentially account for decreased ATPase activity). Gel filtration experiments were conducted to assess the oligomeric states of the PspF 1-275 variants (as described previously (33). For all the single variants, the equilibrium between apparent dimer and hexamer forms is displaced in favor of the dimeric form, suggesting that each substituted residue is functionally important for oligomer formation (data not shown). We tested whether the presence of nucleotide could favor high order oligomer formation, as is the case for WT (33), and observed a similar increase in self-association with all the variants except E97A/E97R, R95A/R95E, and R98E (supplemental Figs. 2S and 3S). In conclusion, these residues (Arg 95 , Glu 97 , and Arg 98 ) impact on nucleotide-dependent oligomerization, suggesting that they play an important role in nucleotide-dependent protein organization.
Swapping Pair Residues Glu 97 and Arg 131 Restores Substrate Binding but Not Substrate Remodeling Activity-Having established the major changes in activities associated with single substitutions, we tested whether formation of salt bridges between the paired residues identified was sufficient to support all, or some, of the nucleotide-dependent activities. Based on the likelihood of intrasubunit communication (Fig. 1), we investigated the roles of Glu 81 , Arg 91 , Glu 97 , and Arg 131 . Recall these residues might relay and amplify conformational changes in the nucleotide binding pocket to L1 and L2. We constructed double, triple, and quadruple alanine and "pair swapping" variants (supplemental Table I) and determined whether these variants retained remodeling activity ( Fig. 2A, lanes 10 -23) but did not detect any. These results suggest requirements beyond just For each variant potentially involved in cis subunit communication, we tested the following: A, remodeling activity (transcription activation, as % of WT; maximal error is 14%); B, substrate interaction activity (binding to 54 using ADP-AlF x , as % of WT; maximal error is 15%); and C, ATPase activity (as % of WT; maximal error is 10%).
simple pairwise interactions between these residues for remodeling activity. However, the swapping pair E97R/R131E is clearly able to interact with the substrate ( 54 ), much better than the corresponding single substitutions, implying that the interaction between Glu 97 and Arg 131 is critical for engagement of L1 with substrate (Fig. 2B, lanes 10 -15).
Glu 97 may have more than one interacting partner (e.g. Glu 97 interacts with Arg 131 in the presence of ATP and with Arg 91 in the presence of ADP, Fig. 1, A and B) prompting analysis of triple and quadruple substitutions (supplemental Table I). Substrate interactions of the E81A/R91A/R131A (2% of WT) and R91A/E97A/R131A (12% of WT) triple alanine variants were detected, but no substrate remodeling activity was observed (Fig. 2, A and B, compare lane 16 with 10 and 12 and lane 20  with 12 and 14).
These data imply that the substrate interaction activity of L1 is tightly regulated. Indeed, the correct presentation of the 54 -interacting motif cannot only be explained by the release of repressive interactions maintaining L1 in an "off" state (see results with variant E81A/R91A/E97A/R131A) but is also an outcome of "positive" interactions leading to an L1 "on" state. We have shown the importance of the Glu 97 / Arg 131 pair for substrate engagement with L1 (swapping var-iant E97R/R131E) demonstrating that L2 residue Arg 131 plays an active role in controlling L1 exposure. In addition, these results suggest that these residues are not only involved in one protein activity but several, probably via different "pair interactions." The failure to identify substitutions that restore full oligomer activity strongly suggests the requirement of more than one subunit in the process, subunits most likely harboring different interacting pairs (due to different nucleotides bound) at the same time.
Glu 97 Is Acting in Cis on Arg 131 -We next addressed whether Glu 97 /Arg 131 communication was occurring in the same subunit (in cis) or between adjacent subunits (in trans). If mixing the two single variants E97R and R131E yields a gain of activity (compared with the starting activity of the single variants), then the residues are involved in intersubunit communication (in trans), whereas no change in activity suggests intrasubunit communication (in cis). Using a fixed PspF 1-275 total concentration, we mixed E97R and R131E and all the possible combinations with E97A/E97R and R131A/R131E at ratios from 6:0 to 0:6 and tested the substrate binding ( Fig. 4A and data not shown), remodeling (Fig. 4B, black bars, and data not shown), and ATPase (Fig. 4B, gray bars, and data not shown) activities. This "doping" experiment did not reveal any changes in substrate interaction or remodeling when compared with the results obtained with the corresponding single substitutions. In contrast, a substrate interaction complex was observed with the double variant E97R/R131E. These results demonstrate that the complementary effect observed in E97R/R131E is due to a "cis" complementation phenomenon (i.e. interactions are occurring in the same subunit). As expected, all the other combinations that were tried failed to change the activities tested (data not shown). In addition, a similar experiment was performed with combinations of single and double variants to try to mimic E81A/R91A/R131A and R91A/E97A/R131A, but no gain of function was observed (data not shown), again suggesting that the increased activities observed in the triple variants are due to in cis communication between these residues. We conclude that the functionally significant communication between Glu 81 , Glu 97 , Arg 91 , and Arg 131 occurs via direct interactions in the same subunit (in cis) to control L1 use.
Swapping Pair Residues Arg 95 /Glu 130 or Arg 98 /Glu 130 Support PspF Activity-Having shown that Glu 97 and Arg 131 in cis interaction is directly involved in coordinating L1-L2, we investigated the role of Arg 95 , Arg 98 , and Glu 130 residues located at the subunit interface. To investigate this putative "in trans communication pathway," we constructed double and triple alanine and pair swapping variants.
Consistent with the single alanine substitutions (Fig. 3), the double alanine variants are completely unable to stably interact with, or remodel, their substrate. In contrast, the "swapping" variants have significant remodeling activity (34% of WT for R95E/E130R and 45% of WT for R98E/E130R) also reflected by their substrate binding ability (Fig. 3B), whereas the corresponding single substitutions were inactive (Fig. 3A, compare  lane 9 with 3 and 7, and lane 11 with 5 and 7). The role of Arg 95 /Glu 130 and Arg 98 /Glu 130 pairs seems to be similar in terms of substrate remodeling activity. Nevertheless, these two pairs can be differentiated based on the intrinsic protein activ- ities (ATPase and oligomerization activities ( Fig. 3C and  supplemental Fig. 3S)). R95E/E130R can form an apparent constitutive hexamer, although R98E/E130R oligomerization is dependent on protein concentration and is nucleotide-sensitive (as is WT, see supplemental Fig. 3S). This result suggests that, in contrast to Arg 95 /Glu 130 , the Arg 98 / Glu 130 interaction is crucial for oligomeric ring organization. The differential ATPase activities of R95E/E130R (42% of WT activity) and R98E/E130R (100% of WT activity) further demonstrate an unequal contribution of these two pairs in the global activity of the hexamer (Fig. 3C). The lack of activity observed with R95E/R98E/E130R seems to be due to a major defect in self-assembly (no oligomer detected in gel filtration, see supplemental Fig. 3S). In contrast, R95A/ R98A/E130A, although unable to interact with and remodel its substrate, exhibited full ATPase activity (Fig. 3C, lane 13). Taken together, these data suggest important roles for the Arg 95 , Arg 98 , and Glu 130 side chains in establishing productive interactions to allow the optimal hexamer activity, potentially acting in trans as explored below.
Arg 95 and Arg 98 Are Acting in Trans on Glu 130 -We established the importance of Arg 95 /Glu 130 and Arg 98 /Glu 130 pairs without knowing whether any direct interactions between these residues involved one or several subunits. As our initial hypothesis was that Glu 130 could interact with trans Arg 95 and/or trans Arg 98 , we tested whether the effect observed in the pair swapping variants was due to the communication between residues from adjacent subunits (in trans) or in the same subunit (in cis). We performed "doping experiments by mixing two single variants at different ratios. If residues are involved in intersubunit communication (trans) we should observe a gain of activity compared with the starting activity of the single variants. The single substitutions R95E, R98E, and E130R are inactive for substrate binding and remodeling, facilitating detection of any gain of function upon mixing. At a fixed PspF 1-275 total concentration, we mixed R95E and E130R or R98E and E130R at varying ratios and analyzed their substrate binding (Fig. 5, A-C), remodeling (Fig. 5, B-D, black bars), and ATPase (Fig. 5, B-D, gray bars) activities. For both R95E and E130R or R98E and E130R pairs, we observed a gain of activity in the mixing experiments, compared with the single substitution alone, maximal at a ratio of 3:3, demonstrating that the interaction between these residues are critical for oligomeric ring activity. Mixing R95A and E130A or R98A and E130A or all the combinations possible with R95A/R95E, R98A/R98E, and E130A/ E130R gave no gain of activity (data not shown). We conclude that the functional interaction between trans Arg 95 /Glu 130 and trans Arg 98 /Glu 130 occurs between two adjacent subunits probably via the interaction (formation of a salt bridge) between the side chains of Glu 130 and either trans Arg 95 or trans Arg 98 .
Differential Role of Arg 95 /Glu 130 and Arg 98 /Glu 130 Pairs in ATPase Activity-As the pair swapping variants R95E/E130R and R98E/E130R exhibit different ATPase activities (42 and 100% of WT, respectively) to evaluate in cis and in trans effects, we examined the extent to which mixing of the corresponding single variants impacted upon ATPase activities. Strikingly, we observed when mixing R95E and E130R from a 1:5 to 5:1 ratio, the apparent ATPase activity remains constant (about 50% of WT) and similar to the double variant R95E/E130R (Figs. 5B and 2D, gray bars). In contrast, when mixing R98E and E130R, we observed the ATPase activity changes with the protein ratios, with an optimal activity observed when using the ratio 3:3 (Fig. 5, B-2D, gray bars). These results clearly demonstrate that Arg 95 and Arg 98 have distinct roles and contributions to the hexameric ring activity.
We conclude that the Arg 98 /Glu 130 interaction is necessary and sufficient to allow full ATPase activity, whereas the Arg 95 / Glu 130 interaction seems to negatively impact on the hexamer ATPase active site. We propose that the 50% ATPase activity observed with R95E/E130R is due to a lack of communication caused by the inability of Arg 98 to interact with Glu 130 from which we infer the following: (i) there is communication between subunits, and (ii) all catalytic sites are not independent. A, native gel autoradiograph shows in lane 1 ADP-AlF x -dependent formation of a stable complex between PspF and 32 P-54 . Lanes 2-8 show the absence of this complex with 54 using different E97R:R131E ratios as indicated. Lane 9 shows the signal obtained with the double variant E97R/R131E. Radioactivity was quantified, and the quantity of complex formed for each mixture is expressed as percent of complex formed with WT. B, same protein mixtures were used in substrate remodeling activities (black bars, transcription activation, as % of WT; maximal error is 14%) and ATPase activity (gray bars, as % of WT; maximal error is 10%).
These outcomes further emphasize the importance of subunitsubunit communication for the ATPase and remodeling activities of the ring.

DISCUSSION
The conserved AAA ϩ protein core has become specialized by the insertion of structural motifs allowing nucleotide-dependent specific substrate binding and remodeling. Understanding the nucleotide-driven mechanisms of action is of basic interest and practical importance. For one type member AAA ϩ protein (PspF), we have now established the functional significance of strategically placed residues for "cis" and "trans" subunit communication in supporting coordination between subunits for nucleotide-dependent substrate remodeling (Fig. 6). Our data establish that subunits of the hexameric ring are not independent for activity and that at least two adjacent subunits need to be coordinated in their function (by trans Arg 95trans Arg 98 -Glu 130 residues) to allow optimal substrate binding and remodeling.
An "in Cis" Pathway for L1-L2 Coordination-In this study, we investigated the roles of Glu 81 , Glu 97 , Arg 91 , and Arg 131 in the putative signaling pathway used for establishing PspF substrate binding and remodeling activities. Glu 81 , Glu 97 , and Arg 131 were suggested to take part in such a pathway but remained without functional characterization (24 -26). Here, we have identified a key residue (Arg 91 ) with a crucial and unexpected role in the coordination of L1 and L2. The location of these residues (Fig. 1) and their conservation (supplemental Fig. 1S) suggested roles in an L1-L2 coordination regulatory pathway. Residues Glu 81 and Arg 91 are located at the extremities of L1, Arg 131 at the beginning of L2 (and adjacent to Glu 130 ), and Glu 97 in the second part of helix 3 between the conserved FIGURE 5. R95E or R98E complement E130R activity in trans. Doping experiments using R95E and E130R or R98E and E130R are shown. The native gel autoradiograph shows the formation of a stable complex between PspF and 32 P-54 in the presence of ADP-AlF x . Lane 1 is the positive control showing the complex formed with WT. Lanes 2-8 show the complex formation using different R95E:E130R (A) or R98E:E130R (C) ratios (as indicated). Lane 9 shows the signal obtained with the double variant corresponding to R95E/E130R or R98E/E130R. Radioactivity was quantified, and the quantity of complex formed for each mixture is expressed as percentage of complex formed with WT. The same protein mixtures for R95E/E130R (B) or R98E/E130R (D) were used in substrate remodeling activities (black bars, transcription activation, as % of WT; maximal error is 14%) and ATPase activity (gray bars, as % of WT; maximal error is 10%).
Arg 95 and Arg 98 residues involved in the "trans" communication pathway.
Substitutions of Arg 91 had (i) no effect on ATPase activity (Fig. 2C, lanes 4 and 5) and (ii) only moderately decreased substrate interaction when the charge was inverted (R91E, Fig. 2B, lanes 4 and 5) but had (iii) a very drastic negative effect on substrate remodeling. This last observation establishes that the role of this residue is crucial in coupling ATPase activity and substrate remodeling after the initial substrate interaction.
In this study, we also directly demonstrate the importance of the conserved pair Glu 97 /Arg 131 on protein activity by restoring some activity in the swapping variant E97R/R131E (Fig. 2, lane  13). The Glu 97 and Arg 131 single substitutions completely abolish ATPase activity and substrate interaction activity (Fig. 2, lanes 6 -9) except for R131A, which retained significant substrate interaction activity but no ATPase (Fig. 2, lane 8). This last result suggests more than one role of the Glu 97 /Arg 131 pair and the involvement of Glu 97 and Arg 131 in interactions with other residues. We propose that Glu 97 could have two roles during substrate remodeling. The first role will be to coordinate L1-L2 (via Arg 131 interaction) to allow L1 to interact with the substrate, and the second will be to permit the coupling of ATPase to substrate remodeling after the initial binding event (via Arg 91 interaction). Unfortunately, because Glu 97 affects the initial substrate-binding interaction (R91E/E97R, Fig. 2, lane  15), subsequent roles are difficult to assay. Nevertheless, our functional data strongly suggest that Glu 97 is pivotal in different interaction pairs for the L1-L2 coordination needed for substrate interaction and remodeling.
Roles of Arg 95 and Arg 98 in Fine-tuning of the Hexameric Ring Activity-Structural data obtained by nucleotide soaking (26) suggested that the Glu 130 side chain orientation might change to allow the trans Arg 98 /Glu 130 interaction to occur in the ATPbound state and the alternative trans Arg 95 /Glu 130 interaction in the ADP-bound state (with the limitations discussed before concerning the model in Fig. 1). Here, we revealed that trans Arg 95 /Glu 130 and trans Arg 98 /Glu 130 pairs are involved in nucleotide-dependent subunit coordination, which is necessary for both substrate binding and remodeling. The results obtained studying the "trans" pathway, which involves residues distant from the substrate interaction motif (contained in L1), demonstrate that more than one subunit is involved in substrate binding and remodeling. Indeed, we established that productive L1 exposure is regulated by the coordination between adjacent subunits.
One question of major interest is the nature of the ATPase cycle taking place in the AAA ϩ oligomeric ring. Four possibilities have been described as follows: stochastic, synchronized, sequential, and rotational ATPase mechanisms (11). The identification and characterization of Arg 95 , Arg 98 , and Glu 130 variants are of special interest because they provide evidence that the PspF ATPase cycle is unlikely to be stochastic (and so different to the proposed mechanism for ClpX (40), a member of the same AAA ϩ clade). This is consistent with our previous observations with PspF, showing stimulation of ATPase activity by ADP and inhibition of ATPase by excess ATP (33).
In addition, and not obvious from the structural model, the contributions of paired residues are different at the level of ATPase activity: Arg 98 /Glu 130 (R98E/E130R 100% of WT) seems crucial for ATPase activity, whereas Arg 95 /Glu 130 (R95E/ E130R 42% of WT) seems more dispensable, consistent with a concerted ATPase mechanism where R95A (giving a locked Arg 98 /Glu 130 interaction) and R98A (giving a locked Arg 95 / Glu 130 interaction) are almost inactive for ATP hydrolysis.
We propose that switching between the trans Arg 98 / Glu 130 (ATP) and trans Arg 95 /Glu 130 (ADP) interactions is central to achieving coordination between adjacent subunits in the hexamer and is required for full ring activities, further demonstrating the subunit coordination requirement for ATPase, substrate interaction, and substrate remodeling activities.
Identification of a New Class of Uncoupling Variant-Two uncoupling variants, R91A and R95A/R98A/E130A, were identified, retaining full ATPase activity but failing to couple nucleotide binding to either substrate remodeling or substrate binding, respectively, although retaining an intact substrateinteracting L1 motif (distinct from all of the other reported uncoupling variants). Despite both failing to remodel their substrate, these two variants affect distinct steps in the remodeling pathway.
Arg 91 is located at the end of L1 and is proposed (in this study) to contribute to controlling L1 exposure. We established that the uncoupling phenotype observed with this variant is due the lack of a productive "movement" of L1 after the initial contact with the target, probably because of the absence of Arg 91 / Glu 97 interaction (Fig. 6). The lack of Arg 95 , Arg 98 , and Glu 130 side chains (R95A/R98A/E130A) leads to a full but completely unproductive ATPase activity. Interestingly, this variant was FIGURE 6. Model of cis and trans communication pathways for coordinated activity. From our experimental data and the previous structural model of the interface between PspF subunits, we propose a mechanism where cis and trans communication pathways are linked together. When ATP is bound in subunit n, Glu 97 interacts with Arg 131 (in L2 n ) and Glu 130 (just adjacent to Arg 131 in L2 n ) with trans Arg 98 (adjacent to Glu 97 nϩ1 ) allowing the exposure of L1 n . These interactions will fully change in the presence of bound ADP. At this point, Glu 97 interacts with Arg 91 (L1 n ), whereas Glu 81 interacts with Arg 131 (L2 n ), which is expected to change the presentation of L1 n from an exposed to a more buried conformation. At the same time the bound nucleotide signal will be propagated to the adjacent subunit by the new interaction made between Glu 130 and trans Arg 95 . not expected to fail to interact with the substrate because the substituted residues are located far from the substrate-interacting motif. This observation fully supports the proposal that nucleotide-dependent coordination between subunits is important for substrate binding, implying that more than one subunit of the oligomeric ring is required for substrate interaction and remodeling. Importantly, and for the first time, we uncovered a class of variants unable to remodel the target 54 but that retained self-association and ATPase activities. This class of mutant is most probably defective in the coordination of conformational change between subunits. Below, we propose that communication pathways like those described for PspF exist and operate in many AAA ϩ proteins.
An Underlying Pathway-Due to the different locations of the particular substrate-associated interaction insertions (L1 and L2 in this study) and because of their different amino acid sequences, it is not anticipated that the exact pairs of residues identified here will be universally found across the AAA ϩ protein family. Nevertheless, the current insights obtained with PspF could more generally support investigation into subunit organization and communication within AAA ϩ ring assemblies. Inspection of some AAA ϩ protein structures (PspF, NtrC1, NtrC4, Cdc6P, ZraR, ClpX, FtsH, BChi, Vps4, LTag, RuvB, and ClpA) revealed sets of interacting amino acid pairs likely to be functionally equivalent to the pairs studied in PspF (supplemental Fig. 4S). The absence of sufficiently high resolution or appropriate nucleotide-bound structures precludes an exhaustive search for other examples, which are anticipated to exist.