Cryo-EM structure of transmembrane AAA+ protease FtsH in the ADP state

AAA+ proteases regulate numerous physiological and cellular processes through tightly regulated proteolytic cleavage of protein substrates driven by ATP hydrolysis. FtsH is the only known family of membrane-anchored AAA+ proteases essential for membrane protein quality control. Although a spiral staircase rotation mechanism for substrate translocation across the FtsH pore has been proposed, the detailed conformational changes among various states have not been clear due to absence of FtsH structures in these states. We report here the cryo-EM structure for Thermotoga maritima FtsH (TmFtsH) in a fully ADP-bound symmetric state. Comparisons of the ADP-state structure with its apo-state and a substrate-engaged yeast YME1 structure show conformational changes in the ATPase domains, rather than the protease domains. A reconstruction of the full-length TmFtsH provides structural insights for the dynamic transmembrane and the periplasmic domains. Our structural analyses expand the understanding of conformational switches between different nucleotide states in ATP hydrolysis by FtsH.

The authors have solved the structure of Thermotoga maritima (Tm) FtsH in an ADP-bound state using cryo-EM. The cryo-EM imaging was conducted using full-length TmFtsH, and the authors solved the ADP-bound structure comprising hexameric ATPase and protease domains. Through a careful selection of 2D classes containing not only the ATPase and protease domains but also the transmembrane and periplasmic domains, the authors were also able to solve the structure of the ADP-bound full-length TmFtsH with low resolution. By comparing their ADP-bound structure with apo-or ATP-bound structures, conformational changes in the ATPase domains, rather than in the protease domains, were observed. When considering the structure of the full-length TmFtsH, the authors found tilted topologies of transmembrane and periplasmic domains with extra densities in the large gap formed by the tilting, suggesting the mechanism of substrate recognition and initial loading. However, the biological relevance of this tilted structure has not been well supported, and the extra densities in the tilted region have not been fully described, even in the figures. Additionally, while the authors mentioned that the conformational changes in the ATPase and protease domains begin on the connecting loops, there is no detailed description provided. Although the manuscript is well written and easy to follow, the authors need to address the specific concerns listed below.
1. Throughout the manuscript, the description in the text and figures is not completely matched (e.g., on page 4, line 4, Fig 3c needs to be replaced with Fig 2c). In addition, there are several typos in the figure legends (e.g., in figure 2, even though there is no panel "g," it is mentioned in the legend.).
2. Page 4, line 15: The authors have used a construct with H423Y mutation (a zinc-bindingdeficient mutant) for cryo-EM analysis and have mentioned it as an active form in the discussion (page 7, line 5). Please show the additional data or provide pertinent explanation to support this.
3. On page 5, line 1, the authors have mentioned that side chain densities were not observed for Arg318', however, the side chain of Arg318' has been marked in figure 3f. In addition, the authors have proposed a potential interaction between Arg318' and phosphate groups in ADP, but no pertinent evidence, such as a distribution of charges in the neighboring region or additional densities around the phosphate groups, has been provided.
4. Page 5, line 28-29: How does this structure provide the basis for substrate loading and translocation processes? 5. In "Apo-ADP state transition" and "ADP-ATP state transition" sections, the authors have mentioned that the structural changes begin on the loops connecting the protease and ATPase domains. However, there is no description of how the structural changes in the loops are propagated to the ATPase domain.
6. In "Full-length FtsH structure" section, the authors have introduced the full-length structure of TmFtsH whose ATPase and protease domains are tilted relative to the lipid bilayer. Is this tilted structure biologically relevant? 7. In the same section, the authors have suggested that the extra densities in the large gap formed by the tilting could originate due to the substrate or the disordered N-terminus of the protein itself. However, considering the low resolution of the full-length structure, especially in the transmembrane domain, it is not clear whether it is reasonable to assume extra densities. It is recommended to provide an additional figure with a description regarding the extra densities.
8. Page 7, line 8: What is the supporting data for this hypothesis? 9. The authors have used the AlphaFold model structure to fit low-resolution densities in the fulllength structure. However, it is unclear how the authors created a hexameric model structure using AlphaFold. Since the AlphaFold-Multimer was released recently, it would be helpful to mention in the method section, which version was used for the modeling and how it was carried out (AlphaFold, AlphaFold-Multimer, AlphaFold with modifications, or AlphaFold with Colab, etc.).
Minor Concerns: 1. Please systematize the usage of the abbreviated letter code for amino acids.
2. Page 4, 2nd paragraph: It would be better to match the flow of the text and order of figures for easy understanding. 3. Figure 1a: Please provide the explanation for PD and TM in the legend. 4. Figure 4: The meaning of cyan color should be noted. 5. Figure 5e: Please add the explanation for Purple colored arrow in the legend. 6. Supplementary Figure 1: Please add symmetry information for the refinement of the structure.
Reviewer #3 (Remarks to the Author): The authors provide a short report of the cryoEM structure of the ATP-dependent protease Thermotoga maritima FtsH to 3.15 angstrom resolution. FtsH is a membrane-anchored protease that can degrade both membrane and soluble proteins, and the authors use in vitro reconstitution into lipid nanodiscs to isolate and characterize the full-length protein in the fully-ADP bound state. This resolution is a significant improvement on a recently published study (Carvalho et al, 2021) visualizing detergent solubilized full-length FtsH. Previous cryoEM structures of mitochondrial homologues of FtsH reported highly asymmetric interactions with both nucleotide and trapped polypeptide substrates. Here, the authors reveal a largely symmetric structure with all six subunits bound to ADP.
The visualization of the fully-ADP state is of moderate interest to researchers in the AAA+ enzyme field and the high resolution of the reconstruction will add another group of states that are available to be occupied by these enzymes. However, it is unclear under what circumstances the enzyme will occupy this fully-ADP state. The authors suggest this may represent a 'resting state' of the enzyme when ATP is not available, but it would be helpful if they could provide some evidence from the literature regarding the relevance of this state under physiological conditions.
The previous cryoEM structures of mitochondrial FtsH-like enzymes lack the transmembrane domains and small domains that sit on the other side of the membrane. In that regard, information on the structure of these domains in the full-length enzyme is useful, albeit these domains are flexible and poorly visualized in this structure. A higher resolution reconstruction of these regions would have made this structure far more interesting.
The microscopy is well-performed, and the validation of the structure is sufficient.
Specific points 1. The authors state that they applied 6-fold symmetry to the reconstruction and that an alternative reconstruction lacking any applied symmetry had slightly lower resolution. How do the positions of the ATPase domains in these two reconstructions compare? Is there evidence of asymmetry in the unconstrained map that is being lost when 6-fold symmetry is applied? 2. The manuscript contains a number of spelling and grammatical errors. It would be improved by a careful revision to remove these errors. e.g. -- Figure 1e is mislabeld in the legend. --Page 3 Line 31 "followed" should be "following" --Page 4 line 7 "much few" should be "many fewer" Reviewer #4 (Remarks to the Author): This paper reports a cryo-EM study of AAA+ protease TmFtsH in ADP-bound states, including a low-resolution reconstruction of the full-length complex. While the work appears to be technically sound, the biological insights obtained are limited and incremental. Particularly, there is lack of biochemical validation of the purified and imaged complexes, thus leaving it to be less well-defined in terms of the functional state of the solved structure. With additional verification and control, the authors indicated that these structures represent the resting state, which is likely but still hypothetical, and weakly discussed the possible meaning of the structural model and how it adds to our knowledge to the system. In this reviewer's opinion, the paper may be enhanced at least by running a few basic assays to test the protease function. If possible, it will greatly help with a few control cryo-EM reconstructions with ATP or substrate added, which would make the structural interpretation and insights more reliable. At the current situation, one can only make bold assumption and vague discussion as to what has happened to the complex and what we can learn from the symmetric structure, when nearly all substrate-bound AAA+ ATPase complex showed asymmetric conformations. Additionally, the authors might want to consider the following issues when revising the paper.
(1) The authors need to clarify the biochemical condition with respect to the nucleotide type and concentration used in the structural study both in the first section of results and in the Methods. Were the full-length TmFtsH purified with the presence or the absence of ADP or ATP in the buffer? Any magnesium ions provided in the buffer? If no, explain the rationale for why no nucleotides were supplied for the complex?
(2) Where do the ADP ligands bound come from? Were they bound endogenously and copurified or added in the late step of sample preparation?
(3) What were the nucleotide states of the full-length TmFtsH reconstruction shown in Fig. 5? (4) Have the authors conducted degradation assay and/or ATPase activity assay to verify the biochemical activity and function of FtsH or FtsH-MSP nanodisc? (5) The discussion of comparison of the ADP-bound cryo-EM structure with those of ADP-bound crystal structures is not clear. There appears to be notable differences in Supplementary Fig. S4. However, no discussion on such structural differences is provided and any explanation on the compatibility of cryo-EM and crystal structures, as well as why the two methods yield different conformations of the same nucleotide states. This concerns whether the current study rectifies the previous crystal structure study and whether the presented structure implicates a different mechanism. (6) Can the authors comment on the possible reason why all six subunits are in symmetric ADPbound state? It also seems to be a simple test if the authors can add ATP to the purified complex to see if their assertion that ATP and substrate binding are needed to break the symmetric is correct. (7) While the authors claim in the abstract that a possible mechanism of substrate recognition and loading, there are no well-phrased passage describing such a mechanism, except for a few sentences speculating highly hypothetical, vaguely conveyed possibilities. These may not be sufficient to support their claim in the abstract. Response: Following the reviewer's suggestion, we moved the designation of ADP to outside the structure and used the same font color as the nucleotide model.  The authors have solved the structure of Thermotoga maritima (Tm) FtsH in an ADP-bound state using cryo-EM. The cryo-EM imaging was conducted using full-length TmFtsH, and the authors solved the ADP-bound structure comprising hexameric ATPase and protease domains. Through a careful selection of 2D classes containing not only the ATPase and protease domains but also the transmembrane and periplasmic domains, the authors were also able to solve the structure of the ADP-bound full-length TmFtsH with low resolution. By comparing their ADP-bound structure with apo-or ATP-bound structures, conformational changes in the ATPase domains, rather than in the protease domains, were observed. When considering the structure of the full-length TmFtsH, the authors found tilted topologies of transmembrane and periplasmic domains with extra densities in the large gap formed by the tilting, suggesting the mechanism of substrate recognition and initial loading. However, the biological relevance of this tilted structure has not been well supported, and the extra densities in the tilted region have not been fully described, even in the figures. Additionally, while the authors mentioned that the conformational changes in the ATPase and protease domains begin on the connecting loops, there is no detailed description provided. Although the manuscript is well written and easy to follow, the authors need to address the specific concerns listed below.

Response:
We thank the reviewer for the insightful comments to help us improve the manuscript. In the revised manuscript, we have addressed these concerns point-by-point (please see responses below).
1. Throughout the manuscript, the description in the text and figures is not completely matched (e.g., on page 4, line 4, Fig 3c needs to be replaced with Fig 2c). In addition, there are several typos in the figure legends (e.g., in figure 2, even though there is no panel "g," it is mentioned in the legend.).

Response:
We have fixed these problems. Thanks! 2. Page 4, line 15: The authors have used a construct with H423Y mutation (a zinc-bindingdeficient mutant) for cryo-EM analysis and have mentioned it as an active form in the discussion (page 7, line 5). Please show the additional data or provide pertinent explanation to support this.

Response:
The H423Y mutant is protease deficient and could not cleave a protein substrate. However, it's ATPase activity is reserved. This H423Y mutant has been reported to capture multiple FtsH substrates (Westphal et al. 2012(Westphal et al. , doi:10.1074. We attempted to add ATP-Mg2+ with the H42Y mutant to form an FtsH-ATP complex. However, in the reconstructed map, we observed only fully bound ADP molecules in the ATPase domains. We think that ATP was hydrolyzed during our sample preparation. We have revised the text and added the reference in the discussion to make it clear about the use of the protease deficient mutant. (p. 8, lines 9-10).
3. On page 5, line 1, the authors have mentioned that side chain densities were not observed for Arg318', however, the side chain of Arg318' has been marked in figure 3f. In addition, the authors have proposed a potential interaction between Arg318' and phosphate groups in ADP, but no pertinent evidence, such as a distribution of charges in the neighboring region or additional densities around the phosphate groups, has been provided.
Response: Due to the absence of side-chain density, the sidechain of R318' was modeled using its rotamer but only its Cα atom position was used for distance measurement. Nevertheless, the densities for the R318' mainchain and ADP phosphate groups (Fig. 3f) can be observed. In the revised Fig. 3f, we have added mainchain densities (orange) for the R318' segment. (Fig. 3 and its caption) 4. Page 5, line 28-29: How does this structure provide the basis for substrate loading and translocation processes?
Response: Thanks for pointing out this overstated hypothesis. We have deleted this statement in the revised text.
5. In "Apo-ADP state transition" and "ADP-ATP state transition" sections, the authors have mentioned that the structural changes begin on the loops connecting the protease and ATPase domains. However, there is no description of how the structural changes in the loops are propagated to the ATPase domain.

Response:
The connecting loops did not trigger the conformational changes in the ATPase domains. Instead, it's a consequence of conformational changes in the ATPase domains relative to protease hexamer. We have revised the text and added two supplementary figures (Fig. S4a, b) to show conformational changes in the ATPase domains. (p 5, lines 1-3; p 5, lines 26-28).
6. In "Full-length FtsH structure" section, the authors have introduced the full-length structure of TmFtsH whose ATPase and protease domains are tilted relative to the lipid bilayer. Is this tilted structure biologically relevant?
Response: We think this tilted structure is biologically relevant. Structurally, this tilted structure could be of relevance in allowing proximity of substrates to the ATPase pore loops, thus promoting substrate recognition. We have revised the text to include a discussion on the biological relevance of the titled structure. (p. 7, lines 24-25).
7. In the same section, the authors have suggested that the extra densities in the large gap formed by the tilting could originate due to the substrate or the disordered N-terminus of the protein itself. However, considering the low resolution of the full-length structure, especially in the transmembrane domain, it is not clear whether it is reasonable to assume extra densities. It is recommended to provide an additional figure with a description regarding the extra densities.
Response: Taking the reviewers' suggestion, we inspected the extra densities carefully. We agree with the reviewer that due to the disorder of the transmembrane TM domains and the connecting loops between the TM and the ATPase domains, it's premature to assume extra densities arising from a substrate. We have revised the text accordingly.
8. Page 7, line 8: What is the supporting data for this hypothesis?
Response: We deleted this hypothetical sentence in the revised Discussion section.
9. The authors have used the AlphaFold model structure to fit low-resolution densities in the full-length structure. However, it is unclear how the authors created a hexameric model structure using AlphaFold. Since the AlphaFold-Multimer was released recently, it would be helpful to mention in the method section, which version was used for the modeling and how it was carried out (AlphaFold, AlphaFold-Multimer, AlphaFold with modifications, or AlphaFold with Colab, etc.).