Binding to nucleosome poises human SIRT6 for histone H3 deacetylation

Joint Public Review:

Smirnova et al. present a cryo-EM structure of human SIRT6 bound to a nucleosome as well as the results from molecular dynamics simulations. The results show that the combined conformational flexibilities of SIRT6 and the N-terminal tail of histone H3 limit the residues with access to the active site, partially explaining the substrate specificity of this sirtuin-class histone deacetylase. Two other groups have recently published cryo-EM structures of SIRT6:nucleosome complexes; this manuscript confirms and complements these previous findings, with the addition of some novel insights into the role of structural flexibility in substrate selection.

This manuscript is the third in a recent series of reports of cryo-EM structures of Sirt6:nucleosome complexes. The main conclusions of the three studies are similar, but this manuscript from Smirnova et al. includes additional molecular dynamics analysis of the histone tails. These studies suggest that part of the specificity for sites on the H3 tail is the result of only this tail having significant access to the active site. The results are partially validated by showing that H3-K27Ac is sometimes found near the active site in the simulations, and is a weak substrate for the deacetylase in vitro. All of the structures show Sirt6 contacting the acidic patch of H2A-H2B, partial displacement of the H2A C-terminal tail, and displacement of the DNA at the entry-exit site to "unclamp" the H3 N-terminal tail. This manuscript provides additional support for the conclusions drawn in the first two published structures, adds molecular dynamics simulations that provide further insight and includes a biochemical assay that helps to resolve an apparent conflict regarding the deacetylation of H3-K27Ac from the other two papers.

Author Response

The following is the authors’ response to the original reviews.

We are grateful to the reviewers for their remarks which significantly improved the paper. Following these remarks we completed the analysis and validation of our cryo-EM data and peformed several biochemical tests to support our conclusions, lending credbility to the paper. Please find our detailed answers bellow each recommendation of the reviewers.

Major recommendations

1. Errors and omissions in the presentation make the manuscript difficult to access.

a) The text should be edited for grammatical errors more carefully

• We corrected the grammatical errors.

b) Figures should be labeled to allow the reader to follow the logic of the presentation and identify the features being discussed. Identification through the color coding (the identity of the histones, the location of zinc fingers, the active site, and so on) would be helpful.

• We labeled the Rossman fold and Zn-finger domains in Figure 1 and described the histone color codes. The active site of SIRT6 is depicted in Figure 4.

1. The recent publications from the Farnung/Cole and Peterson/Tan/Armache labs need to be cited and the results from Smirnova et al. compared and contrasted with those publications explicitly.

• We added the following paragraph to the discussion section:
  “While this manuscript was under review two studies describing the structure of SIRT6-NCP appeared in press (Wang et al., 2023 ; Chio et al., 2023). The conclusion of these papers regarding the position of SIRT6 on the nucleosome and the unwinding of DNA by the enzyme are similar to our findings. We however dissected in addition the movements of SIRT6 on the nucleosome and analyzed via molecular dynamics the conformations of the H3 tail with respect to the SIRT6 active site. Our results point to the importance of the flexibility between the globular domains of SIRT6 and also explain how SIRT6 can access lysines that are much closer to the histone core than H3K9.”

a) Notably, the Peterson/Tan/Armache labs suggest that H3K27 cannot be deacetylated by SIRT6 whereas the Farnung/Cole labs show deacetylation of H3K27 by SIRT6. Do the results of the Smirnova et al. structure help to resolve this situation?

• We performed deacetylation tests of H3K27Ac nucleosomes and show that SIRT6 deacetylate H3K27Ac albeit at somewhat lower efficiency than H3K9Ac. Our molecular dynamics simulations explain how H3K27, which is close to the histone core, can still be reached by SIRT6 active site. We added the following text to the paper: “To lend support to this claim we tested whether SIRT6 can deacetylate residue H3K27 that was first acetylated by SAGA (Supplemental Fig. 7c). We find that indeed SIRT6 could efficiently deactylate H3K27Ac, although at a somewhat slower rate than H3K9Ac. We conclude that partial DNA unwrapping by SIRT6 allows H3-tail conformations that make lysines that are close to the core of H3 accessible to the enzyme.”

b) The Farnung/Cole labs have visualized an intermediate state of deacetylation. How does this compare to the structure presented in this manuscript? Addressing these points would facilitate further research and discussion in the community.

• We believe the resolution of the SIRT6 Rossmann fold precludes addressing these points.

c) Can the authors exclude the possibility that the additional density observed in Supplemental Figure 6 is not coming from the H3 tail, as observed in the two other structures?

• One density is the continuation of the H2A histone tail. We strongly believe that this density corresponds to this tail. The other density indeed can originate from the H3 tail. Therefore, we didn’t model anything inside it.

d) It would be useful to comment on how much flexibility has been observed in the other structures for the SIRT6 interaction with the acidic patch, and also how other acidic-patch binding proteins compare with the results here.

• We refrain from estimating the flexibility observed in the other structures as no such analysis is provided by these papers. Regarding the interaction with the acidic patch we mention that R175 packs against H2B L103 and serves as a classical “arginine anchor motif” and refer the reader to a review on the topic.

e) Does the presence or absence of NAD+ affect the comparisons among the structures?

• NAD+ binding might affect the fine structure of the active site although NAD+ was not observed in crystal stuctures of SIRT6 in its presence. The resolution of this part precludes further addressing this issue.

1. The lack of biochemical validation of conclusions should be acknowledged and the reasoning behind this choice discussed.

• We added experiments to validate our conclusions with biochemical tests. We produced nucleosomes with acetylatexd histone H3 by employing purified SAGA acetyltransferase complex. We isolated SIRT6 where the four residues implicated in interactions with the acidic patch are mutated to alanines (SIRT6-4A). We show that this mutant has very weak interaction with the nucleosome and much lower H3K9Ac deacetylation activity than WT. Similarly SIRT6-3A with mutations in the residues we suggest involved in binding to nucleosomal DNA also shows weak activity and binding to the nucleosome. We added Supplement Figure 7 that depicts the results of these experiments and embedded reference to these results in the approporiate sections of the text. Furthermore, we also show that SIRT6 is active in deacetylating H3K27Ac. This supports our molecular dynamics simulations showing that when SIRT6 binds the nucleosome, H3 tail can assume conformations where H3K27 is accessible by the enzyme’s active site. These results also appear in Supplement Figure 7.

1. The authors nicely analyze and discuss the conformational flexibility of SIRT6 binding. This is an interesting finding, but Fig. 2 does not adequately convey this flexibility.

• We now considerably improved Figure 2. We added panels c and f which depict clearly the movements we observe.

1. The authors need to explain why two cryo-EM datasets were collected but were not merged, and the labeling of the datasets in the Supplemental Table appear to be switched.

• The two datasets were collected with two very different pixel spacing therefore merging the two was possible only in Relion. This process, however, did not improve the resolution of the SIRT6’s Rossmann fold domain. We thank the reviewer to notice the discrepancy in the text and the Supplemental Table 1, it was corrected.

1. Supplemental Figure 4 should be expanded to show additional representative densities with the respective fit of the model. This will allow the reader to better judge the quality of the data. At least the acidic patch interaction, the DNA-SIRT6 interactions, and the H2A should be shown in this context.

• To illustrate the high-resolution features of the structure as well as the key regions we added Supplemental Figure 4.

1. Standard elements of data analysis and validation should be included (angular distribution plots for cryo-EM reconstructions, a 3D FSC sphericity plot, a Q-score and EMRinger score for the cryo-EM data and atomic model, a model-to-map FSC curve). In general, model building is poorly described as it is unclear which maps (or to what degree different maps) were used for this process. This should be clarified in the methods section and in the Supplemental Table 1.

• The model validation and data analysis details were added to Supplemental Figures 2 and 3 as well as in Supplemental Table 1.

1. The provided maps also do not fully recapitulate the path of the H2A tail. The various density maps and PDB provided for this review do not support the final modeled residues of H2A between residues #118/119-123. This affects the validity of figure 3E and the discussion of the proximity of the potential substrates to the active site. The authors should clarify how they inferred that this is the H2A tail rather than the loosely bound SIRT6 Nterminal loop (whose stability could be altered by the presence or absence of NAD+) as suggested by overlaying the relevant crystal structures.

• We added a panel to Supplemental Figure 4 (d) depicting the density where the H2A tail was modelled.

1. The authors should explain how the data produced an asymmetrically oriented complex with a single SIRT6 molecule bound to one face. Were complexes with two SIRT6 molecules excluded? Is supplementary figure 4A the basis for the orientation and is this sufficient for this purpose?

• Complexes with two SIRT6 molecules were present but only at around 1.5 percent of the whole dataset. These images were excluded from the refinement (shown in Supplementary Figure 2). The DNA orientation is depicted in Supplementary Figure 5A. The resolution obtained at the dyad (~2.5Å) allowed us to distinguish purine and pyrimidine bases. The Widom 601 sequence is asymmetric and the densities clearly show that there is only one orientation of the DNA observed with respect to SIRT6.

1. The authors should clarify how supplemental figure 4B supports the conclusion that DNA is unwrapped. The density is not readily visible and docking of a simple DNA model in the ZN-focused map does not clearly rule out the possibility that this density comes from the H3 N-terminal tail.

• We added to this figure the cryo-EM densities used to model the DNA path and the orientation of SIRT6. This image is now Supplemental Figure 5c.

Minor recommendations

1. The scale bar is missing for the 2D classes shown in Supplemental Figure 2.

• We added the scale bar to the image depicting the 2D classes in Supplemental Figure 2.

1. Masked classifications should be shown in the classification tree (Supplemental Figure 2 +3) with the masks shown as a transparent volume.

• We now show the mask used for the 3D classifications of the SIRT6’s Rossman fold domain in Supplemental Figure 2.

1. Supplemental Figure 3 should show the indicated 3D classifications in the classification tree.

• We added the 3D classifications in Supplemental Figure 3.

1. The authors should consider applying local CTF refinement and particle polishing to improve their resolution.

• We did local and global CTF refinements. Polishing didn’t improve the resolution as movie frame alignment was done outside of Relion.

1. The descriptions of the Widome 601 sequence orientation should be less ambiguous, perhaps mentioning the AT-rich and AT-poor arms instead of left and right arms.

• We corrected the text as required.

1. The statement "Such a large change in DNA trajectory is reminiscent of the chromatin-remodeler ATPases or pioneer transcription factors binding to nucleosome but was not observed in other histone modifiers" requires a citation.

• We added approporiate references.

1. The authors should provide a supplemental figure of the nucleosome-SIRT6 and PRC1-nucleosome structure comparison to complement the discussion section.

• We refer the reader to the paper describing the PRC1-nucleosome structure.