Direct observation of the conformational states of formin mDia1 at actin filament barbed ends and along the filament

  1. Julien Maufront
  2. Bérengère Guichard
  3. Lu-Yan Cao
  4. Aurélie Di Cicco
  5. Antoine Jégou
  6. Guillaume Romet-Lemonne
  7. Aurélie Bertin

Reviewed by

Read the full article

Listed in

Log in to save this article

Abstract

Using electron microscopy, two conformations of the mDia1 formin FH2 domains were directly identified in interaction with the barbed ends of actin. These conformations agree with the speculated open and closed conformations of the “stair-stepping” model. In addition, single FH2 dimers encircling the core of actin filaments were identified.

Article activity feed

  1. Version published to 10.1091/mbc.e22-10-0472
  2. Review Commons

    Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Reply to the reviewers

    Reviewer #1 (Evidence, reproducibility and clarity (Required)):

    This study presents a first structural insight on formin mDia bound to actin filaments in physiological conditions. Based mainly negative stain EM, the authors use 2D and 3D class averaging to describe two main configuration of the formin at the filament barbed end. The two configurations support the previously proposed stair-stepping model, which was based on crystal structures, with an open state where the formin binds two actin monomers and a closed state where three monomers are bound. Because the majority of the structures fall in the first, open state, this supports the existence of this intermediate. The authors also show that the orientation of the free FH2 in this open state is somewhat flexible, as several sub-classes with different angles can be distinguished. Finally, they identify, for the first time, formin densities bound along the length of the filament.

    The data is well presented and I don't have any major issue. The only point is that the information that all the initial structural data comes from negative stain EM comes should be put upfront. One gets the feeling that cryoEM is used throughout until one reads the section on cryoEM. Given that the methodology is now also established for cryoEM, it is regrettable that data was not collected with a 300kV microscope, which may have revealed more details of the conformations, but I understand microscope time is hard to come by, and the authors did a remarkable job from negative-stain EM.

    The finding of formin densities binding along the length of the actin filament is very interesting. Besides the previous cited finding, it also reminds of the observations made in yeast where Bni1 (in S. cerevisiae; PMID 17344480) and For3 (in S. pombe; PMID 16782006) where shown to exhibit retrograde movement with polymerizing actin cables in vivo. This would be interesting to consider in the discussion.

    Reviewer #1 (Significance (Required)):

    This study extends our understanding of the mechanism of formin-mediated actin assembly, by providing a first structural observation in physiological conditions. While confirmatory of previously proposed model, but also excludes an alternative model, and offers novel observations of flexibility and binding along the actin filament length. It will be of great interest to researchers on the actin cytoskeleton.

    My expertise is in the actin cytoskeleton and formins, but I am no expert in EM structural analysis.

    We thank reviewer 1 for the very positive comments and for pointing out the relevance of our study for the actin cytoskeleton field. As advised, we now specify upfront in the abstract and in the introduction that most of the presented results were obtained from negative stain electron microscopy. Following the reviewer’s advice, we have enriched the discussion to highlight the retrograde movements of formins in actin cables observed in vivo.

    Reviewer #2 (Evidence, reproducibility and clarity (Required)):

    Maufront et al. have used EM to study the conformation of mDia1 at the barbed end and the core of actin filaments to explain the molecular mechanism of the FH2 dimer processivity at these sites. Based on modelled structural data they tried to describe how the conformational changes in FH2 dimer lead to its partial dissociation, and then association with filaments during the process of translocation coupled to subunit addition at actin filaments barbed ends. This supports a previous study (Otomo et al. 2005, Nature), in which using X-ray crystallography structural data were used to propose a stair-stepping model for Bni1p translocation at the barbed ends during actin polymerization. The model for mDia1 binding to core filaments is also given. Moreover, using EM structure and the previously reported structures of actin (PDB: 5OOE), and actin with formin FH2 dimer (PDB: 1Y64), authors explained the dynamic nature of FH2 dimer at barbed ends of the filaments using the flapping model. But due to the low resolution of their structures ~ 26-29A0, the finer details of actin and the FH2 dimer structure at barbed ends could not be resolved, leaving open questions about the orientation of actin helical twist at this end during elongation. The authors tried several conditions to get high density barbed-end filaments, but that did not collect adequate number of particles, resulting in low number of particles selected for structure modelling purposes. However, to attain more physiologically relevant structure they used cryo-EM, but were successful in capturing only the open conformation structure of FH2 dimer (at low resolution). Thus, due to low resolution of structures the key findings have not added much to what we already know about the mechanism of FH2 dimer translocation during actin polymerization, except that their studies support the stair-stepping model (Otomo et al. 2005, Nature) and not that of "stepping second" model ( Paul and Pollard. 2008, Curr. Bio.). Thus, this manuscript does not merit publication in this journal.

    We thank reviewer 2 for taking the time to read and review our study. However, we respectfully disagree with the statement that our findings “have not added much to what we already know about the mechanism of FH2 dimer translocation during actin polymerization”. As mentioned in our report, collecting EM data for formins in physiological conditions (at the barbed ends of growing filaments), as we do here for the first time, entails limitations on the number of particles one can observe and on the resulting resolution. Despite this rather low resolution, our data allow us to discriminate between two proposed models accounting for the processivity of formin FH2 domains at filament barbed ends. Being able to determine which of two competing models is valid (as the reviewer says we do) does add a lot to what we already know.

    Major comments:

    1. Present study does not provide any new insight about the conformation of the actin dimer at the barbed ends of actin filaments when FH2 domains of formin are bound. This study appears to be more like an extension of previous research (Otomo et al. 2005, Nature), in which the authors used X-ray crystallography data to propose a model for actin filaments elongation by formin bound at the barbed ends.

    As mentioned above, we respectfully disagree with this remark. First, in Otomo et al. 2005, formins are arranged in a crystal into a non-physiological “daisy chain” arrangement around a non-canonical tetramethyl rhodamine-actin filament. Our observations were made in physiological conditions displaying a single formin dimer at the barbed end of a polymerizing filament. Second, the stair stepping model originating from Otomo et al. was only inferred and extrapolated from the crystal structure and not directly observed. Both the open and the closed conformations were speculations, that had never been observed up to now. In our current report we directly visualize these two conformations. Third, the observations of Otomo et al. were obtained using formin Bni1p from yeast, not the mammalian formin mDia1, for which there is little (PDB 1V9B) structural data available describing the structure of a truncated mDia1 in the absence of actin. Finally, in addition to validating the stair-stepping model experimentally, we make unexpected observations that are totally absent from the model derived from Otomo et al. and subsequent studies.

    The low resolution of structures is a major concern.

    As mentioned above, the limited resolution is the price we had to pay for being in physiological conditions, with formins interacting with the barbed ends of growing actin filaments. Nonetheless, this resolution is sufficient to discriminate between the two previously existing models, and to make new observations, beyond these models.

    Given the low resolution of data, how can the authors decide on the number (4) of classes of FH2 domain (in open state) and present them as "continuum of conformations". They stated "details featured in class 4 do not appear as sharp as in class 2". What was the basis of deciding on the sharpness level?

    We agree that this point was unclear, and we thank the reviewer for pointing it out. The choice of the number of sub-classes for the open state is a trade-off between the sharpness (ie signal-to-noise ratio) of the resulting image, which is a direct consequence of the number of particles within each sub-class, and the internal variability within each sub-class. Class 4 might appear more “blurry” because it gathers particles displaying a range of angles. When increasing the number of generated classes in the 2D processing, we observe angular variations of the FH2 domains intermediate to the ones displayed in Figure 3. However, because increasing the number of classes results in averaging less particles per class, the generated classes appeared more noisy or “blurry” and not as “sharp”, as mentioned in the manuscript. Hence, we chose the number of displayed classes so that the signal-to-noise would remain satisfactory and sufficient to be able to determine the relative angle between the two FH2 domains. To make things clearer, “do not appear as sharp” was replaced by “displayed a lower signal-to-noise ratio and thus looked noisier”. The expression “sharp” was replaced by “enough contrast”.

    The authors showed 30Å structure of FH2 domain encircling actin filaments towards their pointed ends, but said nothing about the kind of decoration it could be, a "daisy-chain" or "concentric circle"? Also, they did not mention anything about the orientation of actin helical twist and specific sites of binding. These information would provide new in-depth understanding of how formins binds while diffusing along the filaments.

    The quality is sufficient to distinguish isolated FH2 dimers along the core of actin.

    Accordingly, the FH2 dimers we observed along the core of our actin filaments adopt a conformation similar to that observed at the barbed end, as mentioned in the text (‘concentric circle’). This observation differs from the reported for INF2 which accumulated along filaments and may interact in a ‘daisy-chain‘ manner (Gurel et al, 2014 ; Sharma et al, 2014). From our data, we can thus assume that formins interact with F-actin along the core of filaments similarly to the way they do at the barbed ends, and might translocate in a two-step manner alongside the actin filament. As stated in the manuscript, the actin helical twist could not be deciphered. For docking the crystal structures within our EM envelope, we used the formin-actin contacts described previously in Otomo et al.

    The author stated - "The leading FH2 domain likely provides a first docking intermediate for actin monomers that would help their orientation relative to the barbed end, resulting in a higher actin monomer on-rate". This statement was made on the basis of observing 79% times FH2 in the open state in their data set. This seems like an overstatement because they don't have any direct structural data to support such claim.

    We agree with the reviewer that our statement, taken from the discussion section, is speculative, and we apologize if this was unclear. Our purpose was to propose a plausible mechanism, based on our structural data, since the FH2 domain stands in front of the barbed end in the “open conformation” and since it likely interacts with actin monomers. We have now rephrased our sentence to state more clearly that is a hypothetical mechanism : “We propose that… could provide…”.

    In the Discussion they mentioned "the FH2 dimer would then be "lagging" behind the elongating barbed end if actin twisting back to 180{degree sign} occurs before the addition of actin monomer and this explains the diffusing along the actin filaments". Did authors encounter filaments with two formins bounds to them in their negative stain images? What is their view on this? In current data, they showed structure in which only one FH2 dimer is bound to the pointed ends of actin filaments. Have they tried increasing the concentration of formins to obtain structures with more than one formin is bound towards the pointed ends of actin filaments?

    Following the recommendations from reviewer 2, we have performed an additional analysis and we now show typical examples of filaments observed with a formin along their core, including cases where two formins are observed on the same filament (Supplementary Figure 12). As we now explain in the discussion section, five different mechanisms (including lagging) can be invoked to explain how a formin can be located along the core of the filament. These five mechanisms can all account for the possibility to have more than one formin on the same filament.

    The lagging mechanism, however, is the only one where we would expect that the filaments with a formin along their core are less likely to also have a formin at their barbed end (because the formin at the core spontaneously departed the bare barbed, that was left bare and with a shorter time to load another formin before fixation of the sample). A simple statistical analysis of our data leads to the estimation that 48 ± 7% (n=50) of actin filaments with a formin within their core also display a formin at their barbed ends. This is significantly less than for the global filament population, where 77 ±0.4% (n=10,461) of barbed ends are decorated with formins. This supports the lagging scenario as a likely mechanism putting formins along the core of the filament.

    Regarding the specific suggestion to increase the formin concentration: We did screen different formin concentrations, but with higher concentrations the level of noise due to unbound formins was significantly increased in the image background and impeded a proper analysis. This is why we consistently used 100 nM formins.

    To increase the density of short filaments for sample preparation, the authors used additional actin binding proteins "shown in supplementary Figure 2.C". There is no supplementary Figure 2.C. Moreover, it would be nice if the concentrations of these proteins are mentioned in the text.

    We apologize for this mistake. Supplementary Figure 2.C has now been added and the protein concentrations have been added in the main text.

    Minor comments:

    1. Figure 1 legend needs editing. E is missing in the legend.

    Thanks for noticing this. We have added the missing legend for 1.E.

    1. There is no supplementary Figure 2.C.

    We apologize for this mistake. We have now added supplementary Figure 2.C.

    It is recommended that the authors report the number of particle used during 2D and RELION 3D classifications in the figures. This would help in better understanding of the probability of the conformations mentioned in the text.

    It was mentioned in the text. We have now made this information clearer to the reader.

    Reviewer #2 (Significance (Required)):

    This is the first direct study showing the two (open and closed) conformations of mDia1 FH2 domain at the barbed ends of actin filaments using EM and cryoEM. The study supports the proposed molecular mechanism of FH2 processivity at the barbed ends during filaments elongation using stair-stepping model reported earlier (Otomo et al. 2005, Nature). For the first time, FH2 has been shown to fluctuate between various angles with respect to static actin filaments, and on this basis they propose a flapping model (Fig 5). They explained the whole mechanism using structural proof, but the low resolution of data raises a question about their quality sufficiency to propose this mechanism. The overall novelty of this manuscripts is insufficient for the publication in this journal. Audience having understanding of the actin and actin binding proteins will be interested in this study. Additionally, researcher from the field of structural biology (EM and CryoEM) will be interested. I have been working in the field of actin and actin binding proteins for past 4 years. Over 10 years' experience in protein biochemistry, structural biology and molecular biology.

    We do not fully understand why, on one hand, reviewer 2 indicates that “for the first time, FH2 has been shown to fluctuate between various angles…” and that “Audience having understanding of the actin and actin binding proteins will be interested in this study. Additionally, researcher from the field of structural biology (EM and CryoEM) will be interested.”. On another hand, reviewer 2 states that “The overall novelty of this manuscripts is insufficient for the publication in this journal.”, which seems contradictory with the above statements and comments.

    Regarding novelty, we insist on the fact that we have achieved for the first time the direct observation of FH2 formin domains at a resolution sufficient to discriminate between two distinct models at the barbed ends, as well as to observe the presence of formin mDia1 along the core of actin filaments in conditions where nobody has proposed that this could happen.

    In addition, we have not specified any specific journal within the possible ones from “review commons”, up to now.

    Reviewer #3 (Evidence, reproducibility and clarity (Required)):

    Summary:

    In this manuscript, Julien et al. use negative stain electron microscopy and cryo-EM to show two conformations of the FH2 domain for the formin mDia1 bound to the barbed end of an actin filament. These conformations support the "stair-stepping" model of FH2 domain movement with an elongating actin filament, as previously postulated by Otomo et al. (reference 1). The two states observe correspond to the "open" (~79%) and closed (~21%). The authors also show the conformational variability of the open state suggesting flexibility in this state. Finally, the authors observe FH2 domains encircling the actin filament at a distance from the barbed end, and suggest that the FH2 can diffuse from the barbed end down the filament.

    Major comments:

    1. Novel insights into formin function derived from this structure would raise impact. Issues that could be addressed include the following. Simply adding some lines to the discussion would not really add impact, but additional experimental/modeling work would.

    We agree that comparing the binding mode of different formins on actin filaments, testing the impact of profilin, and assaying FH2 domains in the absence of FH1, as proposed below, would provide a broad set of interesting additional data. However, without claiming that our results can be generalized to all formins in all conditions, we believe that our findings are novel and should be of interest to a large community. The proposed additional experiments/modeling represent an impressive amount of work, and will be carried out in future investigations. We answer these comments in more details below.

    Whether this model really holds true for all FH2 domains. Formin FH2 dimerization and processive filament barbed end elongation are widespread features of formins, which have been evidenced for many organisms from metazoan to plants. Since we could dock the FH2 from yeast formin Bni1p to account for mammalian mDia1, we think the FH2 domain conformations may be conserved enough among species to display similar translocation mechanisms at the barbed ends of actin filaments, using a two-state mechanism. We chose to use the crystal structure from Bni1 formin (PDB 1Y64) because this structure was obtained in the presence of an actin filaments and brings some insights about the formin-actin contacts.

    In order to convince reviewer 3, we superimposed the existing crystal structure of the FH2 mDia1 domain (PDB: 1V9D) with our model and reconstruction and show (Supplementary Figure 12) that the differences are minor. The mDia1 FH2 domains (atomic structures in red, PDB : 1V9D) are aligned with Bni1p FH2 domains (atomic structures in green and blue, PDB : 1Y64) previously fitted into the electron microscopy envelope of a barbed end capped by a formin in the « open state ». The FH2 domains are well aligned with a slight discrepancy in the knob/actin contact regions (blue arrows). This discrepancy most likely results from the absence of actin partners in the crystals obtained with mDia1 FH2 domains. The Bni1p structure thereby most accurately represents the knob/actin contact region. In addition, the folding of the lasso domain around the post domain is resolved in the Bni1p structure. Note here that the Bni1p lasso domains wrap equally well around the Bni1p post domain and the mDia1 post domain (green arrows).

    Whether the % time spent in the open and closed states might dictate the vastly different elongation rates mediated by different formins. For example, mDia1 is considered one of the 'faster' elongators (equivalent to actin alone in the absence of profilin), while fission yeast Cdc12 essentially caps filaments in the absence of profilin. We have discussed this aspect thoroughly in the discussion section to conclude that:” Our direct assessment of the open state occupancy rate thus provides important information on the molecular nature of the formin-barbed end conformations which could not be directly inferred from kinetic measurements, with or without mechanical tension, so far. Considering a gating factor of 0.9 and considering that formin mDia1 spends 79% of the time in the open state, we can compute that the on-rate for monomers would be slightly higher (14% higher) for an mDia1-bearing barbed end in the open state, than for a bare barbed end.”

    We agree that repeating our set of EM experiments and analysis with other formins, like fission yeast Cdc12, would be interesting. However, this would take a long time, and falls out of the scope of our paper.

    Whether the % time spent in the open and closed states varies if filaments are actively elongating in the presence or absence of profilin. We have chosen not to include profilin in our experiments, and to limit the concentration of G-actin, in order to reduce the background in our EM micrographs. Also, a rapid filament elongation would increase the amount of F-actin per barbed end, while a dense population of short filaments is key to obtain accurate data (as we explain in the discussion, paragraph 1, p9).

    We speculate that, by providing a link between the FH1 domains and the filament barbed end, profilin might very well alter the percentage of time spent in the open state, and mitigate lagging as mentioned in the discussion section. Properly addressing the impact of profilin with our EM experiments is very challenging, for the reasons we have explained. It would require further investigations, beyond the scope of this study.

    How this model impacts the interactions of formins with other proteins at the barbed end. For example, capping proteins. We did not include capping proteins (or other additional proteins) because we wanted to avoid increasing the number of particles from diverse nature per field of view, as they constitute a background that is detrimental for the analysis of EM micrographs. We would have add to sort out additional populations in the course of image analysis. We thus only mixed actin and formin in our assays.

    Do these results relate to formin function in disease? Because formins regulate actin polymerization, their malfunction is linked to a variety of diseases. We therefore expect our findings to be useful to researchers in the medical field. However, our study remains in the scope of basic research and primarily aims at understanding the mechanisms of formin-assisted actin polymerization.

    1. The observation that formin FH2 domains can bind filament sides has been made several times. In particular, a structural model of the FH2 domain of the INF2 formin along the side of an actin filament (Gurel et al 2014, PMID 24915113). This publication also references other papers showing other formins binding to filament sides. There are two points to this comment:

    The model in Gurel et al is that the FH2 domain does not slide down the filament from the barbed end. Rather, the FH2 dimer has an appreciable dissociation rate, enabling it to encircle the filament without having to slide. This FH2 dissociation has been observed for another formin that has been shown to bind filament sides, FMNL1 (called FRL1 in the listed publication), in Harris et al 2006 (PMID 16556604). The authors must explain their reasoning for thinking that mDia1's FH2 can slide down the filament from the barbed end. One possibility is to make observations of this FH2 population in filaments that were not sonicated. What is the average distance of FH2s from the barbed end? We thank the reviewer for pointing our attention to this report from Gurel et al. which we now cite. Following this comment, as well as point 6 of reviewer 2, we now discuss the different mechanisms that could lead to our observation of mDia1 along the core of the filament. We provide a new analysis of our data (discussion section), arguing in favor of the lagging mechanism (i.e. ‘sliding down’ from the barbed end), without excluding the competing scenarios. Briefly, we compute that 48 ± 7% (n=50) of actin filaments with a formin within their core also display a formin at their barbed ends. This is significantly less than for the global filament population, where 77 ±0.4% (n=10,461) of barbed ends are decorated with formins. This supports the lagging scenario, which is the only one where a filament with a formin along its core should be less likely to also have a formin at its barbed end.

    The distance of FH2s from the barbed end would provide additional information. However, it is difficult to estimate, since we often to not see the entire filament, and since we do not know which end is the barbed end.

    Interestingly, in some of the works studying formin binding to filament sides, mDia1 was shown to be rather poor in this property. It would be useful to get an idea of what % of the observed FH2s are in the filament core, as opposed to at the barbed end. Along with the additional analysis mentioned in the previous point, we have now estimated that about 8% of actin filaments display a formin within their core. We have added this number in the manuscript (end of the Results section). As a comparison, in our assays, 77% of filament barbed ends bear a formin.

    The authors must reference the past works showing FH2 binding to filament sides, particularly the structural work. At present, no mention of prior work on FH2 side binding is mentioned. As advised, we have now added additional references and more particularly Gurel et al, 2014.

    1. My major technical concern in this manuscript is that the authors use the FH1-FH2-DAD domain of mDia1 for the imaging, but use FH2 structure of Bni1p for 3D characterization (Otomo et al.). Even though Bni1p has been used for functional and structural analysis, mDia1 and Bni1p FH2 domains share low sequence homology. In addition, mDia1 only partially complements loss of Bni1 function in vivo (Moseley et al., 2004 PMID 14657240). Can the authors use the partial structural information of the mDia1 FH2 from Shimada et al 2004 (PDB 1V9D, PMID 14992721)? Alternately, the authors could have used FH2 domain of Bni1p for imaging. At the very least, the authors should explain clearly why they used different proteins for imaging and modeling.

    As mentioned above (please see our response to point 1.a), we chose to use the crystal structure from formin Bni1 (PDB 1Y64) because this structure was obtained in the presence of an actin monomers, and it thus brings some insights about the formin-actin contacts. The existing structures obtained from formin mDia1 does not include actin (full length by EM: Maiti et al, 2012; crystal structure of subdomains (without FH1): Otomo et al., 2010 PLoS one). It thus seems relevant, in the context of our investigations, to use a structure where formin-actin contacts could be at least partially inferred.

    Further, we superimposed the existing crystal structure of the FH2 mDia1 domain (PDB: 1V9D) with our model and reconstruction and show that the differences are minor (please see the figure in our response to point 1.a, above).

    1. The open and closed states are observed from negative staining data. However, the authors can only find one of the states (open) by cryo-EM, which decreases the confidence level of the paper's conclusions. It would be useful for the authors do a little more to try to find the closed conformation by cryo-EM.

    Using Cryo-EM we can already recover the most abundant open conformation.

    Unfortunately, as pointed out here, the number of particles obtained was too low to enable high resolution and reveal the two observed conformations. Indeed, considering a density of ~ 5 barbed ends par micrograph, the collection of tens of thousands of images would have been necessary, which was not realistic regarding the access we have to latest generation microscopes.

    1. It is unclear whether there are additional effects of using FH1-FH2-DAD protein (not FH2 only) for the imaging, as it shows long protrusion at the tip of actin barbed end. To avoid those concerns the authors could use only FH2 domain of mDia1. Also the authors have to note that they used Bni1p structure because there are no published structures of mDia1 so far.

    We had indeed tried to use a construct deprived of the flexible FH1 domain but the lower purity of this construct and the presence of aggregates led to the collection of lower quality EM micrographs. As profilin was not included in our assay, FH1 domains were not involved in actin polymerization at the barbed end and thus remain very flexible and unstructured. Consistently, we did not detect any additional electronic density that could result from the FH1 domains.

    We indeed point out (p5) that “We used the crystal structure from yeast Bni1p FH2 domains in interactions with an actin filament, rather than the existing one from mammalian mDia1 formin FH2 dimer in isolation (PDB 1V9D), because actin-formin contacts are described in the Bni1p structure.” Minor comments:

    1. Figure 1: It would be interesting if imaging is provided for mDia1 bound to filaments which it has nucleated. Would it be possible that binding to pre-formed filaments is different to that for mDia1-nucleated filaments?

    This is a good suggestion for further investigations but it extends beyond the scope of this study: as we explain, our attempts to nucleate filaments from mDia1 lead to lower quality micrographs, and the sonication of preformed filaments was our best option. However, we do not expect the translocation mechanism of FH2 to differ, as a function of the nucleation history of the filament, since the formin interacts with a filament whose elongation it has assisted over several subunits.

    1. Supplementary figure 2: Numbers of things in the S2 is unclear and poorly described in both results and methods. In particular, figure S2A, the definitions of the black and gray lines (steady state actin) is not clear. Are they containing 5% pyrene actin? Is that actin in polymerization buffer or in monomer-actin buffer? Is that actin incubated with actin polymerization buffer for a certain time before measurement of fluorescent intensity? In figure S2B, how the authors calculate the monomer actin concentration? The authors should provide the information in either results or methods part.

    We apologize for the lack of information. Since this is a standard assay, we have now added more details in the Methods section (rather than in the Results section).

    All curves shown in figure S2 were obtained with 5% pyrene actin. The gray curve shows the pyrene fluorescence intensity baseline from 1 µM G-actin monomers, obtained in G-buffer. The black curve is the fluorescence intensity at steady-state of 1 µM actin in polymerizing conditions, (after 1 hour of incubation at room temperature, at 5 µM, the sample was diluted without sonication and left for another hour before measuring the fluorescence intensity).

    The monomeric actin concentrations shown in figure S2B are derived from the intensity level of pyrene at any time point during the experiment, using the simple equations we now present in the Methods section.

    1. Supplementary figure 2 C: The figure and legend are missing in the manuscript. Furthermore, the authors describe that they used Gc-globulin to sequester monomeric actin in solution. Is gc-globulin widely used for actin monomer sequestration?

    Thank you for noticing the missing panel which is now back in place. Indeed, Gc globulin is known to sequester G-actin (Van Baelen, H., R. Bouillon, and P. DeMoor. 1980. “Vitamin D binding protein (Gc-globulin) binds actin”. J. Biol. Chem. 255:2270-2272). This is why we have attempted to use it. We could see a slight effect but we did not want to increase the noise within our images with additional proteins that would have made the analysis more complicated.

    CROSS-CONSULTATION COMMENTS Reviewer #1 mentions that the authors identify formin densities bound along the actin filament for the first time. I agree that the imaging of the mDia1 along the actin filament using electron microscopy is novel, but the concept of formin binding has already been found and studied well with other formins (PMID 16556604, PMID 24915113) and even mDia1 has poor binding activity compared to other formins. It was really nice of the authors to show the mDia1 side filament binding, but I don't think it is a striking finding.

    I have no comment for Reviewer #2.

    Reviewer #3 (Significance (Required)):

    If the EM refinements and 3D rendering techniques are conducted rigorously (which this reviewer is unable to judge), the data support an existing theory of how FH2 domains interact with the actin barbed end. Overall, the data will be of interest in formin field. However, as written the paper confirms an existing model, and does not represent new insight. Impact would be raised by providing insights from these findings that impact formin function or disease.

    We have answered this concern above. The existing models were speculative and not based on direct observations. They relied on data obtained in non-physiological conditions.

    Here, we directly observe two distinct conformations in our structural data, and clearly validate one model over the other. This provides a major advancement in our understanding of formin interaction with actin filaments. In addition, we uncovered an unexpected behavior of formin mDia1, which can readily be found along the core of the filament without the aid of additional proteins, and we propose a mechanism based on our data to account for this observation.

    Another main point is that the observation of FH2 domains bound along an actin filament, while interesting, is not novel. Others have found this for other formins, but those papers are not referenced here.

    The direct binding of formins to the sides of actin filaments is thought to be specific to some particular formins (we now cite additional references in our manuscript, to discuss this point). Formin mDia1, which is a ubiquitous and widely studied mammalian formin (perhaps the most studied), has only been described to diffuse along actin filaments when a capping protein dislodges it from the barbed end (Bombardier et al. Nat Com 2015). Here, we show that formin mDia1 can be found encircling the core of actin filaments, in the absence of any capping protein. This behavior is novel and unexpected. It should open new avenues for research on formin mDia1, as well as on other formins.

  3. Review Commons

    Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Referee #3

    Evidence, reproducibility and clarity

    Summary:

    In this manuscript, Julien et al. use negative stain electron microscopy and cryo-EM to show two conformations of the FH2 domain for the formin mDia1 bound to the barbed end of an actin filament. These conformations support the "stair-stepping" model of FH2 domain movement with an elongating actin filament, as previously postulated by Otomo et al. (reference 1). The two states observe correspond to the "open" (~79%) and closed (~21%). The authors also show the conformational variability of the open state suggesting flexibility in this state. Finally, the authors observe FH2 domains encircling the actin filament at a distance from the barbed end, and suggest that the FH2 can diffuse from the barbed end down the filament.

    Major comments:

    1. Novel insights into formin function derived from this structure would raise impact. Issues that could be addressed include the following. Simply adding some lines to the discussion would not really add impact, but additional experimental/modeling work would.
      • a. Whether this model really holds true for all FH2 domains.
      • b. Whether the % time spent in the open and closed states might dictate the vastly different elongation rates mediated by different formins. For example, mDia1 is considered one of the 'faster' elongators (equivalent to actin alone in the absence of profilin), while fission yeast Cdc12 essentially caps filaments in the absence of profilin.
      • c. Whether the % time spent in the open and closed states varies if filaments are actively elongating in the presence or absence of profilin.
      • d. How this model impacts the interactions of formins with other proteins at the barbed end. For example, capping proteins.
      • e. Do these results relate to formin function in disease?
    2. The observation that formin FH2 domains can bind filament sides has been made several times. In particular, a structural model of the FH2 domain of the INF2 formin along the side of an actin filament (Gurel et al 2014, PMID 24915113). This publication also references other papers showing other formins binding to filament sides. There are two points to this comment:
      • a. The model in Gurel et al is that the FH2 domain does not slide down the filament from the barbed end. Rather, the FH2 dimer has an appreciable dissociation rate, enabling it to encircle the filament without having to slide. This FH2 dissociation has been observed for another formin that has been shown to bind filament sides, FMNL1 (called FRL1 in the listed publication), in Harris et al 2006 (PMID 16556604). The authors must explain their reasoning for thinking that mDia1's FH2 can slide down the filament from the barbed end. One possibility is to make observations of this FH2 population in filaments that were not sonicated. What is the average distance of FH2s from the barbed end?
      • b. Interestingly, in some of the works studying formin binding to filament sides, mDia1 was shown to be rather poor in this property. It would be useful to get an idea of what % of the observed FH2s are in the filament core, as opposed to at the barbed end.
      • c. The authors must reference the past works showing FH2 binding to filament sides, particularly the structural work. At present, no mention of prior work on FH2 side binding is mentioned.
    3. My major technical concern in this manuscript is that the authors use the FH1-FH2-DAD domain of mDia1 for the imaging, but use FH2 structure of Bni1p for 3D characterization (Otomo et al.). Even though Bni1p has been used for functional and structural analysis, mDia1 and Bni1p FH2 domains share low sequence homology. In addition, mDia1 only partially complements loss of Bni1 function in vivo (Moseley et al., 2004 PMID 14657240). Can the authors use the partial structural information of the mDia1 FH2 from Shimada et al 2004 (PDB 1V9D, PMID 14992721)? Alternately, the authors could have used FH2 domain of Bni1p for imaging. At the very least, the authors should explain clearly why they used different proteins for imaging and modeling.
    4. The open and closed states are observed from negative staining data. However, the authors can only find one of the states (open) by cryo-EM, which decreases the confidence level of the paper's conclusions. It would be useful for the authors do a little more to try to find the closed conformation by cryo-EM.
    5. It is unclear whether there are additional effects of using FH1-FH2-DAD protein (not FH2 only) for the imaging, as it shows long protrusion at the tip of actin barbed end. To avoid those concerns the authors could use only FH2 domain of mDia1. Also the authors have to note that they used Bni1p structure because there are no published structures of mDia1 so far.

    Minor comments:

    1. Figure 1: It would be interesting if imaging is provided for mDia1 bound to filaments which it has nucleated. Would it be possible that binding to pre-formed filaments is different to that for mDia1-nucleated filaments?
    2. Supplementary figure 2: Numbers of things in the S2 is unclear and poorly described in both results and methods. In particular, figure S2A, the definitions of the black and gray lines (steady state actin) is not clear. Are they containing 5% pyrene actin? Is that actin in polymerization buffer or in monomer-actin buffer? Is that actin incubated with actin polymerization buffer for a certain time before measurement of fluorescent intensity? In figure S2B, how the authors calculate the monomer actin concentration? The authors should provide the information in either results or methods part.
    3. Supplementary figure 2 C: The figure and legend are missing in the manuscript. Furthermore, the authors describe that they used Gc-globulin to sequester monomeric actin in solution. Is gc-globulin widely used for actin monomer sequestration?

    Referees cross-commenting

    Reviewer #1 mentions that the authors identify formin densities bound along the actin filament for the first time. I agree that the imaging of the mDia1 along the actin filament using electron microscopy is novel, but the concept of formin binding has already been found and studied well with other formins (PMID 16556604, PMID 24915113) and even mDia1 has poor binding activity compared to other formins. It was really nice of the authors to show the mDia1 side filament binding, but I don't think it is a striking finding.

    I have no comment for Reviewer #2.

    Significance

    If the EM refinements and 3D rendering techniques are conducted rigorously (which this reviewer is unable to judge), the data support an existing theory of how FH2 domains interact with the actin barbed end. Overall, the data will be of interest in formin field. However, as written the paper confirms an existing model, and does not represent new insight. Impact would be raised by providing insights from these findings that impact formin function or disease.

    Another main point is that the observation of FH2 domains bound along an actin filament, while interesting, is not novel. Others have found this for other formins, but those papers are not referenced here.

  4. Review Commons

    Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Referee #2

    Evidence, reproducibility and clarity

    Maufront et al. have used EM to study the conformation of mDia1 at the barbed end and the core of actin filaments to explain the molecular mechanism of the FH2 dimer processivity at these sites. Based on modelled structural data they tried to describe how the conformational changes in FH2 dimer lead to its partial dissociation, and then association with filaments during the process of translocation coupled to subunit addition at actin filaments barbed ends. This supports a previous study (Otomo et al. 2005, Nature), in which using X-ray crystallography structural data were used to propose a stair-stepping model for Bni1p translocation at the barbed ends during actin polymerization. The model for mDia1 binding to core filaments is also given. Moreover, using EM structure and the previously reported structures of actin (PDB: 5OOE), and actin with formin FH2 dimer (PDB: 1Y64), authors explained the dynamic nature of FH2 dimer at barbed ends of the filaments using the flapping model. But due to the low resolution of their structures ~ 26-29A0, the finer details of actin and the FH2 dimer structure at barbed ends could not be resolved, leaving open questions about the orientation of actin helical twist at this end during elongation.

    The authors tried several conditions to get high density barbed-end filaments, but that did not collect adequate number of particles, resulting in low number of particles selected for structure modelling purposes. However, to attain more physiologically relevant structure they used cryo-EM, but were successful in capturing only the open conformation structure of FH2 dimer (at low resolution). Thus, due to low resolution of structures the key findings have not added much to what we already know about the mechanism of FH2 dimer translocation during actin polymerization, except that their studies support the stair-stepping model (Otomo et al. 2005, Nature) and not that of "stepping second" model ( Paul and Pollard. 2008, Curr. Bio.). Thus, this manuscript does not merit publication in this journal.

    Major comments:

    1. Present study does not provide any new insight about the conformation of the actin dimer at the barbed ends of actin filaments when FH2 domains of formin are bound. This study appears to be more like an extension of previous research (Otomo et al. 2005, Nature), in which the authors used X-ray crystallography data to propose a model for actin filaments elongation by formin bound at the barbed ends.
    2. The low resolution of structures is a major concern.
    3. Given the low resolution of data, how can the authors decide on the number (4) of classes of FH2 domain (in open state) and present them as "continuum of conformations". They stated "details featured in class 4 do not appear as sharp as in class 2". What was the basis of deciding on the sharpness level?
    4. The authors showed 30Å structure of FH2 domain encircling actin filaments towards their pointed ends, but said nothing about the kind of decoration it could be, a "daisy-chain" or "concentric circle"? Also, they did not mention anything about the orientation of actin helical twist and specific sites of binding. These information would provide new in-depth understanding of how formins binds while diffusing along the filaments.
    5. The author stated - "The leading FH2 domain likely provides a first docking intermediate for actin monomers that would help their orientation relative to the barbed end, resulting in a higher actin monomer on-rate". This statement was made on the basis of observing 79% times FH2 in the open state in their data set. This seems like an overstatement because they don't have any direct structural data to support such claim.
    6. In the Discussion they mentioned "the FH2 dimer would then be "lagging" behind the elongating barbed end if actin twisting back to 180{degree sign} occurs before the addition of actin monomer and this explains the diffusing along the actin filaments". Did authors encounter filaments with two formins bounds to them in their negative stain images? What is their view on this? In current data, they showed structure in which only one FH2 dimer is bound to the pointed ends of actin filaments. Have they tried increasing the concentration of formins to obtain structures with more than one formin is bound towards the pointed ends of actin filaments?
    7. To increase the density of short filaments for sample preparation, the authors used additional actin binding proteins "shown in supplementary Figure 2.C". There is no supplementary Figure 2.C. Moreover, it would be nice if the concentrations of these proteins are mentioned in the text.

    Minor comments:

    1. Figure 1 legend needs editing. E is missing in the legend.
    2. There is no supplementary Figure 2.C.
    3. It is recommended that the authors report the number of particle used during 2D and RELION 3D classifications in the figures. This would help in better understanding of the probability of the conformations mentioned in the text.

    Significance

    This is the first direct study showing the two (open and closed) conformations of mDia1 FH2 domain at the barbed ends of actin filaments using EM and cryoEM. The study supports the proposed molecular mechanism of FH2 processivity at the barbed ends during filaments elongation using stair-stepping model reported earlier (Otomo et al. 2005, Nature). For the first time, FH2 has been shown to fluctuate between various angles with respect to static actin filaments, and on this basis they propose a flapping model (Fig 5). They explained the whole mechanism using structural proof, but the low resolution of data raises a question about their quality sufficiency to propose this mechanism. The overall novelty of this manuscripts is insufficient for the publication in this journal.

    Audience having understanding of the actin and actin binding proteins will be interested in this study. Additionally, researcher from the field of structural biology (EM and CryoEM) will be interested. I have been working in the field of actin and actin binding proteins for past 4 years. Over 10 years' experience in protein biochemistry, structural biology and molecular biology.

  5. Review Commons

    Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Referee #1

    Evidence, reproducibility and clarity

    This study presents a first structural insight on formin mDia bound to actin filaments in physiological conditions. Based mainly negative stain EM, the authors use 2D and 3D class averaging to describe two main configuration of the formin at the filament barbed end. The two configurations support the previously proposed stair-stepping model, which was based on crystal structures, with an open state where the formin binds two actin monomers and a closed state where three monomers are bound. Because the majority of the structures fall in the first, open state, this supports the existence of this intermediate. The authors also show that the orientation of the free FH2 in this open state is somewhat flexible, as several sub-classes with different angles can be distinguished. Finally, they identify, for the first time, formin densities bound along the length of the filament.

    The data is well presented and I don't have any major issue. The only point is that the information that all the initial structural data comes from negative stain EM comes should be put upfront. One gets the feeling that cryoEM is used throughout until one reads the section on cryoEM. Given that the methodology is now also established for cryoEM, it is regrettable that data was not collected with a 300kV microscope, which may have revealed more details of the conformations, but I understand microscope time is hard to come by, and the authors did a remarkable job from negative-stain EM.

    The finding of formin densities binding along the length of the actin filament is very interesting. Besides the previous cited finding, it also reminds of the observations made in yeast where Bni1 (in S. cerevisiae; PMID 17344480) and For3 (in S. pombe; PMID 16782006) where shown to exhibit retrograde movement with polymerizing actin cables in vivo. This would be interesting to consider in the discussion.

    Significance

    This study extends our understanding of the mechanism of formin-mediated actin assembly, by providing a first structural observation in physiological conditions. While confirmatory of previously proposed model, but also excludes an alternative model, and offers novel observations of flexibility and binding along the actin filament length. It will be of great interest to researchers on the actin cytoskeleton.

    My expertise is in the actin cytoskeleton and formins, but I am no expert in EM structural analysis.

  6. Version published to 10.1101/2022.06.28.497909v1 on bioRxiv