Write a comment

A direct computational assessment of vinculin-actin unbinding kinetics reveals catch bonding behavior. Willmor J. Peña Ccoa, Fatemah Mukadum, Aubin Ramon, Guillaume Stirnemann, Glen M. Hocky. bioRxiv 2024.10.10.617580

This PREreview was written after a journal club given by O.L.-C. in the bussilab group meeting. All the members of the group, including external guests, are acknowledged for participating in the discussion and providing feedback that was useful to prepare this report. The corresponding authors of the original manuscript were consulted before posting this report, and did not influence its content.

Summary

The authors use Molecular Dynamics with enhanced sampling techniques to gain insight in the directional catch-bond mechanism of Vinculin tail (Vt) interaction with F-actin. They construct two models of the Vt-actin complex that are hypothesized to represent a weak state and a strong state in a force-activated allostery model of the catch bond. They use enhanced sampling techniques to estimate the free-energy landscape of the unbinding process in each state, as well as the unbinding kinetics, in absence and presence of pulling forces of variable intensities and directions along the actin filament axis. Their results demonstrate higher kinetic stability for the strong state with respect to the hypothesized weak one, with unbinding kinetics in range of experimental expectations, confirming the viability of a “3-state” (2 bound states, 4 unbinding pathways) kinetic model of the bidirectional catch bond observed in single-molecule experiments. They additionally observe an increase in bond lifetimes for moderate constant pulling force (10-20 pN) in both directions, indicating an intrinsic catch-bond behavior within each “allosteric” state that may superimpose to the overall allosteric one. They show how an external pulling force affects the positioning of the H1 α-helix believed to act as a regulatory motif of the weak-to-strong transition, providing a compelling structural hypothesis for a force-induced allosteric mechanism. Finally, they provide molecular insight on the difference in stability between both states, highlighting the role of a C-terminal extension (CTE) and the redistribution of Vt-actin contacts under force.

Comments

• We wonder if the nomenclature “Holo” can be confusing at first glance for the reader, given the historical usage of the holo- and apo- prefixes to designate protein constructs with and without their constitutive prosthetic groups (usually non-proteic cofactors)

• We suggest that the authors make clearer the composition of the Holo and Aligned protein sequences given the different numbering in 6UPW and 1QKR (that we guess is due to the presence of Metavinculin instead of Vinculin), that in our understanding are identical except for the absence of the leading H1 helix.

• In FES calculations, since the constant force (that is, a linear bias) is applied on the same coordinate Q_‖, we expect that the resulting FES could be entirely predicted from the FES at zero force by simple addition of the linear slope, given sufficient exploration of the Q_‖ direction during the OPES-MetaD sampling. This fact could be used by the authors to assess the consistency between the FES computed at different forces. Alternatively, one may want to first aggregate the OPES-MetaD simulations at all forces using appropriate reweighting, and then estimate minimum free-energy paths and free-energy barriers at arbitrary force using the aggregated FES. This approach might lead to a better statistical use of the vast amount of simulations and a smoother estimate of the FES at all forces within the studied range. Finally, the multiple trajectories (20 replicates x 9 force values) could be used in a single bootstrap to assess the statistical uncertainty of the results (see below).

• Given the coarse nature of the set of chosen CVs (Q_‖ and Q_⊥) it is unclear whether the Vt is able to regain its canonical binding site during OPES-MetaD, notably because of free rotation with respect to the actin filament.

  • The authors acknowledge the difficulty in the methods (sec. 5.5) without explicitly stating whether such rebinding events happen at all in their simulations. We believe this is an important piece of information for proper understanding of the FES presented. We would suggest showing time series to clarify how many binding/unbinding events are observed.

  • One might expect that the absence of recrossing lead to a poor estimate of the free energy difference between the bound and unbound states as well as the height of the binding free energy barrier. On the bright side, the estimate of the unbinding barrier – which is the one they are the most interested in – can still be expected to be reliable.

  • The authors suggest to run multiple (20) separate OPES-MetaD simulations to compensate for this limitation. It should be acknowledged that independent runs starting from the bound state will not correct for the systematic bias caused by the absence of rebinding events. A bootstrap on these replicas would anyway estimate their statistical error.

  • We wonder if the use of the more specific CV Q_contact might allow for such recrossing to happen within OPES-MetaD, without the need of aggregating a high number of independent trajectories. In our understanding the authors only used Q_contact to assess the robustness of the free-energy barrier height to the precise choice of the projection space, but did not try to perform OPES-MetaD directly on this CV space, which could be instructive.

• The authors analyze the FES by determining a minimum free-energy path using the “String method” as a post-processing method.

  • The Methods section might benefit from some information about the use of the method, in particular that it is directly applied on the 2D CV-space projected FES (as opposed to a search of a minimum energy path on the full potential energy surface as originally proposed in [50]), and provide details about initialization (choice of end points for the string, number of nodes, initial interpolation) and robustness to these parameters in the converged paths and corresponding barrier estimates.

  • Since the FES are aggregated from 20 independent OPES-MetaD runs, it might be relatively straightforward to estimate errors (for example using bootstrapping) and provide error bars on Fig 2c. We believe this would strengthen the significance of the observed barrier difference.

• One may be concerned about the significance and reliability of the constructed “Aligned” state, since this state was constructed by aligning Vt to another conformation (in what we could refer to as “docking-by-homology”) with little experimental confirmation that such a state is stable in vitro. We understand that the model constitutes the core hypothesis of the whole computational approach, and that the consistency of the computational outcomes with single-molecule experiments themselves validates its plausibility. Nevertheless, it could be argued that lower binding affinity and lifetime are to be expected from a suboptimal binding partner in a suboptimal binding pose. This raises the question of whether the proposed model corresponds to a specific binding mode in reality, or if the results could be reproduced with a different alignment. Is this ruled out by the stability observed in the 500 ns simulations shown in Fig S1?

• In Sec 3.2 §4, the authors say “these results [...] do not quantitatively explain the observed experimental results, since the experimental changes in lifetime shown in Fig. 1D reflect a net 1.4 kcal/mol change in barrier in the negative direction, and 0.6 kcal/mol in the positive direction (if one assumes a constant prefactor the kinetic rate constant)”. It was somewhat unclear to us where these values come from (Are they computed from the fitted 3-state model in SI S1? At a specific value of the pulling force?) and how exactly they are compared to the computed barriers for Holo and Aligned to conclude to a discrepancy (Overestimated?)

• In Sec 3.4 §3, the authors convincingly remark that in two out of five simulations of Holo+H1 state pulled towards the barbed end, the conformation of H2–H5 becomes more similar to the Aligned (unbound Vt) structure, suggesting a first step in the strong → weak allosteric transition. We wonder if (i) the specific contacts made with actin and (ii) the specific intra-domain contacts of H1 with the H2–H5 bundle are also indicative of a displacement toward the Aligned state, since this would be an even stronger argument validating the proposed model.

• Given the relative simplicity of the supposed allosteric motif and the suggestive results of the Holo+H1 simulations, we cannot help but wonder whether the authors also tried to "unfold" the H1 helix in the Aligned model with a N-terminal pulling force since this seems a very natural test to look for equally suggestive indications of a weak → strong allosteric transition under force.

• Typos or writing remarks:

  • Fig 1B: The caption is inconsistent with the figure. In the caption, p₁ denotes the COM of Vt helices H2–H5, p₂ the COM of actin A1/A2 and p₃ the COM of actin A4/A5. On the figure instead (p₁, p₂, p₃) → (p₃, p₁, p₂).

  • Some repetitions that might be elegantly avoided

    • “capture the difficult-to-capture” (abstract)

    • “protein of interest [...] on our molecule of interest” (introduction §1)

    • “takes into account a Boltzmann weighted average over all possible configurations [...] a Boltzmann weighted average over all possible configurations” (introduction §4)

  • Typos

    • “FimH-manose” → “FimH-mannose” (introduction §8)

    • “which is the the direction” (3.1 §1)

    • “These approximate one-dimensional free energy pathways also give us a way to define when the system has crossed into the unfolded state” → Maybe the authors meant “unbound state” (3.1 §1)

    • “not all of the catch bond need come from” → “need to come from” (3.2 §3)

    • “we note that these results are suggestive, they do not quantitatively” → We wonder if the authors intended a formulation along the lines of: “we note that despite being suggestive, these results do not quantitatively” (3.2 §4)

    • “a constant prefactor the kinetic rate constant” → “a constant prefactor for the kinetic rate constant” (3.2 §4)

    • “grafted to our Holo structure random orientation” → “grafted to our Holo structure with a random orientation” (3.4 §2)

    • “TIP3 water” → “TIP3P water” (5.1 §2)

Competing interests

O.L.-C. has completed his PhD under the supervision of one of the corresponding authors (G.S.) with whom he has published several papers in the past 5 years. They have no ongoing collaboration at the time of writing this review. The corresponding authors were asked permission before writing and before publishing the review and accepted with no intervention on the outcome.