Author response:
eLife Assessment
The manuscript by von Velsen et al. offers valuable structural insights into the mitogen-activated protein kinase (MAPK) pathway by providing cryo-EM structures of stabilized MEK1-ERK2 kinase-substrate complexes in inactive, active, and nucleotide-free states, complemented by HDX-MS, SAXS, ITC, crystallography, and molecular dynamics. The work provides solid evidence for the overall architecture of the complex and identifies interaction sites that help explain MAPK pathway specificity. However, some mechanistic conclusions are not yet fully supported, particularly the designation of one state as an active phosphoryl-transfer configuration, the claim that substrate binding releases the MEK1 catalytic machinery, the proposed link to processive phosphorylation, and the extrapolation to disease-associated mutations.
We would like to counter the final statement. We were very careful in our description of state 2, while we describe it as ‘active’ we clearly explain that the resolution of the reconstruction is not sufficient to define all the classical indicators of a kinase active state; however, the map is consistent with the active conformation, the complex is active in vitro, and the MEK1 variant used is the well known DD mutant that is constitutively active. While the A-loop is not observed, this is in agreement with many crystal structures of other DD mutants. We therefore decided to define this state as ‘active’ as the A-loop of ERK2 approaches the active site, the alpha-C helix has moved in and the A-loop of MEK1 no longer occludes the active site - to clarify the state we refer to the classically active confirmation as ‘fully active’. Our supporting data also show that the complex is highly dynamic during turnover, meaning we have captured MEK1 in a number of conformations on the landscape of an active state – we feel that rather than a limitation, this is an important observation in MAP2K studies. Finally, the determination of an 80 kDa complex by cryoEM to resolutions well below 4 Å is a huge technical achievement allowing the first snapshots of the MEK1-ERK2 complex to be visualised.
Regarding the A-helix release – our observation is that the helix becomes less folded on binding of substrate. There are many studies, which we cite, that show that destabilising this helix leads to release of the catalytic machinery, see Mansour et al, 1996, Biochemistry, 35, 15529-15536 and Jindal et al 2017 J. Biol. Chem. 292, 18814-18820 for initial studies. Our observation shows that this is linked to substrate binding – a very relevant new insight that demonstrates the importance of this helix, in addition to many previous studies, but links unfolding to substrate recognition for the first time.
For the mechanism of processive phosphorylation – it has been well established that both processive and distributive mechanisms exist. While the way that a distributive mechanism could work is obvious (complete dissociation of the two proteins), it has not been clear how a MAP2K can remain bound to its substrate and exchange nucleotides. While caution should be employed in interpreting our state 3 structure, it clearly shows what nucleotide exchange when bound to substrate can look like and that this low nucleotide affinity state is linked to disorder in the P-loop, the A-helix and substrate binding via the KIM. We would love to perform experiments that could demonstrate this but cannot at present think of an appropriate method – the reviewers did not suggest a route either.
Finally, for the cancer-causing mutations – there are many studies demonstrating that the mutations lead to a destabilisation of the A-helix. Our study links this to substrate recognition. While this is inference, it seems justified to describe a link between substrate recognition, A-helix unfolding and disease mutations given the large body of literature describing these events.
We are currently performing a series of in-cell activity assays that should strengthen our claims regarding the A-helix and other observations in the structure - the histidine interactions in particular.
Public Reviews:
Reviewer #1 (Public review):
Summary:
This manuscript describes three conformers derived from a complex between ERK2-T185V, a variant of MEK1-DD with the KIM sequence replaced by the KIM from the p38 activator, GRA24, ADP, and AlF4-. The goal was to try to capture the complex in its active state. The results show contacts between the kinases between their N-lobe and their C-lobes that resemble MKK6-p38 complexes previously reported by the authors. Two MEK1-ERK2 conformers (States 1,3) are deemed inactive based on the lack of access of ERK-Y187 to the MEK1 active site, and the absence of ADP bound to MEK1 in State 3, while one conformer (State 2) is deemed active, but not fully active due to disorder in MEK1 activation loop (A-loop) and an essential salt bridge between strand beta3 and helix aC. HDX-MS and SAXS solution measurements and all-atom MD simulations are used to model the mutant complex and variants with WT ERK2. The study concludes that substrate recognition involves low-energy contacts with MEK, allowing substantial protein flexibility within the complex in a manner that may accommodate processive phosphorylation of ERK2.
Strengths:
The strengths of the work are that the findings provide important structural insights for MEK-ERK signaling and protein phosphorylation in general. These are valuable given that atomic resolution structures of kinase-substrate complexes are still limited in number. The authors succeeded in showing key contacts between subunits and conformational variations within the complex.
Weaknesses:
Weaknesses were that some of the conclusions about activity state, dynamics, and effects of ligand binding were less convincing. For example, that State 2 truly represents an active configuration seemed ambiguous, given the absence of Mg2+ and AlF4- in the cryoEM structure and disorder in the activation loop and the K97-E114 salt bridge. Conclusions by SAXS that ADP-AlF4 binding increases active site compaction while increasing local flexibility were not rigorously supported by HDX data, given that the latter were performed without ligand. Sections of the narrative and figures throughout were often confusing, and many assertions were made without clear explanation. Data shown in the supplementary materials were not always described in the Results, even those important for the conclusions. Figure legends and text lacked clear descriptions of specific complexes analyzed. Substantial changes are recommended to improve the readability and clarity of the work.
We thank reviewer #1 for in-depth comments and analysis of our manuscript. However, there is a misunderstanding regarding HDX-MS and SAXS data. First, the HDX-MS data were performed on the ADP.AlF4- inhibited complex - this was not made sufficiently clear in the text, and we will amend this. Secondly, we are not trying to support local flexibility observed in the SAXS data with the HDX data. The HDX data support the interactions observed in the cryoEM structure and demonstrate flexibility in the proline-rich loop, the ERK A-loop and unfolding of the MEK1 A-helix. The SAXS data demonstrate that when the transition state complex is formed, the complex is more compact but flexibility within the complex increases - as observed in the dimensionless Kratky plot, supporting our observations in the cryoEM maps. Therefore, the HDX data and SAXS data are separate observations. We thank reviewer #1 for all the comments and will rewrite the manuscript in order to increase clarity as suggested.
Reviewer #2 (Public review):
Summary:
The authors used Cryo-EM to obtain a complex between MEK1 and ERK2. They used the same method as previously used by the same authors to form a stable complex between MKK6 and p38, an extra-strong KIM replacing the wild-type KIM in MEK1. Three conformers were resolved, with the highest resolution of 3.0 Å. The multiple conformers indicate more flexibility in MEK2 than in ERK2. These data suggest that nucleotide exchange is possible while maintaining MEK1-ERK2 interactions. SAXS and HDX data reinforce the idea of flexibility. They point to interactions between the two N-terminal domains between histidines at the N-terminus of helix C and between the G helices that are maintained in each of the 3 conformers, and sequence and structure suggest these histidines may be a source of specificity in MEK1-ERK2 versus MKK6-p38 interactions. A 2.2 Å structure of a complex between ERK1 (88% identical to ERK2) and the docking peptide used was also presented. Molecular dynamics simulations suggest that the MEK1-ERK2 complex can assume a fully active configuration of MEK1.
Strengths:
This is the first structure of a MEK1-ERK2 complex. The structural data are valuable additions to our understanding of MAP2K-MAPK interactions. The discussion points offered in the results section are palatable. These include the origins of specificity and the idea of flexibility in the MAP2K in support of a processive mechanism for the dual phosphorylation activity of MAP2Ks.
Weaknesses:
(1) This reviewer considers that the abstract is overstated. Specifically, this paper does not reveal the molecular details of phosphoryl transfer, nor does it demonstrate that substrate binding releases the catalytic machinery.
(2) The discussion is in some places not supported by evidence and in others has superfluous text. Examples follow:
- "Once the αG-helix is docked, and the C-lobe histidine triad is in place, the N-lobe interactions must then be fulfilled." The data in this paper does not suggest an order of events.
- "If the substrate MAPK is incorrect, the N-lobe interaction will not be stabilised, preventing alignment of the MAPK A-loop with the MAP2K active site." This statement could be described as obvious.
(3) Much of the discussion is embedded in the results, such that it is difficult to separate new facts offered by the paper from speculation.
We thank reviewer #2 for comments and thorough analysis of our manuscript. We agree that perhaps the abstract should be toned down in terms of claims of an active conformation even though we feel that the combination of the first structure of the MEK1-ERK2 complex combined with MD simulation studies clearly demonstrate how phosphoryl transfer will occur. We are now also performing in-cell assays to support our theory of A-helix regulation. For point 2 we based the order of events on data from Juyoux et al 2023 Science, 381, 1217-1225, where in long-timescale MD simulations and experimentally validated adaptive Markov state model simulations the KIM interaction was the last to dissociate after the alpha-G helix interaction. In our MD simulations of the MEK1-ERK2 complex, the interactions formed by the N-lobe were weaker than those formed by the alpha-G. Indeed, dissociation of the N-lobe was observed in various independent simulations, whereas dissociation of the alpha-G was observed only once.
Assuming that, as in the MKK6-p38a complex, the association proceeds along the reverse of the dominant dissociation pathway, the simulations suggest that the KIM interaction forms first, followed by the alpha-G and finally the N-lobes. While alternative association pathways are possible, this interpretation is consistent with the MD and in line with the main association and dissociation pathway observed for the MKK6-p38a complex. This is additionally supported by the observation that there is no catalytic activity if the KIM is removed, demonstrating this as the first essential recruitment event. We will expand this section to include our arguments.
For the second example, we feel this is rather unfair. The statement that if the His-His interaction is absent, catalysis will be prevented is only obvious if one knows about the His-His interaction - this is the first structure showing pathway-specific interactions in the variable loop regions of a MAPK. If it is obvious, why has no one described these residues as important before?
We have taken on board the comments on the style of the manuscript and will make significant changes as suggested by both reviewers.
