Computational Modeling Needed to Test Covalent Protein Modification Mechanism

[mattodd]

In this project, we think we've seen an unusual reaction - a direct substitution at the nitrogen atom of an iminoquinone. The nucleophile is a residue on a protein, in this case SARS-CoV-2 Nsp12 (RdRp). The substrate molecule is RA-0001351. We are seeing covalent modification of the protein according to mass spectrometry. We don't yet know which residue is doing the modification. The postdoc leading the project, Xin (@qxsml), has isolated a model adduct formed between RA-0001351 and a model nucleophile (protected cysteine) in solution that supports the idea of a direct substitution.

There's very little precedence for substitutions of this kind. All we can find are the following two papers "Exploiting Iminoquinones as Electrophilic at Nitrogen “N+” Synthons for C–N Bond Construction" and "Exploiting N-Centered Umpolung Reactivity of α-Iminomalonates for the Synthesis of N-Sulfenylimines and Sulfonamides.". Although different, these papers mechanistically support our findings. On the face of it, an attack by the nucleophile on the ring would seem more likely. These two papers (1 and 2) reported GSH attacking iminoquinones on the ring, not to mention the well-studied cysteine reactivity with quinone compounds, as shown in this paper. Or perhaps it occurs as part of an addition elimination reaction. This paper published in 1990, reported a similar mechanism where the nucleophile attacked the nitrogen but iminoquinone restored aromaticity. However their reactions were carried out under neutral conditions, whereas ours are under mild basic conditions (pH 8.0 in Gel-Based Assay). Considering the proton on the phenol hydroxyl group is acidic, this pathway also seems possible. Additionally this paper shows that with enough driving force, it's possible to remove aromaticity and reform the iminoquinone ring.

The three possibilities are shown below (A-C), where we've simplified the nucleophile to MeSH and the aromatic ring to methyl. Protons are provided to mimic possible protons available in the protein. We know that route A is not happening (since the protein gains mass that implies loss of the sulfonate leaving group).

The challenge: which mechanism would be predicted to be favoured in aqueous solution (i.e. when not driven by the protein): which is the thermodynamically favoured pathway, and what are the relative sizes of the transition state barriers?

We'd like to include this analysis in the paper we'll publish and so suitably robust solutions can lead to authorship.

Please comment directly below, or (less good) email Xin on ucnvqiu@ucl.ac.uk

[mattodd]

Posted on LinkedIn and X

[rahmanszsaleem]

Posted on LinkedIn and X

[velocirraptor23]

Very interesting!!, Any way we can chat about it? Binding pose and reactivity to be taken in count to explain this. BW

Cesar

[qxsml]

Very interesting!!, Any way we can chat about it? Binding pose and reactivity to be taken in count to explain this. BW

Cesar

Hi Cesar,

Thank you very much for your interest. I wonder if anytime next Tuesday suits you for a discussion? I look forward to diving into the details with you.

Best wishes Xin

[BardiB]

Hi everyone,

Just saw the post on linkedin from Mat through a mutual connection and thought it would be interesting to give this challenge a go. Since setting up calculations for transition states and reaction intermediates takes time, I only looked at thermodynamics for the reactants and products.

Calculations of free energies at the GFN2-xTB level in an implicit water solvent indicate that product 2 is indeed the thermodynamically favored product, with the overall reaction being exothermic by 37.9 kcal/mol. The expected reaction for product 1 is exothermic by 24.8 kcal/mol and in the case for the reaction for product 3, then it is exothermic by 26.6 kcal/mol.

From thermodynamics alone, then product 2 should be favored. However, reaction intermediates and transition states could be higher in energy for product 2 than for product 1 and that needs to be explored.

I do want to note that xTB calculations aren’t the most accurate, but they do take seconds on my laptop (unlike DFT calculations) and give a nice initial idea about the problem being studied.

I think that will be enough for now, but I would love to talk more, and perhaps get into DFT calculations if that is something that would be interesting for everyone :)

Cheers! Bardi

[dehaenw]

J. Org. Chem. 1959, 24, 8, 1154–1155 might be a useful reference: the authors find N-chloro quinone imines form sulfenimines (i.e. the same type of adduct detected here). Unfortunately, as it is a fairly old paper, no mechanistic speculations are made. But at least it seems to be some kind of precedent, just with chloride instead of p-chloro phenylsulfonate as the leaving group.

More recently, also on n chloro quinone imine substrates, similar work was done by a group of Ukrainian authors, some spectral data can be found in those papers.

[p-w-kenny]

Although I don’t have access to the software need to perform a computational study, I think that it could be worth thinking about this fourth potential mechanism:

Here’s an article that might be of interest although it’s not directly relevant: https://doi.org/10.1021/bi990125k

[qxsml]

Hi everyone,

Just saw the post on linkedin from Mat through a mutual connection and thought it would be interesting to give this challenge a go. Since setting up calculations for transition states and reaction intermediates takes time, I only looked at thermodynamics for the reactants and products.

Calculations of free energies at the GFN2-xTB level in an implicit water solvent indicate that product 2 is indeed the thermodynamically favored product, with the overall reaction being exothermic by 37.9 kcal/mol. The expected reaction for product 1 is exothermic by 24.8 kcal/mol and in the case for the reaction for product 3, then it is exothermic by 26.6 kcal/mol.

From thermodynamics alone, then product 2 should be favored. However, reaction intermediates and transition states could be higher in energy for product 2 than for product 1 and that needs to be explored.

I do want to note that xTB calculations aren’t the most accurate, but they do take seconds on my laptop (unlike DFT calculations) and give a nice initial idea about the problem being studied.

I think that will be enough for now, but I would love to talk more, and perhaps get into DFT calculations if that is something that would be interesting for everyone :)

Cheers! Bardi

Hi Bardi,

Thank you very much for sharing your insights and the results of your calculations. It's fascinating to see that product 2 is thermodynamically favored based on the GFN2-xTB level calculations. Your explanation about the exothermic nature of the reactions for each product is very informative.

I wonder how much energy would be released in the conversion from product 3 to product 2. If this process releases a significant amount of energy, could mechanism C be the most plausible one?

Cheers! Xin

[velocirraptor23]

My email is cmendozamartinez001@dundee.ac.uk. Happy to discuss this.

[corinwagen]

I ran a quick calculation (click to view) at the B3LYP-D3BJ/6-31G(d)/CPCM(water) level of theory - found a concerted TS, imaginary freq suggests bond breaking and bond forming are both happening.

Running an optimization downhill from the TS region shows N–O scission which also suggests there's not a stable intermediate.

TS seems very early. Not super surprising that this'd be concerted - cf. recent work on SNAr and the Marcus theory arguments made in that paper - but some explicit H-bonding to the ketone might make this reaction more stepwise. It's often tough to model these sorts of explicit inner-sphere effects in silico though.

[corinwagen]

@qxsml @mattodd happy to discuss more - corin@rowansci.com