19F NMR Fragment Screen Approach for SARS-CoV-2 Helicase, Nsp13.

[TomkUCL]

A summary and discussion of the 19F NMR fragment screen approach towards the SARS-CoV-2 helicase, Nsp13.

Current Objectives as of 11th March 2024:

@TomkUCL:

• [x] 1. Synthesis of fluorinated ATP-analogue (2-fluoro ATP) to study the activity of the SARS-CoV-2 helicase, Nsp13, under physiological conditions via an n-FABS experiment
• [x] 2. Acquiring fragments of interest from Joe Newman (pdb 5rmm and related analogues with biochemical activity). See update comment
• [x] 3. Quality control (1H and 19F NMR) of fragments delivered from the supplier @chriswaudby has the library arrived? and if so can I be getting on with QC?

**@jelenaUCL profile

• [x] 1. Nsp13 protein synthesis and purification.
• [ ] 2. QC of protein, including a description of methods.

@nikita-harvey profile @chriswaudby profile:

• [x] 1. Status of fragment library delivery.
• [x] 3. Listing, ordering, and stock check of required consumables (tubes, buffers, etc).

[TomkUCL]

1) Synthesis of 2-fluoro ATP for studying Nsp13 activity and binding to act as a positive control and measure of ATPase inhibition.

Some background information on the use of 19F NMR to study the biochemical ATP-hydrolysis behaviour of several ATPase enzymes, including the viral Zika NS3 helicase

At pH 7, we are looking at the stabilised form as a tributylammonium salt:

Details of the experimental procedure can be found here in the Signals ELN

tR = 0.45 min. 2F-ATP, [M-H]- expected 524, appears at 523.6. The LCMS looks clean. 2F-ATP reverse phase 2 (5-20_ MeCN), then hexane wash in buffer pH 7.pdf The product is stored in pH 7.0 buffer (tributyl ammonium acetate) at -20degC to prevent degradation and hydrolysis.

1H NMR of the product (teal) vs 2-fluoroadenosine starting material (red) agrees with full conversion to the desired product after two reverse phase purifications and a wash with hexane to remove excess tributylamine base.

19F NMR suggests the product contains 2F-ATP (left) and 2F-ADP (right) in approximately 2:1 ratio.

The 31P NMR (set up by Nikita) indicates around seven phosphorus environments, rather than the expected three, which further indicates hydrolysis.

I am looking for suggestions in terms of stabilising the product. Preferably I'd isolate the lithium salt by adding LiOH, but I am concerned about phospho-ester hydrolysis. Jelena pointed out that thawing and refreezing can be an issue with nucleotide stability, so it will remain frozen at -20C until use.

[TomkUCL]

2) Purchasing fragments of interest from Joe Newman (pdb 5rmm and related analogues with biochemical activity).

Fragments of interest from Joe Newman @Oxford include PDB 5rmm and its para-tolyl analogue, which showed near identical poses by crystallography (shown here below).

Additionally, below are some fluorinated analogues of 5rmm from Joe that showed activity by fluorescence polarisation assay which he would be happy for us to include in the 19F NMR screen:

Update - Monday 11th March 2024

Joe will send some of these fluorinated 5rmm analogues (10uL at 100 mM in DMSO), some with inhibition data and structures, to use as positive controls for the 19F NMR screen. @chriswaudby is this enough?

[chriswaudby]

Update on the library side of things:

• We have 10 mg of compounds, MW 100-300.
• Plan to initially dissolve in dry d6-DMSO at 100 mM - ca. 500 uL. If solubility is limited, we will drop compounds to 50 mM.
• Take 50 uL aliquots for working stocks, and freeze the rest (-80ºC) for storage. Working stocks to be stored at RT under N2.
• Quality control: dilute 5 µL working stocks with 195 µL d6-DMSO (1.25–2.5 mM). Store at RT under N2. Test purity/stability by LCMS at 0, 1, 3, 6 months.
• Initial NMR characterisation: use diluted stocks to prepare samples of individual fragments (50-100 µM in aqueous buffer). Acquire 1D 1H, 1D 19F experiments etc. ERETIC will provide quantitative concentration reference.
• Preparation of cocktails: Mix 50 x 20µL for fragment concentrations of 1-2 mM. This will give final fragment concentrations of 50-100 µM during screening (5% DMSO).

Queries:

• Choice of aqueous buffer. Suggest 25 mM sodium phosphate, 50 mM NaCl, pH 7.0, 5% D2O.
• Solubilising fragments - Jenny Lam's group have microplate shakers, unless there are alternatives? Sonication?

Current activities:

• Nikita testing accuracy of robot, particularly for dispensing DMSO. Have contacted Bruker for further advice.
• Nikita setting up ERETIC qNMR method
• Chris setting up broadband 19F relaxation measurements

[TomkUCL]

@chriswaudby In terms of buffer, the following solution was used in this paper to study Zika NS3 helicase for recording NS3h-ADP-MgF3(Wat)- 19F NMR spectra:

500 µM NS3 helicase (NS3h) was used in a buffer containing 20 mM Tris pH 7.5, 300 mM NaCl, 2mM DTT, 10 mM MgCl2, 20 mM NH4F, and 10 mM ADP/GDP

ADP-MgF3 could be a useful measure of helicase activity, in addition to 2-fluoro-ATP. @jelenaUCL do you recall how much ADP was ordered?

[jelenaUCL]

Regarding buffer, when the protein is present I would prefer 25 mM HEPES pH 7.5, 150 mM NaCl, 0.5 mM TCEP. TCEP can be omitted, but this protein has 26 Cys residues, so I would prefer to keep it. In the optimization step I can try to go down in salt concentration.

[nikita-harvey]

Update on the liquid handling robot - Bruker have now sent me instructions to change the pump calibration settings for DMSO. I'm waiting for an order of non-deuterated DMSO to arrive (which we'll have to use as the wash solvent), then will calibrate the pump. After that we should be able to go ahead with adding DMSO to the compounds, even with custom amounts for each if needed.

[TomkUCL]

Hi @chriswaudby @nikita-harvey, just to check where we stand, we have the compounds in an they just need diluting and LCMS?

• A quick update - Jelena's Nsp13 production is progressing well and we have received the three compounds shown below from Joe to screen.

• The first compound (5MMM fluorinated analogue) showed 53%inhibition @100uM by FP assay, and we have the crystal structure for the non-fluorinated analogue. I think this is our best bet for a positive control for the 19F NMR screen.

• Latest from Joe is that SGC Toronto didn’t find much sign of binding on the SPR but they did say that they tested on their F19 NMR and found some activity (at least one of these). He suspects that the most likely to be active is in well A01 (crystal structure shown above) but I have not seen the data yet, also activity is likely to be relatively weak around 50 uM or higher but might be a good control for establishing the assay.

Some FP curves are shown below.

[nikita-harvey]

@TomkUCL Yes, that's correct. I have now calibrated the robot to handle DMSO accurately, so we should be able to start suspending the standards pretty soon. After that we will need to run them all by NMR before deciding on cocktail formations.

[TomkUCL]

Whilst housebound I have done a focused virtual screen of >9000 compounds based on the 5rmm fluorinated analogue (Well A11) that we hope to use as a positive control compound in the 19F NMR fragment screen.

The hope is that this will generate hit compounds quickly based on fragment hits from the 19F NMR screen using 1-step chemistry reactions following a fragment-growing / combinatorial chemistry approach.

Briefly, a combinatorial library of 10'882 compounds was created using an SNAr reaction scheme of compound A11 with the Enamine SNAr building block library. These were put through Lipinski filters, 3D structures generated from SMILES using Gypsum-DL 1.2.1 then rigid docked against Nsp13 (pdb 5rmm) 5' ssRNA site using Vina 1.2.3.

After docking, a LigGrep filter was then applied to select poses where an oxygen atom was in contact with ASN516 NH (conserved in the crystal structures of 5rmm and its tolyl analogue). 19 compounds were found to pass this filter, which were then subjected to a second, more computationally expensive round of docking. The resulting poses overlayed with the experimental fragment 5rmm, along with the docking scores are shown in the slide deck.

[TomkUCL]

Second LCMS of 2F-ATP in buffer prior to positive control screening shows no indications of degradation. 6.5.2024.pdf

[TomkUCL]

Update 06/06/2024

• Below are the positive control fragments from Joe that we are testing today, plus 2F-ATP as discussed above. Compound A01 is the fluoro analogue of the co-crystalised fragment (PDB ID 5RMM) and has the largest ATPase activity, and so is the most promising. It would be good to get the ATPase binding curves for these compounds and ideally see how they fair in the FRET unwindase assay.

• We diluted each positive control compound in d6-DMSO from 100mM (15uL) to a 5.6mM (100uL) working stock. We will then run a QC in @jelenaUCL's Nsp13 buffer (25.0mM HEPES, 150.0mM NaCl, 0.5 mM TCEP) both with and without protein at a final total volume of 560uL.

• 2F-ATP is in buffer and cannot be dried, so I will dilute this based on full conversion to an approximate concentration of 5.6 mM based on full NMR conversion to ATP/ADP product. I can then work backward by taking a known volume and adding internal standard (e.g. TMS) to quantify, but for now, we just want to see binding to the protein.

We will begin screening at 1.0 uM protein concentration and increase if necessary.

UPDATE 06/06/2024, 17:05

• No fluorine signal was seen in the initial 2F-ATP test, so we rescreened it at 100x higher concentration (approx. 540.0 uM) based on theoretical yield.

• No apparent binding of fragment A01 was seen at 5.6uM and 1.0 uM protein concentration.

UPDATE 07/06/2024

Here are @jelenaUCL 's thoughts regarding yesterday's experiments: nsp13 test samples with 4 potential positive controls.

• One of my objectives was to see if we can push the protein concentration below 5uM. Reason is that we would like to screen, in addition to the apo form, DNA bound form, and with non-hydrolysable ATP analogue. Moreover, there will be also other fluorinated molecules that will need testing.

• Conditions I eventually set on is 4 uM nsp13 + 50 uM compound (in 25 mM HEPES, 150 mM NaCl, 0.5 mM TCEP)

• One of the controls we also wanted to check is just 1D 1H of nsp13 - the problem is that with such low concentrations, we see very little signal for the amide region. Nikita set up a longer experiment o/n and there are some low-intensity peaks. (As a former protein NMR person this to me does not look the greatest, but maybe I am not used to looking at such low conc proteins. Nevertheless, if we are collectively still worried about the folding state of the protein I can sacrifice one aliquot of protein that is at 20uM and do proper 1D 1H spectra. @chriswaudby let me know what are your thoughts?

• We tested 4 compounds - "A1-A3" are from Joe Newman and show some inhibitory effects in biochemical assays and "A4" 2F-ATP made by @TomkUCL .

• From Joe's fragments I do not see any difference when comparing to reference. @TomkUCL @nikita-harvey to double-check the spectra to confirm. We need to deconvolute why we are not seeing binding for these fragments, particularly compound A1.

• For the 2F-ATP we see the appearance of the additional peak and an increase of one of the 2 peaks from reference. What I believe is probably happening is hydrolysis of the ATP by protein. But the open question is how we can prove the binding event with this?

Points of discussion:

1. Is additional protein QC necessary?
2. Deconvolute what is happening with 2F-ATP when it is added to the protein.
3. I know that Chris wanted to screen at much higher concentrations - if that is the worry, we can still try that and I will make an additional protein batch so we have enough for all other future plans.
4. Do we need to increase concentration of buffer to 50 mM for fragment screening in case compounds are very basic/acidic? Or 25 mM should be enough to keep pH constant.

UPDATE 21/06/2024

Comments from @chriswaudby are that from looking at the 19F R2 and R1 measurements, there’s no sign of binding for any compound. All the peaks within the A4 sample relax more quickly but this is not due to the protein - perhaps a consequence of larger molecular weight, or formation of micelles/aggregates?

Chris - if this is true, how do we explain the new left-side peak and the depletion of the middle peak but not the right side in the presence of protein?

I think that the 2F-ATP sample used (A4) could be a mixture of ATP, ADP and AMP based on the 1D 19F NMR. However, the 1H signals are buried under the water from the buffer, so a water suppression sequence might be needed to confirm.

Attempts to purify 2F-ATP further have proven unsuccessful. I will attempt to crystallise it as a Na or Li salt, which should aid the purification. Chris pointed out that we were missing Mg2+ ions in the buffer which the helicase uses to bind ATP so we will need to add cellular levels of MgCl2 to the buffer in the next attempt.

@nikita-harvey and @jelenaUCL measured 1D 1H spectra of nsp13 at higher conc (and without 19F suppression) and concluded that protein looks folded. Buffer: 50 mM HEPES pH 7.2, 150 mM NaCl, 1 mM MgCl2, 0.5 mM TCEP + 10%D2O.

Her overall conclusion is that protein is mostly likely ok, but to further confirm this we need a good positive control which so far does not exist for Nsp13.

FURTHER ACTIONS:

• [ ] @TomkUCL attempt to crystallise 2F-ATP as the Na or Li salt form and recheck purity for rescreening against Nsp13

• [ ] @jelenaUCL or @nikita-harvey Upload 1D protein spectra with an explanation of the protein being folded.

• [ ] @nikita-harvey check 1H spectra of BioNET fragments that seem to be missing fluorine peaks to determine the quality of the library (see https://github.com/StructuralGenomicsConsortium/CNP4-Nsp13-C-terminus-B/issues/46#issuecomment-2152904781 for spreadsheet link).

• [ ] @TomkUCL generate fragment cocktail mixtures in CCPNMR once anomalous fragments from task 3 have been clarified.

• [ ] @TomkUCL @rahmanszsaleem source or identify an alternative positive control (e.g. 2F-AMP-PNP) to confirm something binding to Nsp13 in a biophysical assay without worrying about nucleotide hydrolysis.

Please feel free to add any further tasks you think are required here.

[TomkUCL]

19F spectra of 500 BioNet fluorinated fragment library have been gathered and are being processed. Some spectra lack fluorine signals, will try to identify which ones. The fragments are being divided into mixtures of 50 fragments per cocktail (2.0mM per fragment, 100.0mM total fragment concentration) using ccpnmr AnalysisScreen such that the 19F signals in each mixture do not overlap (>0.5 ppm apart).

[TomkUCL]

UPDATE 13/06/2024

I have looked through the QC spectra for the BioNET 19F fragment library to screen against Nsp13.

Here is the link to the spreadsheet 19F NMR QC.xlsx

Of these, 46 out of 521 (or 8.8%) appear to have no 19F signal, which I have flagged. We can double-check the 1H for these and run a second sample if necessary. Without these, we are left with 495 fluorinated fragments.

[TomkUCL]

Most promising positive controls for Nsp13 with proposed routes. Awaiting quotes and timeframes from CROs for synthesis.

1) Non-hydrolysable, fluorinated ATP analogue (2F-AMP-PMP):

2) Micromolar (ATPase) hits from DEL screen plus fluorinated analogues: