Protocol to perform fragment screening using NMR spectroscopy

Summary

Fragment-based drug design plays an important role in drug discovery. Protein-observed NMR experiments with isotopically labeled samples are used to probe target-ligand interactions and map the ligand-binding sites. Here, we present a protocol to perform fragment screening using NMR spectroscopy. We describe steps for producing ¹⁵N-labeled Kirsten rat sarcoma viral oncogene homolog (KRAS) G12D protein, fragment screening using ¹H-¹⁵N-heteronuclear single quantum coherence (HSQC) experiment, fragment deconvolution, determining binding affinities, and mapping the fragment-binding site. This protocol provides a strategy in fragment screening.

Graphical abstract

Highlights

• •Detailed protocol for producing isotopically labeled proteins
• •Procedure for fragment screening using protein-observed NMR ¹H-¹⁵N-HSQC experiment
• •Methodology for characterizing fragment hits
• •Steps for mapping fragment-binding sites on a target protein

Publisher’s note: Undertaking any experimental protocol requires adherence to local institutional guidelines for laboratory safety and ethics.

Fragment-based drug design plays an important role in drug discovery. Protein-observed NMR experiments with isotopically labeled samples are used to probe target-ligand interactions and map the ligand-binding sites. Here, we present a protocol to perform fragment screening using NMR spectroscopy. We describe steps for producing ¹⁵N-labeled Kirsten rat sarcoma viral oncogene homolog (KRAS) G12D protein, fragment screening using ¹H-¹⁵N-heteronuclear single quantum coherence (HSQC) experiment, fragment deconvolution, determining binding affinities, and mapping the fragment-binding site. This protocol provides a strategy in fragment screening.

Before you begin

Fragment-based drug design is a powerful strategy for developing potent inhibitors against diverse targets.¹ Several small molecule drugs derived from this strategy have been approved.² One of the characteristics of fragments is that the fragment compounds usually bind weakly to a target. Therefore, sensitive biophysical methods such as thermal shift assay have been frequently utilized in fragment screening. Solution NMR spectroscopy is a powerful tool to determine structures of small molecules and macromolecules such as proteins and nucleic acids. Both ligand-observed NMR experiments including, Carr-Purcell-Meiboom-Gill (CPMG), saturation transfer difference (STD) and WaterLOGSY and protein-observed NMR such as Heteronuclear single quantum coherence (HSQC) experiments have been widely used in fragment screening.³ The application of HSQC experiments in fragment screening is limited by several factors because isotopically labeled proteins are required and the targets require to have a high-quality spectrum.⁴ Despite the limitation, the target-observed experiments are very useful in fragment screening because additional information such as fragment binding site can be obtained during screening. Numerous examples show the power of HSQC experiments in fragment screening and the success encourages the application of this method to diverse targets.⁵

Kirsten rat sarcoma viral oncogene homologue (KRAS) gene is a well-studied oncogene and mutates in a series of cancers. KRAS protein contains a guanosine nucleotide-binding domain binding to guanosine diphosphate (GDP) and guanosine triphosphate (GTP) to form different states, resulting in different signals.⁶ Mutations in position 12 were observed in several cancers, making corresponding mutants such as G12C and G12D important targets for drug discovery. KRAS protein was considered as a “undruggable” target due to the lack of a suitable pocket for binding to drug-like small molecules. To overcome such challenge, different strategies were applied in developing KRAS inhibitors. In addition to improving the potency through forming covalent interactions with the cysteine residue in KRAS G12C mutant, fragment-based drug design also results in potent small molecule inhibitors of KRAS G12D.⁷ NMR-based methods were employed in fragment screening against KRAS.⁸ As the yield of recombinant KRAS protein is high, with approximately 10 mg of ¹⁵N-labeled KRAS produced per liter bacterial culture, HSQC-type experiments can be applied to identify fragments for further development. In this protocol, the procedures involved in fragment screening using ¹H-¹⁵N-HSQC experiment are described.

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Bacterial and virus strains
Escherichia coli BL21 (DE3)	Agilent	230132
Chemicals, peptides, and recombinant proteins
Dithiothreitol (DTT)	GoldBio	DTT25
β-mercaptoethanol	Sigma-Aldrich	M6250
Dimethyl sulfoxide (DMSO)	Sigma-Aldrich	D2650
Imidazole	Sigma-Aldrich	56750
Glycerol	Sigma-Aldrich	G5516
Na2HPO4	Sigma-Aldrich	S5136
NaH2PO4	Sigma-Aldrich	S9638
Magnesium sulfate anhydrous	Sigma-Aldrich	63136
Kanamycin sulphate	Sigma-Aldrich	K1377
Magnesium chloride	Sigma-Aldrich	M0250
GDP	Sigma-Aldrich	G7127
Calcium chloride 2-hydrate	VWR	VWRC22317.260
Thiamine hydrochloride	Sigma-Aldrich	T4625
HEPES	1st BASE	BUF-1821
Tris-HCl	1st BASE	BUF-1415
Ethylenediaminetetraacetic acid (EDTA)	1st BASE	BUF-1053
NaCl	Merck	1.06404
15NH4Cl	Cambridge Isotope Laboratory	NLM-467
13C-glucose	Cambridge Isotope Laboratory	CLM-1396
Deuterium oxide (99%)	Cambridge Isotope Laboratory	DLM-4-99
Dimethyl sulfoxide-d6	Cambridge Isotope Laboratory	CDLM-10-50
Fragment library	Enamine	N/A
Ni2+-NTA resin	QIAGEN	30230
HiPrep 16/60 Sephacryl S-200 HR	GE Healthcare	17-1166-01
PD-10 desalting columns	GE Healthcare	17-0851-01
Precision Plus Protein Dual Color Standards	Bio-Rad	1610374
NuPAGE MES SDS running buffer	Invitrogen	NP0002-02
NuPAGE NOVEX Bis-Tris gels	Invitrogen	NP0323BOX
AmiconR Ultra-15, 3 kDa (MWCO)	Merck	UFC900396
NMR tubes	NORELL	S-30HT-7
Long tips	NORELL	NP204
Microtubes	Corning Life Sciences	MCT-150-C
Round-bottom tube	SPL Life Sciences	40014
Conical tube	SPL Life Sciences	50050
Recombinant DNA
pET28a-KRAS G12D	This study	N/A
Software and algorithms
Topspin 2.1	Bruker	N/A
NMRPipe		https://www.ibbr.umd.edu/nmrpipe
NMRviewJ		https://nmrfx.org/nmrfx/nmrviewj
Protein Calculator		https://protcalc.sourceforge.net/
Other
Sonicator	Qsonica	Misonix Q700
Centrifuge	Beckman Coulter	Avanti JXN-26
Centrifuge	Beckman Coulter	Avanti J-25I
Centrifuge	Eppendorf	5810R
Incubator	Infors	Multitron standard
ÄKTA Pure protein purification system	GE Healthcare	ÄKTA Pure

Materials and equipment

Prepare a 10 × salt solution for making minimal medium by weighing the listed items, dissolve in 1 L Milli-Q water and autoclave at 121°C for 30 min. Store the solution at room temperature for 6 months.

Reagent	Final concentration	Amount /L
Na2HPO4	422.6 mM	60 g
KH2PO4	220 mM	30 g
NaCl	85.5 mM	5 g

Prepare the following 1000 × stock solutions in 100 mL Milli-Q water. Filter these solution with a 0.2 μM filter and store at room temperature for 6 months, except that IPTG, DTT and kanamycin should be stored at −20°C for 6 months.

Reagent	Final concentration	Amount
MgSO4	1000 mM	12 g
CaCl2.2H2O	100 mM	1.47 g
Isopropyl β-d-1-thiogalactopyranoside (IPTG)	1000 mM	23.8 g
DTT	1000 mM	15.4 g
Thiamine HCl	297 mM	10 g
Kanamycin	51.5 mM	3 g

Prepare 10 mL of the following solution with carbon and nitrogen sources and filter through a 0.2 μm filter. Store at 4°C for up to four weeks.

Reagent	Final concentration	Amount
15NH4Cl	1840 mM	1 g
Glucose	1110 mM	2 g

Prepare 1 L minimal medium in an autoclaved flask for growing bacterial cells by adding the following components into 876 mL Milli-Q water. Store at 4°C for up to two weeks.

Stock solution	Final concentration	Amount
10 × salt solution		100 mL
1000 ×MgSO4	1 mM	1 mL
1000 ×CaCl2	0.1 mM	1 mL
1000 ×Thiamine HCl	0.3 mM	1 mL
1000 ×Kanamycin	0.0515 mM	1 mL
100 ×15NH4Cl	18.4 mM	10 mL
100 × Glucose	11.1 mM	10 mL

The following buffers are used in protein purification. Resuspension buffer contains 20 mM sodium phosphate, pH 7.8, 500 mM NaCl and 2 mM β-mercaptoethanol. Add the following components to 800 mL Milli-Q water. (Note: add fresh β-mercaptoethanol before the experiment). Washing buffer contains 20 mM sodium phosphate, pH 7.2, 1 M NaCl, 20 mM imidazole and 2 mM β-mercaptoethanol. Add the following components to 780 mL Milli-Q water. (Note: add fresh β-mercaptoethanol before the experiment). Elution buffer contains 20 mM sodium phosphate, pH 6.5, 500 mM NaCl, 500 mM imidazole and 2 mM β-mercaptoethanol. Add the following components to 300 mL Milli-Q water. (Note: add fresh β-mercaptoethanol before the experiment). Gel filtration buffer for protein NMR study contains 20 mM sodium phosphate, pH 7.2, 150 mM NaCl, 4 mM MgCl₂, 5 mM GDP and 1 mM DTT. Add the following components into 940 mL Milli-Q water and degas using a vacuum pump before gel filtration chromatography.

Buffers used in protein purification

Prepare resuspension buffer, filter using a 0.22 μm filter and degas for 10 min. Store at 25°C for up to six weeks.

Note: Add fresh β-mercaptoethanol before experiment.

Reagent	Concentration	Amount
0.2 M Sodium phosphate (pH 7.8)	20 mM	100 mL
5 M NaCl	500 mM	100 mL
14.3 M β-mercaptoethanol	2 mM	140 μL

Prepare washing buffer, filter using a 0.22 μm filter and degas for 10 min. Store at 25°C for up to two weeks.

Note: Add fresh β-mercaptoethanol before experiment.

Reagent	Concentration	Amount
0.2 M Sodium phosphate (pH 7.2)	20 mM	100 mL
5 M NaCl	500 mM	100 mL
14.3 M β-mercaptoethanol	2 mM	140 μL
1 M Imidazole	20 mM	20 mL

Prepare elution buffer, filter using a 0.22 μm filter and degas for 10 min. Store at 25°C for up to two weeks.

Note: Add fresh β-mercaptoethanol before experiment.

Reagent	Concentration	Amount
0.2 M Sodium phosphate (pH 6.5)	20 mM	100 mL
1 M Imidazole (pH 6.5)	500 mM	500 mL
5 M NaCl	500 mM	100 mL
14.3 β-mercaptoethanol	2 mM	140 μL

Prepare gel filtration buffer, filter using a 0.22 μm filter and degas for 10 min. Store at 4°C for up to two weeks.

Note: Add DTT into the buffer before experiment.

Reagent	Concentration	Amount
1 M HEPES (pH 7.3)	20 mM	20 mL
5 M NaCl	150 mM	30 mL
1 M MgCl2	4 mM	4 mL
1 M GDP	5 mM	5 mL
1 M DTT	1 mM	1 mL

Step-by-step method details

Step: Production of ¹⁵N-labeled KRAS G12D bound to GDP

Recombinant KRAS G12D forms complexes with GDP and GTP due to their strong binding affinities. The process involves purifying ¹⁵N-labeled KRAS G12D from bacterial cells and replacing the endogenous GDP or GTP with unlabeled GDP by adding EDTA and then exchanging KRAS-G12D in a buffer containing 5 mM GDP and 4 mM MgCl₂.

Note: The following steps are for purifying protein from a one-liter culture. Additional time will be needed if a larger quantity of protein is required for the experiment.

Duration: 4–5 days.

• 1.The codon optimized cDNA encoding human KRAS G12D (residues 1–169) was synthesized by GenScript and cloned into a pET28a vector using the NdeI and XhoI sites.

Note: The resulting plasmid encodes a recombinant protein with a His tag, a thrombin cleavage site and residues 1–169 of KRAS G12D.

• 2.Mix 1 μL of the expression plasmid encoding KRAS G12D fused with an His tag with 50 μL of BL21(DE3) competent cells (Agilent Technologies) in a 1.5 mL Eppendorf tube and put it on ice for 30 min.
• 3.Heat the mixture at 42°C for 45 s and put the mixture on ice for 1 min.
• 4.Add 200 μL of LB medium and leave the tube at 37°C for 1 h with shaking at 200 rpm.
• 5.Plate the cells onto Luria-Bertani (LB) agar plates supplemented with 30 μg/mL kanamycin and keep the plate at 37°C for 12–17 h.

Pause point: The plate can be stored at 4°C for up to 1 week.

• 6.
  Make 1 L M9 medium in which 1g ¹⁵N-labeled NH₄Cl was used to replace the normal NH₄Cl.

  • a.
    Take 50 mL of medium for seed preparation.
  • b.
    Pick up one to two colonies and inoculate into 50 mL M9 media supplemented with 30 μg/mL kanamycin.
  • c.
    Incubate the cells in a bacterial incubator at 37°C and 200 rpm for 14–18 h.
• 7.Transfer the 50 mL culture to the remaining 950 mL M9 supplemented with 30 μg/mL kanamycin. The culture is kept in a shaker incubator at 37°C and 200 rpm.
• 8.Monitor the absorbance of the culture at 600 nm during incubation.
• 9.When the absorbance at 600 nm reaches 0.6, lower the temperature of the incubator to 18°C and shaking speed to 100 rpm for approximately 30 min, until absorbance at 600 nm reaches 0.7–0.8.
• 10.Add 1 M IPTG to the culture to a final concentration of 1 mM. Protein expression was induced at 18°C and 200 rpm for 16 h.
• 11.Harvest cells by centrifugation at 9,000 ×g for 10 min at 4°C.
• 12.
  Freeze cells for later purification.

  • a.
    Resuspend the 2 g (wet weight) of cells in 40 mL resuspension buffer.
  • b.
    Freeze the cells in liquid nitrogen.
  • c.
    Store the cells at −80°C for long-term storage.

Pause point: The pellet in the buffer can be kept at −80°C or below for up to several months.

• 13.Thaw the cell pellet from −80°C on ice and lyse the cells using sonication in an ice bath for 30 min at 30% power with intervals of 5 s on and 5 s off.
• 14.Centrifuge the cell lysate at 20,000 ×g for 15 min at 4°C. Transfer the supernatant to a 50 mL tube.
• 15.Transfer 4 mL Ni²⁺-NTA resin into a gravity column and wash the resin with 40 mL resuspension buffer.
• 16.Load the supernatant onto the resin and let the solution pass through the resin through gravity. Allow all supernatant to flow through the resin.
• 17.Wash the resin with 40 mL resuspension buffer, followed by 40 mL washing buffer to remove impurities.
• 18.Elute the recombinant protein with 20 mL elution buffer and collect the fractions with 5 mL in each tube.
• 19.Analyze 10 μL sample of each fraction by polyacrylamide gel electrophoresis (PAGE).

Note: Once recombinant protein was identified in the gel, the corresponding fractions can be collected for further purification.

• 20.
  Prepare GDP bound KRAS G12D.

  • a.
    Add EDTA to 10 mM to the protein solution to remove the metal ions in the protein sample and reduce protein’s interactions with endogenous GDP or GTP.

    • i.
      Keep the sample on ice for 1 h.
    • ii.
      Change the buffer to gel filtration buffer using a PD10 column.

      Note: This step will remove endogenous GDP or GTP, metal ions and include new GDP and Mg²⁺ as the buffer contains 5 mM GDP and 4 mM MgCl₂ to produce GDP bound protein.

  • b.
    Pre-equilibrate a HiPrep 16/60 Sephacryl S-200 HR column with two-column volumes of gel filtration buffer with a flow rate of 0.5 mL/min on the ÄKTA pure system.
• 21.Load 5 mL sample into the sample loop and inject the sample into the column. Record absorbance at 280 nm and start collecting the fractions from 30 mL to 100 mL after injection. Analyze the fractions using SDS-PAGE.
• 22.Combine the fractions from a single peak in the chromatography and concentrate using a 3 kDa cutoff concentrator.
• 23.Concentrate the protein to 0.3 mM concentration and flash frozen in liquid nitrogen and stored at −80°C for long-term storage.
• 24.Protein concentration can be determined by measuring its absorbance at 280 nm using a Nanodrop spectrophotometer.

Note: The extinction coefficient of KRAS G12D at 280 is estimated to be 10.36 mM^–1 cm^–1, which is obtained from Protein Calculator webserver.

Pause point: The sample can be stored at −80°C for up to 6 months.

Step: Sample quality analysis by NMR spectroscopy

The quality of the purified sample must be evaluated using NMR spectroscopy. It is essential to confirm that the purified protein is properly folded in solution, which can be determined through 1D proton and 2D ¹H-¹⁵N-HSQC NMR experiments.

Duration: 1–2 h.

• 25.
  To test the quality of the sample by NMR, the following steps are used.

  • a.
    Add 10% D₂O to the protein sample (200 μL for 3 mm NMR tube and 550 μL for 5 mm NMR tube).
  • b.
    Transfer the sample into an NMR tube using a pipette.
  • c.
    load the sample tube into the NMR magnet.
• 26.
  Set up experiment for data acquisition.

  • a.
    Run topspin by clicking the program icon.
  • b.
    Set up the data acquisition temperature to 298K such as using command “edte”.
  • c.
    Equilibrate the sample in the magnet for 10–15 min.
• 27.To start the experiment, generate a new dataset using the command “edc” and execute command “rpar ZGPR to load some basic experimental parameters. Change the pulse program to “zggpw5”.
• 28.
  Prepare for data acquisition.

  • a.
    Start performing lock by running “lock” command, choose solvent-H₂O + D₂O.
  • b.
    Perform tuning and matching, shim using corresponding commands “atma”.
  • c.
    Run “topshim” to perform shimming.

Note: The command names may be different in different systems.

• 29.
  Calibrate the 90° hard pulse for ¹H and set up other parameters.

  • a.
    Run the command “pulsecal”.
  • b.
    Adjust the o1p to 4.7 ppm and this value can be slightly modified using “gs” mode.
  • c.
    Set up the number of scans (ns) to 64 by typing in “ns = 64”. Update the parameters needed in the program.
  • d.
    Adjust receiver gain though running command “rga”.
• 30.
  Start 1D data collection and processing.

  • a.
    Run command “zg” to start data collection.
  • b.
    Process the data using command “efp”.
  • c.
    Adjust phase manually or using command “apk”.
• 31.Zoom in at interested regions of the spectrum and save the spectrum in required format (Figure 1).
• 32.
  Set up an experiment for the 2D experiment.

  • a.
    Use command “edc” to generate a new experiment.
  • b.
    Load pulse sequence for 2D ¹H-¹⁵N-HSQC experiment using command “rpar hsqcetf3gps”.
  • c.
    Update the parameters using commands “getprosol”, and “rga”.
• 33.Set up other parameters such as spectral width (36 ppm), time domain points (2k × 128), carrier frequency (o3p = 118 ppm) and number of scans (ns = 4).
• 34.
  Data acquisition and processing for the 2D experiment.

  • a.
    Acquire the 2D ¹H-¹⁵N-HSQC data by typing command “zg”.
  • b.
    Process the data using command “xfb”.
  • c.
    Adjust phases at two dimensions using the phase correction tab.
• 35.Export the spectrum using the export function in Topspin (Figure 2).

CRITICAL: Please follow the instructions of the NMR facility to make sure that the corresponding important parameters including pulse widths and power levels of different channels are up to date. Before starting the screening experiment, the 2D ¹H-¹⁵N-HSQC experiment of the purified protein must be collected and the cross peaks in the spectrum must be observed.

Figure 1.

The ¹H NMR spectrum of purified KRAS G12D

The dispersion of the peaks demonstrated the folding of protein in solution. The wide dispersion of amide protons at the range of 6–10 ppm and appearance of chemical shifts of methyl protons at −0.5-0 ppm demonstrate the folding of the protein in solution.

Figure 2.

2D ¹H-¹⁵N-HSQC spectrum of GDP bound KRAS G12D

The spectrum was collected on a Bruker 600 MHz at 25°C. The data was processed and visualized. The amide proton and amide cross peaks were observed in the spectrum. The number of the cross peaks is close to the non-proline residue number of KRAS G12D construct. Over 155 out of 165 non-proline cross peaks were observed in the spectrum, suggesting the folding of the protein in solution. The missing peaks may be due to signal overlap or protein dynamics. The quality of the spectrum suggests that protein can be used for further experiments.

Prerequisite: Topspin (Bruker), NMRPipe,⁹ NMRViewJ¹⁰ or other related software on a computer for data processing.

Step: Fragment screening and hit identification

Fragment screening can commence once the protein and fragment library are prepared. The screening process involves several steps, beginning with the use of fragment mixtures to save both experimental time and cost.

Duration: 6–12 days, depending on the library size the sensitivity of the experiments.

• 36.Combine 10 fragments at a concentration of 100 mM to yield a stock mixture containing 10 mM of each fragment. A sample with equal amount of DMSO is used to collect the reference spectrum.
• 37.DMSO and fragment stock mixtures are added into the protein sample and keep at room temperature for approximately 10 min, respectively. Transfer the samples to NMR tubes, respectively.

Note: Shigemi tubes can be used in this stage to save samples.The final sample contains 0.2 mM of KRAS and 1 mM of each fragment.

• 38.Load the NMR tube into the probe. Adjust the parameters for data acquisition.
• 39.Collect the 2D ¹H-¹⁵N-HSQC spectra of KRAS G12D in the presence of different fragment mixtures and process the data.
• 40.Overlay the spectra of KRAS in the presence DMSO and fragment mixtures, respectively. Only the mixtures that can cause chemical shift perturbations or line broadening are considered as positive hits for hit deconvolution in the next step.
• 41.
  Fragment deconvolution to identify fragment hits.

  • a.
    Mix protein with individual fragment in the positive mixtures.
  • b.
    Collect the 2D ¹H-¹⁵N-HSQC spectra, respectively.
  • c.
    Select hit fragments that can induce chemical shift perturbation.
  • d.
    Record the hit information for titration experiment in the next steps (Figure 3).

Figure 3.

Superimposed 2D ¹H-¹⁵N-HSQC spectra of KRAS G12D in the presence of DMSO (black) and a mixture of fragment (red)

Chemical shift perturbation was observed upon addition of the mixture. Deconvolution experiment will be carried out to identify the fragment binding to KRAS G12D.

Step: Titration experiment to determine binding affinity

Timing: 1–5 days depending on the number of hits identified

To determine the binding affinity of the hits, titration experiments can be conducted. Since fragment screening is performed using NMR experiments, the binding affinity of a fragment can be measured using the same technique. The resulting affinity values will be used to rank the identified fragment hits.

• 42.Collect the 2D ¹H-¹⁵N-HSQC spectra of KRAS G12D in the absence and presence of different amounts of fragments. Use the following steps to process the data and determine the binding affinity.
• 43.Type the command “bruker” and read parameters to generate a file named as fid.com. Make sure that NMRPipe is installed in the computer.
• 44.Run the file “fid.com” and process the data to generate a file that can be read by NMRviewJ.

Note: The names of the files can be names as frag_1.nv, frag_2.nv and etc.

• 45.Launch the NMRviewJ by typing the icon on the Desktop or open the data using other software, which depends on the individual user. Read these files that are the spectra of the protein with different amounts of fragments (Figure 4).
• 46.Open the dataset table to view the opened files. Select all the files and save .par files by clicking on the WritePar section to generate a series of .par files which are correlated with the “.nv” files.
• 47.
  Process the parameter files using the following steps.

  • a.
    Open these .par files, respectively.
  • b.
    Include the following line “property ligcon” (concentration of ligand) to the bottom of the file.
  • c.
    Input the concentration of the ligand in the line for the corresponding HSQC experiment.

Note: The unit of the concentration does not need to be included. The value in most cases is the mM of the fragment in the NMR tube. For example, for the spectrum of KRAS was recorded in the present of 0.5 mM fragment, the following line “property ligcon 0.5” can be included in the .par file.

• 48.Save these .par files.
• 49.Go back to NMRviewJ program and read par files in the Dataset Table section to update the ligand information for titration analysis.
• 50.Select all the files in the Dataset Table and click the “Draw” tab to plot overlaid spectra by defining a window with 1 row and 1 column and named with a name such as titration.

Note: The color of the spectra can be changed or set to make sure all the spectra can be differentiated. Zoom in the spectra and adjust the noise level to make sure the spectra can be seen properly.

• 51.Turn on the Attributes of the window with all the titration data. Click the PeakPick tab and select the peak picking.
• 52.Initiate Titration Analysis by starting it from the Analysis tab.

Note: Under the setup tab, input the information of the dataset and protein concentration. The unit of the protein concentration input here must be the same as those of fragments in the .par files. Select one of the peak list that was saved in the previous step. Click the Use Data Props tab to start the application.

• 53.
  Determine binding affinity for a select residue.

  • a.
    Select a peak that exhibited chemical shift perturbations upon addition of different amounts of fragments.
  • b.
    Click on the Measure tab to show the path of fragment induced chemical shift changes of a cross peak.
  • c.
    Click the Fit tab to generate the Kd value.
  • d.
    Repeat this step for more residues to get an averaged Kd value.

Figure 4.

Superimposed 2D ¹H-¹⁵N-HSQC spectra of KRAS in the presence of different amounts of fragment in a titration experiment

The concentration ratios of protein to fragment are shown in different color. Residues exhibited concentration dependent CSPs are labeled with residue name and sequence number.

Step: Map the fragment binding site

Determining the fragment binding site is crucial for selecting hits for further development and understanding their mechanisms of action. The following steps will be employed to identify fragment binding sites on KRAS.

Duration: 1–2 h.

CRITICAL: The assignment of the 2D ¹H-¹⁵N-HSQC spectrum of KRAS G12D can be obtained from the database (BMRB). If the assignment of the target protein is not available, backbone assignment needs to be performed. The protocol for assignment has been reported, which requires a series of 3D experiments.^11,12

• 54.
  Generate a peak list for the ¹H-¹⁵N-HSQC spectrum of free KRAS G12D.

  • a.
    Open the HSQC spectra of free KRAS G12D in NMRviewJ.
  • b.
    Do peak picking for the spectrum of free protein and assign the peak based on the known information.

    Note: If a peak list file is available, change the peak file into the format that can be read by NMRviewJ.

  • c.
    Read the peak list and adjust the peak positions to the center of the cross peaks.
  • d.
    Write the updated peak list of free protein using command “writepks peak1 newpeak1 PL”.
• 55.
  Generate a peak list for the ¹H-¹⁵N-HSQC spectrum of fragment bound-KRAS G12D.

  • a.
    Copy the peak of the free protein into a new name and read this peak list.
  • b.
    Open the spectrum of fragment bound KRAS G12D.
  • c.
    Load the peak onto the spectrum.
  • d.
    Adjust the peak positions to the center of the cross peaks.
  • e.
    Write the updated peak list of the complex using the command “writepks peak2 newpeak2 PL”.
• 56.
  Calculate chemical shift perturbations using the obtained peak lists.

  • a.
    Read the saved peak lists using Microsoft Excel to calculate ligand induced chemical shift perturbations.
  • b.
    Calculate the averaged chemical shift perturbations (Δδ) using the following formula-sqrt (0.5∗(δ_H∗δ_H + (α∗δ_N∗)∗(α∗δ_N∗)), where δ_H and δ_N are the chemical shift changes upon addition of the fragment in ppm for the ¹H and ¹⁵N dimensions, respectively. The value of the coefficient in this study is 0.2.
  • c.
    Make a histogram plot when needed using the plot function in Excel.
  • d.
    Rank the residues based on the values of ligand induced chemical shift perturbations.
• 57.Ligand efficiency (LE) can be determined using the following equation: LE=(-2.303RT)/N∗log(Kd).

Note: R is the ideal gas constant, T is the temperature in Kelvin and N is the number of non-hydrogen atoms in the fragment.

CRITICAL: Backbone resonance assignment of KRAS G12D can be obtained from Biological Magnetic Resonance Bank (BMRB): 27719.¹³ Other software such as SPARKY and CCPNmr can also be used to do resonance assignment, measure binding affnity, and analyze ligand binding sites.

• 58.
  Load the protein structure for mapping fragment binding sites.

  • a.
    Launch PyMOL on a computer.
  • b.
    Read the structure file of a protein. In this case, structure of KRAS is exported using command “fetch 3GFT”.
  • c.
    Show the structure in cartoon mode using command “show cartoon” or in surface mode using command “show surface”.
  • d.
    Set the color of the structure in a preferred color such as white.
• 59.Determine the perturbed amino acids whose chemical shift perturbations above the threshold in the histogram. Highlight the selected residues in a color such as orange in the figure (Figure 5).
• 60.Label the residues on the structure.
• 61.Export the figures in the desired formats such as .png format. Present the figure after modification using software such as Microsoft Office (Figure 5).

CRITICAL: The structure of a protein can be shown in different modes in PyMOL. The output format can be altered based on the experience of different users. Please refer to the manual of PyMOL for more information. The figures exported from PyMOL can be labeled using other applications such as Microsoft PowerPoint. Researchers can determine which structures are used for presentation when many structures are available. It can be a structure determined using experimental methods or homology model.

• 62.Rank the fragment hits based on LE values and binding affinities.

Note: Its inhibitory effect on enzymatic activity is low due to the low binding affinities. Group the fragments into different classes based on the structural similarities when possible.

• 63.Select satisfied hits for further growth.

Note: This can be achieved through different strategies such as structure-based drug design and searching homologues from drug-like libraries.

Figure 5.

Map fragment binding site based on the titration result

(A) The structure of KRAS is shown. The structure of a mutant KRAS in complex with Mg²⁺ and a GTP analogue (PDB: 3GFT) is used. The structure of the GTP analogue is labled as GNP and shown insticks mode. The metal ion is shown as a green sphere. Residues exhibiting chemical shift perturbations are shown as brown spheres.

(B) The surface representation of KRAS. The orientation of the structure is same as (A).

Expected outcomes

Solution NMR spectroscopy is a powerful tool to identify fragments binding weakly to a target. It can also be applied to determine the ligand binding modes such as binding affinity, binding specificity and the ligand binding site when the assignment of the ¹H-¹⁵N HSQC spectrum of a protein is available. This protocol was applied to perform fragment screening against KRAS G12D with aims to develop small molecule inhibitors. This protocol is also applicable to other proteins.

Limitations

Although the ¹H-¹⁵N HSQC experiment is a powerful strategy to perform fragment screening, determining binding affinity and mapping ligand binding site, isotopically labeled proteins are required in the experiment. Therefore, the yield of the target protein needs to be high to reduce experimental cost. In addition, for a protein with a molecular weight more than 30 kDa, the ¹H-¹⁵N HSQC spectrum may be crowded. More complicated sample labeling strategies such as deuteration or site-specific labeling is required. Ligand-observed NMR screening methods should be considered when cost of the target protein preparation is the limitation factor in screening. In addition, if the protein is stable, recycling of the target protein can be considered, which will reduce the cost in protein production. Nonetheless, this is a highly recommended strategy in fragment screening. With the application of sensitive probes , less sample can be used in data acquisition. This method should be considered in fragment-based drug design.

Troubleshooting

Problem 1

Transformation efficiency of the target plasmid is low, resulting in no colonies on the plate (steps 2–5).

Potential solution

• •Increase the amount of the plasmid during transformation or increase the concentration of the plasmid.
• •Incubate for a longer period (2 h) after heat shock and addition of LB medium.

Problem 2

No target protein induction observed (steps 6–12).

Potential solution

• •Choose different bacterial cells for protein expression.
• •Perform a small-scale induction first before carrying out a large-scale protein production.
• •For producing human proteins in bacteria, codon optimization is required as rare codons may affect the expression of the target protein in bacteria.

Problem 3

The yield of the ¹⁵N-labeled protein is low (steps 6–12).

Potential solution

• •Optimize the concentration of IPTG during the induction stage.
• •Optimize the concentration of IPTG to get an improved yield.
• •Make sure the absorbance at 600 nm is not above 1.0 when IPTG is added as higher value suggests that lower carbon and nitrogen sources remain in the medium.
• •The induction temperature needs to be optimized. Higher temperature may result in higher growth rate while protein may aggregate when it is induced under this condition.
• •In addition, leaky expression of some toxic proteins can result in low protein induction. Testing protein expression in different types of bacterial cells is needed before initiating a large-scale protein purification.

Problem 4

Protein is not stable during purification (steps 13–24).

Potential solution

• •Optimize the buffers such as pH and salt concentration used in protein purification.
• •Include reducing agents such as DTT and other reagents such as glycerol to make protein more stable during purification.
• •Add 0.1% sodium azide in the gel filtration buffer to improve the sample stability.
• •Try to perform the purification experiment in a cold room to avoid sample aggregation/degradation due to the high temperature of the environment.

Problem 5

The GDP is not fully exchanged in the protein sample, resulting in a different HSQC spectrum from the published ones (steps 20–23).

Potential solution

Due to the tight binding between the GTP/GDP and KRAS proteins, the endogenous GTP or GDP may not be completely exchanged to GDP. To improve the exchange efficiency, alkaline phosphatase can be added into the protein solution to remove the GDP/GTP bound to KRAS, then repeat the following experiments (steps 20–23).

Problem 6

The signal intensity in the proton NMR spectrum is lower than expected (steps 25–30).

Potential solution

It depends on the nature of the target protein. If the protein forms oligomers or has a high molecular weight (e.g., >30 kDa), consider using more scans or using a deuterated sample to improve signals. Site-specific labeling strategy can also be considered. For a monomeric protein, considering using a protein sample with a higher concentration for data acquisition or increase number of scans. The 90° proton hard pulse needs to be adjusted to gain optimal signal-to-noise. Check if water signal is suppressed efficiently.

Problem 7

The quality of the ¹H-¹⁵N HSQC spectrum is different from expected (steps 31–32).

Potential solution

• •Increase the concentration of sample or test a different pulse program.
• •Make sure that the “tune and match” step is performed, and the hard pulses of different channels are calibrated.
• •Use more scans or different pulse programs such as fast HSQC to increase the signal.
• •The presence of magnesium may also affect the spectrum, using low concentrations of magnesium to improve the signal-to-noise of the spectrum.

Problem 8

No positive mixtures are identified or too many mixtures induce chemical shift changes (steps 25–30).

Potential solution

• •Increase the concentration of fragment mixtures to increase ligand to protein ratios during screening if no hit is identified.
• •Use a library with different fragment structures for screening if it is possible. When hit rate is high, consider increasing the threshold of chemical shift perturbations and only choose the hits binding to the preferred binding site.
• •Determine binding affinities and choosing the hits that bind specifically to a target and with a higher ligand efficiency are important to solve the problem.

Problem 9

The assignment of the protein of interest is not available because no assignment is deposited in BMRB or no publication is available (steps 54–56).

Potential solution

For screening, assignment is not necessary as hits still can be identified based on changes of the spectra. It is important to have assignment of the spectrum to map the binding site. Assignment of the spectrum can be obtained through backbone resonance assignment through additional experiments.^11,12 In addition, some other strategies such as site-directed mutation, competition experiment can be considered to determine the ligand binding site.^14,15

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to the lead contact, CongBao Kang (kang_congbao@eddc.sg).

Technical contact

For all technical questions, please contact Qiwei Huang (huang_qiwei@eddc.sg) and CongBao Kang (kang_congbao@eddc.sg).

Materials availability

This study did not generate new unique reagents.

Data and code availability

The data supporting the current study are available from the corresponding author on request.