Smoothening the Bumpy Road of Covalent Drug Discovery with Protein NMR

Recent years have seen a surge of interest in the use of covalent compounds in small molecule drug discovery. Much of this interest has been driven by the success of covalent compounds for “undruggable” targets such as kRAS. While it is relatively easy to buy or build a library of cysteine reactive compounds and screen them using Mass Spectrometry (MS), does this really provide all the information to decide whether a “hit” is a good starting point for a drug discovery campaign?

Typically, a covalent campaign starts with screening a library for conjugation to the target using MS. From this, it’s possible to determine if a compound reacts with the protein or not and if so, what the stoichiometry is. By applying a technique called LC-MS/MS, it is often, but certainly not always, possible to determine which amino acid residue has been modified. Then, of course, you’ll want to see if the compounds that do react have a biological affect. But where do you go from here? And what if the LC-MS/MS was not able to identify the residue that has been modified?

It's not widely recognized that NMR can be a powerful tool to characterize the interaction of covalent ligands with proteins of pharmacological interest, and yet it is! Here we show that by analyzing relatively easy to obtain, 2D NMR spectra of conjugated proteins, we can detect heterogeneity in the ligand, non-druglike interactions and whether or not there are non-covalent interactions between the two molecules. This latter information may be critical for tuning the specificity of the compound.

So first the “gotcha”. Investigating proteins by NMR most often requires replacement of the natural abundance isotopes 14N and 12C by 15N and 13C respectively. The heavier, non-radioactive isotopes allow the spectra to be “spread out” into more dimensions allowing a more detailed view into the protein at amino acid resolution. The most straight forward method to label the protein is to recombinantly express it in E. coli growing on a media containing these stable isotopes. In addition, the methods described below are most applicable to proteins up to 50-60 kDa. These two requirements do place some restrictions on the range of targets to which the technology can be applied. If this didn’t scare you, what can NMR do for your covalent complex?

Specificity of compound conjugation is always at the forefront when pursuing a covalent approach (see here for example). In most cases, specificity is achieved through non-covalent interactions combined with a warhead that reacts slowly. NMR is readily capable of differentiating a covalent compound that only interacts with the protein via the warhead from one with a more substantial interaction. Figures 1A & 1B show overlaid spectra of apo vs conjugated protein. In 1A, there are only a few peaks that shift between the two spectra indicating that the compound minimally interacts with the protein whereas in 1B, a large number of peaks shift, indicating a much more intimate interaction. The compound in 1B is the clear front-runner and should be prioritized for Hit-to-Lead efforts.

Figure 1. 2D NMR spectra of Covalent Ligand-Protein Complexes: overlaid spectra of apo (red) vs protein conjugated to a covalent ligand (blue). A) Only a few peaks shift between the two spectra indicating a minimal interaction of the compound with the protein. B) A large number of peaks shift showing a good interaction between compound and the protein. The compound is a good candidate for Hit-to-Lead campaign.

Similarly, these simple 2D experiments can also uncover another “enemy” of smooth hit-to-lead efforts: structural heterogeneity. Figure 2 shows the overlaid NMR spectrum of an apo-protein (red) and the protein conjugated to a covalent ligand (blue). Here the casual observer can see that one peak (see the upper zoom panel) in the apo-protein splits into four peaks in the conjugated version. This indicates that there are four possible states of the ligand which may arise from a variety of structural characteristics, such as isomerism, either in the isolated compound or generated upon conjugation. In general, such complexity is undesirable as it reduces non-covalent interaction with the protein. On the plus side, the experienced spectroscopist will also observe that the blue spectrum indicates that the protein remains folded in the conjugated state, clearly a desirable outcome.

These information rich experiments can be run in typically less than 2 hours making it possible to readily characterize batches of up to 20 compounds at a time. In so doing, you can be sure your covalent drug discovery program is heading down a smooth road right from the beginning.

Figure 2. 2D NMR spectrum of a Covalent Ligand-Protein Complex. Overlaid spectra of apo (red) vs protein conjugated to a covalent ligand (blue). The splitting of the single red peak into 4 blue peaks indicates that there are multiple bound states of the ligand.