2.1. Observation of IL-STD NMR in a Binary Protein-Ligand Interaction: Naproxen Binding to Bovine Serum Albumin
Noticeably, our first evidence of IL-STD between adjacent subsites in a protein, rather than coming from a protein binding to two ligands, came from a single-ligand one-protein interaction: the binding of Naproxen (NPX) to bovine serum albumin (BSA). The physiological importance of serum albumins and their remarkable ability to bind endogenous and exogenous compounds have determined the usefulness of BSA as a model for the study of protein-ligand interactions [
9]. In fact, the complex of BSA with NPX, a non-steroidal anti-inflammatory (
Figure S1a), seems to be a very good benchmark system as it has been previously used for optimization of many experimental techniques, particularly ligand-observed NMR experiments [
10,
11].
We first were using the BSA/NPX system to carry out a simple STD NMR investigation into the impact of direct irradiation of a ligand signal (simultaneously with protein irradiation) on the final binding epitope mapping. To that aim, we ran two STD NMR experiments on the BSA/NPX complex, one irradiating at
= 0.60 ppm (standard selective protein saturation), and the other with
= 1.46 ppm, i.e., irradiating right at the signal of the CH
3 group at position α to the carbonyl group. The results demonstrate that direct irradiation of a ligand signal in an STD NMR experiment can have a significant impact on the determined binding epitope mapping, with ligand protons spatially close to the irradiated proton showing up to close to 80% increase in relative STD factors (i.e., in the binding epitope;
Figure S2). To make a quantitative comparison of the binding epitope mappings at the two frequencies (
,
), we define the “multi-frequency STD NMR factor” as:
where “∗” means “direct ligand irradiation” and “0” means standard conditions (selective protein saturation).
measures the relative changes in binding epitope between the two different irradiation frequencies, i.e., what changes occur in the ligand binding epitope when one of its protons is directly saturated simultaneously with the protein.
The
results represented as a bar graph in
Figure 2, left, can be considered arising from a combination of inter-molecular protein-ligand NOEs and intra-molecular ligand NOEs. While the first only takes place in the bound state, the latter occurs in both, the bound and the free state of the ligand. However, the fast chemical exchange required for STD NMR observation leads to very efficient accumulation of saturation in the free state, so that the resulting binding epitope is effectively dominated by the contribution from the bound state (
Figure S2b). In this way, irradiation of the chosen methyl signal of NPX leads to significant increases in relative STDs for protons spatially close to that methyl group in the molecule, i.e., protons Hα, a4 and a6 (
Figure 2, left).
The trend of
values along the molecule (arrows in
Figure 2, left) are globally indicative of a reduced effect of the direct irradiation of the methyl group as a function of the distance to the methyl group. Unexpectedly, we observed a disruption in that trend in the form of an increase of the
value in the protons of the methoxy group which is well remote from the irradiated methyl protons, both from a spatial point of view and from the point of view of chemical shifts (the methoxy group resonates at 3.91 ppm, while we are irradiating at 1.46 ppm). This observation is not justifiable by intra-ligand spin diffusion: indeed, the histogram in
Figure 2, left, clearly show that the
values decrease from protons a4 and a6 to protons a1 and a3, in accordance to their distances from the directly irradiated methyl group in α. Therefore, the methoxy group, which is the furthest from the directly irradiated proton should show the smallest
value. The fact that this is not true can be only explained by inter-ligand spin diffusion from an adjacently bound ligand, as explained below.
This is indeed what it is observed when inspecting the architecture of the binding pockets for NPX in BSA. Previous structural studies have demonstrated that BSA has three NPX binding sites (NPS1, 2, and 3;
Figures S1 and S3) [
12,
13]. Consequently, the observed STD signals can carry structural information on the binding of NPX to three different sites, which would make very difficult to quantitatively interpret the STD NMR binding epitope. However, all the sites are reported to show different kinetics and affinities (see
Supplementary Information and [
10,
14,
15]), so that not all the binding sites will contribute similarly to the intensities of the STD NMR signals. What is more, the observed epitope pattern (
Figure S1c) strongly supports that the major contribution indeed comes from the NPX occupying drug site 1 (NPS3 in
Figure 2, right). This is the only site in which BSA establishes close hydrophobic contacts with NPX all along the ligand molecule (
Figure S3), with both methyl groups showing the lowest amount of saturation transfer, in very good agreement with the NMR observations. In NPS3, the ligand is homogeneously buried within the protein, while the binding modes of Naproxen in NPS1 and NPS2 do not agree with the observed binding epitope (
Figure S1c).
In summary, although NPX might be binding to all the three binding sites, the STD NMR responses are mostly reporting on its binding to site NPS3, due to its most favourable fast exchange kinetics between the free and bound states. In
Figure 2, right, an expansion around sites NPS2 and NPS3 of the crystal structure of the BSA/NPX complex [
16] shows that site NPS3 is adjacent to the NPS2 site. The presence of both ligand molecules in both sites makes the methoxy group of the NPS3-bound NPX molecule account for a very short distance to the α-methyl group of the NPS2-bound NPX molecule (4.4 Å between respective carbon atoms, making all their respective protons to fall within strong NOE distance).
Accordingly, in the STD NMR experiment with direct irradiation of the α-methyl protons of NPX (
= 1.46 ppm), in addition to the intramolecular NOEs in the bound state of NPX leading to strong STD increases in protons α, a4 and a6, the protons of the α-methyl group of the NPS2-bound NPX molecule can very efficiently become extra sources of saturation to the methoxy group of NPS3-bound NPX (
Figure 2, right). This is in excellent agreement with the observed pattern of perturbed binding epitope mapping (
Figure 2 and
Figure S2). Accordingly, the observed relative increase in STD of the NPX methoxy proton does not have an intra-molecular origin, but it is the result of an inter-molecular NOE with the methyl protons of the adjacent ligand molecule in NPS2 (i.e., result from an ILOE). To test the reliability of these observations, we carried out the experiments on the BSA/NPX complex at three different magnetic fields: 500 MHz, 600 MHz and 800 MHz (
Figure S5). The inter-molecular NOE character of the observation in the context of the proposed occupancy of adjacent subsites (NPS2 and NPS3) was further confirmed by observation of an intense ILOE between the protons of the methyl and methoxy groups of NPX in a 2D NOESY experiment (
Figure S6).
In summary, this study of the single-ligand BSA/NPX system constitutes a solid proof-of-concept of the ability of the novel IL-STD NMR approach to detect inter-ligand proximity in the bound state. Importantly, our study on different magnetic fields strongly supports that this effect is independent from B0, so, if resolution allows, the method do not need the use of very high magnetic field spectrometers.
2.2. IL-STD NMR to Study Multi-Ligand Binding: Definitions
We next decided to explore the applicability of IL-STD NMR to study multi-ligand binding systems, which have high relevance for fragment-based drug discovery approaches. Based on the previous results, our hypothesis was that, in a ternary complex of a protein with two weak-affinity ligands binding to adjacent subsites (one so-called “
reporter ligand” in, let’s call it subsite-I, and the other the “
ligand of interest” in subsite-II), a particular multi-frequency STD NMR protocol can be followed to reveal inter-ligand contacts in the bound state, as sketched in
Figure 3.
The protocol involves running a total of 4xSTD NMR experiments (two pairs), each one at a distinct saturation frequency: one pair of experiments are run with selective protein irradiation (
; black cartoons in
Figure 3a,b), and the other pair with selective irradiation (
) on some frequency corresponding to a specific proton on the
reporter ligand (expected to be close to protons of the
ligand of interest). In the
experiment, saturation is simultaneously generated on protein protons, due to the typically large chemical shifts range and broad signals of a protein (red cartoons,
Figure 3c,d). At each frequency, STD NMR experiments are carried out on 2 different samples: a sample containing only the protein and our
ligand of interest in subsite-II (
Figure 3b,d; “binary complex” or “−“ system), and another sample additionally containing the
reporter ligand whose orientation in the binding pocket is known (in subsite-I;
Figure 3a,c; “ternary complex” or “+” system).
For the analysis of IL-STD NMR experiments, the 4 resulting STD NMR outcomes are termed as:
, resulting from any of the 2 experiments with on-resonance irradiation at the frequency of the
reporter ligand protons, and
, resulting from any of the 2 experiments with on-resonance irradiation solely on the protein protons. Additionally, both
and
will include subscripts to distinguish the experiments carried out either on the binary (−) or the ternary (+) complex. It is important to highlight that under δ* irradiation most of the protons of the
reporter ligand will show significant changes in saturation as a consequence of the intra-ligand spin diffusion process in the bound state, as well as the intra-ligand NOE in the free state and, as a consequence, STD intensities on protons of the irradiated ligand will be of no value. In IL-STD NMR experiments, relevant structural information will only be reported for the
ligand of interest in the ternary complex, by the values of what we now call it the Inter-ligand STD NMR factor, η(IL-STD), obtained as:
where
and
are defined as:
For the sake of simplicity, a more detailed description of this equation is not included here, but in the
Supplementary Materials (see also
Tables S2 and S3). Large positive η(IL-STD) values on the
ligand of interest will indicate protons receiving inter-ligand saturation from the irradiated protons of the
reporter ligand (excluding protons receiving intra-ligand NOE or direct irradiation), thus reporting on proximity between two regions of two bound ligands and confirming that they are occupying adjacent subsites.
As mentioned above, similar information is available from tr-NOESY experiments on ternary complexes, through the “ILOEs” (Inter-Ligand NOEs) methodology developed by Pellecchia et al. [
8]. However, the main advantage of IL-STD NMR compared to ILOE is time efficiency, a major concern for pharmaceutical companies. The long experimental times of tr-NOESY experiments and the complications arising from the acquisition of 2D spectra indeed has made ILOE the least popular among ligand based NMR methods in pharmaceutical companies in a recent poll [
16]. Thus, in order to test IL-STD NMR on a biologically and pharmaceutically relevant system, for the next part of this study we chose to prove it on the ternary complex of cholera toxin subunit B (CTB) with two adjacent lead fragments (
Figure 4), a protein-ligand system that has been approached by FBDD [
17].
2.3. IL-STD NMR Applied to Fragments Inhibitors of Cholera Toxin Subunit B (CTB)
We recently carried out a structural study on the binding of several fragments to different subsites of the GM1 binding pocket of CTB [
7]. By a combination of DEEP-STD NMR (DEEP-STD fingerprinting), STD NMR competition experiments and computational tools, we unveiled a hitherto unknown binding site adjacent to the two well-known GM1 binding subsites (i.e., the galactose and sialic acid subsites). We demonstrated that CTB is able to form a ternary complex with 3-nitrophenyl-α-D-galactopyranoside (3NPG) and a new inhibitor,
1 (
Figure 4a) [
17], as proven by the observation of an ILOE cross peak correlating proton Hc,d of 3NPG with proton Htriaz of
1 (
Figure S7) [
7]. A 3D molecular model of the ternary complex (
Figure 4b) was then obtained by superposition of the XRD structure of the CTB/3NPG complex (PDB ID: 1EEI) [
18], with a 3D structure of CTB/
1, obtained by MD simulations validated by CORCEMA-ST [
19].
We tested our proposed IL-STD NMR experiments on the 3NPG/CTB/
1 ternary complex. In this case, the orientation of 3NPG in the galactose subsite is known, and hence 3NPG was considered the
reporter ligand. To probe proximity between protons Hc,d of 3NPG and Htriaz of
1 by IL-STD NMR we then carried out the STD* experiments with irradiation of 3NPG at δ* = 7.27 ppm (Hc,d signal) and the STD
0 experiments at δ
0 = 0 ppm, to check if IL-STD NMR could reproduce the information previously obtained by ILOE [
7] (
Figure 5).
The simple visual inspection of the STD
0 and STD* spectra in
Figure 5 is already very informative. First, in the experiment with irradiation at the aromatic region (δ* = 7.27 ppm), a general decrease of the STD signals was evident, due to the lower number of irradiated protein protons, in comparison with the experiment with aliphatics irradiation. An expected increase was observed for the STD intensities of the Hc,d peak at 7.27 ppm (directly irradiated) and of the aromatic signals of
1 within 0.15 ppm high-field from it (7.21 ppm, 7.16 ppm, 7.12 ppm). Also, the intensity of the Hb proton (7.69 ppm) increased when irradiating the adjacent Hc,d, due to intra-ligand NOE. Notably, proton Htriaz (7.67 ppm) of the ligand of interest,
1, was the only signal from
1 with an increase in STD intensity, not justifiable by direct irradiation or intra-ligand NOE (blue square,
Figure 5). This inter-ligand transfer of magnetization between Hc,d and Htriazole results from their proximity (within NOE distance) in the bound state. The control experiment in the absence of 3NPG (binary CTB/
1 sample), at the bottom section of
Figure 5, clearly shows that, in the absence of 3NPG, the signal intensity of Htriaz decreases when irradiating at 7.27 ppm. This is a proof that the increase in STD intensity observed for this proton in the ternary complex, upon irradiation of the reporter ligand, 3NPG, does not arise from any artefact (e.g., neither from direct irradiation of ligand
1, nor direct irradiation of some protein protons adjacent to
1), but it is a genuine effect of inter-ligand saturation transfer taking place in the bound state.
STD NMR outcomes from the 4xSTD NMR experiments on the CTB system are shown in the top of
Figure 6, along with a schematic representation of the experimental configuration of the 4 different experiments sketched on the bottom. Phenyl protons of
1 were excluded from the analysis as direct irradiation could not be fully discarded. Interestingly, just the presence of 3NPG seems to slightly affect the binding epitope of
1, even when irradiation is selective on the protein (STD
0 values). However, the variations are small, although, noticeably, only the Htriaz of
1 seemed to increase its STD intensity just as a mere consequence of 3NPG being present in the adjacent subsite. We interpret this result as reinforcing the existence of an ILOE that is able to transfer part of the saturation at δ* irradiation from Hc of 3NPG to Htriaz of
1.
The detection and quantitation of inter-ligand STDs was carried out through the determination of the
factor, as defined previously (
Figure 3). Monitoring the
factors of the different protons of
1 clearly showed inter-ligand contacts of protons Htriaz, 4″b, 3″, and 2″ with proton Hc of the 3NPG molecule in the adjacent galactose subsite (
Figure 7, left).