The rise of NMR-integrated fragment-based drug discovery in China

Abstract

Nuclear magnetic resonance (NMR) spectroscopy is an indispensable tool to probe weak protein-ligand interactions, which are key to the hit identification and hit-to-lead evolution in fragment-based drug discovery (FBDD). The integration of NMR technology in FBDD has facilitated the development of a diverse array of candidate compounds and FDA-approved drugs. Here, we summarized the rapid advancement and application of NMR techniques in contemporary China, which serves as a catalyst for the ongoing prosperousness of fragment-derived inhibitors against various targets.

Graphical abstract

1. Introduction

Fragment-based drug discovery (FBDD) has achieved remarkable success in target-based drug discovery over the past two decades [1], with 6 fragment-derived drugs approved and dozens of candidates in the clinic [2]. Unlike high-throughput screening, the core strategy of FBDD involves screening of low molecular weight fragments (MW < 300 Da) against specific targets, enabling a more focused and efficient exploration of chemical space [3]. Upon identification of hits, these fragments are refined through mergidng, growing, or linking strategies to develop lead compounds for further investigation (Fig. 1).

Fig. 1.

The scheme of the strategy of fragment-based drug discovery. The diagram illustrates three parallel strategies—merging, growing, and linking after fragment screening.

The simplicity of these low molecular weight compounds assures a high possibility of fully matching the interactions with a target, leading to increased hit rates [4]. The high ligand efficiency [5] of these hits makes FBDD particularly effective for challenging targets with shallow and wide binding pockets, e.g., “undruggable” proteins or protein-protein interactions [6,7]. These hits are further utilized to assess the druggability of the targets [8,9] and establish rapidly and explicitly the first-round structure-activity relationship (SAR) as an essential guide for subsequent hit-to-lead optimization [10]. When crystal structures of target-hit complexes are unavailable, NMR-derived structural restraints, coupled with molecular docking or dynamics simulations, provide a robust alternative to accelerate hit-to-lead evolution with improved success rates [11].

Over the past decade, advancements in screening and structural biology techniques have led to more than 20 successful fragment-to-lead campaigns annually [12,13]. Due to its efficiency and relatively low cost, FBDD has gained global traction in both academia and industry [14,15], with multiple infrastructures being constructed for fragment-based screening. In recent years, China has emerged as a major contributor to the field of FBDD and NMR research, producing numerous noteworthy achievements. This review summarizes the key NMR techniques used in FBDD, with a particular emphasis on recent progress in China's FBDD research landscape.

2. NMR: a “gold standard” method for FBDD

Due to the inherently weak affinities of fragments, biophysical methods capable of detecting such interactions, particularly NMR, X-ray crystallography, and surface plasmon resonance (SPR), have become essential in FBDD (Fig. 2a) [16]. Among these, NMR offers unique advantages by providing atomic-level structural insights and employing a versatile set of techniques to explore various aspects of target-hit interactions. First, NMR is minimally destructive and invasive with high reproducibility. Data can be collected under near-physiological conditions with controlled variables such as temperature, pH, buffer composition, and atmospheric pressure. Second, NMR excels in structural analysis and quantification of small molecules in solution, making it indispensable for quality control in fragment libraries and fragment-derived compounds. It ensures accurate assessments of purity, stability, solubility, and concentration throughout FBDD workflows. Third, NMR is well-suited to detect interactions in a wide dynamic range from nM to mM. For weak-binding ligand in a fast exchange regime, the dissociation constant (K_D) is determined by the dose-dependent chemical shift perturbations (CSPs) of the protein, which represent the population-averaged chemical shifts in the free and bound states [17]. Although quantification of K_D from CSPs is not feasible in the intermediate to slow exchange scale, the binders may be qualitatively identified through the signal disappearance. Additionally, CSP data allow precise mapping of hit-binding topologies. Finally, NMR is a multi-valued output approach, hence the first-round ligand-observed screening is usually in a cocktail form to enhance the screening throughput with reduced spectral time and amounts of targets. Meanwhile, multiple NMR experiments can be designed to evaluate the alterations of chemical shifts, signal intensities, and even signal signs induced by the binders. The cross-validation by these NMR techniques is crucial for the filtering of false positive or false negative results of fragment-based screening. Therefore, these features collectively establish NMR as a cornerstone of FBDD since its inception decades ago.

Fig. 2.

(a) The widely recruited biophysical techniques in fragment-based drug discovery, including X-ray crystallography (XRC), NMR, SPR, mass spectrometry (MS), isothermal titration calorimetry (ITC), fluorescence polarization (FP), thermal shift assay (TSA), differential scanning fluorimetry (DSF) and microscale thermophoresis (MST). The proportions in the pie chart represent the relative usage of each biophysical method in the FBDD field, calculated by dividing the number of Web of Science publications associated with a specific method by the total number of articles referencing all commonly used biophysical methods included in the analysis. The number of NMR-related (b) or FBDD-related (c) publications from 2000 to 2023 containing China address, retrieved from Web of Science.

In contemporary China, NMR and FBDD studies are undergoing rapid growth, as evidenced by the increasing number of NMR or FBDD-related publications per year (Fig. 2b and c). NMR techniques have been nowadays widely applied to advance pharmaceutical endeavors as evidenced by numerous literatures. This trajectory strongly foresees that NMR will continue to deliver tangible benefits to more and more patients in the coming decade.

3. NMR techniques in FBDD

NMR techniques are heavily involved across different stages of FBDD including fragment library construction, hit screening and validation, structural modeling, and hit-to-lead processes (Fig. 3). In fragment-based screening, NMR techniques are typically classified into ligand-observed and target-observed screening (Fig. 4). The ligand-observed NMR screening is highly versatile, as it is sensitive to detect the ligand signals transferred from the non-isotope-labeled target without limit of molecular weight (Table 1). However, this method may produce false positive results probably due to compound or target aggregation, or false negative results when strong binding falls out of the required fast exchange scale. For instance, the saturation transfer difference (STD) experiment involves subtracting a spectrum in which the protein was selectively saturated, usually irradiating at 0 to −1 ppm, from one recorded without protein saturation. Therefore, only the signals of the binders remain in the STD spectrum, as the binders receive saturation transfer from the protein via spin diffusion through the nuclear Overhauser effect (NOE). The STD signal of the binder is accumulated through exchanging of the saturated ligand in the bound state from the unsaturated one in the free state. Therefore, the STD experiment identifies weak-binding ligands in the fast exchange time scale, usually with a K_D above 10 μM in our fragment screening campaigns [18]. The water-ligand observed via gradient spectroscopy (waterLOGSY) method also relies on NOE magnetization transfer from a water molecule to a ligand in the bound state with the targeted protein [19]. Carre-Purcelle-Meiboome-Gill (CPMG) detects the apparent transverse relaxation rate R₂, which is an averaged R₂ of the small molecule in its free and bound states. The large R₂ value of the small molecule in its bound state enabled the effective detection of the apparent R₂ changes even if the protein concentration is much lower than the ligand concentration during fragment-based screening. However, in cases of slow exchange, the CPMG experiment effectively measures the relaxation rate R₂ of the ligand in its free state, rendering it unsuitable for studying strong-binding ligands [20]. Taken together, ligand-observed NMR screening techniques are often suitable for initial fragment-based screening, and it is beneficial to pair them with other screening methods in later stages to obtain additional validation and more comprehensive insights for promising hit compounds [21]. On the other hand, the protein-observed NMR screening, despite the necessity of isotope-labeled targets and limitation on the molecular weight of the target, is the most robust approach to detect specific ligand binding in the whole affinity range from nM to mM, although quantitative determination of the K_D is only feasible for weak binders. Furthermore, when chemical shift assignments of the target are available, this approach enables precise mapping of ligand-binding sites. Together, these complementary techniques can cross-validate screening results, enhancing reliability and depth in hit identification.

Fig. 3.

NMR techniques employed across FBDD stages. STD; WaterLOGSY; CPMG; FAXS, fluorine chemical shift anisotropy and exchange for screening; HSQC, heteronuclear singular quantum correlation; HMQC, heteronuclear multiple quantum correlation; NOESY, nuclear Overhauser effect spectroscopy; PrOF NMR, protein-observed ¹⁹F NMR; DEEP-STD, differential epitope mapping by STD NMR; ILOE, inter-ligand NOEs; RD, relaxation dispersion; NMR², NMR molecular replacement; RDC, residual dipolar coupling; PRE, paramagnetic relaxation enhancement; PCS, pseudocontact shift.

Fig. 4.

Illustration of the widely used NMR methods in FBDD. The target-ligand binding was determined by either the ligand-observed or the protein-observed spectra. The STD, CPMG and waterLOGSY were the widely used ligand-observed spectra, which detected the appearance, decay or sign inversion of the binder signals, respectively. The HSQC spectra of the ¹³C or ¹⁵N-labeled target protein depicted the CSPs of resonances induced by a hit compound. Signals without significant perturbations were denoted by half red and half black circles.

Table 1.

The ligand or target-observed NMR fragment screening.

Strategy	Method	Advantages	Disadvantages
Ligand-observed screening	STD, CPMG, waterLOGSY, FAXS	High sensitivity; No molecular weight limitation of the target; No labeling required.	Higher rate of false positives or false negatives
Target-observed screening	CSP, PrOF NMR	Sensitive to whole affinity range from nM to mM; Topology of the ligand binding site.	Requires isotope labeling; Molecular weight limitation of the target; Low throughput; Demand for higher protein expression.

At the following structural modeling and hit-to-lead evolution stages, structural restraints can be retrieved from various NMR techniques, e.g., CSPs, PRE, DEEP-STD, and PCS [[22], [23], [24]]. It is particularly valuable when crystal structures are not available due to the weak binding affinities and/or aqueous solubility of the hits.

4. Case studies of NMR FBDD in China

NMR has been proven powerful in FBDD for a wide variety of targets, ranging from classical drug targets such as kinases, proteases, and receptors to more challenging targets like protein-protein interactions, as exemplified in Table 2 for cases studied in China.

Table 2.

Exemplified NMR-integrated FBDD in China.

Type of target	Target	Disease	NMR method	Ref.
enzyme	RNase L	viral infection	STD	[25]
	NDM-1	bacterial infection	1H-15N HSQC	[27]
	BACE1	alzheimer's disease	STD	[26]
reader	AF9	cancer	STD, waterLOGSY,	[37]
			CPMG, 1H-15N HSQC
	TDRD3	cancer	1H-15N HSQC, 15N relaxation dispersion	[35]
			STD, waterLOGSY,
	TRIM24	cancer	CPMG, 1H-15N HSQC	[33]
			STD, waterLOGSY,
	BRD7/9	cancer	CPMG, 1H-15N HSQC	[34]
protein-protein interaction(PPI)	MDA-9	cancer	1H-15N HSQC, PRE	[46]
	PSD-95	schemic stroke	1H-15N HSQC	[47]
	Mcl-1	cancer	1H-15N HSQC	[52]
	PDEδ/RAS	cancer	STD, CPMG	[51]
	RhoGDI2	cancer	STD, waterLOGSY, CPMG,	[50]
			1H-15N HSQC, 3D-NMR, PRE
	LARG/RhoA	cancer	STD, waterLOGSY,	[53]
			1H-15N HSQC
RNA-binding protein	TDP-43	neurodegenerative disease	STD, waterLOGSY,	[54]
			CPMG, 1H-15N HSQC

4.1. FBDD targeting enzymes

Enzymes that catalyze essential biochemical reactions in a variety of biological processes are classical drug targets for human diseases, e.g., metabolic disorders, bacterial and viral infections, and cancers. A promising example is the pseudokinase-endoribonuclease RNase L, a key player in antiviral innate immunity. Tang et al. employed STD-NMR and enzyme activity assays to identify 13 fragments with inhibitory activity [25], laying the groundwork for the development of RNase L inhibitors. Another example is β-secretase (BACE1), an enzyme implicated in Alzheimer's disease that cleaves amyloid precursor protein to generate amyloid-beta (Aβ) peptides. Utilizing STD-NMR, Fang et al. developed non-competitive BACE1 inhibitors that addressed concerns about the toxicity observed with traditional competitive inhibitors in clinical trials [26]. This approach yielded a series of conjugates exhibiting low cytotoxicity and high selectivity. Guo et al. employed target-based NMR to investigate the antibiotic hydrolase NDM-1 [27]. With the backbone chemical shifts of NDM-1 assigned by multidimensional NMR spectroscopy a priori, the residue-by-residue CSPs were retrieved from in ¹H-¹⁵N HSQC spectra of NDM-1 upon ligand titration. The ligand binding site of NDM-1 was hence mapped by CSPs and subsequently facilitated the elucidation of ligand binding modes using CSP-guided molecular docking simulations. An optimized molecule with antibacterial activity was hence discovered with potentials to overcome bacterial drug resistance.

4.2. FBDD targeting epigenetic readers

Epigenetic readers that recognize specific histone posttranslational modifications are emerging targets, particularly in anticancer treatment [28]. These readers recognize specific epigenetic markers to initiate downstream effects. Dysregulation of these processes is frequently linked to diseases, such as cancer [29,30]. Notably, small molecule drugs targeting readers of acetylation modifications, such as BET bromodomain inhibitors, have been approved in cancer treatment [31,32]. More recently, Liu et al. has identified several hits against TRIM24, a bromodomain-containing protein implicated in cancer, using ligand-based NMR screening (waterLOGSY, CPMG, ¹H, STD) [33]. Subsequent CSP analysis in the ¹⁵N-labeled HSQC spectrum of TRIM24 bromodomain facilitated the determination of the binding sites, dissociation constants, which in combination of the ligand efficiencies prioritized the hits to further optimization. Similarly, multiple new hits have been identified using NMR fragment-based screening [34]. In addition, Liu et al. uncovered 14 chemotypes of TDRD3 [35], a Tudor domain-containing protein involved in cancer cell proliferation. The structural plasticity of the TDRD3 Tudor domain revealed by their studies is key to future optimization. In the stress granules, multiple methylation sites of G3BP1 can be simultaneously bound by multiple TDRD3 molecules. Such multivalent interactions lead to the enrichment of TDRD3 in the highly concentrated condensate phase, with spatially neighboring Tudor domains. Inspired by this, the hit compound was further covalently conjugated to a bivalent inhibitor, resulting in a more effective blockade of the multivalent interactions involved in the assembly of stress granules [36]. Four hits of the AF9 YEATS domain, a reader for lysine acylation, have also been identified using both ligand-observed (waterLOGSY, CPMG, ¹H, and STD) and protein-observed NMR screening (¹H-¹⁵N HSQC) approaches, ten inhibitors with conserved pharmacophores were then designed and synthesized for further investigations [37].

4.3. FBDD targeting protein-protein interactions

Protein-protein interactions network within cells [38] regulate intricate biological processes such as signaling, cell proliferation, and apoptosis [39,40]. Unlike the well-defined pockets of classical enzyme targets, PPI interfaces are often flat and shallow, making it difficult for a strong binding of drug-like compounds to block the protein-protein interactions [41]. PPIs also lack natural ligands as a molecular blueprint [42], posing further challenges for lead design and discovery. Therefore, PPIs have long been considered as high-hanging fruit in drug discovery despite of their essential roles in pathogenesis. PPIs have recently garnered attention as promising drug targets [43,44], facilitated by the advance of FBDD. The PPI “hotspots” can be probed by low-molecular-weight fragments, which may further match the grooves at the PPI interface through fragment ligation and growth [45].

In China, researchers are actively pursuing inhibitors targeting PPIs. For instance, Tang et al. conducted NMR screening of two PDZ domains of MDA-9, guided by the observation of CSPs [46]. They successfully identified four novel hits, and the K_D of these hits, which are in the fast exchange binding regime, were determined through NMR titration. Molecular docking and PRE experiments were employed to characterize the binding conformations of two hits. Similarly, Zang et al. utilized ¹H-¹⁵N HSQC to screen against the PDZ domain of PSD-95 [47]. In another study, three hits have been uncovered as ligands of Rho guanine-nucleotide dissociation inhibitor 2 (RhoGDI2), which interacts with Rho GTPase to regulate cancer cell metastasis [48,49]. The protein-observed and ligand-observed NMR spectroscopy were all recruited in this study to characterize the binding modes of these hits, assisted by PRE and molecular docking, thus laid a solid structural basis for subsequent lead evolution [50]. Another successful case of FBDD was the discovery of PDEδ inhibitors using ligand-based NMR screening approaches, including STD and CPMG. Based on the scaffold of one of these hits, a highly active PDEδ inhibitor was ultimately designed to block the PDEδ-RAS interaction to inhibit ras-associated cancer cell proliferation and downstream RAS signaling pathway [51].

5. Summary and perspectives

Over the past two decades, FBDD has flourished and yielded significant outcomes, establishing itself as a pivotal approach in the field of small molecule drug discovery. Chinese scientists are collaboratively engaging with their global counterparts to enhance the efficacy of small molecule drugs. NMR technology, playing an indispensable role in fragment screening and identification, provides valuable structural insights into ligand-target interactions, guiding optimization from hit to lead. Its distinctive advantages have positioned NMR as one of the most preferred methods for FBDD.

Despite the widespread application of NMR technology across diverse targets in China, its predominant use is centered on fragment screening. The majority of research efforts have primarily focused on hit identification, highlighting a substantial challenge in the optimization of hits to lead. To address this challenge, the integration of NMR technology with structural biology and computational biology tools is imperative to harness their complementary strengths. This strategic amalgamation aims to propel the discovery of lead compounds, particularly for challenging targets. With continuous innovation in the field, FBDD is poised to overcome existing challenges, offering new opportunities for the discovery of novel drugs that will ultimately benefit patients.

CRediT authorship contribution statement

Zihuan Li: Writing – original draft. Lei Wang: Writing – review & editing. Jia Gao: Writing – review & editing. Ke Ruan: Writing – review & editing.

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Ke Ruan is an editorial board member of Magnetic Resonance Letters but was not involved in the editorial review or the decision to publish this article.

Biographies

Ke Ruan received the B.S. and M.S. degrees from Peking University, and Ph.D. degree from Johns Hopkins University. He is currently a professor at Division of Life sciences and Medicine, University of Science and Technology of China. His lab is focused on fragment-based anticancer drug discovery targeting protein-protein interactions, protein dynamics, as well as the molecular mechanisms and dynamic interactions of biomolecular condensates.

Jia Gao earned her Ph.D. from the University of Science and Technology of China and is currently an Associate Researcher at the same institution. Her work focuses on identifying lead compounds targeting cancer-related proteins, with an emphasis on post-translational modifications and splicing factors. She has authored over 20 articles in peer-reviewed journals, including Angewandte Chemie International Edition.